Autoware is the world\u2019s leading open-source software project for autonomous driving. Autoware is built on Robot Operating System (ROS) and enables commercial deployment of autonomous driving in a broad range of vehicles and applications.
This page is a collection of the links to the competitions that are related to the Autoware Foundation.
Title Status Description Ongoing Autoware / TIER IV Challenge 2023 Date: May 15, 2023 - Nov. 1st, 2023 As one of the main contributors of Autoware, TIER IV has been facing many difficult challenges through development, and TIER IV would like to sponsor a challenge to solve such engineering challenges. Any researchers, students, individuals or organizations are welcome to participate and submit their solution to any of the challenges we propose. Ongoing Japan Automotive AI Challenge 2023 Registration: June 5, 2023 - July 14, 2023 Qualifiers: July 3, 2023 - Aug. 31, 2023 Finals: Nov. 12, 2023 In this competition, we focus on challenging tasks posed by autonomous driving in factory environments and aim to develop Autoware-based AD software that can overcome them. The qualifiers use the digital twin autonomous driving simulator AWSIM to complete specific tasks within a virtual environment. Teams that make it to the finals have the opportunity to run their software on actual vehicles in a test course in Japan."},{"location":"autoware-competitions/#proposing-new-competition","title":"Proposing New Competition","text":"
If you want add a new competition to this page, please propose it in a TSC meeting and get confirmation from the AWF.
Thank you for your interest in contributing! Autoware is supported by people like you, and all types and sizes of contribution are welcome.
As a contributor, here are the guidelines that we would like you to follow for Autoware and its associated repositories.
Code of Conduct
What should I know before I get started?
Autoware concepts
Contributing to open source projects
How can I get help?
How can I contribute?
Participate in discussions
Join a working group
Report bugs
Make a pull request
Like Autoware itself, these guidelines are being actively developed and suggestions for improvement are always welcome! Guideline changes can be proposed by creating a discussion in the Ideas category.
"},{"location":"contributing/#code-of-conduct","title":"Code of Conduct","text":"
To ensure the Autoware community stays open and inclusive, please follow the Code of Conduct.
If you believe that someone in the community has violated the Code of Conduct, please make a report by emailing conduct@autoware.org.
"},{"location":"contributing/#what-should-i-know-before-i-get-started","title":"What should I know before I get started?","text":""},{"location":"contributing/#autoware-concepts","title":"Autoware concepts","text":"
To gain a high-level understanding of Autoware's architecture and design, the following pages provide a brief overview:
Autoware architecture
Autoware concepts
For experienced developers, the Autoware interfaces and individual component pages should also be reviewed to understand the inputs and outputs for each component or module at a more detailed level.
"},{"location":"contributing/#contributing-to-open-source-projects","title":"Contributing to open source projects","text":"
If you are new to open source projects, we recommend reading GitHub's How to Contribute to Open Source guide for an overview of why people contribute to open source projects, what it means to contribute and much more besides.
"},{"location":"contributing/#how-can-i-get-help","title":"How can I get help?","text":"
Do not open issues for general support questions as we want to keep GitHub issues for confirmed bug reports. Instead, open a discussion in the Q&A category. For more details on the support mechanisms for Autoware, refer to the Support guidelines.
Note
Issues created for questions or unconfirmed bugs will be moved to GitHub discussions by the maintainers.
"},{"location":"contributing/#how-can-i-contribute","title":"How can I contribute?","text":""},{"location":"contributing/#discussions","title":"Discussions","text":"
You can contribute to Autoware by facilitating and participating in discussions, such as:
Proposing a new feature to enhance Autoware
Joining an existing discussion and expressing your opinion
Organizing discussions for other contributors
Answering questions and supporting other contributors
The various working groups within the Autoware Foundation are responsible for accomplishing goals set by the Technical Steering Committee. These working groups are open to everyone, and joining a particular working group will allow you to gain an understanding of current projects, see how those projects are managed within each group and to contribute to issues that will help progress a particular project.
To see the schedule for upcoming working group meetings, refer to the Autoware Foundation events calendar.
Before you report a bug, please search the issue tracker for the appropriate repository. It is possible that someone has already reported the same issue and that workarounds exist. If you can't determine the appropriate repository, ask the maintainers for help by creating a new discussion in the Q&A category.
When reporting a bug, you should provide a minimal set of instructions to reproduce the issue. Doing so allows us to quickly confirm and focus on the right problem.
If you want to fix the bug by yourself that will be appreciated, but you should discuss possible approaches with the maintainers in the issue before submitting a pull request.
Creating an issue is straightforward, but if you happen to experience any problems then create a Q&A discussion to ask for help.
You can submit pull requests for small changes such as:
Minor documentation updates
Fixing spelling mistakes
Fixing CI failures
Fixing warnings detected by compilers or analysis tools
Making small changes to a single package
If your pull request is a large change, the following process should be followed:
Create a GitHub Discussion to propose the change. Doing so allows you to get feedback from other members and the Autoware maintainers and to ensure that the proposed change is in line with Autoware's design philosophy and current development plans. If you're not sure where to have that conversation, then create a new Q&A discussion.
Create an issue following consensus in the discussions
Create a pull request to implement the changes that references the Issue created in step 2
Create documentation for the new addition (if relevant)
Examples of large changes include:
Adding a new feature to Autoware
Adding a new documentation page or section
For more information on how to submit a good pull request, have a read of the pull request guidelines and don't forget to review the required license notations!
Autoware is licensed under Apache License 2.0. Thus all contributions will be licensed as such as per clause 5 of the Apache License 2.0:
5. Submission of Contributions. Unless You explicitly state otherwise,\n any Contribution intentionally submitted for inclusion in the Work\n by You to the Licensor shall be under the terms and conditions of\n this License, without any additional terms or conditions.\n Notwithstanding the above, nothing herein shall supersede or modify\n the terms of any separate license agreement you may have executed\n with Licensor regarding such Contributions.\n
Here is an example copyright header to add to the top of a new file:
Copyright [first year of contribution] The Autoware Contributors\nSPDX-License-Identifier: Apache-2.0\n
We don't write copyright notations of each contributor here. Instead, we place them in the NOTICE file like the following.
This product includes code developed by [company name].\nCopyright [first year of contribution] [company name]\n
Let us know if your legal department has a special request for the copyright notation.
Currently, the old formats explained here are also acceptable. Those old formats can be replaced by this new format if the original authors agree. Note that we won't write their copyrights to the NOTICE file unless they agree with the new format.
Also, it is encouraged to apply Clang-Tidy to each file. For the usage, see Applying Clang-Tidy to ROS packages.
Note that not all rules are covered by Clang-Tidy.
"},{"location":"contributing/coding-guidelines/languages/cpp/#style-rules","title":"Style rules","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#include-header-files-in-the-defined-order-required-partially-automated","title":"Include header files in the defined order (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale","title":"Rationale","text":"
Due to indirect dependencies, the include system of C++ makes different behaviors if the header order is different.
To reduce unintended bugs, local header files should come first.
If there are any reasons for defining macros at a specific position, write a comment before the macro.
// Compliant\n#include \"my_header.hpp\"\n\n#include \"my_package/foo.hpp\"\n\n#include <package1/foo.hpp>\n#include <package2/bar.hpp>\n\n#include <std_msgs/msg/header.hpp>\n\n#include <iostream>\n#include <vector>\n\n// For the foo bar reason, the FOO_MACRO must be defined here.\n#define FOO_MACRO\n#include <foo/bar.hpp>\n
"},{"location":"contributing/coding-guidelines/languages/cpp/#use-lower-snake-case-for-function-names-required-partially-automated","title":"Use lower snake case for function names (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_1","title":"Rationale","text":"
It is consistent with the C++ standard library.
It is consistent with other programming languages such as Python and Rust.
"},{"location":"contributing/coding-guidelines/languages/cpp/#use-upper-camel-case-for-enum-names-required-partially-automated","title":"Use upper camel case for enum names (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_2","title":"Rationale","text":"
"},{"location":"contributing/coding-guidelines/languages/cpp/#use-lower-snake-case-for-constant-names-required-partially-automated","title":"Use lower snake case for constant names (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_3","title":"Rationale","text":"
"},{"location":"contributing/coding-guidelines/languages/cpp/#count-acronyms-and-contractions-of-compound-words-as-one-word-required-partially-automated","title":"Count acronyms and contractions of compound words as one word (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_4","title":"Rationale","text":"
To clarify the boundaries of words when acronyms are consecutive.
ROS 2 logging is a powerful tool for understanding and debugging ROS nodes.
This page focuses on how to design console logging in Autoware and shows several practical examples. To comprehensively understand how ROS 2 logging works, refer to the logging documentation.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#logging-use-cases-in-autoware","title":"Logging use cases in Autoware","text":"
Developers debug code by seeing the console logs.
Vehicle operators take appropriate risk-avoiding actions depending on the console logs.
Log analysts analyze the console logs that are recorded in rosbag files.
To efficiently support these use cases, clean and highly visible logs are required. For that, several rules are defined below.
DEBUG: Use this level to show debug information for developers. Note that logs with this level is hidden by default.
INFO: Use this level to notify events (cyclic notifications during initialization, state changes, service responses, etc.) to operators.
WARN: Use this level when a node can continue working correctly, but unintended behaviors might happen.
For example, \"path optimization failed but the previous data can be used\", \"the localization score is low\", etc.
ERROR: Use this level when a node can't continue working correctly, and unintended behaviors would happen.
For example, \"path optimization failed and the path is empty\", \"the vehicle will trigger an emergency stop\", etc.
FATAL: Use this level when the entire system can't continue working correctly, and the system must be stopped.
For example, \"the vehicle control ECU doesn't respond\", \"the system storage crashed\", etc.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#filter-out-unnecessary-logs-by-setting-logging-options-required-non-automated","title":"Filter out unnecessary logs by setting logging options (required, non-automated)","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#rationale_1","title":"Rationale","text":"
Some third-party nodes such as drivers may not follow the Autoware's guidelines. If the logs are noisy, unnecessary logs should be filtered out.
Use the --log-level {level} option to change the minimum level of logs to be displayed:
<launch>\n<!-- This outputs only FATAL level logs. -->\n<node pkg=\"demo_nodes_cpp\" exec=\"talker\" ros_args=\"--log-level fatal\" />\n</launch>\n
If you want to disable only specific output targets, use the --disable-stdout-logs, --disable-rosout-logs, and/or --disable-external-lib-logs options:
<launch>\n<!-- This outputs to rosout and disk. -->\n<node pkg=\"demo_nodes_cpp\" exec=\"talker\" ros_args=\"--disable-stdout-logs\" />\n</launch>\n
<launch>\n<!-- This outputs to stdout. -->\n<node pkg=\"demo_nodes_cpp\" exec=\"talker\" ros_args=\"--disable-rosout-logs --disable-external-lib-logs\" />\n</launch>\n
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#use-throttled-logging-when-the-log-is-unnecessarily-shown-repeatedly-required-non-automated","title":"Use throttled logging when the log is unnecessarily shown repeatedly (required, non-automated)","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#rationale_2","title":"Rationale","text":"
If tons of logs are shown on the console, people miss important message.
The following cases are acceptable even if it's not throttled.
The message is really worth displaying every time.
The message level is DEBUG.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#do-not-depend-on-rclcppnode-in-core-library-classes-but-depend-only-on-rclcpplogginghpp-advisory-non-automated","title":"Do not depend on rclcpp::Node in core library classes but depend only on rclcpp/logging.hpp (advisory, non-automated)","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#rationale_3","title":"Rationale","text":"
Core library classes, which contain reusable algorithms, may also be used for non-ROS platforms. When porting libraries to other platforms, fewer dependencies are preferred.
// Compliant\n#include <rclcpp/logging.hpp>\n\nclass FooCore {\npublic:\nexplicit FooCore(const rclcpp::Logger & logger) : logger_(logger) {}\n\nvoid process() {\nRCLCPP_INFO(logger_, \"message\");\n}\n\nprivate:\nrclcpp::Logger logger_;\n};\n\n// Compliant\n// Note that logs aren't published to `/rosout` if the logger name is different from the node name.\n#include <rclcpp/logging.hpp>\n\nclass FooCore {\nvoid process() {\nRCLCPP_INFO(rclcpp::get_logger(\"foo_core_logger\"), \"message\");\n}\n};\n\n\n// Non-compliant\n#include <rclcpp/node.hpp>\n\nclass FooCore {\npublic:\nexplicit FooCore(const rclcpp::NodeOptions & node_options) : node_(\"foo_core_node\", node_options) {}\n\nvoid process() {\nRCLCPP_INFO(node_.get_logger(), \"message\");\n}\n\nprivate:\nrclcpp::Node node_;\n};\n
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#tips","title":"Tips","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#use-rqt_console-to-filter-logs","title":"Use rqt_console to filter logs","text":"
To filter logs, using rqt_console is useful:
ros2 run rqt_console rqt_console\n
For more details, refer to ROS 2 Documentation.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#useful-marco-expressions","title":"Useful marco expressions","text":"
To debug program, sometimes you need to see which functions and lines of code are executed. In that case, you can use __FILE__, __LINE__ and __FUNCTION__ macro:
The commonly used coordinate systems include the world coordinate system, the vehicle coordinate system, and the sensor coordinate system.
The world coordinate system is a fixed coordinate system that defines the physical space in the environment where the vehicle is located.
The vehicle coordinate system is the vehicle's own coordinate system, which defines the vehicle's position and orientation in the world coordinate system.
The sensor coordinate system is the sensor's own coordinate system, which is used to define the sensor's position and orientation in the vehicle coordinate system.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#how-coordinates-are-used-in-autoware","title":"How coordinates are used in Autoware","text":"
In Autoware, coordinate systems are typically used to represent the position and movement of vehicles and obstacles in space. Coordinate systems are commonly used for path planning, perception and control, can help the vehicle decide how to avoid obstacles and to plan a safe and efficient path of travel.
Transformation of sensor data
In Autoware, each sensor has a unique coordinate system and their data is expressed in terms of the coordinates. In order to correlate the independent data between different sensors, we need to find the position relationship between each sensor and the vehicle body. Once the installation position of the sensor on the vehicle body is determined, it will remain fixed during running, so the offline calibration method can be used to determine the precise position of each sensor relative to the vehicle body.
ROS TF2
The TF2 system maintains a tree of coordinate transformations to represent the relationships between different coordinate systems. Each coordinate system is given a unique name and they are connected by coordinate transformations. How to use TF2, refer to the TF2 tutorial.
earth: earth coordinate system describe the position of any point on the earth in terms of geodetic longitude, latitude, and altitude. In Autoware, the earth frame is only used in the GnssInsPositionStamped message.
map: map coordinate system is used to represent the location of points on a local map. Geographical coordinate system are mapped into plane rectangular coordinate system using UTM or MGRS. The map frame`s axes point to the East, North, Up directions as explained in Coordinate Axes Conventions.
base_link: vehicle coordinate system, the origin of the coordinate system is the center of the rear axle of the vehicle.
imu, lidar, gnss, radar: these are sensor frames, transfer to vehicle coordinate system through mounting relationship.
camera_link: camera_link is ROS standard camera coordinate system .
camera_optical_link: camera_optical_link is image standard camera coordinate system.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#estimating-the-base_link-frame-by-using-the-other-sensors","title":"Estimating the base_link frame by using the other sensors","text":"
Generally we don't have the localization sensors physically at the base_link frame. So various sensors localize with respect to their own frames, let's call it sensor frame.
We introduce a new frame naming convention: x_by_y:
We cannot directly get the sensor frame. Because we would need the EKF module to estimate the base_link frame first.
Without the EKF module the best we can do is to estimate Map[map] --> sensor_by_sensor --> base_link_by_sensor using this sensor.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#example-by-the-gnssins-sensor","title":"Example by the GNSS/INS sensor","text":"
For the integrated GNSS/INS we use the following frames:
The gnss_ins_by_gnss_ins frame is obtained by the coordinates from GNSS/INS sensor. The coordinates are converted to map frame using the gnss_poser node.
Finally gnss_ins_by_gnss_ins frame represents the position of the gnss_ins estimated by the gnss_ins sensor in the map.
Then by using the static transformation between gnss_ins and the base_link frame, we can obtain the base_link_by_gnss_ins frame. Which represents the base_link estimated by the gnss_ins sensor.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#how-they-can-be-created","title":"How they can be created","text":"
Calibration of sensor
The conversion relationship between every sensor coordinate system and base_link can be obtained through sensor calibration technology. How to calibrating your sensors refer to this link calibrating your sensors.
Localization
The relationship between the base_link coordinate system and the map coordinate system is determined by the position and orientation of the vehicle, and can be obtained from the vehicle localization result.
Geo-referencing of map data
The geo-referencing information can get the transformation relationship of earth coordinate system to local map coordinate system.
Place configuration files such as node parameters. For ROS parameters, use the extension .param.yaml. For non-ROS parameters, use the extension .yaml.
Rationale: Since ROS parameters files are type-sensitive, they should not be the target of some code formatters and linters. In order to distinguish the file type, we use different file extensions.
Place header files exposed to other packages. Do not place files directly under the include directory, but place files under the directory with the package name. This directory is used for mostly library headers. Note that many headers do not need to be placed here. It is enough to place the headers under the src directory.
Autoware use ROS 2 launch system to startup the software. Please see the official documentation to get a basic understanding about ROS 2 Launch system if you are not familiar with it.
"},{"location":"contributing/coding-guidelines/ros-nodes/launch-files/#guideline","title":"Guideline","text":""},{"location":"contributing/coding-guidelines/ros-nodes/launch-files/#the-organization-of-launch-files-in-autoware","title":"The organization of launch files in Autoware","text":"
Autoware mainly has two repositories related to launch file organization: the autoware.universe and the autoware_launch.
the autoware.universe contains the code of the main Autoware modules, and its launch directory is responsible for launching the nodes of each module. Autoware software stack is organized based on the architecture, so you may find that we try to match the launch structure similar to the architecture (splitting of files, namespace). For example, the tier4_map_launch subdirectory corresponds to the map module, so do the other tier4_*_launch subdirectories.
The autoware_launch is a repository referring to autoware.universe. The mainly purpose of introducing this repository is to provide the general entrance to start the Autoware software stacks, i.e, calling the launch file of each module.
The autoware.launch.xml is the basic launch file for road driving scenarios.
As can be seen from the content, the entire launch file is divided into several different modules, including Vehicle, System, Map, Sensing, Localization, Perception, Planning, Control, etc. By setting the launch_* argument equals to true or false , we can determine which modules to be loaded.
The logging_simulator.launch.xml is often used together with the recorded ROS bag to debug if the target module (e.g, Sensing, Localization or Perception) functions normally.
The planning_simulator.launch.xml is based on the Planning Simulator tool, mainly used for testing/validation of Planning module by simulating traffic rules, interactions with dynamic objects and control commands to the ego vehicle.
The e2e_simulator.launch.xml is the launcher for digital twin simulation environment.
"},{"location":"contributing/coding-guidelines/ros-nodes/launch-files/#add-a-new-package-in-autoware","title":"Add a new package in Autoware","text":"
If a newly created package has executable node, we expect sample launch file and configuration within the package, just like the recommended structure shown in previous directory structure page.
In order to automatically load the newly added package when starting Autoware, you need to make some necessary changes to the corresponding launch file. For example, if using ICP instead of NDT as the pointcloud registration algorithm, you can modify the autoware.universe/launch/tier4_localization_launch/launch/pose_estimator/pose_estimator.launch.xml file to load the newly added ICP package.
Another purpose of introducing the autoware_launch repository is to facilitate the parameter management of Autoware. Thinking about this situation: if we want to integrate Autoware to a specific vehicle and modify parameters, we have to fork autoware.universe which also has a lot of code other than parameters and is frequently updated by developers. By integrating these parameters in autoware_launch, we can customize the Autoware parameters just by forking autoware_launch repository. Taking the localization module as an examples:
all the \u201claunch parameters\u201d for localization component is listed in the files under autoware_launch/autoware_launch/config/localization.
the \"launch parameters\" file paths are set in the autoware_launch/autoware_launch/launch/components/tier4_localization_component.launch.xml file.
in autoware.universe/launch/tier4_localization_launch/launch, the launch files loads the \u201claunch parameters\u201d if the argument is given in the parameter configuration file. You can still use the default parameters in each packages to launch tier4_localization_launch within autoware.universe.
All the fields by default have the following units depending on their types:
type default unit distance meter (m) angle radians (rad) time second (s) speed m/s velocity m/s acceleration m/s\u00b2 angular vel. rad/s angular accel. rad/s\u00b2
If a field in a message has any of these default units, don't add any suffix or prefix denoting the type.
type non-default unit suffix distance nanometer _nm distance micrometer _um distance millimeter _mm distance kilometer _km angle degree (deg) _deg time nanosecond _ns time microsecond _us time millisecond _ms time minute _min time hour (h) _hour velocity km/h _kmph
If a unit that you'd like to use doesn't exist here, create an issue/PR to add it to this list.
"},{"location":"contributing/coding-guidelines/ros-nodes/message-guidelines/#message-field-types","title":"Message field types","text":"
For list of types supported by the ROS interfaces see here.
Also copied here for convenience:
Message Field Type C++ equivalent boolboolbyteuint8_tcharcharfloat32floatfloat64doubleint8int8_tuint8uint8_tint16int16_tuint16uint16_tint32int32_tuint32uint32_tint64int64_tuint64uint64_tstringstd::stringwstringstd::u16string"},{"location":"contributing/coding-guidelines/ros-nodes/message-guidelines/#arrays","title":"Arrays","text":"
On top of the message, briefly explain what the message contains and/or what it is used for. For an example, see sensor_msgs/msg/Imu.msg.
If necessary, add line comments before the fields that explain the context and/or meaning.
For simple fields like x, y, z, w you might not need to add comments.
Even though it is not strictly checked, try not to pass 100 characters in a line.
Example:
# Number of times the vehicle performed an emergency brake\nuint32 count_emergency_brake\n\n# Seconds passed since the last emergency brake\nuint64 duration\n
Autoware ROS nodes have declared parameters which values are provided during the node start up in the form of a parameter file. All the expected parameters with corresponding values should exist in the parameter file. Depending on the application, the parameter values might need to be modified.
Find more information on parameters from the official ROS documentation:
A ROS package which uses the declare_parameter(...) function should:
use the declare_parameter(...) with out a default value
create a parameter file
create a schema file
The rationale behind this workflow is to have a verified single source of truth to pass to the ROS node and to be used in the web documentation. The approach reduces the risk of using invalid parameter values and makes maintenance of documentation easier. This is achieved by:
declare_parameter(...) throws an exception if an expected parameter is missing in the parameter file
the schema validates the parameter file in the CI and renders a parameter table, as depicted in the graphics below
As there is no default_value provided, the function throws an exception if a parameter were to be missing in the provided *.param.yaml file. Use a type from the C++ Type column in the table below for the declare_parameter(...) function, replacing INSERT_TYPE.
ParameterType Enum C++ Type PARAMETER_BOOLboolPARAMETER_INTEGERint64_tPARAMETER_DOUBLEdoublePARAMETER_STRINGstd::stringPARAMETER_BYTE_ARRAYstd::vector<uint8_t>PARAMETER_BOOL_ARRAYstd::vector<bool>PARAMETER_INTEGER_ARRAYstd::vector<int64_t>PARAMETER_DOUBLE_ARRAYstd::vector<double>PARAMETER_STRING_ARRAYstd::vector<std::string>
The table has been derived from Parameter Type and Parameter Value.
See example: Lidar Apollo Segmentation TVM Nodes declare function
The parameter file is minimal as there is no need to provide the user with additional information, e.g., description or type. This is because the associated schema file provides the additional information. Use the template below as a starting point for a ROS node.
Note: /** is used instead of the explicit node namespace, this allows the parameter file to be passed to a ROS node which has been remapped.
To adapt the template to the ROS node, replace each INSERT_PARAMETER_..._NAME and INSERT_PARAMETER_..._VALUE for all parameters. Each declare_parameter(...) takes one parameter as input. All the parameter files should have the .param.yaml suffix so that the auto-format can be applied properly.
Autoware has the following two types of parameter files for ROS packages:
Node parameter file
Node parameter files store the default parameters provided for each package in Autoware.
For example, the parameter of behavior_path_planner
All nodes in Autoware must have a parameter file if ROS parameters are declared in the node.
For FOO_package, the parameter is expected to be stored in FOO_package/config.
The launch file for individual packages must load node parameter by default:
When a user creates a launch package for the user's vehicle, the user should copy node parameter files for the nodes that are called in the launch file as \"launch parameter files\".
Launch parameter files are then customized specifically for user's vehicle.
For example, the customized parameter of behavior_path_planner stored under autoware_launch
The examples for launch parameter files are stored under autoware_launch.
JSON Schema is used the validate the parameter file(s) ensuring that it has the correct structure and content. Using JSON Schema for this purpose is considered best practice for cloud-native development. The schema template below shall be used as a starting point when defining the schema for a ROS node.
The schema file path is INSERT_PATH_TO_PACKAGE/schema/ and the schema file name is INSERT_NODE_NAME.schema.json. To adapt the template to the ROS node, replace each INSERT_... and add all parameters 1..N.
See example: Lidar Apollo Segmentation TVM Nodes schema
Parameters have several attributes, some are required and some optional. The optional attributes are highly encouraged when applicable, as they provide useful information about a parameter and can ensure the value of the parameter is within its bounds.
a tested and verified value, see JSON Schema default
bound(s)
type dependent, e.g., integer, range and size
"},{"location":"contributing/coding-guidelines/ros-nodes/parameters/#tips-and-tricks","title":"Tips and Tricks","text":"
Using well established standards enables the use of conventional tooling. Below is an example of how to link a schema to the parameter file(s) using VS Code. This enables a developer with convenient features such as auto-complete and parameter bound validation.
In the root directory of where the project is hosted, create a .vscode folder with two files; extensions.json containing
The RedHat YAML extension enables validation of YAML files using JSON Schema and the \"yaml.schemas\" setting associates the *.schema.json file with all *.param.yaml files in the config/ folder.
ROS allows topics, parameters and nodes to be namespaced which provides the following benefits:
Multiple instances of the same node type will not cause naming clashes.
Topics published by a node can be automatically namespaced with the node's namespace providing a meaningful and easily-visible connection.
Keeps from cluttering the root namespace.
Helps to maintain separation-of-concerns.
This page focuses on how to use namespaces in Autoware and shows some useful examples. For basic information on topic namespaces, refer to this tutorial.
"},{"location":"contributing/coding-guidelines/ros-nodes/topic-namespaces/#how-topics-should-be-named-in-node","title":"How topics should be named in node","text":"
Autoware divides the node into the following functional categories, and adds the start namespace for the nodes according to the categories.
localization
perception
planning
control
sensing
vehicle
map
system
When a node is run in a namespace, all topics which that node publishes are given that same namespace. All nodes in the Autoware stack must support namespaces by avoiding practices such as publishing topics in the global namespace.
In general, topics should be namespaced based on the function of the node which produces them and not the node (or nodes) which consume them.
Classify topics as input or output topics based on they are subscribed or published by the node. In the node, input topic is named input/topic_name and output topic is named output/topic_name.
Configure the topic in the node's launch file. Take the joy_controller node as an example, in the following example, set the input and output topics and remap topics in the joy_controller.launch.xml file.
Visualization or debug purpose topics should have ~/debug/ namespace.
e.g., In node obstacle_avoidance_planner, in order to debug or visualizing topics, using topic names of type ~/debug/topic_name to publish information.
Contributions to Autoware's documentation are welcome, and the same principles described in the contribution guidelines should be followed. Small, limited changes can be made by forking this repository and submitting a pull request, but larger changes should be discussed with the community and Autoware maintainers via GitHub Discussion first.
Examples of small changes include:
Fixing spelling or grammatical mistakes
Fixing broken links
Making an addition to an existing, well-defined page, such as the Troubleshooting guide.
Examples of larger changes include:
Adding new pages with a large amount of detail, such as a tutorial
Re-organization of the existing documentation structure
You should refer to the Google developer documentation style guide as much as possible. Reading the Highlights page of that guide is recommended, but if not then the key points below should be noted.
Use standard American English spelling and punctuation.
Use sentence case for document titles and section headings.
Use descriptive link text.
Write short sentences that are easy to understand and translate.
"},{"location":"contributing/documentation-guidelines/#tips","title":"Tips","text":""},{"location":"contributing/documentation-guidelines/#how-to-preview-your-modification","title":"How to preview your modification","text":"
There are two ways to preview your modification on a documentation website.
"},{"location":"contributing/documentation-guidelines/#1-using-github-actions-workflow","title":"1. Using GitHub Actions workflow","text":"
Follow the steps below.
Create a pull request to the repository.
Add the deploy-docs label from the sidebar (See below figure).
Wait for a couple of minutes, and the github-actions bot will notify the URL for the pull request's preview.
"},{"location":"contributing/documentation-guidelines/#2-running-an-mkdocs-server-in-your-local-environment","title":"2. Running an MkDocs server in your local environment","text":"
Instead of creating a PR, you can use the mkdocs command to build Autoware's documentation websites on your local computer. Assuming that you are using Ubuntu OS, run the following to install the required libraries.
Autoware uses the fork-and-pull model. For more details about the model, refer to GitHub Docs.
The following is a general example of the pull request workflow based on the fork-and-pull model. Use this workflow as a reference when you contribute to Autoware.
Create an issue.
Discuss the approaches to the issue with maintainers.
Confirm the support guidelines before creating an issue.
Follow the discussion guidelines when you discuss with other contributors.
Create a fork repository. (for the first time only)
Write code in your fork repository according to the approach agreed upon in the issue.
Write the tests and documentation as appropriate.
Follow the coding guidelines guidelines when you write code.
Follow the Testing guidelines guidelines when you write tests.
Follow the Documentation guidelines guidelines when you write documentation.
Follow the commit guidelines when you commit your changes.
Test the code.
It is recommended that you summarize the test results, because you will need to explain the test results in the later review process.
If you are not sure what tests should be done, discuss them with maintainers.
Create a pull request.
Follow the pull request rules when you create a pull request.
Wait for the pull request to be reviewed.
The reviewers will review your code following the review guidelines.
Not only the reviewers, but also the author is encouraged to understand the review guidelines.
If CI checks have failed, fix the errors.
Address the review comments pointed out by the reviewers.
If you don't understand the meaning of a review comment, ask the reviewers until you understand it.
Fixing without understanding the reason is not recommended because the author should be responsible for the final content of their own pull request.
If you don't agree with a review comment, ask the reviewers for a rational reason.
The reviewers are obligated to make the author understand the meanings of each comment.
After you have done with the review comments, re-request a review to the reviewers and back to 6.
If there are no more new review comments, the reviewers will approve the pull request and proceed to 8.
Merge the pull request.
Anyone with write access can merge the pull request if there is no special request from maintainers.
The author is encouraged to merge the pull request to feel responsible for their own pull request.
If the author does not have write access, ask the reviewers or maintainers.
"},{"location":"contributing/pull-request-guidelines/#pull-request-rules","title":"Pull request rules","text":""},{"location":"contributing/pull-request-guidelines/#use-an-appropriate-pull-request-template-required-non-automated","title":"Use an appropriate pull request template (required, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale","title":"Rationale","text":"
The unified style of descriptions by templates can make reviews efficient.
There are two types of templates. Select one based on the following condition.
Standard change:
Complexity:
New features or significant updates.
Requires deeper understanding of the codebase.
Impact:
Affects multiple parts of the system.
Basically includes minor features, bug fixes and performance improvement.
Needs testing before merging.
Small change:
Complexity:
Documentation, simple refactoring, or style adjustments.
Easy to understand and review.
Impact:
Minimal effect on the system.
Quicker merge with less testing needed.
"},{"location":"contributing/pull-request-guidelines/#steps-to-use-an-appropriate-pull-request-template","title":"Steps to use an appropriate pull request template","text":"
Select the appropriate template, as shown in this video.
Read the selected template carefully and fill the required content.
Check the checkboxes during a review.
There are pre-review checklist and post-review checklist for the author.
"},{"location":"contributing/pull-request-guidelines/#set-appropriate-reviewers-after-creating-a-pull-request-required-partially-automated","title":"Set appropriate reviewers after creating a pull request (required, partially automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_1","title":"Rationale","text":"
Pull requests must be reviewed by appropriate reviewers to keep the quality of the codebase.
For most ROS packages, reviewers will be automatically assigned based on the maintainer information in package.xml.
If no reviewer is assigned automatically, assign reviewers manually following the instructions in GitHub Docs.
You can find the reviewers by seeing the .github/CODEOWNERS file of the repository.
If you are not sure the appropriate reviewers, ask @autoware-maintainers.
If you have no rights to assign reviewers, mention reviewers instead.
"},{"location":"contributing/pull-request-guidelines/#apply-conventional-commits-to-the-pull-request-title-required-automated","title":"Apply Conventional Commits to the pull request title (required, automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_2","title":"Rationale","text":"
Conventional Commits can generate categorized changelogs, for example using git-cliff.
For the repositories that contain code (most repositories), use the definition of conventional-commit-types for the type.
For documentation repositories such as autoware-documentation, use the following definition:
feat
Add new pages.
Add contents to the existing pages.
fix
Fix the contents in the existing pages.
refactor
Move contents to different pages.
docs
Update documentation for the documentation repository itself.
build
Update the settings of the documentation site builder.
! (breaking changes)
Remove pages.
Change the URL of pages.
perf and test are generally unused. Other types have the same meaning as the code repositories.
"},{"location":"contributing/pull-request-guidelines/#add-the-related-component-names-to-the-scope-of-conventional-commits-advisory-non-automated","title":"Add the related component names to the scope of Conventional Commits (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_3","title":"Rationale","text":"
It helps contributors find pull requests that are relevant to them.
For ROS packages, adding the package name or component name is good.
feat(trajectory_follower): add an awesome feature\nrefactor(planning, control): use common utils\n
"},{"location":"contributing/pull-request-guidelines/#keep-a-pull-request-small-advisory-non-automated","title":"Keep a pull request small (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_4","title":"Rationale","text":"
Small pull requests are easy to understand for reviewers.
Small pull requests are easy to revert for maintainers.
Avoid developing two features in one pull request.
Avoid mixing different types (feat, fix, refactor, etc.) of changes in the same commit.
"},{"location":"contributing/pull-request-guidelines/#remind-reviewers-if-there-is-no-response-for-more-than-a-week-advisory-non-automated","title":"Remind reviewers if there is no response for more than a week (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_5","title":"Rationale","text":"
It is the author's responsibility to care about their own pull request until it is merged.
Autoware has several checks for a pull request. The results are shown at the bottom of the pull request page as below.
If the \u274c mark is shown, click the Details button and investigate the failure reason.
If the Required mark is shown, you cannot merge the pull request unless you resolve the error. If not, it is optional, but preferably it should be fixed.
The following sections explain about common CI checks in Autoware. Note that some repositories may have different settings.
The Developer Certificate of Origin (DCO) is a lightweight way for contributors to certify that they wrote or otherwise have the right to submit the code they are contributing to the project.
This workflow checks whether the pull request fulfills DCO. You need to confirm the required items and commit with git commit -s.
For more information, refer to the GitHub App page.
pre-commit is a tool to run formatters or linters when you commit.
This workflow checks whether the pull request has no error with pre-commit.
In the workflow pre-commit.ci - pr is enabled in the repository, it will automatically fix errors by pre-commit.ci as many as possible. If there are some errors remain, fix them manually.
You can run pre-commit in your local environment by the following command:
pre-commit run -a\n
Or you can install pre-commit to the repository and automatically run it before committing:
pre-commit install\n
Since it is difficult to detect errors with no false positives, some jobs are split into another config file and marked as optional. To check them, use the --config option:
pre-commit run -a --config .pre-commit-config-optional.yaml\n
This workflow detects spelling mistakes using CSpell with our dictionary file. You can submit pull requests to tier4/autoware-spell-check-dict to update the dictionary.
Since it is difficult to detect errors with no false positives, it is an optional workflow, but it is preferable to remove spelling mistakes as many as possible.
This workflow checks colcon build and colcon test for the pull request. To make the CI faster, it doesn't check all packages but only modified packages and the dependencies.
This workflow is the ARM64 version of build-and-test-differential. You need to add the ARM64 label to run this workflow.
For reference information, since ARM machines are not supported by GitHub-hosted runners, we use self-hosted runners prepared by the AWF. For the details about self-hosted runners, refer to GitHub Docs.
"},{"location":"contributing/pull-request-guidelines/commit-guidelines/#use-dash-case-for-the-separator-of-branch-names-advisory-non-automated","title":"Use dash-case for the separator of branch names (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/commit-guidelines/#rationale_1","title":"Rationale","text":"
If you have already submitted a pull request, you do not have to change the branch name because you need to re-create a pull request, which is noisy and a waste of time. Be careful from the next time.
"},{"location":"contributing/pull-request-guidelines/review-tips/","title":"Review tips","text":""},{"location":"contributing/pull-request-guidelines/review-tips/#review-tips","title":"Review tips","text":""},{"location":"contributing/pull-request-guidelines/review-tips/#toggle-annotations-or-review-comments-in-the-diff-view","title":"Toggle annotations or review comments in the diff view","text":"
There might be some annotations or review comments in the diff view during your review.
To toggle annotations, press the A key.
Before:
After:
To toggle review comments, press the I key.
For other keyboard shortcuts, refer to GitHub Docs.
"},{"location":"contributing/pull-request-guidelines/review-tips/#view-code-in-the-web-based-visual-studio-code","title":"View code in the web-based Visual Studio Code","text":"
You can open Visual Studio Code from your browser to view code in a rich UI. To use it, press the . key on any repository or pull request.
For more detailed usage, refer to github/dev.
"},{"location":"contributing/pull-request-guidelines/review-tips/#check-out-the-branch-of-a-pull-request-quickly","title":"Check out the branch of a pull request quickly","text":"
If you want to check out the branch of a pull request, it's generally troublesome with the fork-and-pull model.
# Copy the user name and the fork URL.\ngit remote add {user-name} {fork-url}\ngit checkout {user-name}/{branch-name}\ngit remote rm {user-name} # To clean up\n
Instead, you can use GitHub CLI to simplify the steps, just run gh pr checkout {pr-number}.
You can copy the command from the top right of the pull request page.
An integration test is defined as the phase in software testing where individual software modules are combined and tested as a group. Integration tests occur after unit tests, and before validation tests.
The input to an integration test is a set of independent modules that have been unit tested. The set of modules is tested against the defined integration test plan, and the output is a set of properly integrated software modules that is ready for system testing.
"},{"location":"contributing/testing-guidelines/integration-testing/#value-of-integration-testing","title":"Value of integration testing","text":"
Integration tests determine if independently developed software modules work correctly when the modules are connected to each other. In ROS 2, the software modules are called nodes. Testing a single node is a special type of integration test that is commonly referred to as component testing.
Integration tests help to find the following types of errors:
Incompatible interactions between nodes, such as non-matching topics, different message types, or incompatible QoS settings.
Edge cases that were not touched by unit testing, such as a critical timing issue, network communication delays, disk I/O failures, and other such problems that can occur in production environments.
Issues that can occur while the system is under high CPU/memory load, such as malloc failures. This can be tested using tools like stress and udpreplay to test the performance of nodes with real data.
With ROS 2, it is possible to program complex autonomous-driving applications with a large number of nodes. Therefore, a lot of effort has been made to provide an integration-test framework that helps developers test the interaction of ROS 2 nodes.
Doing so adds smoke tests that ensure that a node can be:
Launched with a default parameter file.
Terminated with a standard SIGTERM signal.
For the full API documentation, refer to the package design page.
Note
This API is not suitable for all smoke test cases. It cannot be used when a specific file location (eg: for a map) is required to be passed to the node, or if some preparation needs to be conducted before node launch. In such cases use the manual solution from the component test section below.
"},{"location":"contributing/testing-guidelines/integration-testing/#integration-test-with-a-single-node-component-test","title":"Integration test with a single node: component test","text":"
The simplest scenario is a single node. In this case, the integration test is commonly referred to as a component test.
To add a component test to an existing node, you can follow the example of the lanelet2_map_loader in the map_loader package (added in this PR).
In package.xml, add:
<test_depend>ros_testing</test_depend>\n
In CMakeLists.txt, add or modify the BUILD_TESTING section:
Then a launch description is created to launch the node under test. Note that the test_map.osm file path is found and passed to the node, something that cannot be done with the smoke testing API:
@pytest.mark.launch_test\ndef generate_test_description():\n\n lanelet2_map_path = os.path.join(\n get_package_share_directory(\"map_loader\"), \"test/data/test_map.osm\"\n )\n\n lanelet2_map_loader = Node(\n package=\"map_loader\",\n executable=\"lanelet2_map_loader\",\n parameters=[{\"lanelet2_map_path\": lanelet2_map_path}],\n )\n\n context = {}\n\n return (\n LaunchDescription(\n [\n lanelet2_map_loader,\n # Start test after 1s - gives time for the map_loader to finish initialization\n launch.actions.TimerAction(\n period=1.0, actions=[launch_testing.actions.ReadyToTest()]\n ),\n ]\n ),\n context,\n )\n
Note
Since the node need time to process the input lanelet2 map, we use a TimerAction to delay the start of the test by 1s.
In the example above, the context is empty but it can be used to pass objects to the test cases.
You can find an example of using the context in the ROS 2 context_launch_test.py test example.
Finally, a test is executed after the node executable has been shut down (post_shutdown_test). Here we ensure that the node was launched without error and exited cleanly.
@launch_testing.post_shutdown_test()\nclass TestProcessOutput(unittest.TestCase):\n def test_exit_code(self, proc_info):\n # Check that process exits with code 0: no error\n launch_testing.asserts.assertExitCodes(proc_info)\n
"},{"location":"contributing/testing-guidelines/integration-testing/#running-the-test","title":"Running the test","text":"
Continuing the example from above, first build your package:
The simple test described in Integration test with a single node: component test can be extended in numerous directions, such as testing a node's output.
"},{"location":"contributing/testing-guidelines/integration-testing/#testing-the-output-of-a-node","title":"Testing the output of a node","text":"
To test while the node is running, create an active test by adding a subclass of Python's unittest.TestCase to *launch.test.py. Some boilerplate code is required to access output by creating a node and a subscription to a particular topic, e.g.
import unittest\n\nclass TestRunningDataPublisher(unittest.TestCase):\n\n @classmethod\n def setUpClass(cls):\n cls.context = Context()\n rclpy.init(context=cls.context)\n cls.node = rclpy.create_node(\"test_node\", context=cls.context)\n\n @classmethod\n def tearDownClass(cls):\n rclpy.shutdown(context=cls.context)\n\n def setUp(self):\n self.msgs = []\n sub = self.node.create_subscription(\n msg_type=my_msg_type,\n topic=\"/info_test\",\n callback=self._msg_received\n )\n self.addCleanup(self.node.destroy_subscription, sub)\n\n def _msg_received(self, msg):\n # Callback for ROS 2 subscriber used in the test\n self.msgs.append(msg)\n\n def get_message(self):\n startlen = len(self.msgs)\n\n executor = rclpy.executors.SingleThreadedExecutor(context=self.context)\n executor.add_node(self.node)\n\n try:\n # Try up to 60 s to receive messages\n end_time = time.time() + 60.0\n while time.time() < end_time:\n executor.spin_once(timeout_sec=0.1)\n if startlen != len(self.msgs):\n break\n\n self.assertNotEqual(startlen, len(self.msgs))\n return self.msgs[-1]\n finally:\n executor.remove_node(self.node)\n\n def test_message_content():\n msg = self.get_message()\n self.assertEqual(msg, \"Hello, world\")\n
Unit testing is the first phase of testing and is used to validate units of source code such as classes and functions. Typically, a unit of code is tested by validating its output for various inputs. Unit testing helps ensure that the code behaves as intended and prevents accidental changes of behavior.
Autoware uses the ament_cmake framework to build and run tests. The same framework is also used to analyze the test results.
ament_cmake provides several convenience functions to make it easy to register tests in a CMake-based package and to ensure that JUnit-compatible result files are generated. It currently supports a few different testing frameworks like pytest, gtest, and gmock.
In order to prevent tests running in parallel from interfering with each other when publishing and subscribing to ROS topics, it is recommended to use commands from ament_cmake_ros to run tests in isolation.
See below for an example of using ament_add_ros_isolated_gtest with colcon test. All other tests follow a similar pattern.
"},{"location":"contributing/testing-guidelines/unit-testing/#create-a-unit-test-with-gtest","title":"Create a unit test with gtest","text":"
In my_cool_pkg/test, create the gtest code file test_my_cool_pkg.cpp:
Next add an entry under BUILD_TESTING in the CMakeLists.txt to compile the test source files:
if(BUILD_TESTING)\n\nament_add_ros_isolated_gtest(test_my_cool_pkg test/test_my_cool_pkg.cpp)\ntarget_link_libraries(test_my_cool_pkg ${PROJECT_NAME})\ntarget_include_directories(test_my_cool_pkg PRIVATE src) # For private headers.\n...\nendif()\n
This automatically links the test with the default main function provided by gtest. The code under test is usually in a different CMake target (${PROJECT_NAME} in the example) and its shared object for linking needs to be added. If the test source files include private headers from the src directory, the directory needs to be added to the include path using target_include_directories() function.
To register a new gtest item, wrap the test code with the macro TEST (). TEST () is a predefined macro that helps generate the final test code, and also registers a gtest item to be available for execution. The test case name should be in CamelCase, since gtest inserts an underscore between the fixture name and the class case name when creating the test executable.
gtest/gtest.h also contains predefined macros of gtest like ASSERT_TRUE(condition), ASSERT_FALSE(condition), ASSERT_EQ(val1,val2), ASSERT_STREQ(str1,str2), EXPECT_EQ(), etc. ASSERT_* will abort the test if the condition is not satisfied, while EXPECT_* will mark the test as failed but continue on to the next test condition.
Info
More information about gtest and its features can be found in the gtest repo.
In the demo CMakeLists.txt, ament_add_ros_isolated_gtest is a predefined macro in ament_cmake_ros that helps simplify adding gtest code. Details can be viewed in ament_add_gtest.cmake.
Look in the ~/workspace/log/test_<date>/<package_name> directory for all the raw test commands, std_out, and std_err. There is also the ~/workspace/log/latest_*/ directory containing symbolic links to the most recent package-level build and test output.
To print the tests' details while the tests are being run, use the --event-handlers console_cohesion+ option to print the details directly to the console:
$ colcon test --event-handlers console_cohesion+ --packages-select my_cool_pkg\n\n...\ntest 1\n Start 1: test_my_cool_pkg\n\n1: Test command: /usr/bin/python3 \"-u\" \"~/workspace/install/share/ament_cmake_test/cmake/run_test.py\" \"~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml\" \"--package-name\" \"my_cool_pkg\" \"--output-file\" \"~/workspace/build/my_cool_pkg/ament_cmake_gtest/test_my_cool_pkg.txt\" \"--command\" \"~/workspace/build/my_cool_pkg/test_my_cool_pkg\" \"--gtest_output=xml:~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml\"\n1: Test timeout computed to be: 60\n1: -- run_test.py: invoking following command in '~/workspace/src/my_cool_pkg':\n1: - ~/workspace/build/my_cool_pkg/test_my_cool_pkg --gtest_output=xml:~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml\n1: [==========] Running 1 test from 1 test case.\n1: [----------] Global test environment set-up.\n1: [----------] 1 test from test_my_cool_pkg\n1: [ RUN ] test_my_cool_pkg.test_hello\n1: Hello World\n1: [ OK ] test_my_cool_pkg.test_hello (0 ms)\n1: [----------] 1 test from test_my_cool_pkg (0 ms total)\n1:\n1: [----------] Global test environment tear-down\n1: [==========] 1 test from 1 test case ran. (0 ms total)\n1: [ PASSED ] 1 test.\n1: -- run_test.py: return code 0\n1: -- run_test.py: inject classname prefix into gtest result file '~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml'\n1: -- run_test.py: verify result file '~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml'\n1/5 Test #1: test_my_cool_pkg ................... Passed 0.09 sec\n\n...\n\n100% tests passed, 0 tests failed out of 5\n\nLabel Time Summary:\ncopyright = 0.49 sec*proc (1 test)\ncppcheck = 0.20 sec*proc (1 test)\ngtest = 0.05 sec*proc (1 test)\nlint_cmake = 0.18 sec*proc (1 test)\nlinter = 1.34 sec*proc (4 tests)\nxmllint = 0.47 sec*proc (1 test)\n\nTotal Test time (real) = 7.91 sec\n...\n
The GNSS data is available in sensor_msgs/msg/NavSatFix message type.
But also the Applanix raw messages are also included in applanix_msgs/msg/NavigationPerformanceGsof50 and applanix_msgs/msg/NavigationSolutionGsof49 message types. In order to be able to play back these messages, you need to build and source the applanix_msgs package.
# Create a workspace and clone the repository\nmkdir -p ~/applanix_ws/src && cd \"$_\"\ngit clone https://github.com/autowarefoundation/applanix.git\ncd ..\n\n# Build the workspace\ncolcon build --symlink-install --packages-select applanix_msgs\n\n# Source the workspace\nsource ~/applanix_ws/install/setup.bash\n\n# Now you can play back the messages\n
Also make sure to source Autoware Universe workspace too.
# Install awscli\n$ sudo apt update && sudo apt install awscli -y\n\n# This will download the entire dataset to the current directory.\n# (About 10.9GB of data)\n$ aws s3 sync s3://autoware-files/collected_data/2022-08-22_leo_drive_isuzu_bags/ ./2022-08-22_leo_drive_isuzu_bags --no-sign-request\n\n# Optionally,\n# If you instead want to download a single bag file, you can get a list of the available files with following:\n$ aws s3 ls s3://autoware-files/collected_data/2022-08-22_leo_drive_isuzu_bags/ --no-sign-request\n PRE all-sensors-bag1_compressed/\n PRE all-sensors-bag2_compressed/\n PRE all-sensors-bag3_compressed/\n PRE all-sensors-bag4_compressed/\n PRE all-sensors-bag5_compressed/\n PRE all-sensors-bag6_compressed/\n PRE driving_20_kmh_2022_06_10-16_01_55_compressed/\n PRE driving_30_kmh_2022_06_10-15_47_42_compressed/\n\n# Then you can download a single bag file with the following:\naws s3 sync s3://autoware-files/collected_data/2022-08-22_leo_drive_isuzu_bags/all-sensors-bag1_compressed/ ./all-sensors-bag1_compressed --no-sign-request\n
"},{"location":"datasets/#autocoreai-lidar-ros-2-bag-file-and-pcap","title":"AutoCore.ai - lidar ROS 2 bag file and pcap","text":"
This dataset contains pcap files and ros2 bag files from Ouster OS1-64 Lidar. The pcap file and ros2 bag file is recorded in the same time with slight difference in duration.
Autoware provides the runtimes and technology components by open-source software. The runtimes are based on the Robot Operating System (ROS). The technology components are provided by contributors, which include, but are not limited to:
Sensing
Camera Component
LiDAR Component
RADAR Component
GNSS Component
Computing
Localization Component
Perception Component
Planning Component
Control Component
Logging Component
System Monitoring Component
Actuation
DBW Component
Tools
Simulator Component
Mapping Component
Remote Component
ML Component
Annotation Component
Calibration Component
"},{"location":"design/#concern-assumption-and-limitation","title":"Concern, Assumption, and Limitation","text":"
The downside of the microautonomy architecture is that the computational performance of end applications is sacrificed due to its data path overhead attributed to functional modularity. In other words, the trade-off characteristic of the microautonomy architecture exists between computational performance and functional modularity. This trade-off problem can be solved technically by introducing real-time capability. This is because autonomous driving systems are not really designed to be real-fast, that is, low-latency computing is nice-to-have but not must-have. The must-have feature for autonomous driving systems is that the latency of computing is predictable, that is, the systems are real-time. As a whole, we can compromise computational performance to an extent that is predictable enough to meet the given timing constraints of autonomous driving systems, often referred to as deadlines of computation.
The Autoware concepts page describes the design philosophy of Autoware. Readers (service providers and all Autoware users) will learn the basic concepts underlying Autoware development, such as microautonomy and the Core/Universe architecture.
The Autoware architecture page describes an overview of each module that makes up Autoware. Readers (all Autoware users) will gain a high-level picture of how each module that composes Autoware works.
The Autoware interfaces page describes in detail the interface of each module that makes up Autoware. Readers (intermediate developers) will learn how to add new functionality to Autoware and how to integrate their own modules with Autoware.
The current Autoware is defined to be a layered architecture that clarifies each module's role and simplifies the interface between them. By doing so:
Autoware's internal processing becomes more transparent.
Collaborative development is made easier because of the reduced interdependency between modules.
Users can easily replace an existing module (e.g. localization) with their own software component by simply wrapping their software to fit in with Autoware's interface.
Note that the initial focus of this architecture design was solely on driving capability, and so the following features were left as future work:
Autoware's architecture consists of the following six stacks. Each linked page contains a more detailed set of requirements and use cases specific to that stack:
A diagram showing Autoware's nodes in the default configuration can be found on the Node diagram page. Detailed documents for each node are available in the Autoware Universe docs.
Note that Autoware configurations are scalable / selectable and will vary depending on the environment and required use cases.
This document presents the design concept of the Control Component. The content is as follows:
Autoware Control Design
Outlining the policy for Autoware's control, which deals with only general information for autonomous driving systems and provides generic control commands to the vehicle.
Vehicle Adaptation Design
Describing the policy for vehicle adaptation, which utilizes adapter mechanisms to standardize the characteristics of the vehicle's drive system and integrate it with Autoware.
Control Feature Design
Demonstrating the features provided by Autoware's control.
Presenting the approach towards the functions installed in the vehicle such as ABS.
"},{"location":"design/autoware-architecture/control/#autoware-control-design","title":"Autoware Control Design","text":"
The Control Component generates the control signal to which the Vehicle Component subscribes. The generated control signals are computed based on the reference trajectories from the Planning Component.
The Control Component consists of two modules. The trajectory_follower module generates a vehicle control command to follow the reference trajectory received from the planning module. The command includes, for example, the desired steering angle and target speed. The vehicle_command_gate is responsible for filtering the control command to prevent abnormal values and then sending it to the vehicle. This gate also allows switching between multiple sources such as the MRM (minimal risk maneuver) module or some remote control module, in addition to the trajectory follower.
The Autoware control system is designed as a platform for automated driving systems that can be compatible with a diverse range of vehicles.
The control process in Autoware uses general information (such as target acceleration and deceleration) and no vehicle-specific information (such as brake pressure) is used. Hence it can be adjusted independently of the vehicle's drive interface enabling easy integration or performance tuning.
Furthermore, significant differences that affect vehicle motion constraints, such as two-wheel steering or four-wheel steering, are addressed by switching the control vehicle model, achieving control specialized for each characteristic.
Autoware's control module outputs the necessary information to control the vehicle as a substitute for a human driver. For example, the control command from the control module looks like the following:
Note that vehicle-specific values such as pedal positions and low-level information such as individual wheel rotation speeds are excluded from the command.
Autoware is designed to be an autonomous driving platform able to accommodate vehicles with various drivetrain types.
This is an explanation of how Autoware handles the standardization of systems with different vehicle drivetrain. The interfaces for vehicle drivetrain are diverse, including steering angle, steering angular velocity, steering torque, speed, accel/brake pedals, and brake pressure. To accommodate these differences, Autoware adds an adapter module between the control component and the vehicle interface. This module performs the conversion between the proprietary message types used by the vehicle (such as brake pressure) and the generic types used by Autoware (such as desired acceleration). By providing this conversion information, the differences in vehicle drivetrain can be accommodated.
If the information is not known in advance, an automatic calibration tool can be used. Calibration will occur within limited degrees of freedom, generating the information necessary for the drivetrain conversion automatically.
This configuration is summarized in the following diagram.
"},{"location":"design/autoware-architecture/control/#examples-of-several-vehicle-interfaces","title":"Examples of several vehicle interfaces","text":"
This is an example of the several drivetrain types in the vehicle interface.
Vehicle Lateral interface Longitudinal interface Note Lexus Steering angle Accel/brake pedal position Acceleration lookup table conversion for longitudinal JPN TAXI Steering angle Accel/brake pedal position Acceleration lookup table conversion for longitudinal GSM8 Steering EPS voltage Acceleration motor voltage, Deceleration brake hydraulic pressure lookup table and PID conversion for lateral and longitudinal YMC Golfcart Steering angle Velocity Logiee yaw rate Velocity F1 TENTH Steering angle Motor RPM interface code"},{"location":"design/autoware-architecture/control/#control-feature-design","title":"Control Feature Design","text":"
The following lists the features provided by Autoware's Control/Vehicle component, as well as the conditions and assumptions required to utilize them effectively.
The proper operation of the ODD is limited by factors such as whether the functions are enabled, delay time, calibration accuracy and degradation rate, and sensor accuracy.
Feature Description\u3000 Requirements/Assumptions Note \u3000Limitation for now Lateral Control Control the drivetrain system related to lateral vehicle motion Trying to increase the number of vehicle types that can be supported in the future. Only front-steering type is supported. Longitudinal Control Control the drivetrain system related to longitudinal vehicle motion Slope Compensation Supports precise vehicle motion control on slopes Gradient information can be obtained from maps or sensors attached to the chassis If gradient information is not available, the gradient is estimated from the vehicle's pitch angle. Delay Compensation Controls the drivetrain system appropriately in the presence of time delays The drivetrain delay information is provided in advance If there is no delay information, the drivetrain delay is estimated automatically (automatic calibration). However, the effect of delay cannot be completely eliminated, especially in scenarios with sudden changes in speed. Only fixed delay times can be set for longitudinal and lateral drivetrain systems separately. It does not accommodate different delay times for the accelerator and brake. Drivetrain IF Conversion (Lateral Control) Converts the drivetrain-specific information of the vehicle into the drivetrain information used by Autoware (e.g., target steering angular velocity \u2192 steering torque) The conversion information is provided in advance If there is no conversion information, the conversion map is estimated automatically (automatic calibration). The degree of freedom for conversion is limited (2D lookup table + PID FB). Drivetrain IF Conversion (Longitudinal Control) Converts the drivetrain-specific information of the vehicle into the drivetrain information used by Autoware (e.g., target acceleration \u2192 accelerator/brake pedal value) The conversion information is provided in advance If there is no conversion information, the conversion map is estimated automatically (automatic calibration). The degree of freedom for conversion is limited (2D lookup table + PID FB). Automatic Calibration Automatically estimates and applies values such as drivetrain IF conversion map and delay time. The drivetrain status can be obtained (must) Anomaly Detection Notifies when there is a discrepancy in the calibration or unexpected drivetrain behavior The drivetrain status can be obtained (must) Steering Zero Point Correction Corrects the midpoint of the steering to achieve appropriate steering control The drivetrain status can be obtained (must) Steering Deadzone Correction Corrects the deadzone of the steering to achieve appropriate steering control The steering deadzone parameter is provided in advance If the parameter is unknown, the deadzone parameter is estimated from driving information Not available now Steering Deadzone Estimation Dynamically estimates the steering deadzone from driving data Not available now Weight Compensation Performs appropriate vehicle control according to weight Weight information can be obtained from sensors If there is no weight sensor, estimate the weight from driving information. Currently not available Weight Estimation Dynamically estimates weight from driving data Currently not available
The list above does not cover wheel control systems such as ABS commonly used in vehicles. Regarding these features, the following considerations are taken into account.
"},{"location":"design/autoware-architecture/control/#integration-with-vehicle-side-functions","title":"Integration with vehicle-side functions","text":"
ABS (Anti-lock Brake System) and ESC (Electric Stability Control) are two functions that may be pre-installed on a vehicle, directly impacting its controllability. The control modules of Autoware assume that both ABS and ESC are installed on the vehicle and their absence may cause unreliable controls depending on the target ODD. For example, with low-velocity driving in a controlled environment, these functions are not necessary.
Also, note that this statement does not negate the development of ABS functionality in autonomous driving systems.
"},{"location":"design/autoware-architecture/control/#autoware-capabilities-and-vehicle-requirements","title":"Autoware Capabilities and Vehicle Requirements","text":"
As an alternative to human driving, autonomous driving systems essentially aim to handle tasks that humans can perform. This includes not only controlling the steering wheel, accel, and brake, but also automatically detecting issues such as poor brake response or a misaligned steering angle. However, this is a trade-off, as better vehicle performance will lead to superior system behavior, ultimately affecting the design of ODD.
On the other hand, for tasks that are not typically anticipated or cannot be handled by a human driver, processing in the vehicle ECU is expected. Examples of such scenarios include cases where the brake response is clearly delayed or when the vehicle rotates due to a single-side tire slipping. These tasks are typically handled by ABS or ESC.
Localization aims to estimate vehicle pose, velocity, and acceleration.
Goals:
Propose a system that can estimate vehicle pose, velocity, and acceleration for as long as possible.
Propose a system that can diagnose the stability of estimation and send a warning message to the error-monitoring system if the estimation result is unreliable.
Design a vehicle localization function that can work with various sensor configurations.
Non-goals:
This design document does not aim to develop a localization system that
is infallible in all environments
works outside of the pre-defined ODD (Operational Design Domain)
has better performance than is required for autonomous driving
This section shows example sensor configurations and their expected performances. Each sensor has its own advantages and disadvantages, but overall performance can be improved by fusing multiple sensors.
The vehicle is located in a structure-rich environment, such as an urban area
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable","title":"Situations that can make the system unstable","text":"
The vehicle is placed in a structure-less environment, such as a rural landscape, highway, or tunnel
Environmental changes have occurred since the map was created, such as snow cover or the construction/destruction of buildings.
Surrounding objects are occluded
The car is surrounded by objects undetectable by LiDAR, e.g., glass windows, reflections, or absorption (dark objects)
The environment contains laser beams at the same frequency as the car's LiDAR sensor(s)
The system can estimate the vehicle location on the point cloud map with the error of ~10cm.
The system is operable at night.
"},{"location":"design/autoware-architecture/localization/#3d-lidar-or-camera-vector-map","title":"3D-LiDAR or Camera + Vector Map","text":""},{"location":"design/autoware-architecture/localization/#expected-situation_1","title":"Expected situation","text":"
Road with clear white lines and loose curvatures, such as a highway or an ordinary local road.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_1","title":"Situations that can make the system unstable","text":"
White lines are scratchy or covered by rain or snow
Tight curvature such as intersections
Large reflection change of the road surface caused by rain or paint
The vehicle is placed in an open environment with few to no surrounding objects, such as a rural landscape.
"},{"location":"design/autoware-architecture/localization/#situation-that-can-make-the-system-unstable","title":"Situation that can make the system unstable","text":"
GNSS signals are blocked by surrounding objects, e.g., tunnels or buildings.
The vehicle is placed in an environment with rich visual features, such as an urban area.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_2","title":"Situations that can make the system unstable","text":"
The vehicle is placed in a texture-less environment.
The vehicle is surrounded by other objects.
The camera observes significant illumination changes, such as those caused by sunshine, headlights from other vehicles or when approaching the exit of a tunnel.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_3","title":"Situations that can make the system unstable","text":"
The vehicle is running on a slippery or bumpy road, which can cause incorrect observations of wheel speed.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_4","title":"Situations that can make the system unstable","text":"
IMUs have a bias1 that is dependent on the surrounding temperature, and can cause incorrect sensor observation or odometry drift.
The vehicle is placed in an environment with low magnetic noise
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_5","title":"Situations that can make the system unstable","text":"
The vehicle is placed in an environment with high magnetic noise, such as one containing buildings or structures with reinforced steel or other materials that generate electromagnetic waves.
The car is placed in an environment with magnetic markers installed.
"},{"location":"design/autoware-architecture/localization/#situations-where-the-system-becomes-unstable","title":"Situations where the system becomes unstable","text":"
By implementing different modules, various sensor configurations and algorithms can be used.
The localization system can start pose estimation from an ambiguous initial location.
The system can produce a reliable initial location estimation.
The system can manage the state of the initial location estimation (uninitialized, initializable, or non-initializable) and can report to the error monitor.
Two architectures are defined, \"Required\" and \"Recommended\". However, the \"Required\" architecture only contains the inputs and outputs necessary to accept various localization algorithms. To improve the reusability of each module, the required components are defined in the \"Recommended\" architecture section along with a more detailed explanation.
Produces the vehicle velocity, angular velocity, acceleration, angular acceleration, and their covariances
It is possible to create a single module for both twist and acceleration or to create two separate modules - the choice of architecture is up to the developer
The twist estimator produces velocity and angular velocity from internal sensor observation
The accel estimator produces acceleration and angular acceleration from internal sensor observations
Monitors and guarantees the stability and reliability of pose estimation by fusing information obtained from multiple localization modules
Reports error status to the error monitor
"},{"location":"design/autoware-architecture/localization/#tf-tree","title":"TF tree","text":"frame meaning earth ECEF (Earth Centered Earth Fixed\uff09 map Origin of the map coordinate (ex. MGRS origin) viewer User-defined frame for rviz base_link Reference pose of the ego-vehicle (projection of the rear-axle center onto the ground surface) sensor Reference pose of each sensor
Developers can optionally add other frames such as odom or base_footprint as long as the tf structure above is maintained.
The localization module should provide pose, velocity, and acceleration for control, planning, and perception.
Latency and stagger should be sufficiently small or adjustable such that the estimated values can be used for control within the ODD (Operational Design Domain).
The localization module should produce the pose on a fixed coordinate frame.
Sensors should be independent of each other so that they can be easily replaced.
The localization module should provide a status indicating whether or not the autonomous vehicle can operate with the self-contained function or map information.
Tools or manuals should describe how to set proper parameters for the localization module
Valid calibration parameters should be provided to align different frame or pose coordinates and sensor timestamps.
To maintain sufficient pose estimation performance for safe operation, the following metrics are considered:
Safety
The distance traveled within the ODD where pose estimation met the required accuracy, divided by the overall distance traveled within the ODD, as a percentage.
The anomaly detection rate for situations where the localization module cannot estimate pose within the ODD
The accuracy of detecting when the vehicle goes outside of the ODD, as a percentage.
Computational load
Latency
"},{"location":"design/autoware-architecture/localization/#5-interface-and-data-structure","title":"5. Interface and Data Structure","text":""},{"location":"design/autoware-architecture/localization/#6-concerns-assumptions-and-limitations","title":"6. Concerns, Assumptions, and Limitations","text":""},{"location":"design/autoware-architecture/localization/#prerequisites-of-sensors-and-inputs","title":"Prerequisites of sensors and inputs","text":""},{"location":"design/autoware-architecture/localization/#sensor-prerequisites","title":"Sensor prerequisites","text":"
Input data is not defective.
Internal sensor observation such as IMU continuously keeps the proper frequency.
Input data has correct and exact time stamps.
Estimated poses can be inaccurate or unstable if the timestamps are not exact.
Sensors are correctly mounted with exact positioning and accessible from TF.
If the sensor positions are inaccurate, estimation results may be incorrect or unstable.
A sensor calibration framework is required to properly obtain the sensor positions.
Autoware relies on high-definition point cloud maps and vector maps of the driving environment to perform various tasks such as localization, route planning, traffic light detection, and predicting the trajectories of pedestrians and other vehicles.
This document describes the design of map component of Autoware, including its requirements, architecture design, features, data formats, and interface to distribute map information to the rest of autonomous driving stack.
Map should provide two types of information to the rest of the stack:
Semantic information about roads as a vector map
Geometric information about the environment as a point cloud map (optional)
A vector map contains highly accurate information about a road network, lane geometry, and traffic lights. It is required for route planning, traffic light detection, and predicting the trajectories of other vehicles and pedestrians.
A 3D point cloud map is primarily used for LiDAR-based localization and part of perception in Autoware. In order to determine the current position and orientation of the vehicle, a live scan captured from one or more LiDAR units is matched against a pre-generated 3D point cloud map. Therefore, an accurate point cloud map is crucial for good localization results. However, if the vehicle has an alternate localization method with enough accuracy, for example using camera-based localization, point cloud map may not be required to use Autoware.
In addition to above two types of maps, Autoware also requires a supplemental file for specifying the coordinate system of the map in geodetic system.
This diagram describes the high-level architecture of Map component in Autoware.
The Map component consists of the following sub-components:
Point Cloud Map Loading: Load and publish point cloud map
Vector Map Loading: Load and publish vector map
Projection Loading: Load and publish projection information for conversion between local coordinate (x, y, z) and geodetic coordinate (latitude, longitude, altitude)
"},{"location":"design/autoware-architecture/map/#4-component-interface","title":"4. Component interface","text":""},{"location":"design/autoware-architecture/map/#input-to-the-map-component","title":"Input to the map component","text":"
From file system
Point cloud map and its metadata file
Vector map
Projection information
"},{"location":"design/autoware-architecture/map/#output-from-the-map-component","title":"Output from the map component","text":"
To Sensing
Projection information: Used to convert GNSS data from geodetic coordinate system to local coordinate system
To Localization
Point cloud map: Used for LiDAR-based localization
Vector map: Used for localization methods based on road markings, etc
To Perception
Point cloud map: Used for obstacle segmentation by comparing LiDAR and point cloud map
Vector map: Used for vehicle trajectory prediction
To Planning
Vector map: Used for behavior planning
To API layer
Projection information: Used to convert localization results from local coordinate system to geodetic coordinate system
The point cloud map must be supplied as a file with the following requirements:
The point cloud map must be projected on the same coordinate defined in map_projection_loader in order to be consistent with the lanelet2 map and other packages that converts between local and geodetic coordinates. For more information, please refer to the readme of map_projection_loader.
It must be in the PCD (Point Cloud Data) file format, but can be a single PCD file or divided into multiple PCD files.
Each point in the map must contain X, Y, and Z coordinates.
An intensity or RGB value for each point may be optionally included.
It must cover the entire operational area of the vehicle. It is also recommended to include an additional buffer zone according to the detection range of sensors attached to the vehicle.
Its resolution should be at least 0.2 m to yield reliable localization results.
It can be in either local or global coordinates, but must be in global coordinates (georeferenced) to use GNSS data for localization.
For more details on divided map format, please refer to the readme of map_loader in Autoware Universe.
Note
Three global coordinate systems are currently supported by Autoware, including Military Grid Reference System (MGRS), Universal Transverse Mercator (UTM), and Japan Rectangular Coordinate System. However, MGRS is a preferred coordinate system for georeferenced maps. In a map with MGRS coordinate system, the X and Y coordinates of each point represent the point's location within the 100,000-meter square, while the Z coordinate represents the point's elevation.
The vector cloud map must be supplied as a file with the following requirements:
It must be in Lanelet2 format, with additional modifications required by Autoware.
It must contain the shape and position information of lanes, traffic lights, stop lines, crosswalks, parking spaces, and parking lots.
Except at the beginning or end of a road, each lanelet in the map must be correctly connected to its predecessor, successors, left neighbor, and right neighbor.
Each lanelet in the map must contain traffic rule information including its speed limit, right of way, traffic direction, associated traffic lights, stop lines, and traffic signs.
It must cover the entire operational area of the vehicle.
Note that the diagram is for reference. We are planning to update this diagram every release and may have old information between the releases. If you wish to check the latest node diagram use rqt_graph after launching the Autoware.
The Planning component generates the trajectory message that will be subscribed to by the Control component based on the environmental state obtained from the Localization and the Perception components.
The goal of the Planning component is to generate a trajectory (path and velocity) of the ego vehicle that is safe and well-regulated while satisfying the given mission.
Goals:
The basic functions are provided so that a simple ODD can be defined.
The functionality is modularized to accommodate the third-party components. That is, a complicated or realistic ODD needs not be defined by the basic functions provided by default.
The capability is extensible with the third-party components or the decision of human operators.
The mechanism and policy are separated to allow the system or operators to change the behavior of the ego vehicle. Ultimately speaking, the policy can be set to crash an obstacle and the mechanism always follows. Otherwise, the system is not safe from the design point of view.
Non-goals:
The Planning component is not self-contained but can be extended with third parties.
The Planning component is not aimed at the complete functionality and capability.
The Planning component is not designed to always outperform human drivers.
The Planning component is not capable of \u201cnever crashes\u201d.
This diagram describes the high-level architecture of the Planning Component.
The Planning component consists of the following sub-components:
Mission Planning: Calculates the route based on the given goal and map information.
Scenario Planning: Determines the trajectory based on the current scenario, such as Lane Driving or Parking.
Lane Driving: Calculates the trajectory for driving within constructed lanes.
Behavior Planner: Calculates suitable path based on safety considerations and traffic rules.
Motion Planner: Calculates suitable trajectory for the vehicle by taking into account safety factors, vehicle motion considerations, and instructions from the behavior planner.
Parking: Calculates the trajectory for parking in unstructured areas.
Validation: Verifies the safety of the trajectory.
Each component contains some modules that can be dynamically loaded and unloaded based on the situation. For instance, the Behavior Planning component includes modules such as lane change, intersection, and crosswalk modules.
Our planning components are built based on the microautonomy architecture with Autoware. We adopt a modular system framework where the tasks are implemented as modules that can be dynamically loaded and unloaded to achieve different features depending on the given use cases.
This section describes the inputs and outputs of the Planning Component and of its internal modules. See the Planning Component Interface page for the current implementation.
"},{"location":"design/autoware-architecture/planning/#input-to-the-planning-component","title":"Input to the planning component","text":"
From Map
Vector map: Contains all static information about the environment, including lane connection information for route planning, lane geometry for generating a reference path, and traffic rule-related information.
From Perception
Detected object information: Provides real-time information about objects that cannot be known in advance, such as pedestrians and other vehicles. The Planning Component plans maneuvers to avoid collisions with these objects.
Detected obstacle information: Supplies real-time information about the location of obstacles, which is more primitive than Detected Object and used for emergency stops and other safety measures.
Occupancy map information: Offers real-time information about the presence of pedestrians and other vehicles and occluded area information.
Traffic light recognition result: Provides the current state of each traffic light in real time. The Planning Component extracts relevant information for the planned path and determines whether to stop at intersections.
From Localization
Vehicle motion information: Includes the ego vehicle's position, velocity, acceleration, and other motion-related data.
From System
Operation mode: Indicates whether the vehicle is operating in Autonomous mode.
From Human Machine Interface (HMI)
Feature execution: Allows for executing/authorizing autonomous driving operations, such as lane changes or entering intersections, by human operators.
From API Layer
Goal: Represents the final position that the Planning Component aims to reach.
Checkpoint: Represents a midpoint along the route to the destination. This is used during route calculation.
Velocity limit: Sets the maximum speed limit for the vehicle.
"},{"location":"design/autoware-architecture/planning/#output-from-the-planning-component","title":"Output from the planning component","text":"
To Control
Trajectory: Provides a smooth sequence of pose, twist, and acceleration that the Control Component must follow. The trajectory is typically 10 seconds long with a 0.1-second resolution.
Turn Signals: Controls the vehicle's turn indicators, such as right, left, hazard, etc. based on the planned maneuvers.
To System
Diagnostics: Reports the state of the Planning Component, indicating whether the processing is running correctly and whether a safe plan is being generated.
To Human Machine Interface (HMI)
Feature execution availability: Indicates the status of operations that can be executed or are required, such as lane changes or entering intersections.
Trajectory candidate: Shows the potential trajectory that will be executed after the user's execution.
To API Layer
Planning factors: Provides information about the reasoning behind the current planning behavior. This may include the position of target objects to avoid, obstacles that led to the decision to stop, and other relevant information.
"},{"location":"design/autoware-architecture/planning/#internal-interface-in-the-planning-component","title":"Internal interface in the planning component","text":"
Mission Planning to Scenario Planning
Route: Offers guidance for the path that needs to be followed from the starting point to the destination. This path is determined based on information such as lane IDs defined on the map. At the route level, it doesn't explicitly indicate which specific lanes to take, and the route can contain multiple lanes.
Behavior Planning to Motion Planning
Path: Provides a rough position and velocity to be followed by the vehicle. These path points are usually defined with an interval of about 1 meter. Although other interval distances are possible, it may impact the precision or performance of the planning component.
Drivable area: Defines regions where the vehicle can drive, such as within lanes or physically drivable areas. It assumes that the motion planner will calculate the final trajectory within this defined area.
Scenario Planning to Validation
Trajectory: Defines the desired positions, velocities, and accelerations which the Control Component will try to follow. Trajectory points are defined at intervals of approximately 0.1 seconds based on the trajectory velocities.
Validation to Control Component
Trajectory: Same as above but with some additional safety considerations.
"},{"location":"design/autoware-architecture/planning/#how-to-add-new-modules-wip","title":"How to add new modules (WIP)","text":"
As mentioned in the goal session, this planning module is designed to be extensible by third-party components. For specific instructions on how to add new modules and expand its functionality, please refer to the provided documentation or guidelines (WIP).
"},{"location":"design/autoware-architecture/planning/#supported-functions","title":"Supported Functions","text":"Feature Description Requirements Figure Route Planning Plan route from the ego vehicle position to the destination. Reference implementation is in Mission Planner, enabled by launching the mission_planner node. - Lanelet map (driving lanelets) Path Planning from Route Plan path to be followed from the given route. Reference implementation is in Behavior Path Planner. - Lanelet map (driving lanelets) Obstacle Avoidance Plan path to avoid obstacles by steering operation. Reference implementation is in Avoidance, Obstacle Avoidance Planner. Enable flag in parameter: launch obstacle_avoidance_planner true Demonstration Video - objects information Path Smoothing Plan path to achieve smooth steering. Reference implementation is in Obstacle Avoidance Planner. Demonstration Video - Lanelet map (driving lanelet) Narrow Space Driving Plan path to drive within the drivable area. Furthermore, when it is not possible to drive within the drivable area, stop the vehicle to avoid exiting the drivable area. Reference implementation is in Obstacle Avoidance Planner. Demonstration Video - Lanelet map (high-precision lane boundaries) Lane Change Plan path for lane change to reach the destination. Reference implementation is in Lane Change. Demonstration Video - Lanelet map (driving lanelets) Pull Over Plan path for pull over to park at the road shoulder. Reference implementation is in Goal Planner. Demonstration Videos: Simple Pull Over Arc Forward Pull Over Arc Backward Pull Over - Lanelet map (shoulder lane) Pull Out Plan path for pull over to start from the road shoulder. Reference implementation is in Pull Out Module. Demonstration Video: Simple Pull Out Backward Pull Out - Lanelet map (shoulder lane) Path Shift Plan path in lateral direction in response to external instructions. Reference implementation is in Side Shift Module. - None Obstacle Stop Plan velocity to stop for an obstacle on the path. Reference implementation is in Obstacle Stop Planner, Obstacle Cruise Planner. launch obstacle_stop_planner and enable flag: TODO, launch obstacle_cruise_planner and enable flag: TODO Demonstration Video - objects information Obstacle Deceleration Plan velocity to decelerate for an obstacle located around the path. Reference implementation is in Obstacle Stop Planner, Obstacle Cruise Planner. Demonstration Video - objects information Adaptive Cruise Control Plan velocity to follow the vehicle driving in front of the ego vehicle. Reference implementation is in Obstacle Stop Planner, Obstacle Cruise Planner. - objects information Decelerate for cut-in vehicles Plan velocity to avoid a risk for cutting-in vehicle to ego lane. Reference implementation is in Obstacle Cruise Planner. - objects information Surround Check at starting Plan velocity to prevent moving when an obstacle exists around the vehicle. Reference implementation is in Surround Obstacle Checker. Enable flag in parameter: use_surround_obstacle_check true in tier4_planning_component.launch.xml Demonstration Video - objects information Curve Deceleration Plan velocity to decelerate the speed on a curve. Reference implementation is in Motion Velocity Smoother. - None Curve Deceleration for Obstacle Plan velocity to decelerate the speed on a curve for a risk of obstacle collision around the path. Reference implementation is in Obstacle Velocity Limiter. Demonstration Video - objects information - Lanelet map (static obstacle) Crosswalk Plan velocity to stop or decelerate for pedestrians approaching or walking on a crosswalk. Reference implementation is in Crosswalk Module. Demonstration Video - objects information - Lanelet map (pedestrian crossing) Intersection Oncoming Vehicle Check Plan velocity for turning right/left at intersection to avoid a risk with oncoming other vehicles. Reference implementation is in Intersection Module. Demonstration Video - objects information - Lanelet map (intersection lane and yield lane) Intersection Blind Spot Check Plan velocity for turning right/left at intersection to avoid a risk with other vehicles or motorcycles coming from behind blind spot. Reference implementation is in Blind Spot Module. Demonstration Video - objects information - Lanelet map (intersection lane) Intersection Occlusion Check Plan velocity for turning right/left at intersection to avoid a risk with the possibility of coming vehicles from occlusion area. Reference implementation is in Intersection Module. Demonstration Video - objects information - Lanelet map (intersection lane) Intersection Traffic Jam Detection Plan velocity for intersection not to enter the intersection when a vehicle is stopped ahead for a traffic jam. Reference implementation is in Intersection Module. Demonstration Video - objects information - Lanelet map (intersection lane) Traffic Light Plan velocity for intersection according to a traffic light signal. Reference implementation is in Traffic Light Module. Demonstration Video - Traffic light color information Run-out Check Plan velocity to decelerate for the possibility of nearby objects running out into the path. Reference implementation is in Run Out Module. Demonstration Video - objects information Stop Line Plan velocity to stop at a stop line. Reference implementation is in Stop Line Module. Demonstration Video - Lanelet map (stop line) Occlusion Spot Check Plan velocity to decelerate for objects running out from occlusion area, for example, from behind a large vehicle. Reference implementation is in Occlusion Spot Module. Demonstration Video - objects information - Lanelet map (private/public lane) No Stop Area Plan velocity not to stop in areas where stopping is prohibited, such as in front of the fire station entrance. Reference implementation is in No Stopping Area Module. - Lanelet map (no stopping area) Merge from Private Area to Public Road Plan velocity for entering the public road from a private driveway to avoid a risk of collision with pedestrians or other vehicles. Reference implementation is in Merge from Private Area Module. - objects information - Lanelet map (private/public lane) WIP Speed Bump Plan velocity to decelerate for speed bumps. Reference implementation is in Speed Bump Module. Demonstration Video - Lanelet map (speed bump) Detection Area Plan velocity to stop at the corresponding stop when an object exist in the designated detection area. Reference implementation is in Detection Area Module. Demonstration Video - Lanelet map (detection area) No Drivable Lane Plan velocity to stop before exiting the area designated by ODD (Operational Design Domain) or stop the vehicle if autonomous mode started in out of ODD lane. Reference implementation is in No Drivable Lane Module. - Lanelet map (no drivable lane) Collision Detection when deviating from lane Plan velocity to avoid conflict with other vehicles driving in the another lane when the ego vehicle is deviating from own lane. Reference implementation is in Out of Lane Module. - objects information - Lanelet map (driving lane) WIP Parking Plan path and velocity for given goal in parking area. Reference implementation is in Free Space Planner. Demonstration Video - objects information - Lanelet map (parking area) Autonomous Emergency Braking (AEB) Perform an emergency stop if a collision with an object ahead is anticipated. It is noted that this function is expected as a final safety layer, and this should work even in the event of failures in the Localization or Perception system. Reference implementation is in Out of Lane Module. - Primitive objects Minimum Risk Maneuver (MRM) Provide appropriate MRM (Minimum Risk Maneuver) instructions when a hazardous event occurs. For example, when a sensor trouble found, send an instruction for emergency braking, moderate stop, or pulling over to the shoulder, depending on the severity of the situation. Reference implementation is in TODO - TODO WIP Trajectory Validation Check the planned trajectory is safe. If it is unsafe, take appropriate action, such as modify the trajectory, stop sending the trajectory or report to the autonomous driving system. Reference implementation is in Planning Validator. - None Running Lane Map Generation Generate lane map from localization data recorded in manual driving. Reference implementation is in WIP - None WIP Running Lane Optimization Optimize the centerline (reference path) of the map to make it smooth considering the vehicle kinematics. Reference implementation is in Static Centerline Optimizer. - Lanelet map (driving lanes) WIP"},{"location":"design/autoware-architecture/planning/#reference-implementation","title":"Reference Implementation","text":"
The following diagram describes the reference implementation of the Planning component. By adding new modules or extending the functionalities, various ODDs can be supported.
Note that some implementation does not adhere to the high-level architecture design and require updating.
For more details, please refer to the design documents in each package.
mission_planner: calculate route from start to goal based on the map information.
behavior_path_planner: calculates path and drivable area based on the traffic rules.
lane_following
lane_change
avoidance
pull_over
pull_out
side_shift
behavior_velocity_planner: calculates max speed based on the traffic rules.
detection_area
blind_spot
cross_walk
stop_line
traffic_light
intersection
no_stopping_area
virtual_traffic_light
occlusion_spot
run_out
no_drivable_lane
obstacle_avoidance_planner: calculate path shape under obstacle and drivable area constraints
surround_obstacle_checker: keeps the vehicle being stopped when there are obstacles around the ego-vehicle. It works only when the vehicle is stopped.
obstacle_stop_planner: When there are obstacles on or near the trajectory, it calculates the maximum velocity of the trajectory points depending on the situation: stopping, slowing down, or adaptive cruise (following the car).
stop
slow_down
adaptive_cruise
costmap_generator: generates a costmap for path generation from dynamic objects and lane information.
freespace_planner: calculates trajectory considering the feasibility (e.g. curvature) for the freespace scene. Algorithms are described here.
scenario_selector : chooses a trajectory according to the current scenario.
external_velocity_limit_selector: takes an appropriate velocity limit from multiple candidates.
motion_velocity_smoother: calculates final velocity considering velocity, acceleration, and jerk constraints.
"},{"location":"design/autoware-architecture/planning/#important-parameters","title":"Important Parameters","text":"Package Parameter Type Description obstacle_stop_plannerstop_planner.stop_position.max_longitudinal_margin double distance between the ego and the front vehicle when stopping (when cruise_planner_type:=obstacle_stop_planner) obstacle_cruise_plannercommon.safe_distance_margin double distance between the ego and the front vehicle when stopping (when cruise_planner_type:=obstacle_cruise_planner) behavior_path_planneravoidance.avoidance.lateral.lateral_collision_margin double minimum lateral margin to obstacle on avoidance behavior_path_planneravoidance.avoidance.lateral.lateral_collision_safety_buffer double additional lateral margin to obstacle if possible on avoidance obstacle_avoidance_planneroption.enable_outside_drivable_area_stop bool If set true, a stop point will be inserted before the path footprint is outside the drivable area."},{"location":"design/autoware-architecture/planning/#notation","title":"Notation","text":""},{"location":"design/autoware-architecture/planning/#1-self-crossing-road-and-overlapped","title":"[1] self-crossing road and overlapped","text":"
To support the self-crossing road and overlapped road in the opposite direction, each planning module has to meet the specifications
Currently, the supported modules are as follows.
lane_following (in behavior_path_planner)
detection_area (in behavior_velocity_planner)
stop_line (in behavior_velocity_planner)
virtual_traffic_light (in behavior_velocity_planner)
obstacle_avoidance_planner
obstacle_stop_planner
motion_velocity_smoother
"},{"location":"design/autoware-architecture/planning/#2-size-of-path-points","title":"[2] Size of Path Points","text":"
Some functions do not support paths with only one point. Therefore, each modules should generate the path with more than two path points.
Abstraction of data formats to enable usage of sensors from various vendors
Perform common/primitive sensor data processing required by each component
"},{"location":"design/autoware-architecture/sensing/#inputs","title":"Inputs","text":""},{"location":"design/autoware-architecture/sensing/#input-types","title":"Input types","text":"Sensor Data Message Type Point cloud (Lidars, depth cameras, etc.) sensor_msgs/msg/PointCloud2.msg Image (RGB, monochrome, depth, etc. cameras) sensor_msgs/msg/Image.msg Radar scan radar_msgs/msg/RadarScan.msg Radar tracks radar_msgs/msg/RadarTracks.msg GNSS-INS position sensor_msgs/msg/NavSatFix.msg GNSS-INS orientation autoware_sensing_msgs/GnssInsOrientationStamped.msg GNSS-INS velocity geometry_msgs/msg/TwistWithCovarianceStamped.msg GNSS-INS acceleration geometry_msgs/msg/AccelWithCovarianceStamped.msg Ultrasonics sensor_msgs/msg/Range.msg"},{"location":"design/autoware-architecture/sensing/#design-by-data-types","title":"Design by data-types","text":"
GNSS/INS data pre-processing design
Image pre-processing design
Point cloud pre-processing design
Radar data pre-processing design
Ultrasonics data pre-processing design
"},{"location":"design/autoware-architecture/sensing/data-types/gnss-ins-data/","title":"GNSS/INS data pre-processing design","text":""},{"location":"design/autoware-architecture/sensing/data-types/gnss-ins-data/#gnssins-data-pre-processing-design","title":"GNSS/INS data pre-processing design","text":"
In the ideal case, the driver is expected to output a point cloud with the PointXYZIRCADT point type.
name datatype derived description XFLOAT32false X position YFLOAT32false Y position ZFLOAT32false Z position I (intensity) UINT8false Measured reflectivity, intensity of the point R (return type) UINT8false Laser return type for dual return lidars C (channel) UINT16false Vertical channel id of the laser that measured the point A (azimuth) FLOAT32trueatan2(Y, X), Horizontal angle from the front of the lidar to the point D (distance) FLOAT32truehypot(X, Y, Z), Euclidean distance of the point to lidar T (time) UINT32false Nanoseconds passed since the time of the header when this point was measured
Note
A (azimuth) and D (distance) fields are derived fields. They are provided by the driver to reduce the computational load on some parts of the perception stack.
Note
If the Motion Distortion Corrector Filter won't be used, the T (time) field can be omitted, PointXYZIRCAD point type can be used.
Warning
Autoware will support conversion from PointXYZI to PointXYZIRC or PointXYZIRCAD (with channel and return is set to 0) for prototyping purposes. However, this conversion is not recommended for production use since it's not efficient.
We will use following ranges for intensity, compatible with the VLP16 User Manual:
Quoting from the VLP-16 User Manual:
For each laser measurement, a reflectivity byte is returned in addition to distance. Reflectivity byte values are segmented into two ranges, allowing software to distinguish diffuse reflectors (e.g. tree trunks, clothing) in the low range from retroreflectors (e.g. road signs, license plates) in the high range. A retroreflector reflects light back to its source with a minimum of scattering. The VLP-16 provides its own light, with negligible separation between transmitting laser and receiving detector, so retroreflecting surfaces pop with reflected IR light compared to diffuse reflectors that tend to scatter reflected energy.
Diffuse reflectors report values from 0 to 100 for reflectivities from 0% to 100%.
Retroreflectors report values from 101 to 255, where 255 represents an ideal reflection.
In a typical point cloud without retroreflectors, all intensity points will be between 0 and 100.
Retroreflective Gradient road sign, Image Source
But in a point cloud with retroreflectors, the intensity points will be between 0 and 255.
"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#intensity-mapping-for-other-lidar-brands","title":"Intensity mapping for other lidar brands","text":""},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#hesai-pandarxt16","title":"Hesai PandarXT16","text":"
Hesai Pandar XT16 User Manual
This lidar has 2 modes for reporting reflectivity:
Linear mapping
Non-linear mapping
If you are using linear mapping mode, you should map from [0, 255] to [0, 100] when constructing the point cloud.
If you are using non-linear mapping mode, you should map (hesai to autoware)
Software User Manual v2.0.0 for all Ouster sensors
In the manual it is stated:
Reflectivity [16 bit unsigned int] - sensor Signal Photons measurements are scaled based on measured range and sensor sensitivity at that range, providing an indication of target reflectivity. Calibration of this measurement has not currently been rigorously implemented, but this will be updated in a future firmware release.
So it is advised to map the 16 bit reflectivity to [0, 100] range.
I couldn't get the english user manual, link of website
In a user manual I was able to find it says:
Byte 7 represents echo strength, and the value range is 0-255. (Echo strength can reflect the energy reflection characteristics of the measured object in the actual measurement environment. Therefore, the echo strength can be used to distinguish objects with different reflection characteristics.)
So it is advised to map the [0, 255] to [0, 100] range.
Various lidars support multiple return modes. Velodyne lidars support Strongest and Last return modes.
In the PointXYZIRCT and PointXYZIRC types, R field represents return mode with an UINT8.
R (return type) Description 0 Unknown / Not Marked 1 Strongest 2 Last"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#channel","title":"Channel","text":"
The channel field is used to identify the vertical channel of the laser that measured the point. In various lidar manuals or literature, it can also be called laser id, ring, laser line.
For Velodyne VLP-16, there are 16 channels. Default order of channels in drivers are generally in firing order.
In the PointXYZIRCT and PointXYZIRC types, C field represents the vertical channel id with an UINT16.
"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#solid-state-and-petal-pattern-lidars","title":"Solid state and petal pattern lidars","text":"
Warning
This section is subject to change. Following are suggestions and open for discussion.
For solid state lidars that have lines, assign row number as the channel id.
In lidar point clouds, each point measurement can have its individual time stamp. This information can be used to eliminate the motion blur that is caused by the movement of the lidar during the scan.
The header contains a Time field. The time field has 2 components:
Field Type Description secint32 Unix time (seconds elapsed since January 1, 1970) nanosecuint32 Nanoseconds elapsed since the sec field
The header of the point cloud message is expected to have the time of the earliest point it has.
Note
The sec field is int32 in ROS 2 humble. The largest value it can represent is 2^31 seconds, it is subject to year 2038 problems. We will wait for actions on ROS 2 community side.
More info at: https://github.com/ros2/rcl_interfaces/issues/85
"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#individual-point-time","title":"Individual point time","text":"
Each PointXYZIRCT point type has the T field for representing the nanoseconds passed since the first-shot point of the point cloud.
To calculate exact time each point was shot, the T nanoseconds are added to the header time.
Note
The T field is uint32 type. The largest value it can represent is 2^32 nanoseconds, which equates to roughly 4.29 seconds. Usual point clouds don't last more than 100ms for full cycle. So this field should be enough.
"},{"location":"design/autoware-architecture/sensing/data-types/radar-data/","title":"Radar data pre-processing design","text":""},{"location":"design/autoware-architecture/sensing/data-types/radar-data/#radar-data-pre-processing-design","title":"Radar data pre-processing design","text":"
Warning
Under Construction
"},{"location":"design/autoware-architecture/sensing/data-types/ultrasonics-data/","title":"Ultrasonics data pre-processing design","text":""},{"location":"design/autoware-architecture/sensing/data-types/ultrasonics-data/#ultrasonics-data-pre-processing-design","title":"Ultrasonics data pre-processing design","text":"
The Vehicle Interface component provides an interface between Autoware and a vehicle that passes control signals to the vehicle\u2019s drive-by-wire system and receives vehicle information that is passed back to Autoware.
The Vehicle Interface component converts Autoware commands to a vehicle-specific format and converts vehicle status in a vehicle-specific format to Autoware messages.
The interface between Autoware and the Vehicle component is abstracted and independent of hardware.
The interface is extensible such that additional vehicle-specific commands can be easily added. For example, headlight control.
Non-goals:
Accuracy of responses from the vehicle will not be defined, but example accuracy requirements from reference designs are provided as examples.
The Vehicle Interface component consists of the following components:
A Raw Vehicle Command Converter component that will pass through vehicle commands from the Control component if velocity/acceleration control is supported by the drive-by-wire system. Otherwise, the Control commands will be modified according to the control method (eg: converting a target acceleration from the Control component to a vehicle specific accel/brake pedal value through the use of an acceleration map)
A Vehicle Interface component (vehicle specific) that acts as an interface between Autoware and a vehicle to communicate control signals and to obtain information about the vehicle (steer output, tyre angle etc)
Each component contains static nodes of Autoware, while each module can be dynamically loaded and unloaded (corresponding to C++ classes). The mechanism of the Vehicle Interface component is depicted by the following figures:
The Vehicle Interface component can provide the following features in functionality and capability:
Basic functions
Converting Autoware control commands to vehicle specific command
Converting vehicle specific status information (velocity, steering) to Autoware status message
Diagnostics
List available features
Provide a warning if the Control component tries to use a feature that is not available in the Vehicle Interface component
Additional functionality and capability features may be added, depending on the vehicle hardware. Some example features are listed below:
Safety features
Disengage autonomous driving via manual intervention.
This can be done through the use of an emergency disengage button, or by a safety driver manually turning the steering wheel or pressing the brake
Optional controls
Turn indicator
Handbrake
Headlights
Hazard lights
Doors
Horn
Wipers
"},{"location":"design/autoware-architecture/vehicle/#4-interface-and-data-structure","title":"4. Interface and Data Structure","text":"
The interface of the Vehicle Interface component for other components running in the same process space to access the functionality and capability of the Vehicle Interface component is defined as follows.
From Control
Actuation Command
target acceleration, braking, and steering angle
From Planning
Vehicle Specific Commands (optional and a separate message for each type)
Shift
Door
Wiper
etc
From the vehicle
Vehicle status messages
Vehicle-specific format messages for conversion into Autoware-specific format messages
Velocity status
Steering status (optional)
Shift status (optional)
Turn signal status (optional)
Actuation status (optional)
The output interface of the Vehicle Interface component:
Vehicle control messages to the vehicle
Control signals to drive the vehicle
Depends on the vehicle type/protocol, but should include steering and velocity commands at a minimum
Vehicle status messages to Autoware
Actuation Status
Acceleration, brake, steering status
Vehicle odometry (output to Localization)
Vehicle twist information
Control mode
Information about whether the vehicle is under autonomous control or manual control
Shift status (optional)
Vehicle shift status
Turn signal status (optional)
Vehicle turn signal status
The data structure for the internal representation of semantics for the objects and trajectories used in the Vehicle Interface component is defined as follows:
"},{"location":"design/autoware-architecture/vehicle/#5-concerns-assumptions-and-limitations","title":"5. Concerns, Assumptions, and Limitations","text":"
Concerns
Architectural trade-offs and scalability
Assumptions
-
Limitations
"},{"location":"design/autoware-architecture/vehicle/#6-examples-of-accuracy-requirements-by-odd","title":"6. Examples of accuracy requirements by ODD","text":""},{"location":"design/autoware-concepts/","title":"Autoware concepts","text":""},{"location":"design/autoware-concepts/#autoware-concepts","title":"Autoware concepts","text":"
Autoware is the world\u2019s first open-source software for autonomous driving systems. Autoware provides value for both The technology developers of autonomous driving systems can create new components based on Autoware. The service operators of autonomous driving systems, on the other hand, can select appropriate technology components with Autoware. This is enabled by the microautonomy architecture that modularizes its software stack into the core and universe subsystems (modules).
Autoware uses a pipeline architecture to enable the development of autonomous driving systems. The pipeline architecture used in Autoware consists of components similar to three-layer-architecture. And they run in parallel. There are 2 main modules: the Core and the Universe. The components in these modules are designed to be extensible and reusable. And we call it microautonomy architecture.
The Core module contains basic runtimes and technology components that satisfy the basic functionality and capability of sensing, computing, and actuation required for autonomous driving systems. AWF develops and maintains the Core module with their architects and leading members through their working groups. Anyone can contribute to the Core but the PR(Pull Request) acceptance criteria is more strict compared to the Universe.
The Universe modules are extensions to the Core module that can be provided by the technology developers to enhance the functionality and capability of sensing, computing, and actuation. AWF provides the base Universe module to extend from. A key feature of the microautonomy architecture is that the Universe modules can be contributed to by any organization and individual. That is, you can even create your Universe and make it available for the Autoware community and ecosystem. AWF is responsible for quality control of the Universe modules through their development process. As a result, there are multiple types of the Universe modules - some are verified and validated by AWF and others are not. It is up to the users of Autoware which Universe modules are selected and integrated to build their end applications.
The interface design is the most essential piece of the microautonomy architecture, which is classified into internal and external interfaces. The component interface is designed for the components in a Universe module to communicate with those in other modules, including the Core module, within Autoware internally. The AD(Autonomous Driving) API, on the other hand, is designed for the applications of Autoware to access the technology components in the Core and Universe modules of Autoware externally. Designing solid interfaces, the microautonomy architecture is made possible with AWF's partners, and at the same time is made feasible for the partners.
A grand challenge of the microautonomy architecture is to achieve real-time capability, which guarantees all the technology components activated in the system to predictably meet timing constraints (given deadlines). In general, it is difficult, if not impossible, to tightly estimate the worst-case execution times (WCETs) of components.
In addition, it is also difficult, if not impossible, to tightly estimate the end-to-end latency of components connected by a DAG. Autonomous driving systems based on the microautonomy architecture, therefore, must be designed to be fail-safe but not never-fail. We accept that the timing constraints may be violated (the given deadlines may be missed) as far as the overrun is taken into account. The overrun handlers are two-fold: (i) platform-defined and (ii) user-defined. The platform-defined handler is implemented as part of the platform by default, while the user-defined handler can overwrite it or add a new handler to the system. This is what we call \u201cfail-safe\u201d on a timely basis.
"},{"location":"design/autoware-concepts/#requirements-and-roadmap","title":"Requirements and roadmap","text":"
Goals:
All open-source
Use case driven
Real-time (predictable) framework with overrun handling
Code quality
"},{"location":"design/autoware-concepts/difference-from-ai-and-auto/","title":"How is Autoware Core/Universe different from Autoware.AI and Autoware.Auto?","text":""},{"location":"design/autoware-concepts/difference-from-ai-and-auto/#how-is-autoware-coreuniverse-different-from-autowareai-and-autowareauto","title":"How is Autoware Core/Universe different from Autoware.AI and Autoware.Auto?","text":"
Autoware is the world's first \"all-in-one\" open-source software for self-driving vehicles. Since it was first released in 2015, there have been multiple releases made with differing underlying concepts, each one aimed at improving the software.
Autoware.AI is the first distribution of Autoware that was released based on ROS 1. The repository contains a variety of packages covering different aspects of autonomous driving technologies - sensing, actuation, localization, mapping, perception and planning.
While it was successful in attracting many developers and contributions, it was difficult to improve Autoware.AI's capabilities for a number of reasons:
A lack of concrete architecture design leading to a lot of built-up technical debt, such as tight coupling between modules and unclear module responsibility.
Differing coding standards for each package, with very low test coverage.
Furthermore, there was no clear definition of the conditions under which an Autoware-enabled autonomous vehicle could operate, nor of the use cases or situations supported (eg: the ability to overtake a stationary vehicle).
From the lessons learned from Autoware.AI development, a different development process was taken for Autoware.Auto to develop a ROS 2 version of Autoware.
Warning
Autoware.AI is currently in maintenance mode and will reach end-of-life at the end of 2022.
Autoware.Auto is the second distribution of Autoware that was released based on ROS 2. As part of the transition to ROS 2, it was decided to avoid simply porting Autoware.AI from ROS 1 to ROS 2. Instead, the codebase was rewritten from scratch with proper engineering practices, including defining target use cases and ODDs (eg: Autonomous Valet Parking [AVP], Cargo Delivery, etc.), designing a proper architecture, writing design documents and test code.
Autoware.Auto development seemed to work fine initially, but after completing the AVP and and Cargo Delivery ODD projects, we started to see the following issues:
The barrier to new engineers was too high.
A lot of work was required to merge new features into Autoware.Auto, and so it was difficult for researchers and students to contribute to development.
As a consequence, most Autoware.Auto developers were from companies in the Autoware Foundation and so there were very few people who were able to add state-of-the-art features from research papers.
Making large-scale architecture changes was too difficult.
To try out experimental architecture, there was a very large overhead involved in keeping the main branch stable whilst also making sure that every change satisfied the continuous integration requirements.
In order to address the issues with Autoware.Auto development, the Autoware Foundation decided to create a new architecture called Autoware Core/Universe.
Autoware Core carries over the original policy of Autoware.Auto to be a stable and well-tested codebase. Alongside Autoware Core is a new concept called Autoware Universe, which acts as an extension of Autoware Core with the following benefits:
Users can easily replace a Core component with a Universe equivalent in order to use more advanced features, such as a new Localization or Perception algorithm.
Code quality requirements for Universe are more relaxed to make it easier for new developers, students and researchers to contribute, but will still be stricter than the requirements for Autoware.AI.
Any advanced features added to Universe that are useful to the wider Autoware community will be reviewed and considered for potential inclusion in the main Autoware Core codebase.
This way, the primary requirement of having a stable and safe autonomous driving system can be achieved, whilst simultaneously enabling access to state-of-the-art features created by third-party contributors. For more details about the design of Autoware Core/Universe, refer to the Autoware concepts documentation page.
Autoware defines three categories of interfaces. The first one is Autoware AD API for operating the vehicle from outside the autonomous driving system such as the Fleet Management System (FMS) and Human Machine Interface (HMI) for operators or passengers. The second one is Autoware component interface for components to communicate with each other. The last one is the local interface used inside the component.
The components of Autoware are connected via the component interface. Each component uses the interface to provide functionality and to access other components. AD API implementation is also a component. Since the functional elements required for AD API are defined as the component interface, other components do not need to consider AD API directly. Tools for evaluation and debugging, such as simulators, access both AD API and the component interface.
The component interface has a hierarchical specification. The top-level architecture consists of some components. Each component has some options of the next-level architecture. Developers select one of them when implementing the component. The simplest next-level architecture is monolithic. This is an all-in-one and black box implementation, and is suitable for small group development, prototyping, and very complex functions. Others are arbitrary architecture consists of sub-components and have advantages for large group development. A sub-component can be combined with others that adopt the same architecture. Third parties can define and publish their own architecture and interface for open source development. It is desirable to propose them for standardization if they are sufficiently evaluated.
As shown in the table below, interfaces are classified into four communication methods to define their behavior. Function Call is a request-response communication and is used for processing that requires immediate results. The others are publish-subscribe communication. Notification is used to process data that changes with some event, typically a callback. Streams handle continuously changing data. Reliable Stream expects all data to arrive without loss, Realtime Stream expects the latest data to arrive with low delay.
These methods are provided as services or topics of ROS since Autoware is developed using ROS and mainly communicates with its packages. On the other hand, FMS and HMI are often implemented without ROS, Autoware is also expected to communicate with applications that do not use ROS. It is wasteful for each of these applications to have an adapter for Autoware, and a more suitable means of communication is required. HTTP and MQTT are suggested as additional options because these protocols are widely used and can substitute the behavior of services and topics. In that case, text formats such as JSON where field names are repeated in an array of objects, are inefficient and it is necessary to consider the serialization.
The name of the interface must be /<component name>/api/<interface name>, where <component name> is the name of the component. For an AD API component, omit this part and start with /api. The <interface name> is an arbitrary string separated by slashes. Note that this rule causes a restriction that the namespace api must not be used as a name other than AD API and the component interface.
The following are examples of correct interface names for AD API and the component interface:
/api/autoware/state
/api/autoware/engage
/planning/api/route/set
/vehicle/api/status
The following are examples of incorrect interface names for AD API and the component interface:
It is recommended to log the interface for analysis of vehicle behavior. If logging is needed, rosbag is available for topics, and use logger in rclcpp or rclpy for services. Typically, create a wrapper for service and client classes that logs when a service is called.
For each API, consider the restrictions such as following and describe them if necessary.
Services:
response time
pre-condition
post-condition
execution order
concurrent execution
Topics:
recommended delay range
maximum delay
recommended frequency range
minimum frequency
default frequency
"},{"location":"design/autoware-interfaces/#data-structure","title":"Data structure","text":""},{"location":"design/autoware-interfaces/#data-type-definition","title":"Data type definition","text":"
Do not share the types in AD API unless they are obviously the same to avoid changes in one API affecting another. Also, implementation-dependent types, including the component interface, should not be used in AD API for the same reason. Use the type in AD API in implementation, or create the same type and copy the data to convert the type.
"},{"location":"design/autoware-interfaces/#constants-and-enumeration","title":"Constants and enumeration","text":"
Since ROS don't support enumeration, use constants instead. The default value of type such as zero and empty string should not be used to detect that a variable is unassigned. Alternatively, assign it a dedicated name to indicate that it is undefined. If one type has multiple enumerations, comment on the correspondence between constants and variables. Do not use enumeration values directly, as assignments are subject to change when the version is updated.
Clarify what the timestamp indicates. for example, send time, measurement time, update time, etc. Consider having multiple timestamps if necessary. Use std_msgs/msg/Header when using ROS transform. Also consider whether the header is common to all data, independent for each data, or additional timestamp is required.
The interfaces whose communication method is Function Call use a common response status to unify the error format. These interfaces should include a variable of ResponseStatus with the name status in the response. See autoware_adapi_v1_msgs/msg/ResponseStatus for details.
"},{"location":"design/autoware-interfaces/#concerns-assumptions-and-limitations","title":"Concerns, assumptions and limitations","text":"
The applications use the version information provided by AD API to check compatibility. Unknown versions are also treated as available as long as the major versions match (excluding major version 0). Compatibility between AD API and the component interface is assumed to be maintained by the version management system.
If an unintended behavior of AD API is detected, the application should take appropriate action. Autoware tries to keep working as long as possible, but it is not guaranteed to be safe. Safety should be considered for the entire system, including the applications.
"},{"location":"design/autoware-interfaces/ad-api/","title":"Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/#autoware-ad-api","title":"Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/#overview","title":"Overview","text":"
Autoware AD API is the interface for operating the vehicle from outside the autonomous driving system. See here for the overall interface design of Autoware.
The user stories are service scenarios that AD API assumes. AD API is designed based on these scenarios. Each scenario is realized by a combination of use cases described later. If there are scenarios that cannot be covered, please discuss adding a user story.
Use cases are partial scenarios derived from the user story and generically designed. Service providers can combine these use cases to define user stories and check if AD API can be applied to their own scenarios.
Request to cooperate (RTC) is a feature that enables a human operator to support the decision in autonomous driving mode. Autoware usually drives the vehicle using its own decisions, but the operator may prefer to make their decisions in experiments and complex situations.
The planning component manages each situation that requires decision as a scene. Each scene has an ID that doesn't change until the scene is completed or canceled. The operator can override the decision of the target scene using this ID. In practice, the user interface application can hides the specification of the ID and provides an abstracted interface to the operator.
For example, in the situation in the diagram below, vehicle is expected to make two lane changes and turning left at the intersection. Therefore the planning component generates three scene instances for each required action, and each scene instance will wait for the decision to be made, in this case \"changing or keeping lane\" and \"turning left or waiting at the intersection\". Here Autoware decides not to change lanes a second time due to the obstacle, so the vehicle will stop there. However, operator could overwrite that decision through RTC function and force the lane change so that vehicle could reach to it's goal. Using RTC, the operator can override these decisions to continue driving the vehicle to the goal.
Modules that support RTC have the operator decision and cooperation policy in addition to the module decision as shown below. These modules use the merged decision that is determined by these values when planning vehicle behavior. See decisions section for details of these values. The cooperation policy is used when there is no operator decision and has a default value set by the system settings. If the module supports RTC, these information are available in velocity factors or steering factors as cooperation status.
This is an example sequence that overrides the scene decision to force a lane change. It is for the second scene in the diagram in the architecture section. Here let's assume the cooperation policy is set to optional, see the decisions section described later for details.
A planning module creates a scene instance with unique ID when approaching a place where a lane change is needed.
The scene instance generates the module decision from the current situation. In this case, the module decision is not to do a lane change due to the obstacle.
The scene instance generates the merged decision. At this point, there is no operator decision yet, so it is based on the module decision.
The scene instance plans the vehicle to keep the lane according to the merged decision.
The scene instance sends a cooperation status.
The operator receives the cooperation status.
The operator sends a cooperation command to override the module decision and to do a lane change.
The scene instance receives the cooperation command and update the operator decision.
The scene instance updates the module decision from the current situation.
The scene instance updates the merged decision. It is based on the operator decision received.
The scene instance plans the vehicle to change the lane according to the merged decision.
The merged decision is determined by the module decision, operator decision, and cooperation policy, each of which takes the value shown in the table below.
The meanings of these values are as follows. Note that the cooperation policy is common per module, so changing it will affect all scenes in the same module.
Value Description deactivate An operator/module decision to plan vehicle behavior with priority on safety. activate An operator/module decision to plan vehicle behavior with priority on driving. autonomous An operator decision that follows the module decision. none An initial value for operator decision, indicating that there is no operator decision yet. required A policy that requires the operator decision to continue driving. optional A policy that does not require the operator decision to continue driving.
The following flow is how the merged decision is determined.
This is an example of cooperation for lane change module. The behaviors by the combination of decisions are as follows.
Operator decision Policy Module decision Description deactivate - - The operator instructs to keep lane regardless the module decision. So the vehicle keeps the lane by the operator decision. activate - - The operator instructs to change lane regardless the module decision. So the vehicle changes the lane by the operator decision. autonomous - deactivate The operator instructs to follow the module decision. So the vehicle keeps the lane by the module decision. autonomous - activate The operator instructs to follow the module decision. So the vehicle changes the lane by the module decision. none required - The required policy is used because no operator instruction. So the vehicle keeps the lane by the cooperation policy. none optional deactivate The optional policy is used because no operator instruction. So the vehicle keeps the lane by the module decision. none optional activate The optional policy is used because no operator instruction. So the vehicle change the lane by the module decision."},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/","title":"Fail-safe","text":""},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/#fail-safe","title":"Fail-safe","text":""},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/#related-api","title":"Related API","text":"
This API manages the behavior related to the abnormality of the vehicle. It provides the state of Request to Intervene (RTI), Minimal Risk Maneuver (MRM) and Minimal Risk Condition (MRC). As shown below, Autoware has the gate to switch between the command during normal operation and the command during abnormal operation. For safety, Autoware switches the operation to MRM when an abnormality is detected. Since the required behavior differs depending on the situation, MRM is implemented in various places as a specific mode in a normal module or as an independent module. The fail-safe module selects the behavior of MRM according to the abnormality and switches the gate output to that command.
The MRM state indicates whether MRM is operating. This state also provides success or failure. Generally, MRM will switch to another behavior if it fails.
State Description NONE MRM is not operating. OPERATING MRM is operating because an abnormality has been detected. SUCCEEDED MRM succeeded. The vehicle is in a safe condition. FAILED MRM failed. The vehicle is still in an unsafe condition."},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/#behavior","title":"Behavior","text":"
There is a dependency between MRM behaviors. For example, it switches from a comfortable stop to a emergency stop, but not the other way around. This is service dependent. Autoware supports the following transitions by default.
State Description NONE MRM is not operating or is operating but no special behavior is required. COMFORTABLE_STOP The vehicle will stop quickly with a comfortable deceleration. EMERGENCY_STOP The vehicle will stop immediately with as much deceleration as possible."},{"location":"design/autoware-interfaces/ad-api/features/interface/","title":"Interface","text":""},{"location":"design/autoware-interfaces/ad-api/features/interface/#interface","title":"Interface","text":""},{"location":"design/autoware-interfaces/ad-api/features/interface/#related-api","title":"Related API","text":"
This API provides the interface version of the set of AD APIs. It follows Semantic Versioning in order to provide an intuitive understanding of the changes between versions.
This API manages the initialization of localization. Autoware requires a global pose as the initial guess for localization.
"},{"location":"design/autoware-interfaces/ad-api/features/localization/#states","title":"States","text":"State Description UNINITIALIZED Localization is not initialized. Waiting for a global pose as the initial guess. INITIALIZING Localization is initializing. INITIALIZED Localization is initialized. Initialization can be requested again if necessary."},{"location":"design/autoware-interfaces/ad-api/features/motion/","title":"Motion","text":""},{"location":"design/autoware-interfaces/ad-api/features/motion/#motion","title":"Motion","text":""},{"location":"design/autoware-interfaces/ad-api/features/motion/#related-api","title":"Related API","text":"
This API manages the current behavior of the vehicle. Applications can notify the vehicle behavior to the people around and visualize it for operator and passengers.
The motion state manages the stop and start of the vehicle. Once the vehicle has stopped, the state will be STOPPED. After this, when the vehicle tries to start (is still stopped), the state will be STARTING. In this state, calling the start API changes the state to MOVING and the vehicle starts. This mechanism can add processing such as announcements before the vehicle starts. Depending on the configuration, the state may transition directly from STOPPED to MOVING.
State Description STOPPED The vehicle is stopped. STARTING The vehicle is stopped, but is trying to start. MOVING The vehicle is moving. BRAKING (T.B.D.) The vehicle is decelerating strongly."},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/","title":"Operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#operation-mode","title":"Operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#related-api","title":"Related API","text":"
As shown below, Autoware assumes that the vehicle interface has two modes, Autoware control and direct control. In direct control mode, the vehicle is operated using devices such as steering and pedals. If the vehicle does not support direct control mode, it is always treated as Autoware control mode. Autoware control mode has four operation modes.
Mode Description Stop Keep the vehicle stopped. Autonomous Autonomously control the vehicle. Local Manually control the vehicle from nearby with some device such as a joystick. Remote Manually control the vehicle from a web application on the cloud. "},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#states","title":"States","text":""},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#autoware-control-flag","title":"Autoware control flag","text":"
The flag is_autoware_control_enabled indicates if the vehicle is controlled by Autoware. The enable and disable APIs can be used if the control can be switched by software. These APIs will always fail if the vehicle does not support mode switching or is switched by hardware.
"},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#operation-mode-and-change-flags","title":"Operation mode and change flags","text":"
The state operation_mode indicates what command is used when Autoware control is enabled. The flags change_to_* can be used to check if it is possible to transition to each mode.
Since Autoware may not be able to guarantee safety, such as switching to autonomous mode during overspeed. There is the flag is_in_transition for this situation and it will be true when changing modes. The operator who changed the mode should ensure safety while this flag is true. The flag will be false when the mode change is complete.
This API manages the planned behavior of the vehicle. Applications can notify the vehicle behavior to the people around and visualize it for operator and passengers.
The velocity factors is an array of information on the behavior that the vehicle stops or slows down. Each factor has a behavior type which is described below. Some behavior types have sequence and details as additional information.
Behavior Description surrounding-obstacle There are obstacles immediately around the vehicle. route-obstacle There are obstacles along the route ahead. intersection There are obstacles in other lanes in the path. crosswalk There are obstacles on the crosswalk. rear-check There are obstacles behind that would be in a human driver's blind spot. user-defined-attention-area There are obstacles in the predefined attention area. no-stopping-area There is not enough space beyond the no stopping area. stop-sign A stop by a stop sign. traffic-signal A stop by a traffic signal. v2x-gate-area A stop by a gate area. It has enter and leave as sequences and v2x type as details. merge A stop before merging lanes. sidewalk A stop before crossing the sidewalk. lane-change A lane change. avoidance A path change to avoid an obstacle in the current lane. emergency-operation A stop by emergency instruction from the operator.
Each factor also provides status, poses in the base link frame, and distance from that pose. As the vehicle approaches the stop position, this factor appears with a status of APPROACHING. And when the vehicle reaches that position and stops, the status will be STOPPED. The pose indicates the stop position, or the base link if the stop position cannot be calculated.
The steering factors is an array of information on the maneuver that requires use of turn indicators, such as turning left or right. Each factor has a behavior type which is described below and steering direction. Some behavior types have sequence and details as additional information.
Behavior Description intersection A turning left or right at an intersection. lane-change A lane change. avoidance A path change to avoid an obstacle. It has a sequence of change and return. start-planner T.B.D. goal-planner T.B.D. emergency-operation A path change by emergency instruction from the operator.
Each factor also provides status, poses in the base link frame, and distances from that poses. As the vehicle approaches the position to start steering, this factor appears with a status of APPROACHING. And when the vehicle reaches that position, the status will be TURNING. The poses indicate the start and end position of the section where the status is TURNING.
In cases such as lane change and avoidance, the vehicle will start steering at any position in the range depending on the situation. For these types, the section where the status is TURNING will be updated dynamically and the poses will follow that.
This API manages destination and waypoints. Note that waypoints are not like stops and just points passing through. In other words, Autoware does not support the route with multiple stops, the application needs to split it up and switch them. There are two ways to set the route. The one is a generic method that uses pose, another is a map-dependent.
"},{"location":"design/autoware-interfaces/ad-api/features/routing/#states","title":"States","text":"State Description UNSET The route is not set. Waiting for a route request. SET The route is set. ARRIVED The vehicle has arrived at the destination. CHANGING Trying to change the route. Not implemented yet."},{"location":"design/autoware-interfaces/ad-api/features/routing/#goal-modification","title":"Goal modification","text":"
Autoware tries to look for an alternate goal when goal is unreachable (e.g., when there is an obstacle on the given goal). When setting a route from the API, applications can choose whether they allow Autoware to adjust goal pose in such situation. When set false, Autoware may get stuck until the given goal becomes reachable.
This feature is available if the vehicle provides a software interface for the doors. It can be used to create user interfaces for passengers or to control sequences at bus stops.
Each door in a vehicle is assigned an array index. This assignment is vehicle dependent. The layout API returns this information. The description field is a string to display in the user interface, etc. This is an arbitrary string and is not recommended to use for processing in applications. Use the roles field to know doors for getting on and off. Below is an example of the information returned by the layout API.
Index Description Roles 0 front right - 1 front left GET_ON 2 rear right GET_OFF 3 rear left GET_ON, GET_OFF"},{"location":"design/autoware-interfaces/ad-api/features/vehicle-doors/#status","title":"Status","text":"
The status API provides an array of door status. This array order is consistent with the layout API.
Use the command API to control doors. Unlike the status and layout APIs, array index do not correspond to doors. The command has a field to specify the target door index.
This is an estimate of the vehicle kinematics. The vehicle position is necessary for applications to schedule dispatches. Also, using velocity and acceleration, applications can find vehicles that need operator assistance, such as stuck or brake suddenly.
This is the status provided by the vehicle. The indicators and steering are mainly used for visualization and remote control. The remaining energy can be also used for vehicle scheduling.
The vehicle dimensions are used to know the actual distance between the vehicle and objects because the vehicle position in kinematics is the coordinates of the base link. This is necessary for visualization when supporting vehicles remotely.
"},{"location":"design/autoware-interfaces/ad-api/list/","title":"List of Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/list/#list-of-autoware-ad-api","title":"List of Autoware AD API","text":"
Get the MRM state. For details, see the fail-safe.
"},{"location":"design/autoware-interfaces/ad-api/list/api/fail_safe/mrm_state/#message","title":"Message","text":"Name Type Description state uint16 The state of MRM operation. behavior uint16 The currently selected behavior of MRM."},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/","title":"/api/interface/version","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/#apiinterfaceversion","title":"/api/interface/version","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_version_msgs/srv/InterfaceVersion
"},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/#response","title":"Response","text":"Name Type Description major uint16 major version minor uint16 minor version patch uint16 patch version"},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/","title":"/api/localization/initialization_state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/#apilocalizationinitialization_state","title":"/api/localization/initialization_state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/LocalizationInitializationState
Get the initialization state of localization. For details, see the localization.
"},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/#message","title":"Message","text":"Name Type Description state uint16 A value of the localization initialization state."},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/","title":"/api/localization/initialize","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#apilocalizationinitialize","title":"/api/localization/initialize","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/InitializeLocalization
Request to initialize localization. For details, see the localization.
"},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#request","title":"Request","text":"Name Type Description pose geometry_msgs/msg/PoseWithCovarianceStamped[<=1] A global pose as the initial guess. If omitted, the GNSS pose will be used."},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/","title":"/api/motion/accept_start","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/#apimotionaccept_start","title":"/api/motion/accept_start","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/#status","title":"Status","text":"
"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/","title":"/api/motion/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/#apimotionstate","title":"/api/motion/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/#status","title":"Status","text":"
Get the motion state. For details, see the motion state.
"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/#message","title":"Message","text":"Name Type Description state uint16 A value of the motion state."},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/","title":"/api/operation_mode/change_to_autonomous","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/#apioperation_modechange_to_autonomous","title":"/api/operation_mode/change_to_autonomous","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/","title":"/api/operation_mode/change_to_local","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/#apioperation_modechange_to_local","title":"/api/operation_mode/change_to_local","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/","title":"/api/operation_mode/change_to_remote","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/#apioperation_modechange_to_remote","title":"/api/operation_mode/change_to_remote","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/","title":"/api/operation_mode/change_to_stop","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/#apioperation_modechange_to_stop","title":"/api/operation_mode/change_to_stop","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/","title":"/api/operation_mode/disable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/#apioperation_modedisable_autoware_control","title":"/api/operation_mode/disable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/","title":"/api/operation_mode/enable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/#apioperation_modeenable_autoware_control","title":"/api/operation_mode/enable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/","title":"/api/operation_mode/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/#apioperation_modestate","title":"/api/operation_mode/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/OperationModeState
Get the operation mode state. For details, see the operation mode.
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/#message","title":"Message","text":"Name Type Description mode uint8 The selected command for Autoware control. is_autoware_control_enabled bool True if vehicle control by Autoware is enabled. is_in_transition bool True if the operation mode is in transition. is_stop_mode_available bool True if the operation mode can be changed to stop. is_autonomous_mode_available bool True if the operation mode can be changed to autonomous. is_local_mode_available bool True if the operation mode can be changed to local. is_remote_mode_available bool True if the operation mode can be changed to remote."},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/","title":"/api/perception/objects","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/#apiperceptionobjects","title":"/api/perception/objects","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/DynamicObjectArray
Get the recognized objects array with label, shape, current position and predicted path For details, see the perception.
"},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/#message","title":"Message","text":"Name Type Description objects.id unique_identifier_msgs/msg/UUID The UUID of each object objects.existence_probability float64 The probability of the object exits objects.classification autoware_adapi_v1_msgs/msg/ObjectClassification[] The type of the object recognized and the confidence level objects.kinematics autoware_adapi_v1_msgs/msg/DynamicObjectKinematics Consist of the object pose, twist, acceleration and the predicted_paths objects.shape shape_msgs/msg/SolidPrimitive escribe the shape of the object with dimension, and polygon"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/","title":"/api/planning/steering_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/#apiplanningsteering_factors","title":"/api/planning/steering_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/#status","title":"Status","text":"
Latest Version: not released
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/SteeringFactorArray
Get the steering factors, sorted in ascending order of distance. For details, see the planning factors.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/#message","title":"Message","text":"Name Type Description factors.pose geometry_msgs/msg/Pose[2] The base link pose related to the steering factor. factors.distance float32[2] The distance from the base link to the above pose. factors.direction uint16 The direction of the steering factor. factors.status uint16 The status of the steering factor. factors.behavior string The behavior type of the steering factor. factors.sequence string The sequence type of the steering factor. factors.detail string The additional information of the steering factor. factors.cooperation autoware_adapi_v1_msgs/msg/CooperationStatus[<=1] The cooperation status if the module supports."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/","title":"/api/planning/velocity_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/#apiplanningvelocity_factors","title":"/api/planning/velocity_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/#status","title":"Status","text":"
Latest Version: not released
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/VelocityFactorArray
Get the velocity factors, sorted in ascending order of distance. For details, see the planning factors.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/#message","title":"Message","text":"Name Type Description factors.pose geometry_msgs/msg/Pose The base link pose related to the velocity factor. factors.distance float32 The distance from the base link to the above pose. factors.status uint16 The status of the velocity factor. factors.behavior string The behavior type of the velocity factor. factors.sequence string The sequence type of the velocity factor. factors.detail string The additional information of the velocity factor. factors.cooperation autoware_adapi_v1_msgs/msg/CooperationStatus[<=1] The cooperation status if the module supports."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/","title":"/api/planning/cooperation/get_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/#apiplanningcooperationget_policies","title":"/api/planning/cooperation/get_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/GetCooperationPolicies
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status policies.behavior string The type of the target behavior. policies.sequence string The type of the target sequence. policies.policy uint8 The type of the cooporation policy."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/","title":"/api/planning/cooperation/set_commands","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#apiplanningcooperationset_commands","title":"/api/planning/cooperation/set_commands","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/SetCooperationCommands
Set the operator's decision for cooperation. For details, see the cooperation.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#request","title":"Request","text":"Name Type Description commands.uuid unique_identifier_msgs/msg/UUID The ID in the cooperation status. commands.cooperator autoware_adapi_v1_msgs/msg/CooperationDecision The operator's decision."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/","title":"/api/planning/cooperation/set_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#apiplanningcooperationset_policies","title":"/api/planning/cooperation/set_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/SetCooperationPolicies
Set the default decision that is used instead when the operator's decision is undecided. For details, see the cooperation.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#request","title":"Request","text":"Name Type Description policies.behavior string The type of the target behavior. policies.sequence string The type of the target sequence. policies.policy uint8 The type of the cooporation policy."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/","title":"/api/routing/clear_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/#apiroutingclear_route","title":"/api/routing/clear_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/#status","title":"Status","text":"
"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/","title":"/api/routing/route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/#apiroutingroute","title":"/api/routing/route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/#status","title":"Status","text":"
Get the route with the waypoint segments in lanelet format. It is empty if route is not set.
"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/#message","title":"Message","text":"Name Type Description header std_msgs/msg/Header header for pose transformation data autoware_adapi_v1_msgs/msg/RouteData[<=1] The route in lanelet format"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/set_route/","title":"/api/routing/set_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/set_route/#apiroutingset_route","title":"/api/routing/set_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/set_route/#status","title":"Status","text":"
Get the route state. For details, see the routing.
"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/state/#message","title":"Message","text":"Name Type Description state uint16 A value of the route state."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/","title":"/api/vehicle/dimensions","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/#apivehicledimensions","title":"/api/vehicle/dimensions","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/GetVehicleDimensions
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status dimensions autoware_adapi_v1_msgs/msg/VehicleDimensions vehicle dimensions"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/","title":"/api/vehicle/kinematics","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/#apivehiclekinematics","title":"/api/vehicle/kinematics","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/#status","title":"Status","text":"
Latest Version: not released
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/VehicleKinematics
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/#message","title":"Message","text":"Name Type Description geographic_pose geographic_msgs/msg/GeoPointStamped The longitude and latitude of the vehicle. If the map uses local coordinates, it will not be available. pose geometry_msgs/msg/PoseWithCovarianceStamped The pose with covariance from the base link. twist geometry_msgs/msg/TwistWithCovarianceStamped Vehicle current twist with covariance. accel geometry_msgs/msg/AccelWithCovarianceStamped Vehicle current acceleration with covariance."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/","title":"/api/vehicle/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/#apivehiclestatus","title":"/api/vehicle/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/#status","title":"Status","text":"
Latest Version: not released
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/VehicleStatus
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/#message","title":"Message","text":"Name Type Description gear autoware_adapi_v1_msgs/msg/Gear Gear status. turn_indicators autoware_adapi_v1_msgs/msg/TurnIndicators Turn indicators status, only either left or right will be enabled. hazard_lights autoware_adapi_v1_msgs/msg/HazardLights Hazard lights status. steering_tire_angle float64 Vehicle current tire angle in radian. energy_percentage float32 Battery percentage or fuel percentage, it will depends on the vehicle."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/","title":"/api/vehicle/doors/command","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#apivehicledoorscommand","title":"/api/vehicle/doors/command","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/SetDoorCommand
Set the door command. This API is only available if the vehicle supports software door control.
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#request","title":"Request","text":"Name Type Description doors.index uint32 The index of the target door. doors.command uint8 The command for the target door."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/","title":"/api/vehicle/doors/layout","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/#apivehicledoorslayout","title":"/api/vehicle/doors/layout","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/GetDoorLayout
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status doors.roles uint8[] The roles of the door in the service the vehicle provides. doors.description string The description of the door for display in the interface."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/","title":"/api/vehicle/doors/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/#apivehicledoorsstatus","title":"/api/vehicle/doors/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/#status","title":"Status","text":"
Latest Version: not released
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/DoorStatusArray
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/#message","title":"Message","text":"Name Type Description doors.status uint8 current door status"},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/#user-story-of-bus-service","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/#overview","title":"Overview","text":"
This user story is a bus service that goes around the designated stops.
"},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/#scenario","title":"Scenario","text":"Step Operation Use Case 1 Startup the autonomous driving system. Launch and terminate 2 Drive the vehicle from the garage to the waiting position. Change the operation mode 3 Enable autonomous control. Change the operation mode 4 Drive the vehicle to the next bus stop. Drive to the designated position 5 Get on and off the vehicle. Get on and get off 6 Return to step 4 unless it's the last bus stop. 7 Drive the vehicle to the waiting position. Drive to the designated position 8 Drive the vehicle from the waiting position to the garage. Change the operation mode 9 Shutdown the autonomous driving system. Launch and terminate"},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/#user-story-of-bus-service","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/#overview","title":"Overview","text":"
This user story is a taxi service that picks up passengers and drives them to their destination.
"},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/#scenario","title":"Scenario","text":"Step Operation Use Case 1 Startup the autonomous driving system. Launch and terminate 2 Drive the vehicle from the garage to the waiting position. Change the operation mode 3 Enable autonomous control. Change the operation mode 4 Drive the vehicle to the position to pick up. Drive to the designated position 5 Get on the vehicle. Get on and get off 6 Drive the vehicle to the destination. Drive to the designated position 7 Get off the vehicle. Get on and get off 8 Drive the vehicle to the waiting position. Drive to the designated position 9 Return to step 4 if there is another request. 10 Drive the vehicle from the waiting position to the garage. Change the operation mode 11 Shutdown the autonomous driving system. Launch and terminate"},{"location":"design/autoware-interfaces/ad-api/types/","title":"Types of Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/types/#types-of-autoware-ad-api","title":"Types of Autoware AD API","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationCommand/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationDecision
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationCommand/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationDecision/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationDecision/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationPolicy/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationPolicy/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationStatus/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationDecision
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationStatus/#this-type-is-used-by","title":"This type is used by","text":"
uint8 OPEN = 1\nuint8 CLOSE = 2\n\nuint32 index\nuint8 command\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorCommand/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorCommand/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorLayout/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorLayout/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatus/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatus/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatusArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DoorStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatusArray/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObject/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DynamicObject
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectArray/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectKinematics/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DynamicObjectPath
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectKinematics/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectPath/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectPath/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/HazardLights/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/HazardLights/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/LocalizationInitializationState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/LocalizationInitializationState/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MotionState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MotionState/#this-type-is-used-by","title":"This type is used by","text":"
builtin_interfaces/Time stamp\n\n# For common use\nuint16 UNKNOWN = 0\n\n# For state\nuint16 NORMAL = 1\nuint16 MRM_OPERATING = 2\nuint16 MRM_SUCCEEDED = 3\nuint16 MRM_FAILED = 4\n\n# For behavior\nuint16 NONE = 1\nuint16 EMERGENCY_STOP = 2\nuint16 COMFORTABLE_STOP = 3\n\nuint16 state\nuint16 behavior\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MrmState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MrmState/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ObjectClassification/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ObjectClassification/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/OperationModeState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/OperationModeState/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ResponseStatus/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ResponseStatus/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/Route/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/RouteData
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/Route/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteData/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/RouteSegment
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteData/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteOption/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteOption/#this-type-is-used-by","title":"This type is used by","text":"
int64 id\nstring type # The same id may be used for each type.\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RoutePrimitive/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RoutePrimitive/#this-type-is-used-by","title":"This type is used by","text":"
autoware_adapi_v1_msgs/RoutePrimitive preferred\nautoware_adapi_v1_msgs/RoutePrimitive[] alternatives # Does not include the preferred primitive.\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteSegment/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/RoutePrimitive
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteSegment/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteState/#this-type-is-used-by","title":"This type is used by","text":"
# constants for common use\nuint16 UNKNOWN = 0\n\n# constants for direction\nuint16 LEFT = 1\nuint16 RIGHT = 2\nuint16 STRAIGHT = 3\n\n# constants for status\nuint16 APPROACHING = 1\nuint16 TURNING = 3\n\n# variables\ngeometry_msgs/Pose[2] pose\nfloat32[2] distance\nuint16 direction\nuint16 status\nstring behavior\nstring sequence\nstring detail\nautoware_adapi_v1_msgs/CooperationStatus[<=1] cooperation\n\n\n\n# deprecated constants for type\nuint16 INTERSECTION = 1\nuint16 LANE_CHANGE = 2\nuint16 AVOIDANCE_PATH_CHANGE = 3\nuint16 AVOIDANCE_PATH_RETURN = 4\nuint16 STATION = 5\nuint16 PULL_OUT = 6 # Deprecated. Use START_PLANNER.\nuint16 START_PLANNER = 6\nuint16 PULL_OVER = 7 # Deprecated. Use GOAL_PLANNER.\nuint16 GOAL_PLANNER = 7\nuint16 EMERGENCY_OPERATION = 8\n\n# deprecated constants for status\nuint16 TRYING = 2\n\n# deprecated variables\nuint16 type\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactor/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactor/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactorArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/SteeringFactor
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactorArray/#this-type-is-used-by","title":"This type is used by","text":"
# constants\nuint8 UNKNOWN = 0\nuint8 DISABLE = 1\nuint8 LEFT = 2\nuint8 RIGHT = 3\n\nuint8 status\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/TurnIndicators/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/TurnIndicators/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleDimensions/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleDimensions/#this-type-is-used-by","title":"This type is used by","text":"
# Geographic point, using the WGS 84 reference ellipsoid.\n# This data will be invalid If Autoware does not provide projection information between geographic coordinates and local coordinates.\ngeographic_msgs/GeoPointStamped geographic_pose\n\n# Local coordinate from the autoware\ngeometry_msgs/PoseWithCovarianceStamped pose\ngeometry_msgs/TwistWithCovarianceStamped twist\ngeometry_msgs/AccelWithCovarianceStamped accel\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleKinematics/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleKinematics/#this-type-is-used-by","title":"This type is used by","text":"
builtin_interfaces/Time stamp\nautoware_adapi_v1_msgs/Gear gear\nautoware_adapi_v1_msgs/TurnIndicators turn_indicators\nautoware_adapi_v1_msgs/HazardLights hazard_lights\nfloat64 steering_tire_angle\nfloat32 energy_percentage # Battery percentage or fuel percentage, it will depends on the vehicle.\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleStatus/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/Gear
autoware_adapi_v1_msgs/msg/HazardLights
autoware_adapi_v1_msgs/msg/TurnIndicators
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleStatus/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactor/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactor/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactorArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/VelocityFactor
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactorArray/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/AcceptStart/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/AcceptStart/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ChangeOperationMode/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ChangeOperationMode/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ClearRoute/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ClearRoute/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetCooperationPolicies/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationPolicy
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetCooperationPolicies/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetDoorLayout/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DoorLayout
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetDoorLayout/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetVehicleDimensions/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
autoware_adapi_v1_msgs/msg/VehicleDimensions
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetVehicleDimensions/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/InitializeLocalization/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/InitializeLocalization/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationCommands/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationCommand
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationCommands/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationPolicies/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationPolicy
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationPolicies/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetDoorCommand/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DoorCommand
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetDoorCommand/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoute/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
autoware_adapi_v1_msgs/msg/RouteOption
autoware_adapi_v1_msgs/msg/RouteSegment
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoute/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoutePoints/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
autoware_adapi_v1_msgs/msg/RouteOption
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoutePoints/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_version_msgs/srv/InterfaceVersion/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_version_msgs/srv/InterfaceVersion/#this-type-is-used-by","title":"This type is used by","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/use-cases/change-operation-mode/","title":"Change the operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/change-operation-mode/#change-the-operation-mode","title":"Change the operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/change-operation-mode/#related-api","title":"Related API","text":"
"},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/","title":"Drive to the designated position","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/#drive-to-the-designated-position","title":"Drive to the designated position","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/#related-api","title":"Related API","text":"
Driving
Route
"},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/#sequence","title":"Sequence","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/","title":"Get on and get off","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/#get-on-and-get-off","title":"Get on and get off","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/#related-api","title":"Related API","text":"
Door
"},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/#sequence","title":"Sequence","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/initialize-pose/","title":"Initialize the pose","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/initialize-pose/#initialize-the-pose","title":"Initialize the pose","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/initialize-pose/#related-api","title":"Related API","text":"
"},{"location":"design/autoware-interfaces/ad-api/use-cases/launch-terminate/","title":"Launch and terminate","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/launch-terminate/#launch-and-terminate","title":"Launch and terminate","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/launch-terminate/#related-api","title":"Related API","text":"
AD API provides current vehicle status for remote monitoring, visualization for passengers, etc. Use the API below depending on the data you want to monitor.
The vehicle status provides basic information such as kinematics, indicators, and dimensions. This allows a remote operator to know the position and velocity of the vehicle. For applications such as FMS, it can help find vehicles that need assistance, such as vehicles that are stuck or brake suddenly. It is also possible to determine the actual distance to an object from the vehicle dimensions.
The planning factors provides the planning status of the vehicle. HMI can use this to warn of sudden movements of the vehicle, and to share the stop reason with passengers for comfortable driving.
The perception provides the objects detected by Autoware. HMI can use this to visualize objects around the vehicle.
"},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/","title":"Vehicle operation","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/#vehicle-operation","title":"Vehicle operation","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/#request-to-intervene","title":"Request to intervene","text":"
Request to intervene (RTI) is a feature that requires the operator to switch to manual driving mode. It is also called Take Over Request (TOR). Interfaces for RTI are currently being discussed. For now assume that manual driving is requested if the MRM state is not NORMAL. See fail-safe for details.
"},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/#request-to-cooperate","title":"Request to cooperate","text":"
Request to cooperate (RTC) is a feature that the operator supports the decision in autonomous driving mode. Autoware usually drives the vehicle using its own decisions, but the operator may prefer to make their own decisions in complex situations. Since RTC only overrides the decision and does not need to change operation mode, the vehicle can continue autonomous driving, unlike RTC. See cooperation for details.
Actuation status of the ego vehicle for acceleration, steering, and brake.
TODO This represents the reported physical efforts exerted by the vehicle actuators. Published by the Vehicle Interface.
ActuationStatus (github discussion).
builtin_interfaces::msg::Time stamp
float32 acceleration
float32 steering
"},{"location":"design/autoware-interfaces/components/control/#output","title":"Output","text":""},{"location":"design/autoware-interfaces/components/control/#vehicle-control-command","title":"Vehicle Control Command","text":"
A motion signal to drive the vehicle, achieved by the low-level controller in the vehicle layer. Used by the Vehicle Interface.
Current velocity of the ego vehicle, published by the vehicle interface.
autoware_auto_vehicle_msgs/msg/VelocityReport
std_msgs/msg/Header header;
float longitudinal_velocity;
float lateral_velocity;
float heading_rate;
Before the velocity input localization interface, module vehicle_velocity_converter converts message type autoware_auto_vehicle_msgs/msg/VelocityReport to geometry_msgs/msg/TwistWithCovarianceStamped.
Current pose, velocity and acceleration of ego, calculated from localization interface.
Note: Kinematic state contains pose, velocity and acceleration. In the future, pose, velocity and acceleration will not be used as output for localization.
Autoware relies on high-definition point cloud maps and vector maps of the driving environment to perform various tasks. Before launching Autoware, you need to load the pre-created map files.
It loads a Lanelet2 file and publishes the map data as autoware_auto_mapping_msgs/msg/HADMapBin message. The lan/lon coordinates are projected onto the MGRS coordinates.
This page provides specific specifications about the Interface of the Planning Component. Please refer to the planning architecture design document for high-level concepts and data flow.
TODO: The detailed definitions (meanings of elements included in each topic) are not described yet, need to be updated.
"},{"location":"design/autoware-interfaces/components/planning/#input","title":"Input","text":""},{"location":"design/autoware-interfaces/components/planning/#from-map-component","title":"From Map Component","text":"Name Topic Type Description Vector Map /map/vector_map autoware_auto_mapping_msgs/msg/HADMapBin Map of the environment where the planning takes place."},{"location":"design/autoware-interfaces/components/planning/#from-localization-component","title":"From Localization Component","text":"Name Topic Type Description Vehicle Kinematic State /localization/kinematic_state nav_msgs/msg/Odometry Current position, orientation and velocity of ego. Vehicle Acceleration /localization/acceleration geometry_msgs/msg/AccelWithCovarianceStamped Current acceleration of ego.
TODO: acceleration information should be merged into the kinematic state.
"},{"location":"design/autoware-interfaces/components/planning/#from-perception-component","title":"From Perception Component","text":"Name Topic Type Description Objects /perception/object_recognition/objects autoware_auto_perception_msgs/msg/PredictedObjects Set of perceived objects around ego that need to be avoided or followed when planning a trajectory. This contains semantics information such as a object class (e.g. vehicle, pedestrian, etc) or a shape of the objects. Obstacles /perception/obstacle_segmentation/pointcloud sensor_msgs/msg/PointCloud2 Set of perceived obstacles around ego that need to be avoided or followed when planning a trajectory. This only contains a primitive information of the obstacle. No shape nor velocity information. Occupancy Grid Map /perception/occupancy_grid_map/map nav_msgs/msg/OccupancyGrid Contains the presence of obstacles and blind spot information (represented as UNKNOWN). Traffic Signal /perception/traffic_light_recognition/traffic_signals autoware_auto_perception_msgs/msg/TrafficSignalArray Contains the traffic signal information such as a color (green, yellow, read) and an arrow (right, left, straight).
TODO: The type of the Obstacles information should not depend on the specific sensor message type (now PointCloud). It needs to be fixed.
"},{"location":"design/autoware-interfaces/components/planning/#from-api","title":"From API","text":"Name Topic Type Description Max Velocity /planning/scenario_planning/max_velocity_default autoware_adapi_v1_msgs/srv/SetRoutePoints Indicate the maximum value of the vehicle speed plan Operation Mode /system/operation_mode/state autoware_adapi_v1_msgs/msg/OperationModeState Indicates the current operation mode (automatic/manual, etc.). Route Set /planning/mission_planning/set_route autoware_adapi_v1_msgs/srv/SetRoute Indicates to set the route when the vehicle is stopped. Route Points Set /planning/mission_planning/set_route_points autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to set the route with points when the vehicle is stopped. Route Change /planning/mission_planning/change_route autoware_adapi_v1_msgs/srv/SetRoute Indicates to change the route when the vehicle is moving. Route Points Change /planning/mission_planning/change_route_points autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to change the route with points when the vehicle is moving. Route Clear /planning/mission_planning/clear_route autoware_adapi_v1_msgs/srv/ClearRoute Indicates to clear the route information. MRM Route Set Points /planning/mission_planning/mission_planner/srv/set_mrm_route autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to set the emergency route. MRM Route Clear /planning/mission_planning/mission_planner/srv/clear_mrm_route autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to clear the emergency route."},{"location":"design/autoware-interfaces/components/planning/#output","title":"Output","text":""},{"location":"design/autoware-interfaces/components/planning/#to-control","title":"To Control","text":"Name Topic Type Description Trajectory /planning/trajectory autoware_auto_planning_msgs/msg/Trajectory A sequence of space and velocity and acceleration points to be followed by the controller. Turn Indicator /planning/turn_indicators_cmd autoware_auto_vehicle_msgs/msg/TurnIndicatorsCommand Turn indicator signal to be followed by the vehicle. Hazard Light /planning/hazard_lights_cmd autoware_auto_vehicle_msgs/msg/HazardLightsCommand Hazard light signal to be followed by the vehicle."},{"location":"design/autoware-interfaces/components/planning/#to-system","title":"To System","text":"Name Topic Type Description Diagnostics /planning/hazard_lights_cmd diagnostic_msgs/msg/DiagnosticArray Diagnostic status of the Planning component reported to the System component."},{"location":"design/autoware-interfaces/components/planning/#to-api","title":"To API","text":"Name Topic Type Description Path Candidate /planning/path_candidate/* autoware_auto_planning_msgs/msg/Path The path Autoware is about to take. Users can interrupt the operation based on the path candidate information. Steering Factor /planning/steering_factor/* autoware_adapi_v1_msgs/msg/SteeringFactorArray Information about the steering maneuvers performed by Autoware (e.g., steering to the right for a right turn, etc.) Velocity Factor /planning/velocity_factors/* autoware_adapi_v1_msgs/msg/VelocityFactorArray Information about the velocity maneuvers performed by Autoware (e.g., stop for an obstacle, etc.)"},{"location":"design/autoware-interfaces/components/planning/#planning-internal-interface","title":"Planning internal interface","text":"
This section explains the communication between the different planning modules shown in the Planning Architecture Design.
"},{"location":"design/autoware-interfaces/components/planning/#from-mission-planning-to-scenario-planning","title":"From Mission Planning to Scenario Planning","text":"Name Topic Type Description Route /planning/mission_planning/route autoware_planning_msgs/msg/LaneletRoute A sequence of lane IDs on a Lanelet map, from the starting point to the destination."},{"location":"design/autoware-interfaces/components/planning/#from-behavior-planning-to-motion-planning","title":"From Behavior Planning to Motion Planning","text":"Name Topic Type Description Path /planning/scenario_planning/lane_driving/behavior_planning/path autoware_auto_planning_msgs/msg/Path A sequence of approximate vehicle positions for driving, along with information on the maximum speed and the drivable areas. Modules receiving this message are expected to make changes to the path within the constraints of the drivable areas and the maximum speed, generating the desired final trajectory."},{"location":"design/autoware-interfaces/components/planning/#from-scenario-planning-to-validation","title":"From Scenario Planning to Validation","text":"Name Topic Type Description Trajectory /planning/scenario_planning/trajectory autoware_auto_planning_msgs/msg/Trajectory A sequence of precise vehicle positions, speeds, and accelerations required for driving. It is expected that the vehicle will follow this trajectory."},{"location":"design/autoware-interfaces/components/sensing/","title":"Sensing","text":""},{"location":"design/autoware-interfaces/components/sensing/#sensing","title":"Sensing","text":"
The Vehicle Interface receives the Vehicle Signal Commands and Vehicle Control Commands and publishes the vehicle status. It also communicates with vehicle by the vehicle-specific protocol.
The Gate switches multiple Vehicle Control Commands. These signals include autonomous diving command, joystick, remote control, and emergency operation, etc. The Adapter converts generalized control command (target steering, steering rate, velocity, acceleration, jerk) into vehicle-specific control values (steering-torque, wheel-torque, voltage, pressure, accel pedal position, etc).
Commands for various elements of the vehicle unrelated to motion. Published by the Planning module.
"},{"location":"design/autoware-interfaces/components/vehicle-interface/#outputs","title":"Outputs","text":""},{"location":"design/autoware-interfaces/components/vehicle-interface/#vehicle-signal-reports","title":"Vehicle Signal Reports","text":"
Reports for various elements of the vehicle unrelated to motion. Published by the Vehicle Interface.
Actuation status of the ego vehicle for acceleration, steering, and brake. This represents the reported physical efforts exerted by the vehicle actuators. Published by the Vehicle Interface.
Actuation command sent to the ego vehicle. This represents the requested physical efforts to be exerted by the vehicle actuators. Published by the Vehicle Interface as generated by the adapter.
Defining temporal performance metrics on components
An example procedure for adding and evaluating a new node
TODO: Write the following contents.
Create an Autoware package
etc.
"},{"location":"how-to-guides/integrating-autoware/overview/","title":"Overview","text":""},{"location":"how-to-guides/integrating-autoware/overview/#overview","title":"Overview","text":""},{"location":"how-to-guides/integrating-autoware/overview/#requirement-prepare-your-real-vehicle-hardware","title":"Requirement: prepare your real vehicle hardware","text":"
Prerequisites for the vehicle:
An onboard computer that satisfies the Autoware installation prerequisites
The following devices attached
Drive-by-wire interface
LiDAR
Optional: Inertial measurement unit
Optional: Camera
Optional: GNSS
"},{"location":"how-to-guides/integrating-autoware/overview/#1-creating-your-autoware-meta-repository","title":"1. Creating your Autoware meta-repository","text":"
Create your Autoware meta-repository. One easy way is to fork autowarefoundation/autoware and clone it. For how to fork a repository, refer to GitHub Docs.
If you set up multiple types of vehicles, adding a suffix like \"autoware.vehicle_A\" or \"autoware.vehicle_B\" is recommended.
"},{"location":"how-to-guides/integrating-autoware/overview/#2-creating-the-your-vehicle-and-sensor-description","title":"2. Creating the your vehicle and sensor description","text":"
Next, you need to create description packages that define the vehicle and sensor configuration of your vehicle.
Create the following two packages:
YOUR_VEHICLE_launch (see here for example)
YOUR_SENSOR_KIT_launch (see here for example)
Once created, you need to update the autoware.repos file of your cloned Autoware repository to refer to these two description packages.
Create a launch file (refer to the sample vehicle launch package for example). If you have multiple vehicles with the same hardware setup, you can specify vehicle_id to distinguish them.
"},{"location":"how-to-guides/integrating-autoware/overview/#adapt-your_sensor_kit_description-for-autoware-launching-system","title":"Adapt YOUR_SENSOR_KIT_description for autoware launching system","text":""},{"location":"how-to-guides/integrating-autoware/overview/#at-your_sensor_kit_description","title":"At YOUR_SENSOR_KIT_description","text":"
Define URDF and extrinsic parameters for all the sensors here (refer to the sample sensor kit description package for example). Note that you need to calibrate extrinsic parameters for all the sensors beforehand.
Create launch/sensing.launch.xml that launches the interfaces of all the sensors on the vehicle. (refer to the sample sensor kit launch package for example).
Note
At this point, you are now able to run Autoware's Planning Simulator to do a basic test of your vehicle and sensing packages. To do so, you need to build and install Autoware using your cloned repository. Follow the steps for either Docker or source installation (starting from the dependency installation step) and then run the following command:
"},{"location":"how-to-guides/integrating-autoware/overview/#3-create-a-vehicle_interface-package","title":"3. Create a vehicle_interface package","text":"
You need to create an interface package for your vehicle. The package is expected to provide the following two functions.
Receive command messages from vehicle_cmd_gate and drive the vehicle accordingly
Send vehicle status information to Autoware
You can find detailed information about the requirements of the vehicle_interface package in the Vehicle Interface design documentation. You can also refer to TIER IV's pacmod_interface repository as an example of a vehicle interface package.
You need both a pointcloud map and a vector map in order to use Autoware. For more information on map design, please click here.
"},{"location":"how-to-guides/integrating-autoware/overview/#create-a-pointcloud-map","title":"Create a pointcloud map","text":"
Use third-party tools such as a LiDAR-based SLAM (Simultaneous Localization And Mapping) package to create a pointcloud map in the .pcd format. For more information, please click here.
Click the 2D Nav Goal button in the toolbar, or hit the G key
In the 3D View pane, click and hold the left mouse button, and then drag to set the direction for the goal pose. If successful, you will see the calculated planning path on RViz.
You can also engage via RViz with \"AutowareStatePanel\". The panel can be found in Panels > Add New Panel > tier4_state_rviz_plugin > AutowareStatePanel.
Now the vehicle should drive along the calculated path!
"},{"location":"how-to-guides/integrating-autoware/overview/#6-tune-parameters-for-your-vehicle-environment","title":"6. Tune parameters for your vehicle & environment","text":"
You may need to tune your parameters depending on the domain in which you will operate your vehicle.
Autoware requires a pointcloud map and a vector map for the vehicle's operating environment. (Check the map design documentation page for the detailed specification).
This page explains how users can create maps that can be used for Autoware.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/#creating-a-point-cloud-map","title":"Creating a point cloud map","text":"
Traditionally, a Mobile Mapping System (MMS) is used in order to create highly accurate large-scale point cloud maps. However, since a MMS requires high-end sensors for precise positioning, its operational cost can be very expensive and may not be suitable for a relatively small driving environment. Alternatively, a Simultaneous Localization And Mapping (SLAM) algorithm can be used to create a point cloud map from recorded LiDAR scans. Some of the useful open-source SLAM implementations are listed in this page.
If you prefer proprietary software that is easy to use, you can try a fully automatic mapping tool from MAP IV, Inc., MapIV Engine. They currently provide a trial license for Autoware users free of charge.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/#creating-a-vector-map","title":"Creating a vector map","text":"
The easiest way to create an Autoware-compatible vector map is to use Vector Map Builder, a free web-based tool provided by TIER IV, Inc.. Vector Map Builder allows you to create lanes and add additional regulatory elements such as stop signs or traffic lights using a point cloud map as a reference.
For open-source software options, MapToolbox is a plugin for Unity specifically designed to create Lanelet2 maps for Autoware. Although JOSM is another open-source tool that can be used to create Lanelet2 maps, be aware that a number of modifications must be done manually to make the map compatible with Autoware. This process can be tedious and time-consuming, so the use of JOSM is not recommended.
If it is not possible to create HD maps yourself, you can use a mapping service from the following Autoware-compatible map providers instead:
MAP IV, Inc.
AISAN TECHNOLOGY CO., LTD.
TomTom
The table below shows each company's mapping technology and the types of HD maps they support.
Company Mapping technology Available maps MAP IV, Inc. SLAM Point cloud and vector maps AISAN TECHNOLOGY CO., LTD. MMS Point cloud and vector maps TomTom MMS Vector map*
Note
Maps provided by TomTom use their proprietary AutoStream format, not Lanelet2. The open-source AutoStreamForAutoware tool can be used to convert an AutoStream map to a Lanelet2 map. However, the converter is still in its early stages and has some known limitations.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/converting-utm-to-mgrs-map/","title":"Converting UTM maps to MGRS map format","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/converting-utm-to-mgrs-map/#converting-utm-maps-to-mgrs-map-format","title":"Converting UTM maps to MGRS map format","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/converting-utm-to-mgrs-map/#overview","title":"Overview","text":"
If you want to use MGRS (Military Grid Reference System) format in Autoware, you need to convert UTM (Universal Transverse Mercator) map to MGRS format. In order to do that, we will use UTM to MGRS pointcloud converter ROS 2 package provided by Leo Drive.
After the installation of converter tool, we need to define northing, easting and ellipsoid height of local UTM map origin in pc_utm_to_mgrs_converter.param.yaml. For example, you can use latitude, longitude and altitude values in the navsatfix message from your GNSS/INS sensor.
After that, you need to convert latitude and longitude values to northing and easting values. You can use any converter on the internet for converting latitude longitude values to UTM. (i.e., UTMconverter)
Now, we are ready to update pc_utm_to_mgrs_converter.param.yaml, example for our navsatfix message:
/**:\n ros__parameters:\n # Northing of local origin\n- Northing: 4520550.0\n+ Northing: 4542871.33\n\n # Easting of local origin\n- Easting: 698891.0\n+ Easting: 658659.84\n\n # Elipsoid Height of local origin\n- ElipsoidHeight: 47.62\n+ ElipsoidHeight: 74.28\n
Lastly, we will update input and pointcloud the map path in pc_utm_to_mgrs_converter.launch.xml:
The conversion process will be started, you should see Saved <YOUR-MAP-POINTS-SIZE> data points saved to <YOUR-OUTPUT-MAP-PATH> message on your terminal. So, MGRS format pointcloud map saved on your output map directory.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/","title":"Creating a vector map","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/#creating-a-vector-map","title":"Creating a vector map","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/#overview","title":"Overview","text":"
In this section, we will explain how to create Lanelet2 maps with TIER IV's vector map builder tool.
If you want to look at another method, MapToolbox is a plugin for Unity specifically designed to create Lanelet2 maps for Autoware. We didn't recommend JOSM since it needs modifications for Autoware usage. So, this process can be tedious and time-consuming.
You need a TIER IV account for using vector map builder tool. So, if you have not before, please create a TIER IV account in order to use vector map builder tool.
You can follow these pages for creating a Lanelet2 map and its regulatory elements.
Behavior velocity planner's crosswalk module plans velocity to stop or decelerate for pedestrians approaching or walking on a crosswalk. In order to operate that, we will add crosswalk attribute to our lanelet2 map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/crosswalk/#creating-a-crosswalk-attribute","title":"Creating a crosswalk attribute","text":"
In order to create a crosswalk on your map, please follow these steps:
Click Abstraction button on top panel.
Select Crosswalk from the panel.
Click and draw crosswalk on your pointcloud map.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/crosswalk/#testing-created-crosswalk-with-planning-simulator","title":"Testing created crosswalk with planning simulator","text":"
After the completing of creating the map, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Click 2D Pose Estimate button on rviz or press P and give a pose for initialization.
Click 2D Goal Pose button on rviz or press G and give a pose for goal point.
We need to add pedestrians to crosswalk, so activate interactive pedestrians from Tool Properties panel on rviz.
After that, please press Shift, then click right click button for inserting pedestrians.
You can control inserted pedestrian via dragging right click.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/","title":"Creating a Lanelet","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#creating-a-lanelet","title":"Creating a Lanelet","text":"
At this page, we will explain how to create a simple lanelet on your pointcloud map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#creating-a-lanelet2","title":"Creating a Lanelet2","text":"
Firstly, we need to import our pointcloud map to vector map builder tool:
Please click File.
Then, click Import PCD.
Click Browse and select your .pcd file.
You will display the point cloud on your vector map builder tool after the upload is complete:
Uploaded pointcloud map file on vector map builder
Now, we are ready to create lanelet2 map on our pointcloud map:
Please click Create.
Then, click Create Lanelet2Maps.
Please fill your map name
Please fill your MGRS zone. (At tutorial_vehicle, MGRS grid zone: 35T - MGRS 100,000-meter square: PF)
Click Create.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#creating-a-simple-lanelet","title":"Creating a simple lanelet","text":"
In order to create a simple lanelet on your map, please follow these steps:
CLick Lanelet2Maps on the bar
Enable Lanelet mode via selecting Lanelet.
Then, you can click the pointcloud map to create lanelet.
If your lanelet is finished, you can disable Lanelet.
If you want to change your lanelet width, click lanelet --> Change Lanelet Width, then you can enter the lanelet width.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#join-two-lanelets","title":"Join two lanelets","text":"
In order to join two lanelets, please follow these steps:
Please create two distinct lanelet.
Select a Lanelet, then press Shift and select other lanelet.
Now, you can see Join Lanelets button, just press it.
In order to add (join) two or more lanelets to another lanelet, please follow these steps:
Create multiple lanelets.
You can join the first two lanelets like the steps before.
Please check end points ids of first lanelet.
Then you need to change these ids with third lanelet's start point. (Please change with selecting linestring of lanelet)
You will see two next lanes of the first lanelet will be appeared.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#change-speed-limit-of-lanelet","title":"Change Speed Limit Of Lanelet","text":"
In order to change the speed limit of lanelet, please follow these steps:
Select the lanelet where the speed limit will be changed
Set speed limit on the right panel.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#test-lanelets-with-planning-simulator","title":"Test lanelets with planning simulator","text":"
After the completing of creating lanelets, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Click 2D Pose Estimate button on rviz or press P and give a pose for initialization.
Click 2D Goal Pose button on rviz or press G and give a pose for goal point.
Testing our created vector map with planning simulator"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/","title":"Stop Line","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/#stop-line","title":"Stop Line","text":"
Behavior velocity planner's stop line module plans velocity to stop right before stop lines and restart driving after stopped. In order to operate that, we will add stop line attribute to our lanelet2 map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/#creating-a-stop-line-regulatory-element","title":"Creating a stop line regulatory element","text":"
In order to create a stop line on your pointcloud map, please follow these steps:
Please select lanelet to add stop line.
Click Abstraction button on top panel.
Select Stop Line from the panel.
Click on the desired area for inserting stop line.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/#testing-created-the-stop-line-element-with-planning-simulator","title":"Testing created the stop line element with planning simulator","text":"
After the completing of creating the map, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Click 2D Pose Estimate button on rviz or press P and give a pose for initialization.
Click 2D Goal Pose button on rviz or press G and give a pose for goal point.
You can see the stop line marker on the rviz screen.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/","title":"Available Open Source SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/#available-open-source-slam","title":"Available Open Source SLAM","text":"
This page provides the list of available open source Simultaneous Localization And Mapping (SLAM) implementation that can be used to generate a point cloud (.pcd) map file.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/#selecting-which-implementation-to-use","title":"Selecting which implementation to use","text":"
Lidar odometry drifts accumulatively as time goes by and there is solutions to solve that problem such as graph optimization, loop closure and using gps sensor to decrease accumulative drift error. Because of that, a SLAM algorithm should have loop closure feature, graph optimization and should use gps sensor. Additionally, some of the algorithms are using IMU sensor to add another factor to graph for decreasing drift error. While some of the algorithms requires 9-axis IMU sensor strictly, some of them requires only 6-axis IMU sensor or not even using the IMU sensor. Before choosing an algorithm to create maps for Autoware please consider these factors depends on your sensor setup or expected quality of generated map.
Commonly used open-source SLAM implementations are lidarslam-ros2 (LiDAR, IMU*) and LIO-SAM (LiDAR, IMU, GNSS). The required sensor data for each algorithm is specified in the parentheses, where an asterisk (*) indicates that such sensor data is optional. For supported LiDAR models, please check the GitHub repository of each algorithm. While these ROS 2-based SLAM implementations can be easily installed and used directly on the same machine that runs Autoware, it is important to note that they may not be as well-tested or as mature as ROS 1-based alternatives.
The notable open-source SLAM implementations that are based on ROS 1 include hdl-graph-slam (LiDAR, IMU*, GNSS*), LeGO-LOAM (LiDAR, IMU*), LeGO-LOAM-BOR (LiDAR), and LIO-SAM (LiDAR, IMU, GNSS).
Most of these algorithms already have a built-in loop-closure and pose graph optimization. However, if the built-in, automatic loop-closure fails or does not work correctly, you can use Interactive SLAM to adjust and optimize a pose graph manually.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/#list-of-third-party-slam-implementations","title":"List of Third Party SLAM Implementations","text":"Package Name Explanation Repository Link Loop Closure Sensors ROS Version Dependencies FAST-LIO-LC A computationally efficient and robust LiDAR-inertial odometry package with loop closure module and graph optimization https://github.com/yanliang-wang/FAST_LIO_LC ✓ LidarIMUGPS [Optional] ROS 1 ROS MelodicPCL >= 1.8Eigen >= 3.3.4GTSAM >= 4.0.0 FAST_LIO_SLAM FAST_LIO_SLAM is the integration of FAST_LIO and SC-PGO which is scan context based loop detection and GTSAM based pose-graph optimization https://github.com/gisbi-kim/FAST_LIO_SLAM ✓ LidarIMUGPS [Optional] ROS 1 PCL >= 1.8Eigen >= 3.3.4 FD-SLAM FD_SLAM is Feature&Distribution-based 3D LiDAR SLAM method based on Surface Representation Refinement. In this algorithm novel feature-based Lidar odometry used for fast scan-matching, and used a proposed UGICP method for keyframe matching https://github.com/SLAMWang/FD-SLAM ✓ LidarIMU [Optional]GPS ROS 1 PCLg2oSuitesparse hdl_graph_slam An open source ROS package for real-time 6DOF SLAM using a 3D LIDAR. It is based on 3D Graph SLAM with NDT scan matching-based odometry estimation and loop detection. It also supports several graph constraints, such as GPS, IMU acceleration (gravity vector), IMU orientation (magnetic sensor), and floor plane (detected in a point cloud) https://github.com/koide3/hdl_graph_slam ✓ LidarIMU [Optional]GPS [Optional] ROS 1 PCLg2oOpenMP IA-LIO-SAM IA_LIO_SLAM is created for data acquisition in unstructured environment and it is a framework for Intensity and Ambient Enhanced Lidar Inertial Odometry via Smoothing and Mapping that achieves highly accurate robot trajectories and mapping https://github.com/minwoo0611/IA_LIO_SAM ✓ LidarIMUGPS ROS 1 GTSAM ISCLOAM ISCLOAM presents a robust loop closure detection approach by integrating both geometry and intensity information https://github.com/wh200720041/iscloam ✓ Lidar ROS 1 Ubuntu 18.04ROS MelodicCeresPCLGTSAMOpenCV LeGO-LOAM-BOR LeGO-LOAM-BOR is improved version of the LeGO-LOAM by improving quality of the code, making it more readable and consistent. Also, performance is improved by converting processes to multi-threaded approach https://github.com/facontidavide/LeGO-LOAM-BOR ✓ LidarIMU ROS 1 ROS MelodicPCLGTSAM LIO_SAM A framework that achieves highly accurate, real-time mobile robot trajectory estimation and map-building. It formulates lidar-inertial odometry atop a factor graph, allowing a multitude of relative and absolute measurements, including loop closures, to be incorporated from different sources as factors into the system https://github.com/TixiaoShan/LIO-SAM ✓ LidarIMUGPS [Optional] ROS 1ROS 2 PCLGTSAM Optimized-SC-F-LOAM An improved version of F-LOAM and uses an adaptive threshold to further judge the loop closure detection results and reducing false loop closure detections. Also it uses feature point-based matching to calculate the constraints between a pair of loop closure frame point clouds and decreases time consumption of constructing loop frame constraints https://github.com/SlamCabbage/Optimized-SC-F-LOAM ✓ Lidar ROS 1 PCLGTSAMCeres SC-A-LOAM A real-time LiDAR SLAM package that integrates A-LOAM and ScanContext. https://github.com/gisbi-kim/SC-A-LOAM ✓ Lidar ROS 1 GTSAM >= 4.0 SC-LeGO-LOAM SC-LeGO-LOAM integrated LeGO-LOAM for lidar odometry and 2 different loop closure methods: ScanContext and Radius search based loop closure. While ScanContext is correcting large drifts, radius search based method is good for fine-stitching https://github.com/irapkaist/SC-LeGO-LOAM ✓ LidarIMU ROS 1 PCLGTSAM"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/","title":"FAST_LIO_LC","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/#fast_lio_lc","title":"FAST_LIO_LC","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/#what-is-fast_lio_lc","title":"What is FAST_LIO_LC?","text":"
A computationally efficient and robust LiDAR-inertial odometry package with loop closure module and graph optimization.
mkdir -p ~/ws_fastlio_lc/src\n cd ~/ws_fastlio_lc/src\n git clone https://github.com/gisbi-kim/FAST_LIO_SLAM.git\n git clone https://github.com/Livox-SDK/livox_ros_driver\n cd ..\n catkin_make\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/#2-set-parameters","title":"2) Set parameters","text":"
After downloading the repository, change topic and sensor settings on the config file (workspace/src/FAST_LIO_LC/FAST_LIO/config/ouster64_mulran.yaml) with the lidar topic name in your bag file.
For imu-lidar compatibility, extrinsic matrices from calibration must be changed.
To enable auto-save, pcd_save_enable must be 1 from the launch file (workspace/src/FAST_LIO_LC/FAST_LIO/launch/mapping_ouster64_mulran.launch).
# open new terminal: run FAST-LIO\nroslaunch fast_lio mapping_ouster64.launch\n\n# open the other terminal tab: run SC-PGO\nroslaunch aloam_velodyne fastlio_ouster64.launch\n\n# play bag file in the other terminal\nrosbag play RECORDED_BAG.bag --clock\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/","title":"FAST_LIO_SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/#fast_lio_slam","title":"FAST_LIO_SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/#what-is-fast_lio_slam","title":"What is FAST_LIO_SLAM?","text":"
FAST_LIO_SLAM is the integration of FAST_LIO and SC-PGO which is scan context based loop detection and GTSAM based pose-graph optimization.
mkdir -p ~/catkin_fastlio_slam/src\n cd ~/catkin_fastlio_slam/src\n git clone https://github.com/gisbi-kim/FAST_LIO_SLAM.git\n git clone https://github.com/Livox-SDK/livox_ros_driver\n cd ..\n catkin_make\n source devel/setup.bash\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/#2-set-parameters","title":"2) Set parameters","text":"
Set imu and lidar topic on Fast_LIO/config/ouster64.yaml
# terminal 1: run FAST-LIO2\nroslaunch fast_lio mapping_ouster64.launch\n\n # open the other terminal tab: run SC-PGO\ncd ~/catkin_fastlio_slam\n source devel/setup.bash\n roslaunch aloam_velodyne fastlio_ouster64.launch\n\n # play bag file in the other terminal\nrosbag play xxx.bag -- clock --pause\n
You may have an interest in this version of FAST-LIO + Loop closure, implemented by yanliang-wang
Maintainer: Giseop Kim (paulgkim@kaist.ac.kr)
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/","title":"FD-SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/#fd-slam","title":"FD-SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/#what-is-fd-slam","title":"What is FD-SLAM?","text":"
FD_SLAM is Feature&Distribution-based 3D LiDAR SLAM method based on Surface Representation Refinement. In this algorithm novel feature-based Lidar odometry used for fast scan-matching, and used a proposed UGICP method for keyframe matching.
/hdl_graph_slam/dump (hdl_graph_slam/DumpGraph)\n- save all the internal data (point clouds, floor coeffs, odoms, and pose graph) to a directory.\n\n/hdl_graph_slam/save_map (hdl_graph_slam/SaveMap)\n- save the generated map as a PCD file.\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/#3-set-parameters","title":"3) Set parameters","text":"
All the configurable parameters are listed in launch/****.launch as ros params.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/hdl-graph-slam/","title":"hdl_graph_slam","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/hdl-graph-slam/#hdl_graph_slam","title":"hdl_graph_slam","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/hdl-graph-slam/#what-is-hdl_graph_slam","title":"What is hdl_graph_slam?","text":"
An open source ROS package for real-time 6DOF SLAM using a 3D LIDAR. It is based on 3D Graph SLAM with NDT scan matching-based odometry estimation and loop detection. It also supports several graph constraints, such as GPS, IMU acceleration (gravity vector), IMU orientation (magnetic sensor), and floor plane (detected in a point cloud).
Kenji Koide, Jun Miura, and Emanuele Menegatti, A Portable 3D LIDAR-based System for Long-term and Wide-area People Behavior Measurement, Advanced Robotic Systems, 2019 [link].
[Active Intelligent Systems Laboratory, Toyohashi University of Technology, Japan] [Mobile Robotics Research Team, National Institute of Advanced Industrial Science and Technology (AIST), Japan]
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/ia-lio-slam/","title":"IA-LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/ia-lio-slam/#ia-lio-sam","title":"IA-LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/ia-lio-slam/#what-is-ia-lio-sam","title":"What is IA-LIO-SAM?","text":"
IA_LIO_SLAM is created for data acquisition in unstructured environment and it is a framework for Intensity and Ambient Enhanced Lidar Inertial Odometry via Smoothing and Mapping that achieves highly accurate robot trajectories and mapping.
# open new terminal: run IA_LIO\n source devel/setup.bash\n roslaunch lio_sam mapping_ouster64.launch\n\n # play bag file in the other terminal\n rosbag play RECORDED_BAG.bag --clock\n
Thank you for citing IA-LIO-SAM(./config/doc/KRS-2021-17.pdf) if you use any of this code.
Part of the code is adapted from LIO-SAM (IROS-2020).
@inproceedings{legoloam2018shan,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Shan, Tixiao and Englot, Brendan},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
IA-LIO-SAM is based on LIO-SAM (T. Shan, B. Englot, D. Meyers, W. Wang, C. Ratti, and D. Rus. LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping).
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/","title":"ISCLOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#iscloam","title":"ISCLOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#what-is-iscloam","title":"What is ISCLOAM?","text":"
ISCLOAM presents a robust loop closure detection approach by integrating both geometry and intensity information.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#build-and-run","title":"Build and Run","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#1-clone-repository","title":"1. Clone repository","text":"
cd ~/catkin_ws/src\ngit clone https://github.com/wh200720041/iscloam.git\ncd ..\ncatkin_make -j1\nsource ~/catkin_ws/devel/setup.bash\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#2-set-parameter","title":"2. Set Parameter","text":"
Change the bag location and sensor parameters on launch files.
if you would like to generate the map of environment at the same time, you can run
roslaunch iscloam iscloam_mapping.launch\n
Note that the global map can be very large, so it may takes a while to perform global optimization, some lag is expected between trajectory and map since they are running in separate thread. More CPU usage will happen when loop closure is identified.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#ground-truth-comparison","title":"Ground Truth Comparison","text":"
If you use this work for your research, you may want to cite the paper below, your citation will be appreciated
@inproceedings{wang2020intensity,\n author={H. {Wang} and C. {Wang} and L. {Xie}},\n booktitle={2020 IEEE International Conference on Robotics and Automation (ICRA)},\n title={Intensity Scan Context: Coding Intensity and Geometry Relations for Loop Closure Detection},\n year={2020},\n volume={},\n number={},\n pages={2095-2101},\n doi={10.1109/ICRA40945.2020.9196764}\n}\n
Thanks for A-LOAM and LOAM(J. Zhang and S. Singh. LOAM: Lidar Odometry and Mapping in Real-time) and LOAM_NOTED.
Author: Wang Han, Nanyang Technological University, Singapore
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/","title":"LeGO-LOAM-BOR","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/#lego-loam-bor","title":"LeGO-LOAM-BOR","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/#what-is-lego-loam-bor","title":"What is LeGO-LOAM-BOR?","text":"
LeGO-LOAM-BOR is improved version of the LeGO-LOAM by improving quality of the code, making it more readable and consistent. Also, performance is improved by converting processes to multi-threaded approach.
cd ~/catkin_ws/src\ngit clone https://github.com/facontidavide/LeGO-LOAM-BOR.git\ncd ..\ncatkin_make\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/#2-set-parameters","title":"2) Set parameters","text":"
Thank you for citing our LeGO-LOAM paper if you use any of this code:
@inproceedings{legoloam2018,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Tixiao Shan and Brendan Englot},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/","title":"LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#lio-sam","title":"LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#what-is-lio-sam","title":"What is LIO-SAM?","text":"
A framework that achieves highly accurate, real-time mobile robot trajectory estimation and map-building. It formulates lidar-inertial odometry atop a factor graph, allowing a multitude of relative and absolute measurements, including loop closures, to be incorporated from different sources as factors into the system.
*Robosense lidars aren't supported officially, but their Helios series can be used as Velodyne lidars.
The system architecture of LIO-SAM method described in the following diagram, please look at the official repository for getting more information.
System Architecture of LIO-SAM
We are using Robosense Helios 5515 and CLAP B7 sensor on tutorial_vehicle, so we will use these sensors for running LIO-SAM.
Additionally, LIO-SAM tested with Applanix POS LVX and Hesai Pandar XT32 sensor setup. Some additional information according to the sensors will be provided in this page.
After the building of LIO-SAM, we need to record ROS 2 Bag file with including necessary topics for LIO-SAM. The necessary topics are described in the config file on LIO-SAM.
ROS 2 Bag example for LIO-SAM with Robosense Helios and CLAP B7
We will update sensor settings also. Since Robosense Lidars aren't officially supported, we will set our 32-channel Robosense Helios 5515 lidar as Velodyne:
The mapping direction is towards to the going direction in the real world. If LiDAR sensor is backwards, according to the direction you are moving, then you need to change the extrinsicRot too. Unless the IMU tries to go in the wrong direction, and it may occur problems.
For example, in our Applanix POS LVX and Hesai Pandar XT32 setup, IMU direction was towards to the going direction and LiDAR direction has 180 degree difference in Z-axis according to the IMU direction. In other words, they were facing back to each other. The tool may need a transformation for IMU for that.
In that situation, the calibration parameters changed as this:
If you are set configurations and create bag file for LIO-SAM, you can launch LIO-SAM with:
ros2 launch lio_sam run.launch.py\n
The rviz2 screen will be open, then you can play your bag file:
ros2 bag play <YOUR-BAG-FILE>\n
If the mapping process is finished, you can save map with calling this service:
ros2 service call /lio_sam/save_map lio_sam/srv/SaveMap \"{resolution: 0.2, destination: <YOUR-MAP-DIRECTORY>}\"\n
Here is the video for demonstration of LIO-SAM mapping in our campus environment:
The output map format is local UTM, we will change local UTM map to MGRS format for tutorial_vehicle. Also, if you want change UTM to MGRS for autoware, please follow convert-utm-to-mgrs-map page.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#example-result","title":"Example Result","text":"Sample Map Output for our Campus Environment"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#paper","title":"Paper","text":"
Thank you for citing LIO-SAM (IROS-2020) if you use any of this code.
@inproceedings{liosam2020shan,\n title={LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping},\n author={Shan, Tixiao and Englot, Brendan and Meyers, Drew and Wang, Wei and Ratti, Carlo and Rus Daniela},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={5135-5142},\n year={2020},\n organization={IEEE}\n}\n
Part of the code is adapted from LeGO-LOAM.
@inproceedings{legoloam2018shan,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Shan, Tixiao and Englot, Brendan},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
LIO-SAM is based on LOAM (J. Zhang and S. Singh. LOAM: Lidar Odometry and Mapping in Real-time).
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/","title":"Optimized-SC-F-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#optimized-sc-f-loam","title":"Optimized-SC-F-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#what-is-optimized-sc-f-loam","title":"What is Optimized-SC-F-LOAM?","text":"
An improved version of F-LOAM and uses an adaptive threshold to further judge the loop closure detection results and reducing false loop closure detections. Also it uses feature point-based matching to calculate the constraints between a pair of loop closure frame point clouds and decreases time consumption of constructing loop frame constraints.
In this folder, Ground Truth information, optimized pose information, F-LOAM pose information and time information are stored
mkdir -p ~/message/Scans\n\nChange line 383 in the laserLoopOptimizationNode.cpp to your own \"message\" folder path\n
(Do not forget to rebuild your package)
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#3-set-parameters","title":"3) Set parameters","text":"
Set LIDAR topic and LIDAR properties on 'sc_f_loam_mapping.launch'
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#example-result","title":"Example Result","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#results-on-kitti-sequence-00-and-sequence-05","title":"Results on KITTI Sequence 00 and Sequence 05","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#comparison-of-trajectories-on-kitti-dataset","title":"Comparison of trajectories on KITTI dataset","text":"
Test on KITTI sequence You can download the sequence 00 and 05 datasets from the KITTI official website and convert them into bag files using the kitti2bag open source method.
00: 2011_10_03_drive_0027 000000 004540
05: 2011_09_30_drive_0018 000000 002760
See the link: https://github.com/ethz-asl/kitti_to_rosbag
Thanks for SC-A-LOAM(Scan context: Egocentric spatial descriptor for place recognition within 3d point cloud map) and F-LOAM(F-LOAM : Fast LiDAR Odometry and Mapping).
@misc{https://doi.org/10.48550/arxiv.2204.04932,\n doi = {10.48550/ARXIV.2204.04932},\n\n url = {https://arxiv.org/abs/2204.04932},\n\n author = {Liao, Lizhou and Fu, Chunyun and Feng, Binbin and Su, Tian},\n\n keywords = {Robotics (cs.RO), FOS: Computer and information sciences, FOS: Computer and information sciences},\n\n title = {Optimized SC-F-LOAM: Optimized Fast LiDAR Odometry and Mapping Using Scan Context},\n\n publisher = {arXiv},\n\n year = {2022},\n\n copyright = {arXiv.org perpetual, non-exclusive license}\n}\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/","title":"SC-A-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/#sc-a-loam","title":"SC-A-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/#what-is-sc-a-loam","title":"What is SC-A-LOAM?","text":"
A real-time LiDAR SLAM package that integrates A-LOAM and ScanContext.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/#4-saving-as-pcd-file","title":"4) Saving as PCD file","text":"
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-lego-loam/","title":"SC-LeGO-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-lego-loam/#sc-lego-loam","title":"SC-LeGO-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-lego-loam/#what-is-sc-lego-loam","title":"What is SC-LeGO-LOAM?","text":"
SC-LeGO-LOAM integrated LeGO-LOAM for lidar odometry and 2 different loop closure methods: ScanContext and Radius search based loop closure. While ScanContext is correcting large drifts, radius search based method is good for fine-stitching.
@INPROCEEDINGS { gkim-2018-iros,\n author = {Kim, Giseop and Kim, Ayoung},\n title = { Scan Context: Egocentric Spatial Descriptor for Place Recognition within {3D} Point Cloud Map },\n booktitle = { Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems },\n year = { 2018 },\n month = { Oct. },\n address = { Madrid }\n}\n
and
@inproceedings{legoloam2018,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Shan, Tixiao and Englot, Brendan},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
In some cases, for example, when your created point cloud map is either too dense or too large (i.e., exceeding 300 MB), you may want to downsample it for improved computational and memory efficiency. Also, you can consider using dynamic map loading with partial loading, please check map_loader package for more information.
At tutorial_vehicle implementation we will use the whole map, so we will downsample it with using CloudCompare.
Please check the official page for installing options.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/pointcloud-map-downsampling/#downsampling-a-pointcloud-map","title":"Downsampling a pointcloud map","text":"
There are three subsampling methods on CloudCompare, we are using Space method for subsampling, but you can use other methods if you want.
Please open CloudCompare and drag your pointcloud to here, then you can select your pointcloud map by just clicking on the map at the DB tree panel.
Then you can click subsample button on the top panel.
CloudCompare
Please select on your subsample method, we will use space for tutorial_vehicle.
Then you can select options. For example, we need to determine minimum space between points. (Please be careful in this section, subsampling is depending on your map size, computer performance, etc.) We will set this value 0.2 for tutorial_vehicle's map.
Pointcloud subsampling
After the subsampling process is finished, you should select pointcloud on the DB Tree panel as well.
Select your downsampled pointcloud
Now, you can save your downsampled pointcloud with ctrl + s or you can click save button from File bar. Then, this pointcloud can be used by autoware.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/","title":"Calibrating your sensors","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#calibrating-your-sensors","title":"Calibrating your sensors","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#overview","title":"Overview","text":"
Autoware expects to have multiple sensors attached to the vehicle as input to perception, localization, and planning stack. These sensors must be calibrated correctly and their positions must be defined using either urdf files (as in sample_sensor_kit) or as tf launch files.
Navigation2 provides a good tutorial for camera internal calibration.
AutoCore provides a light-weight tool.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#lidar-lidar-calibration","title":"Lidar-lidar calibration","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#lidar-lidar-calibration-tool-from-autocore","title":"Lidar-Lidar Calibration tool from Autocore","text":"
LL-Calib on GitHub, provided by AutoCore, is a lightweight toolkit for online/offline 3D LiDAR to LiDAR calibration. It's based on local mapping and \"GICP\" method to derive the relation between main and sub lidar. Information on how to use the tool, troubleshooting tips and example rosbags can be found at the above link.
Developed by MathWorks, The Lidar Camera Calibrator app enables you to interactively estimate the rigid transformation between a lidar sensor and a camera.
SensorsCalibration toolbox v0.1: One more open source method for Lidar-camera calibration. This is a project for LiDAR to camera calibration,including automatic calibration and manual calibration
Developed by AutoCore, an easy-to-use lightweight toolkit for Lidar-camera-calibration is proposed. Only in three steps, a fully automatic calibration will be done.
Developed by APRIL Lab at Zhejiang University in China, the LI-Calib calibration tool is a toolkit for calibrating the 6DoF rigid transformation and the time offset between a 3D LiDAR and an IMU, based on continuous-time batch optimization. IMU-based cost and LiDAR point-to-surfel (surfel = surface element) distance are minimized jointly, which renders the calibration problem well-constrained in general scenarios.
AutoCore has forked the original LI-Calib tool and overwritten the Lidar input for more general usage. Information on how to use the tool, troubleshooting tips and example rosbags can be found at the LI-Calib fork on GitHub.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/","title":"Creating vehicle and sensor description","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/#creating-vehicle-and-sensor-description","title":"Creating vehicle and sensor description","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/#introduction","title":"Introduction","text":"
Resolves the positions of sensors with base_link as the parent frame and defines the positions and orientations based on sensors_calibration.yaml in individual_params.
In Autoware, <YOUR_SENSOR_KIT_description>/config/sensors_calibration.yaml is not used.
A sensor_kit refers to a subset that includes multiple sensors, and sensor_kit_base_link is the name of its frame. The positions and orientations within the kit are defined in sensor_kit.xacro.
Resolves the positions of sensors with sensor_kit_base_link as the parent and defines the positions and orientations based on sensor_kit_calibration.yaml in individual_params.
In Autoware, <YOUR_SENSOR_KIT_description>/config/sensor_kit_calibration.yaml is not used.
The individual_parameter is where parameters referenced by sensors.xacro and sensor_kit.xacro are stored. As the name imply, it is intended to manage parameters for multiple individual instances.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/#introduction-to-various-parameters","title":"Introduction to Various Parameters","text":"
vehicle_interface.launch.xml is the launch file related to the drive system. Please modify it according to the configuration of your vehicle's drive system.
If you are operating multiple vehicles, use the vehicle_id to switch to the corresponding configuration for each vehicle.
pointcloud_preprocessor.launch.py is the launch file to convert the raw sensor data. For more information, please click here.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/","title":"Creating a vehicle interface for an Ackermann kinematic model","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#creating-a-vehicle-interface-for-an-ackermann-kinematic-model","title":"Creating a vehicle interface for an Ackermann kinematic model","text":"
This page introduces a module vehicle interface and explains how to implement it.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#what-is-a-vehicle-interface","title":"What is a vehicle interface","text":"
Vehicle interface is an interface that connects the control commands and your vehicle's control device. Autoware publishes control commands such as:
Velocity control
Steering control
Car light commands
Then, the vehicle interface converts these commands into actuation such like:
Motor and brake activation
Steering wheel operation
Lighting control
So think of the vehicle interface as a module that runs the vehicle's control device to realize the input commands provided by Autoware.
An example of inputs and outputs for vehicle interface
This page shows you a brief explanation how to implement your vehicle interface, but you can see further information of vehicle interface in the \"design\" page.
Note that there is no package named \"vehicle interface\" prepared in Autoware. It is a necessary package to actuate your vehicle, but you have to create one by yourself since it is very specific to your vehicle's control device.
For example, if you are using a by-wire kit PACMod, a vehicle interface named pacmod_interface published by TIER IV, Inc. is available. However, if you have constructed something original and haven't found an open source vehicle interface applicable, you have to implement your own vehicle interface from scratch.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#how-to-implement-a-vehicle-interface","title":"How to implement a vehicle interface","text":"
The following instructions describe how to create a vehicle interface.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#1-create-a-directory-for-vehicle-interface","title":"1. Create a directory for vehicle interface","text":"
It is recommended to create your vehicle interface at <your-autoware-dir>/src/vehicle/external
cd <your-autoware-dir>/src/vehicle/external\n
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#2-install-or-implement-your-own-vehicle-interface","title":"2. Install or implement your own vehicle interface","text":"
If there is an already complete vehicle interface package (like pacmod_interface), you can install it to your environment. If not, you have to implement your own vehicle interface by yourself. Let's create a new package by ros2 pkg create. The following example will show you how to create a vehicle interface package named my_vehicle_interface.
Then, you should write your implementation of vehicle interface in my_vehicle_interface/src. Again, since this implementation is so specific to the control device of your vehicle, it is beyond the scope of this document to describe how to implement your vehicle interface in detail. Here are some factors that might be considered.
Subscription of control command topics from Autoware
Communication between the vehicle interface and your vehicle's control device
Modification of control values if needed
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#3-prepare-a-launch-file","title":"3. Prepare a launch file","text":"
After you implement your vehicle interface or you want to debug it by launching it, create a launch file of your vehicle interface, and include it to vehicle_interface.launch.xml.
Do not get confused. First, you need to create a launch file for your own vehicle interface module (like my_vehicle_interface.launch.xml) and then include that to vehicle_interface.launch.xml which exists in another directory. Here are the details.
Add a launch directory in the my_vehicle_interface directory, and create a launch file of your own vehicle interface in it. Take a look at Creating a launch file in the ROS 2 documentation.
Next, go to <your-autoware-dir>/src/vehicle, copy the directory /sample_vehicle_launch/, and paste it to the same place (which means it should be lined up with external and sample_vehicle_launch).
You have to rename each \"sample_vehicle\" to something else. For example, if you want to rename \"sample_vehicle\" to \"my_vehicle_name\", you need to change the following. Note that it is restricted to keep the \"_launch\" and \"_description\" part.
After you rename your directories, rename each \"sample_vehicle\" to \"my_vehicle_name\" in the source code.
my_vehicle_name_description/CMakeLists.txt
my_vehicle_name_description/package.xml
my_vehicle_name_description/urdf/vehicle.xacro (there are two parts)
my_vehicle_name_launch/CMakeLists.txt
my_vehicle_name_launch/package.xml
README.md
Include your launch file to my_vehicle_name_launch/my_vehicle_name_launch/launch/vehicle_interface.launch.xml by opening it and add the include terms like below.
Finally, your directory structure may look like below. Most of the files are omitted for clarity, but the files shown here needs modification as said in the previous and current process.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#4-build-the-vehicle-interface-package-and-the-launch-package","title":"4. Build the vehicle interface package and the launch package","text":"
Build three packages my_vehicle_interface, my_vehicle_name_launch and my_vehicle_name_description by colcon build, or you can just build the entire Autoware if you have done other things.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#5-when-you-launch-autoware","title":"5. When you launch Autoware","text":"
Finally, you are done implementing your vehicle interface module! Be careful that you need to launch Autoware with the proper vehicle_model option like the example below. This example is launching planning simulator.
You can subdivide your vehicle interface into smaller packages if you want. Then your directory structure may look like below (not the only way though). Do not forget to launch all packages in my_vehicle_interface.launch.xml.
If you are using a vehicle interface and launch package from a open git repository, or created your own as a git repository, it is highly recommended to add those repositories to your autoware.repos file which is located to directly under your autoware folder like the example below. You can specify the branch or commit hash by the version tag.
autoware.repos
# vehicle (this section should be somewhere in autoware.repos and add the below)\nvehicle/my_vehicle_name_launch:\ntype: git\nurl: https://github.com/<repository-name-A>/my_vehicle_name_launch.git\nversion: main\nvehicle/external/my_vehicle_interface:\ntype: git\nurl: https://github.com/<repository-name-B>/my_vehicle_interface.git\nversion: main\n
Then you can import your entire environment easily to another local device by using the vcs import command. (See the source installation guide)
Autoware now supports control inputs for vehicles based on an Ackermann kinematic model. This section introduces you a brief concept of Ackermann kinematic model and explains how Autoware controls it.
If your vehicle does not suit the Ackermann kinematic model, you have to modified the control commands. Another document gives you an example how to convert your Ackermann kinematic model control inputs into a differential drive model.
The basic style of Ackermann kinematic model has four wheels with an Ackermann link on the front, and it is powered by the rear wheels. The key point of Ackermann kinematic model is that the axes of all wheels intersect at a same point, which means all wheels will trace a circular trajectory with a different radii but a common center point (See the figure below). Therefore, this model has a great advantage that it minimizes the slippage of the wheels, and prevent tires to get worn soon.
In general, Ackermann kinematic model accepts the longitudinal speed \\(v\\) and the steering angle \\(\\phi\\) as inputs. In autoware, \\(\\phi\\) is positive if it is steered counter clockwise, so the steering angle in the figure below is actually negative.
The basic style of an Ackermann kinematic model. The left figure shows a vehicle facing straight forward, while the right figure shows a vehicle steering to the right."},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#control","title":"Control","text":"
Autoware publishes a ROS 2 topic named control_cmd from several types of publishers. A control_cmd topic is a AckermannControlCommand type message that contains
See the AckermannLateralCommand.idl and LongitudinalCommand.idl for details.
The vehicle interface should realize these control commands through your vehicle's control device.
Moreover, Autoware also provides brake commands, light commands, and more (see vehicle interface design), so the vehicle interface module should be applicable to these commands as long as there are devices available to handle them.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/","title":"Customizing for differential drive vehicle","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#customizing-for-differential-drive-vehicle","title":"Customizing for differential drive vehicle","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#1-introduction","title":"1. Introduction","text":"
Currently, Autoware assumes that vehicles use an Ackermann kinematic model with Ackermann steering. Thus, Autoware adopts the Ackermann command format for the Control module's output (see the AckermannDrive ROS message definition for an overview of Ackermann commands, and the AckermannControlCommands struct used in Autoware for more details).
However, it is possible to integrate Autoware with a vehicle that follows a differential drive kinematic model, as commonly used by small mobile robots.
One simple way of using Autoware with a differential drive vehicle is to create a vehicle_interface package that translates Ackermann commands to differential drive commands. Here are two points that you need to consider:
Create vehicle_interface package for differential drive vehicle
Set an appropriate wheel_base
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#21-create-a-vehicle_interface-package-for-differential-drive-vehicle","title":"2.1 Create a vehicle_interface package for differential drive vehicle","text":"
An Ackermann command in Autoware consists of two main control inputs:
steering angle (\\(\\omega\\))
velocity (\\(v\\))
Conversely, a typical differential drive command consists of the following inputs:
left wheel velocity (\\(v_l\\))
right wheel velocity (\\(v_r\\))
So, one way in which an Ackermann command can be converted to a differential drive command is by using the following equations:
\\[ v_l = v - \\frac{l\\omega}{2}, v_r = v + \\frac{l\\omega}{2} \\]
where \\(l\\) denotes wheel tread.
For information about other factors that need to be considered when creating a vehicle_interface package, refer to the vehicle_interface component page.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#22-set-an-appropriate-wheel_base","title":"2.2 Set an appropriate wheel_base","text":"
A differential drive robot does not necessarily have front and rear wheels, which means that the wheelbase (the horizontal distance between the axles of the front and rear wheels) cannot be defined. However, Autoware expects wheel_base to be set in vehicle_info.param.yaml with some value. Thus, you need to set a pseudo value for wheel_base.
The appropriate pseudo value for wheel_base depends on the size of your vehicle. Setting it to be the same value as wheel_tread is one possible choice.
Warning
If the wheel_base value is set too small then the vehicle may behave unexpectedly. For example, the vehicle may drive beyond the bounds of a calculated path.
Conversely, if wheel_base is set too large, the vehicle's range of motion will be restricted. The reason being that Autoware's Planning module will calculate an overly conservative trajectory based on the assumed vehicle length.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#3-known-issues","title":"3. Known issues","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#motion-model-incompatibility","title":"Motion model incompatibility","text":"
Since Autoware assumes that vehicles use a steering system, it is not possible to take advantage of the flexibility of a differential drive system's motion model.
For example, when planning a parking maneuver with the freespace_planner module, Autoware may drive the differential drive vehicle forward and backward, even if the vehicle can be parked with a simpler trajectory that uses pure rotational movement.
"},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/","title":"Creating Autoware meta-repository","text":""},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#creating-autoware-meta-repository","title":"Creating Autoware meta-repository","text":""},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#what-is-meta-repository","title":"What is Meta-repository?","text":"
A meta-repository is a repository that manages multiple repositories, and Autoware is one of them. It serves as a centralized control point for referencing, configuring, and versioning other repositories.
By using Ansible and VCS, you can automatically set up your Autoware. autoware.repos file manages the configuration of multiple repositories.
Note: VCS stands for Version Control System, such as Git or Subversion.
"},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#how-to-create-and-customize-your-autoware-meta-repository","title":"How to create and customize your autoware meta-repository","text":""},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#1-create-autoware-repository","title":"1. Create autoware repository","text":"
If you want to integrate Autoware into your vehicle, the first step is to create an Autoware meta-repository.
One easy way is to fork autowarefoundation/autoware and clone it. For how to fork a repository, refer to GitHub Docs.
If you set up multiple types of vehicles, adding a suffix like autoware.vehicle_A or autoware.vehicle_B is recommended
"},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#2-customize-your-autowarerepos-for-your-environment","title":"2. Customize your autoware.repos for your environment","text":"
You need to customize autoware.repos for your own vehicle's Autoware.
For example, if you want to customize the parameters in your individual_params or autoware_launch package to fit your vehicle, you can modify the configuration of each package and use them accordingly.
Please edit the parameters in Autoware's autoware_individual_params and autoware_launch packages to match your vehicle's specific requirements, as these packages provide sample parameters and may not be tailored to your vehicle by default.
If you want to fork autoware_individual_params and make modifications, it would be as follows:
Example: If you fork individual_params and rename autoware_individual_params.vehicle_A:
Please refer to the following documentation link for instructions on how to create and customize each vehicle_interface:
creating-vehicle-and-sensor-description
creating-vehicle-interface-package
customizing-for-differential-drive-model
Please remember to add all your custom packages, such as interfaces and descriptions, to your autoware.repos to ensure that your packages are properly included and managed within the Autoware repository.
It is possible to specify which components to launch using command-line arguments. For example, if you don't need to launch perception, planning, and control for localization debug, you can launch the following:
The basic command-line options are documented in autoware.launch.xml.
There are options available to switch between different methods for some component. For example, by specifying pose_source/twist_source or perception_mode, you can switch localization and perception methods, respectively. These options allow you to choose the desired algorithms or sensor configurations for the respective functionalities.
For options on eagleye component, please refer to the sub-pages.
Click the 2D Nav Goal button in the toolbar, or hit the G key
In the 3D View pane, click and hold the left mouse button, and then drag to set the direction for the goal pose. If successful, you will see the calculated planning path on RViz.
You can also engage via RViz with \"AutowareStatePanel\". The panel can be found in Panels > Add New Panel > tier4_state_rviz_plugin > AutowareStatePanel.
Once the route is computed, the \"AUTO\" button becomes active. Pressing the AUTO button engages the autonomous driving mode.
Now the vehicle should drive along the calculated path!
During the autonomous driving, the StatePanel appears as shown in the image below. Pressing the \"STOP\" button allows you to stop the vehicle.
By specifying the perception_mode, users can switch between different sensor configurations for perception. This allows you to choose the specific sensor setup that you want to use for the perception tasks.
Current localization launcher implemented by TIER IV supports multiple localization methods, both pose estimators and twist estimators. tier4_localization_component.launch.xml has two arguments to select which estimators to launch:
pose_source : an argument to select pose_estimator, currently supporting ndt (default), yabloc, and eagleye
twist_source : an argument to select twist_estimator, currently supporting gyro_odom (default), and eagleye
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/#ndt-scan-matcher-a-lidar-and-pointcloud-map-based-pose-estimator-default","title":"NDT scan matcher: a LiDAR and pointcloud map based pose estimator (default)","text":"
By default, Autoware launches ndt_scan_matcher for pose estimator. In order to launch this explicitly, you need to specify as follows:
Note that currently twist_source is set to Gyro Odometer as default, so you can skip this argument.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/#yabloc-a-camera-and-vector-map-based-pose-estimator","title":"YabLoc: a camera and vector map based pose estimator","text":"
You can use YabLoc as a camera-based localization method. For more details on YabLoc, please refer to the README of YabLoc in autoware.universe.
To use YabLoc as a pose_estimator, add pose_source:=yabloc when launching Autoware. By default, the pose_source is set to ndt. By specifying this command-line argument, YabLoc nodes will be automatically launched while the NDT nodes will not be started.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/#eagleye-a-gnss-imu-wheel-odometry-based-pose-and-twist-estimator","title":"Eagleye: a GNSS & IMU & wheel odometry based pose and twist estimator","text":"
You can use Eagleye as a GNSS & IMU & wheel odometry-based localization method. For more details on Eagleye, please refer to the Eagleye.
Eagleye has a function for position estimation and twist estimation, namely pose_estimator and twist_estimator, respectively. When running Eagleye in twist_estimator mode with other pose_estimator such as ndt_scan_matcher, Eagleye is still helpful since it can improve scan matching by providing accurate twists using GNSS doppler.
You can use Eagleye by specifying the pose_source and twist_source accordingly through command-line arguments.
Example of using Eagleye as the pose twist estimator:
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/","title":"Eagleye","text":""},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#using-eagleye-with-autoware","title":"Using Eagleye with Autoware","text":"
This page will show you how to set up Eagleye in order to use it with Autoware. For the details of the integration proposal, please refer to this discussion.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#what-is-eagleye","title":"What is Eagleye?","text":"
Eagleye is an open-source GNSS/IMU-based localizer initially developed by MAP IV. Inc. It provides a cost-effective alternative to LiDAR and point cloud-based localization by using low-cost GNSS and IMU sensors to provide vehicle position, orientation, and altitude information.
Eagleye can be utilized in the Autoware localization stack in two ways:
Feed only twist into the EKF localizer.
Feed both twist and pose from Eagleye into the EKF localizer (twist can also be used with regular gyro_odometry).
Note: RTK positioning is required when using Eagleye as the pose estimator. On the other hand, it is not mandatory when using it as the twist estimator.
Eagleye requires latitude/longitude height and doppler velocity generated by the GNSS receiver. Your GNSS ROS driver must publish the following messages:
sensor_msgs/msg/NavSatFix: This message contains latitude, longitude, and height information.
geometry_msgs/msg/TwistWithCovarianceStamped: This message contains gnss doppler velocity information.
Eagleye has been tested with the following example GNSS ROS drivers: ublox_gps and septentrio_gnss_driver. The settings needed for each of these drivers are as follows:
GNSS ROS drivers modification ublox_gps No additional settings are required. It publishes sensor_msgs/msg/NavSatFix and geometry_msgs/msg/TwistWithCovarianceStamped required by Eagleye with default settings. septentrio_gnss_driver Set publish.navsatfix and publish.twist in the config file gnss.yaml to true"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#parameter-modifications-for-integration-into-your-vehicle","title":"Parameter Modifications for Integration into Your Vehicle","text":""},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#topic-name-topic-type","title":"topic name & topic type","text":"
The users must correctly specify input topics for GNSS latitude, longitude, and height , GNSS doppler speed , IMU , and vehicle speed in the eagleye_config.yaml.
Also, the frequency of GNSS and IMU must be set in eagleye_config.yaml
common:\nimu_rate: 50\ngnss_rate: 5\n
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#conversion-from-fix-to-pose","title":"Conversion from fix to pose","text":"
The parameters for converting sensor_msgs/msg/NavSatFix to geometry_msgs/msg/PoseWithCovarianceStamped is listed in fix2pose.yaml. If you use a different geoid or projection type, change these parameters.
The other parameters are described here. Basically, these do not need to be changed .
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#notes-on-initialization","title":"Notes on initialization","text":"
Eagleye requires an initialization process for proper operation. Without initialization, the output for twist will be in the raw value, and the pose data will not be available.
The first step is static initialization, which involves allowing the Eagleye to remain stationary for approximately 5 seconds after startup to estimate the yaw-rate offset.
The next step is dynamic initialization, which involves running the Eagleye in a straight line for approximately 30 seconds. This process estimates the scale factor of wheel speed and azimuth angle.
Once dynamic initialization is complete, the Eagleye will be able to provide corrected twist and pose data.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#how-to-check-the-progress-of-initialization","title":"How to check the progress of initialization","text":"
TODO
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#note-on-georeferenced-maps","title":"Note on georeferenced maps","text":"
Note that the output position might not appear to be in the point cloud maps if you are using maps that are not properly georeferenced. In the case of a single GNSS antenna, initial position estimation (dynamic initialization) can take several seconds to complete after starting to run in an environment where GNSS positioning is available.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/","title":"Evaluating the controller performance","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#evaluating-the-controller-performance","title":"Evaluating the controller performance","text":"
This page shows how to use control_performance_analysis package to evaluate the controllers.
control_performance_analysis is the package to analyze the tracking performance of a control module and monitor the driving status of the vehicle.
If you need more detailed information about package, refer to the control_performance_analysis.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#how-to-use","title":"How to use","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#before-driving","title":"Before Driving","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#1-firstly-you-need-to-launch-autoware-you-can-also-use-this-tool-with-real-vehicle-driving","title":"1. Firstly you need to launch Autoware. You can also use this tool with real vehicle driving","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#2-initialize-the-vehicle-and-send-goal-position-to-create-route","title":"2. Initialize the vehicle and send goal position to create route","text":"
If you have any problem with launching Autoware, please see the tutorials page.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#3-launch-the-control_performance_analysis-package","title":"3. Launch the control_performance_analysis package","text":"
After this command, you should be able to see the driving monitor and error variables in topics.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#4-run-the-plotjuggler-in-sourced-terminal","title":"4. Run the PlotJuggler in sourced terminal","text":"
source ~/autoware/install/setup.bash\n
ros2 run plotjuggler plotjuggler\n
If you do not have PlotJuggler in your computer, please refer here for installation guideline.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#5-increase-the-buffer-size-maximum-is-100-and-import-the-layout-from-autowareuniversecontrolcontrol_performance_analysisconfigcontroller_monitorxml","title":"5. Increase the buffer size (maximum is 100), and import the layout from /autoware.universe/control/control_performance_analysis/config/controller_monitor.xml","text":"
After import the layout, please specify the topics that are listed below.
/localization/kinematic_state
/vehicle/status/steering_status
/control_performance/driving_status
/control_performance/performance_vars
Please mark the If present, use the timestamp in the field [header.stamp] box, then select the OK.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#6-now-you-can-start-to-driving-you-should-see-all-the-performance-and-driving-variables-in-plotjuggler","title":"6. Now, you can start to driving. You should see all the performance and driving variables in PlotJuggler","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#after-driving","title":"After Driving","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#1-you-can-export-the-statistical-output-and-all-data-to-compare-and-later-usage","title":"1. You can export the statistical output and all data to compare and later usage","text":"
With statistical data, you can export the all statistical values like (min, max, average) to compare the controllers.
You can also export all data to later use. To investigate them again, after launch PlotJuggler, import the .cvs file from data section.
You can plot the vehicle position. Select the two curve (keeping CTRL key pressed) and Drag & Drop them using the RIGHT Mouse button. Please visit the Help -> Cheatsheet in PlotJuggler to see more tips about it.
If you see too much noised curve in plots, you can adjust the odom_interval and low_pass_filter_gain from here to avoid noised data.
Autoware should be real-time system when integrated to a service. Therefore, the response time of each callback should be as small as possible. If Autoware appears to be slow, it is imperative to conduct performance measurements and implement improvements based on the analysis. However, Autoware is a complex software system comprising numerous ROS 2 nodes, potentially complicating the process of identifying bottlenecks. To address this challenge, we will discuss methods for conducting detailed performance measurements for Autoware and provide case studies. It is worth noting that multiple factors can contribute to poor performance, such as scheduling and memory allocation in the OS layer, but our focus in this page will be on user code bottlenecks. The outline of this section is as follows:
Improvement is impossible without precise measurements. To measure the performance of the application code, it is essential to eliminate any external influences. Such influences include interference from the operating system and CPU frequency fluctuations. Scheduling effects also occur when core resources are shared by multiple threads. This section outlines a technique for accurately measuring the performance of the application code for a specific node. Though this section only discusses the case of Linux on Intel CPUs, similar considerations should be made in other environments.
To eliminate the influence of scheduling, the node being measured should operate independently, using the same logic as when the entire Autoware system is running. To accomplish this, record all input topics of the node to be measured while the whole Autoware system is running. To achieve this objective, a tool called ros2_single_node_replayer has been prepared.
Details on how to use the tool can be found in the README. This tool records the input topics of a specific node during the entire Autoware operation and replays it in a single node with the same logic. The tool relies on the ros2 bag record command, and the recording of service/action is not supported as of ROS 2 Humble, so nodes that use service/action as their main logic may not work well.
Isolated cores running the node to be measured must meet the following conditions.
Fix CPU frequency and disable turbo boost
Minimize timer interruptions
Offload RCU (Read Copy Update) callback
Isolate the paired core if hyper-threading enabled
To fulfill these conditions on Linux, a custom kernel build with the following kernel configurations is required. You can find many resources to instruct you on how to build a custom Linux kernel (like this one). Note that even if full tickless is enabled, timer interrupts are generated for scheduling if more than two tasks exist in one core.
# Enable CONFIG_NO_HZ_FULL\n-> General setup\n-> Timers subsystem\n-> Timer tick handling (Full dynticks system (tickless))\n(X) Full dynticks system (tickless)\n\n# Allows RCU callback processing to be offloaded from selected CPUs\n# (CONFIG_RCU_NOCB_CPU=y)\n-> General setup\n-> RCU Subsystem\n-*- Offload RCU callback processing from boot-selected CPUs\n
Additionally, the kernel boot parameters need to be set as follows.
In the above configuration, for example, the node to be measured is assumed to run on core 2, and core 8, which is a hyper-threading pair, is also being isolated. Appropriate decisions on which cores to run the measurement target and which nodes to isolate need to be made based on the cache and core layout of the measurement machine. You can easily check if it is properly configured by running cat /proc/softirqs. Since intel_pstate=disable is specified in the kernel boot parameter, userspace can be specified in the scaling governor.
cat /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor // ondemand\nsudo sh -c \"echo userspace > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor\"\n
This allows you to freely set the desired frequency within a defined range.
sudo sh -c \"echo <freq(kz)> > /sys/devices/system/cpu/cpu2/cpufreq/scaling_setspeed\"\n
Turbo Boost needs to be switched off on Intel CPUs, which is often overlooked.
sudo sh -c \"echo 0 > /sys/devices/system/cpu/cpufreq/boost\"\n
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-real-time-performance/#run-single-node-separately","title":"Run single node separately","text":"
Following the instructions in the ros2_single_node_replayer README, start the node and play the dedicated rosbag created by the tool. Before playing the rosbag, appropriately set the CPU affinity of the thread on which the node runs, so it is placed on the isolated core prepared.
taskset --cpu-list -p <target cpu> <pid>\n
To avoid interference in the last level cache, minimize the number of other applications running during the measurement.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-real-time-performance/#measurement-and-visualization","title":"Measurement and visualization","text":"
To visualize the performance of the measurement target, embed code for logging timestamps and performance counter values in the target source code. To achieve this objective, a tool called pmu_analyzer has been prepared.
Details on how to use the tool can be found in the README. This tool can measure the turnaround time of any section in the source code, as well as various performance counters.
In this section, we will present several case studies that demonstrate the performance improvements. These examples not only showcase our commitment to enhancing the system's efficiency but also serve as a valuable resource for developers who may face similar challenges in their own projects. The performance improvements discussed here span various components of the Autoware system, including sensing modules and planning modules. There are tendencies for each component regarding which points are becoming bottlenecks. By examining the methods, techniques, and tools employed in these case studies, readers can gain a better understanding of the practical aspects of optimizing complex software systems like Autoware.
First, we will explain the procedure for performance improvement, taking the node ring_outlier_filter as an example. Refer to the Pull Request for details.
The following figure is a time-series plot of the turnaround time of the main processing part of ring_outlier_filter, analyzed as described in the \"Performance Measurement\" section above.
The horizontal axis indicates the number of callbacks called (i.e., callback index), and the vertical axis indicates the turnaround time.
When analyzing the performance of the sensing module from the viewpoint of performance counter, pay attention to instructions, LLC-load-misses, LLC-store-misses, cache-misses, and minor-faults.
Analysis of the performance counter shows that the largest fluctuations come from minor-faults (i.e., soft page faults), the second largest from LLC-store-misses and LLC-load-misses (i.e., cache misses in the last level cache), and the slowest fluctuations come from instructions (i.e., message data size fluctuations). For example, when we plot minor-faults on the horizontal axis and turnaround time on the vertical axis, we can see the following dominant proportional relationship.
To achieve zero soft page faults, heap allocations must only be made from areas that have been first touched in advance. We have developed a library called heaphook to avoid soft page faults while running Autoware callback. If you are interested, refer to the GitHub discussion and the issue.
To reduce LLC misses, it is necessary to reduce the working set and to use cache-efficient access patterns.
In the sensing component, which handles large message data such as LiDAR point cloud data, minimizing copying is important. A callback that takes sensor data message types as input and output should be written in an in-place algorithm as much as possible. This means that in the following pseudocode, when generating output_msg from input_msg, it is crucial to avoid using buffers as much as possible to reduce the number of memory copies.
To improve cache efficiency, implement an in-place style as much as possible, instead of touching memory areas sporadically. In ROS applications using PCL, the code shown below is often seen.
void callback(const sensor_msgs::PointCloud2ConstPtr &input_msg) {\npcl::PointCloud<PointT>::Ptr input_pcl(new pcl::PointCloud<PointT>);\npcl::fromROSMsg(*input_msg, *input_pcl);\n\n// Algorithm is described for point cloud type of pcl\npcl::PointCloud<PointT>::Ptr output_pcl(new pcl::PointCloud<PointT>);\nfill_pcl(*input_pcl, *output_pcl);\n\nauto output_msg = allocate_msg<sensor_msgs::PointCloud2>(output_size);\npcl::toROSMsg(*output_pcl, *output_msg);\npublish(std::move(output_msg));\n}\n
To use the PCL library, fromROSMsg() and toROSMsg() are used to perform message type conversion at the beginning and end of the callback. This is a wasteful copying process and should be avoided. We should eliminate unnecessary type conversions by removing dependencies on PCL (e.g., https://github.com/tier4/velodyne_vls/pull/39). For large message types such as map data, there should be only one instance in the entire system in terms of physical memory.
First, we will pick up detection_area module in behavior_velocity_planner node, which tends to have long turnaround time. We have followed the performance analysis steps above to obtain the following graph. Axises are the same as the graphs in the sensing case study.
Using pmu_analyzer tool to further identify the bottleneck, we have found that the following multiple loops were taking up a lot of processing time:
for ( area : detection_areas )\nfor ( point : point_clouds )\nif ( boost::geometry::within(point, area) )\n// do something with O(1)\n
It checks whether each point cloud is contained in each detection area. Let N be the size of point_clouds and M be the size of detection_areas, then the computational complexity of this program is O(N^2 * M), since the complexity of within is O(N). Here, given that most of the point clouds are located far away from a certain detection area, a certain optimization can be achieved. First, calculate the minimum enclosing circle that completely covers the detection area, and then check whether the points are contained in that circle. Most of the point clouds can be quickly ruled out by this method, we don\u2019t have to call the within function in most cases. Below is the pseudocode after optimization.
for ( area : detection_areas )\ncircle = calc_minimum_enclosing_circle(area)\nfor ( point : point_clouds )\nif ( point is in circle )\nif ( boost::geometry::within(point, area) )\n// do something with O(1)\n
By using O(N) algorithm for minimum enclosing circle, the computational complexity of this program is reduced to almost O(N * (N + M)) (note that the exact computational complexity does not really change). If you are interested, refer to the Pull Request.
Similar to this example, in the planning component, we take into consideration thousands to tens of thousands of point clouds, thousands of points in a path representing our own route, and polygons representing obstacles and detection areas in the surroundings, and we repeatedly create paths based on them. Therefore, we access the contents of the point clouds and paths multiple times using for-loops. In most cases, the bottleneck lies in these naive for-loops. Here, understanding Big O notation and reducing the order of computational complexity directly leads to performance improvements.
"},{"location":"how-to-guides/others/add-a-custom-ros-message/","title":"Add a custom ROS message","text":""},{"location":"how-to-guides/others/add-a-custom-ros-message/#add-a-custom-ros-message","title":"Add a custom ROS message","text":""},{"location":"how-to-guides/others/add-a-custom-ros-message/#overview","title":"Overview","text":"
During the Autoware development, you will probably need to define your own messages. Read the following instructions before adding a custom message.
Message in autoware_msgs define interfaces of Autoware Core.
If a contributor wishes to make changes or add new messages to autoware_msgs, they should first create a new discussion post under the Design category.
Any other minor or proposal messages used for internal communication within a component(such as planning) should be defined in another repository.
tier4_autoware_msgs is an example of that.
The following is a simple tutorial of adding a message package to autoware_msgs. For the general ROS 2 tutorial, see Create custom msg and srv files.
"},{"location":"how-to-guides/others/add-a-custom-ros-message/#how-to-create-custom-message","title":"How to create custom message","text":"
Make sure you are in the Autoware workspace, and then run the following command to create a new package. As an example, let's create a package to define sensor messages.
Create a package
cd ./src/core/autoware_msgs\nros2 pkg create --build-type ament_cmake autoware_sensing_msgs\n
Create custom messages
You should create .msg files and place them in the msg directory.
NOTE: The initial letters of the .msg and .srv files must be capitalized.
As an example, let's make .msg files GnssInsOrientation.msg and GnssInsOrientationStamped.msg to define GNSS/INS orientation messages:
We need to replace <buildtool_depend>ament_cmake</buildtool_depend> with <buildtool_depend>ament_cmake_auto</buildtool_depend> in the package.xml file.
Build the custom message package
You can build the package in the root of your workspace, for example by running the following command:
Now the GnssInsOrientationStamped message will be discoverable by other packages in Autoware.
"},{"location":"how-to-guides/others/add-a-custom-ros-message/#how-to-use-custom-messages-in-autoware","title":"How to use custom messages in Autoware","text":"
You can use the custom messages in Autoware by following these steps:
Add dependency in package.xml.
For example, <depend>autoware_sensing_msgs</depend>.
Include the .hpp file of the relevant message in the code.
For example, #include <autoware_sensing_msgs/msg/gnss_ins_orientation_stamped.hpp>.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/","title":"Advanced usage of colcon","text":""},{"location":"how-to-guides/others/advanced-usage-of-colcon/#advanced-usage-of-colcon","title":"Advanced usage of colcon","text":"
This page shows some advanced and useful usage of colcon. If you need more detailed information, refer to the colcon documentation.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#common-mistakes","title":"Common mistakes","text":""},{"location":"how-to-guides/others/advanced-usage-of-colcon/#do-not-run-from-other-than-the-workspace-root","title":"Do not run from other than the workspace root","text":"
It is important that you always run colcon build from the workspace root because colcon builds only under the current directory. If you have mistakenly built in a wrong directory, run rm -rf build/ install/ log/ to clean the generated files.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#do-not-unnecessarily-overlay-workspaces","title":"Do not unnecessarily overlay workspaces","text":"
colcon overlays workspaces if you have sourced the setup.bash of other workspaces before building a workspace. You should take care of this especially when you have multiple workspaces.
Run echo $COLCON_PREFIX_PATH to check whether workspaces are overlaid. If you find some workspaces are unnecessarily overlaid, remove all built files, restart the terminal to clean environment variables, and re-build the workspace.
For more details about workspace overlaying, refer to the ROS 2 documentation.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#cleaning-up-the-build-artifacts","title":"Cleaning up the build artifacts","text":"
colcon sometimes causes errors of because of the old cache. To remove the cache and rebuild the workspace, run the following command:
rm -rf build/ install/\n
In case you know what packages to remove:
rm -rf {build,install}/{package_a,package_b}\n
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#selecting-packages-to-build","title":"Selecting packages to build","text":"
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#changing-the-default-configuration-of-colcon","title":"Changing the default configuration of colcon","text":"
Create $COLCON_HOME/defaults.yaml to change the default configuration.
mkdir -p ~/.colcon\ncat << EOS > ~/.colcon/defaults.yaml\n{\n\"build\": {\n\"symlink-install\": true\n}\n}\n
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/","title":"An example procedure for adding and evaluating a new node","text":""},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#an-example-procedure-for-adding-and-evaluating-a-new-node","title":"An example procedure for adding and evaluating a new node","text":""},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#overview","title":"Overview","text":"
This page provides a guide for evaluating Autoware when a new node is implemented, especially about developing a novel localization node.
The workflow involves initial testing and rosbag recording using a real vehicle or AWSIM, implementing the new node, subsequent testing using the recorded rosbag, and finally evaluating with a real vehicle or AWSIM.
It is assumed that the method intended for addition has already been verified well with public datasets and so on.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#1-running-autoware-in-its-standard-configuration","title":"1. Running Autoware in its standard configuration","text":"
First of all, it is important to be able to run the standard Autoware to establish a basis for performance and behavior comparison.
Autoware constantly incorporates new features. It is crucial to initially confirm that it operates as expected with the current version, which helps in problem troubleshooting.
In this context, AWSIM is presumed. Therefore, AWSIM simulator can be useful. If you are using actual hardware, please refer to the How-to guides.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#2-recording-a-rosbag-using-autoware","title":"2. Recording a rosbag using Autoware","text":"
Before developing a new node, it is recommended to record a rosbag in order to evaluate. If you need a new sensor, you should add it to your vehicle or AWSIM.
In this case, it is recommended to save all topics regardless of whether they are necessary or not. For example, in Localization, since the initial position estimation service is triggered by the input to rviz and the GNSS topic, the initial position estimation does not start when playing back data unless those topics are saved.
Consider the use of the mcap format if data capacity becomes a concern.
It is worth noting that using ros2 bag record increases computational load and might affect performance. After data recording, verifying the smooth flow of sensor data and unchanged time series is advised. This verification can be accomplished, for example, by inspecting the image data with rqt_image_view during ros2 bag play.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#3-developing-the-new-node","title":"3. Developing the new node","text":"
When developing a new node, it could be beneficial to reference a package that is similar to the one you intend to create.
It is advisable to thoroughly read the Design page, contemplate the addition or replacement of nodes in Autoware, and then implement your solution.
For example, a node doing NDT, a LiDAR-based localization method, is ndt_scan_matcher. If you want to replace this with a different approach, implement a node which produces the same topics and provides the same services.
ndt_scan_matcher is launched as pose_estimator, so it is necessary to replace the launch file as well.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#4-evaluating-by-a-rosbag-based-simulator","title":"4. Evaluating by a rosbag-based simulator","text":"
Once the new node is implemented, it is time to evaluate it. logging_simulator is a tool of how to evaluate the new node using the rosbag captured in step 2.
When you run the logging_simulator, you can set planning:=false or control:=false to disable the launch of specific component nodes.
After launching logging_simulator, the rosbag file obtained in step 2 should be replayed using ros2 bag play <rosbag_file>.
If you remap the topics related to the localization that you want to verify this time, Autoware will use the data it is calculating this time instead of the data it recorded. Also, using the --topics option of ros2 bag play, you can publish only specific topics in rosbag.
There is ros2bag_extensions available to filter the rosbag file and create a new rosbag file that contains only the topics you need.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#5-evaluating-in-a-realtime-environment","title":"5. Evaluating in a realtime environment","text":"
Once you have sufficiently verified the behavior in the logging_simulator, let's run it as Autoware with new nodes added in the realtime environment.
To debug Autoware, the method described at debug-autoware is useful.
For reproducibility, you may want to fix the GoalPose. In such cases, consider using the tier4_automatic_goal_rviz_plugin.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#6-sharing-the-results","title":"6. Sharing the results","text":"
If your implementation works successfully, please consider a pull request to Autoware.
It is also a good idea to start by presenting your ideas in Discussion at Show and tell.
For localization, YabLoc's Proposal may provide valuable insights.
"},{"location":"how-to-guides/others/applying-clang-tidy-to-ros-packages/","title":"Applying Clang-Tidy to ROS packages","text":""},{"location":"how-to-guides/others/applying-clang-tidy-to-ros-packages/#applying-clang-tidy-to-ros-packages","title":"Applying Clang-Tidy to ROS packages","text":"
The essential thing for debug is to print the program information clearly, which can quickly judge the program operation and locate the problem. Autoware uses ROS 2 logging tool to print debug messages, how to design console logging refer to tutorial Console logging.
"},{"location":"how-to-guides/others/debug-autoware/#using-ros-tools-debug-autoware","title":"Using ROS tools debug Autoware","text":""},{"location":"how-to-guides/others/debug-autoware/#using-command-line-tools","title":"Using command line tools","text":"
ROS 2 includes a suite of command-line tools for introspecting a ROS 2 system. The main entry point for the tools is the command ros2, which itself has various sub-commands for introspecting and working with nodes, topics, services, and more. How to use the ROS 2 command line tool refer to tutorial CLI tools.
Rviz2 is a port of Rviz to ROS 2. It provides a graphical interface for users to view their robot, sensor data, maps, and more. You can run Rviz2 tool easily by:
rviz2\n
When Autoware launch the simulators, the Rviz2 tool is opened by default to visualize the autopilot graphic information.
RQt is a graphical user interface framework that implements various tools and interfaces in the form of plugins. You can run any RQt tools/plugins easily by:
rqt\n
This GUI allows you to choose any available plugins on your system. You can also run plugins in standalone windows. For example, RQt Console:
ros2_graph can be used to generate mermaid description of ROS 2 graphs to add on your markdown files.
It can also be used as a colorful alternative to rqt_graph even though it would require some tool to render the generated mermaid diagram.
It can be installed with:
pip install ros2-graph\n
Then you can generate a mermaid description of the graph with:
ros2_graph your_node\n\n# or like with an output file\nros2_graph /turtlesim -o turtle_diagram.md\n\n# or multiple nodes\nros2_graph /turtlesim /teleop_turtle\n
When your ROS 2 setup is not running as expected, you can check its settings with the ros2doctor tool.
ros2doctor checks all aspects of ROS 2, including platform, version, network, environment, running systems and more, and warns you about possible errors and reasons for issues.
It's as simple as just running ros2 doctor in your terminal.
It has the ability to list \"Subscribers without publishers\" for all topics in the system.
And this information can help you find if a necessary node isn't running.
For more details, see the following official documentation for Using ros2doctor to identify issues.
"},{"location":"how-to-guides/others/debug-autoware/#using-a-debugger-with-breakpoints","title":"Using a debugger with breakpoints","text":"
Many IDE(e.g. Visual Studio Code, CLion) supports debugging C/C++ executable with GBD on linux platform. The following lists some references for using the debugger:
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/","title":"Defining temporal performance metrics on components","text":""},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#defining-temporal-performance-metrics-on-components","title":"Defining temporal performance metrics on components","text":""},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#motivation-to-defining-temporal-performance-metrics","title":"Motivation to defining temporal performance metrics","text":""},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#objective-of-the-page","title":"Objective of the page","text":"
This page introduces policies to define metrics to evaluate temporal performance on components of Autoware. The term \"temporal performance\" is often used throughout the page in order to distinguish between functional performance, which referred to as accuracy as well, and time-related performance.
It is expected that most algorithms employed for Autoware are executed with as high frequency and short response time as possible. In order to achieve safe autonomous driving, one of the desired outcomes is no time gap between perceived and actual situation. The time gap is commonly referred to as delay. If the delay is significant, the system may determine trajectory and maneuver based on outdated situation. Consequently, if the actual situation differs from the perceived one due to the delay, the system may make unexpected decisions.
As mentioned above, this page presents the policies to define metrics. Besides, the page contains lists of sample metrics that are crucial for the main functionalities of Autoware: Localization, Perception, Planning, and Control.
Note
Other functionalities, such as system components for diagnosis, are excluded currently. However they will be taken into account in the near future.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#contribution-of-the-temporal-performance-metrics","title":"Contribution of the temporal performance metrics","text":"
Temporal performance metrics are important for evaluating Autoware. These metrics are particularly useful for assessing delays caused by new algorithms and logic. They can be employed when comparing the temporal performance of software on a desktop computer with that on a vehicle during the vehicle integration phase.
In addition, these metrics are useful for designers and evaluators of middleware, operating systems, and computers. They are selected based on user and product requirements. One of these requirements is to provide sufficient temporal performance for executing Autoware. \"Sufficient temporal performance\" is defined as a temporal performance requirement, but it can be challenging to define the requirement because it varies depending on the product type, Operational Design Domain (ODD), and other factors. Then, this page specifically focuses on temporal performance metrics rather than requirements.
Temporal performance metrics are important for evaluating the reliability of Autoware. However, ensuring the reliability of Autoware requires consideration of not only temporal performance metrics but also other metrics.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#tools-for-evaluating-the-metrics","title":"Tools for evaluating the metrics","text":"
There are several tools available for evaluating Autoware according to the metrics listed in the page. For example, both CARET and ros2_tracing are recommended options when evaluating Autoware on Linux and ROS 2. If you want to measure the metrics with either of these tools, refer to the corresponding user guide for instructions. It's important to note that if you import Autoware to a platform other than Linux and ROS 2, you will need to choose a supported tool for evaluation.
Note
TIER IV plans to measure Autoware, which is running according to the tutorial, and provide a performance evaluation report periodically. An example of such a report can be found here, although it may not include all of the metrics listed.
The page does not aim to provide instructions on how to use these tools or measure the metrics. Its primary focus is on the metrics themselves, as they are more important than the specific tools used. These metrics retain their relevance regardless of the employed platform.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#policies-to-define-temporal-performance-metrics","title":"Policies to define temporal performance metrics","text":"
As mentioned above, the configuration of Autoware varies by the product type, ODD, and other factors. The variety of configurations makes it difficult to define the uniform metrics for evaluating Autoware. However, the policies used to define them are basically reused even when the configuration changes. Each of temporal performance metrics is categorized into two types: execution frequency and response time. Although there are many types of metrics, such as communication latency, the only two types are considered for simplicity. Execution frequency is observed using rate of Inter-Process Communication (IPC) messages. You will find an enormous number of messages in Autoware, but you don't have to take care of all. Some messages might be critical to functionality and they should be chosen for evaluation. Response time is duration elapsed through a series of processing. A series of processing is referred to as a path. Response time is calculated from timestamps of start and end of a path. Although many paths can be defined in Autoware, you have to choose significant paths.
As a hint, here are some characteristics of message and path in order to choose metrics.
Messages and paths on boundaries where observed values from sensors are consumed
Messages and paths on boundaries of functions, e.g., a boundary of perception and planning
Messages and paths on boundaries where timer-based frequency is switched
Messages and paths on boundaries where two different messages are synchronized and merged
Messages that must be transmitted at expected frequency, e.g., vehicle command messages
Those hints would be helpful for most configurations but there may be exclusions. Defining metrics precisely requires an understanding of configuration.
In addition, it is recommended that metrics be determined incrementally from the architectural level to the detailed design and implementation level. Mixing metrics at different levels of granularity can be confusing.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#list-of-sample-metrics","title":"List of sample metrics","text":"
This section demonstrates how to define metrics according to the policies explained and has lists of the metrics for Autoware launched according to the tutorial. The section is divided into multiple subsections, each containing a model diagram and an accompanying list that explains the important temporal performance metrics. Each model is equipped with checkpoints that serve as indicators for these metrics.
The first subsection presents the top-level temporal performance metrics, which are depicted in the abstract structure of Autoware as a whole. The detailed metrics are not included in the model as they would add complexity to it. Instead, the subsequent section introduces the detailed metrics. The detailed metrics are subject to more frequent updates compared to the top-level ones, which is another reason for categorizing them separately.
Each list includes a column for the reference value. The reference value represents the observed value of each metric when Autoware is running according to the tutorial. It is important to note that the reference value is not a required value, meaning that Autoware does not necessarily fail in the tutorial execution if certain metrics do not fulfill the reference value.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#top-level-temporal-performance-metrics-for-autoware","title":"Top-level temporal performance metrics for Autoware","text":"
The diagram below introduces the model for top-level temporal performance metrics.
The following three policies assist in selecting the top-level performance metrics:
Splitting Autoware based on components that consume observed values, such as sensor data, and considering the processing frequency and response time around these components
Dividing Autoware based on the entry point of Planning and Control and considering the processing frequency and response time around these components
Showing the minimum metrics for the Vehicle Interface, as they may vary depending on the target vehicle
Additionally, it is assumed that algorithms are implemented as multiple nodes and function as a pipeline processing system.
ID Representation in the model Metric meaning Related functionality Reference value Reason to choose it as a metric Note AWOV-001 Message rate from CPA #9 to CPA #18 Update rate of result from Prediction to Planning. Perception 10 Hz Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. AWOV-002 Response time from CPA #0 to CPA #20 via CPA #18 Response time in main body of Perception. Perception N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is used if delay compensation is disabled in Tracking. AWOV-003 Response time from CPA #7 to CPA #20 Response time from Tracking output of Tracking to its data consumption in Planning. Perception N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is used if delay compensation is enabled in Tracking. AWOV-004 Response time from CPA #0 to CPA #6 Duration to process pointcloud data in Sensing and Detection. Perception N/A Tracking relies on detection to provide real-time and up-to-date sensed data for accurate tracking. The metric is used if delay compensation is enabled in Tracking. AWOV-005 Message rate from CPA #4 to CPA #5 Update rate of Detection result received by Tracking. Perception 10 Hz Tracking relies on detection to provide real-time and up-to-date sensed data for accurate tracking. AWOV-006 Response time from CPA #0 to CPA #14 Response time from output of observed data from LiDARs to its consumption in EKF Localizer via NDT Scan Matcher. Localization N/A EKF Localizer relies on fresh and up-to-date observed data from sensors for accurate estimation of self pose. AWOV-007 Message rate from CPA #11 to CPA #13 Update rate of pose estimated by NDT Scan Matcher. Localization 10 Hz EKF Localizer relies on fresh and up-to-date observed data from sensors for accurate estimation of self pose. AWOV-008 Message rate from CPA #15 to CPA #12 Update rate of feed backed pose estimated by EKF Localizer. Localization 50 Hz NDT Scan Matcher relies on receiving estimated pose from EKF Localizer smoothly for linear interpolation. AWOV-009 Message rate from CPA #17 to CPA #19 Update rate of Localization result received by Planning. Localization 50 Hz Planning relies on Localization to update the estimated pose frequently. AWOV-010 Response time from CPA #20 to CPA #23 Processing time from beginning of Planning to consumption of Trajectory message in Control. Planning N/A A vehicle relies on Planning to update trajectory within a short time frame to achieve safe driving behavior. AWOV-011 Message rate from CPA #21 to CPA #22 Update rate of Trajectory message from Planning. Planning 10 Hz A vehicle relies on Planning to update trajectory frequently to achieve safe driving behavior. AWOV-012 Message rate from CPA #24 to CPA #25 Update rate of Control command. Control 33 Hz Control stability and comfort relies on sampling frequency of Control. AWOV-013 Message rate between CPA #26 and Vehicle Communication rate between Autoware and Vehicle. Vehicle Interface N/A A vehicle requires Autoware to communicate with each other at predetermined frequency. Temporal performance requirement varies depending on vehicle type.
Note
There is an assumption that each of sensors, such as LiDARs and cameras, outputs a set of pointcloud with a timestamp. CPA #0 is observed with the timestamp. If the sensors are not configured to output the timestamp, the time when Autoware receives the pointcloud is used instead. That is represented by CPA #1 in the model. The detailed metrics employs the idea as well.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#detailed-temporal-performance-metrics-for-perception","title":"Detailed temporal performance metrics for Perception","text":"
The diagram below introduces the model for temporal performance metrics for Perception.
The following two policies assist in selecting the performance metrics:
Regarding the frequency and response time at which Recognition results from Object Recognition and Traffic Light Recognition are consumed in Planning
Splitting Perception component on merging points of data from multiple processing paths and considering the frequency and response time around that point
The following list shows the temporal performance metrics for Perception.
ID Representation in the model Metric meaning Related functionality Reference value Reason to choose it as a metric Note APER-001 Message rate from CPP #2 to CPP #26 Update rate of Traffic Light Recognition. Traffic Light Recognition 10 Hz Planning relies on fresh and up-to-date perceived data from Traffic Light Recognition for making precise decisions. APER-002 Response time from CPP #0 to CPP #30 Response time from camera input to consumption of the result in Planning. Traffic Light Recognition N/A Planning relies on fresh and up-to-date perceived data from Traffic Light Recognition for making precise decisions. APER-003 Message rate from CPP #25 to CPP #28 Update rate of result from Prediction (Object Recognition) to Planning. Object Recognition 10 Hz Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is same as AWOV-001. APER-004 Response time from CPP #6 to CPP #30 Response time from Tracking output of Tracking to its data consumption in Planning. Object Recognition N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is same as AWOV-002 and used if delay compensation is disabled in Tracking. APER-005 Response time from CPP #23 to CPP #30 Response time from Tracking output of Tracking to its data consumption in Planning. Object Recognition N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is same as AWOV-003 and used if delay compensation is enabled in Tracking. APER-006 Response time from CPP #6 to CPP #21 Duration to process pointcloud data in Sensing and Detection. Object Recognition N/A Tracking relies on Detection to provide real-time and up-to-date perceived data. The metrics is same as AWOV-004 and used if delay compensation is enabled in Tracking. APER-007 Message rate from CPP #20 to CPP #21 Update rate of Detection result received by Tracking. Object Recognition 10 Hz Tracking relies on detection to provide real-time and up-to-date sensed data for accurate tracking. The metric is same as AWOV-005 APER-008 Message rate from CPP #14 to CPP #19 Update rate of data sent from Sensor Fusion. Object Recognition 10 Hz Association Merger relies on the data to be updated at expected frequency for data synchronization. APER-009 Message rate from CPP #16 to CPP #19 Update rate of data sent from Detection by Tracker. Object Recognition 10 Hz Association Merger relies on the data to be updated at expected frequency for data synchronization. APER-010 Message rate from CPP #18 to CPP #19 Update rate of data sent from Validation Object Recognition. 10 Hz Association Merger relies on the data to be updated at expected frequency for data synchronization. APER-011 Response time from CPP #6 to CPP #19 via CPP #14 Response time to consume data sent from Sensor Fusion after LiDARs output pointcloud. Object Recognition N/A Association Merger relies on fresh and up-to-date data for data synchronization. APER-012 Response time from CPP #6 to CPP #19 via CPP #16 Response time to consume data sent from Detection by Tracker after LiDARs output pointcloud. Object Recognition N/A Association Merger relies on fresh and up-to-date data for data synchronization. APER-013 Response time from CPP #6 to CPP #19 via CPP #18 Response time to consume data sent from Validator after LiDARs output pointcloud. Object Recognition N/A Association Merger relies on fresh and up-to-date data for data synchronization. APER-014 Message rate from CPP #10 to CPP #13 Update rate of data sent from Clustering. Object Recognition 10 Hz Sensor Fusion relies on the data to be updated at expected frequency for data synchronization. APER-015 Message rate from CPP #5 to CPP #13 Update rate of data sent from Camera-based Object detection. Object Recognition 10 Hz Sensor Fusion relies on the data to be updated at expected frequency for data synchronization. APER-016 Response time from CPP #6 to CPP #13 Response time to consume data sent from Clustering after LiDARs output pointcloud. Object Recognition N/A Sensor Fusion relies on fresh and up-to-date data for data synchronization. APER-017 Response time from CPP #3 to CPP #13 Response time to consume data sent from Camera-based Object detection after Cameras output images. Object Recognition N/A Sensor Fusion relies on fresh and up-to-date data for data synchronization. APER-018 Message rate from CPP #10 to CPP #17 Update rate of data sent from Clustering. Object Recognition 10 Hz Validator relies on the data to be updated at expected frequency for data synchronization. It seems similar to APER-014, but the topic message is different. APER-019 Message rate from CPP #12 to CPP #17 Update rate of data sent from DNN-based Object Recognition. Object Recognition 10 Hz Validator relies on the data to be updated at expected frequency for data synchronization. APER-020 Response time from CPP #6 to CPP #17 via CPP #10 Response time to consume data sent from Clustering after LiDARs output pointcloud. Object Recognition N/A Validator relies on fresh and update-date data for data synchronization. It seems similar to APER-015, but the topic message is different. APER-021 Response time from CPP #6 to CPP #17 via CPP #12 Response time to consume data sent from DNN-based Object Recognition after LiDARs output pointcloud. Object Recognition N/A Validator relies on fresh and update-date data for data synchronization."},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#detailed-temporal-performance-metrics-for-paths-between-obstacle-segmentation-and-planning","title":"Detailed temporal performance metrics for Paths between Obstacle segmentation and Planning","text":"
Obstacle segmentation, which is a crucial part of Perception, transmits data to Planning. The figure below illustrates the model that takes into account performance metrics related to Obstacle segmentation and Planning.
Note
Both the Obstacle grid map and Obstacle segmentation transmit data to multiple sub-components of Planning. However, not all of these sub-components are described in the model. This is because our primary focus is on the paths from LiDAR to Planning via Obstacle segmentation.
The following list shows the temporal performance metrics around Obstacle segmentation and Planning.
ID Representation in the model Metric meaning Related functionality Reference value Reason to choose it as a metric Note OSEG-001 Message rate from CPS #4 to CPS #7 Update rate of Occupancy grid map received by Planning (behavior_path_planner) Obstacle segmentation 10 Hz Planning relies on Occupancy grid map to be updated frequently and smoothly for creating accurate trajectory. OSEG-002 Response time from CPS #0 to CPS #9 via CPS #7 Response time to consume Occupancy grid map after LiDARs output sensing data. Obstacle segmentation N/A Planning relies on fresh and up-to-date perceived data from Occupancy grid map for creating accurate trajectory.. OSEG-003 Message rate from CPS #6 to CPS #11 Update rate of obstacle segmentation received by Planning (behavior_velocity_planner). Obstacle segmentation 10 Hz Planning relies on Obstacle segmentation to be updated frequently and smoothly for creating accurate trajectory. OSEG-004 Response time from CPS #0 to CPS #13 via CPS #11 Response time to consume Obstacle segmentation after LiDARs output sensing data. Obstacle segmentation N/A Planning relies on fresh and up-to-date perceived data from Obstacle segmentation for creating accurate trajectory.."},{"location":"how-to-guides/others/determining-component-dependencies/","title":"Determining component dependencies","text":""},{"location":"how-to-guides/others/determining-component-dependencies/#determining-component-dependencies","title":"Determining component dependencies","text":"
For any developers who wish to try and deploy Autoware as a microservices architecture, it is necessary to understand the software dependencies, communication, and implemented features of each ROS package/node.
As an example, the commands necessary to determine the dependencies for the Perception component are shown below.
Autoware manages dependent package versions in autoware.repos. For example, let's say you make a branch in autoware.universe and add new features. Suppose you update other dependencies with vcs pull after cutting a branch from autoware.universe. Then the version of autoware.universe you are developing and other dependencies will become inconsistent, and the entire Autoware build will fail. We recommend saving the dependent package versions by executing the following command when starting the development.
vcs export src --exact > my_autoware.repos\n
"},{"location":"how-to-guides/others/reducing-start-delays/","title":"Reducing start delays on real vehicles","text":""},{"location":"how-to-guides/others/reducing-start-delays/#reducing-start-delays-on-real-vehicles","title":"Reducing start delays on real vehicles","text":"
In simulation, the ego vehicle reacts nearly instantly to the control commands generated by Autoware. However, with a real vehicle, some delays occur that may make ego feel less responsive.
This page presents start delays experienced when using Autoware on a real vehicle. We define the start delay as the time between (a) when Autoware decides to make the ego vehicle start and (b) when the vehicle actually starts moving. More precisely:
(a) is the time when the speed or acceleration command output by Autoware switches to a non-zero value.
(b) is the time when the measured velocity of the ego vehicle switches to a positive value.
"},{"location":"how-to-guides/others/reducing-start-delays/#start-delay-with-manual-driving","title":"Start delay with manual driving","text":"
First, let us look at the start delay when a human is driving.
The following figure shows the start delay when a human driver switches the gear from parked to drive and instantly releases the brake to push the throttle pedal and make the velocity of the vehicle increase.
There are multiple things to note from this figure.
Brake (red): despite the driver instantly releasing the brake pedal, we see that the measured brake takes around 150ms to go from 100% to 0%.
Gear (orange): the driver switches gear before releasing the brake pedal, but the gear is measured to switch after the brake is released.
Throttle (green) and velocity (blue): the driver pushes the throttle pedal and the vehicle is measured to start moving around 500ms later.
To guarantee passenger comfort, some Autoware modules implement filters on the jerk of the vehicle, preventing sudden changes in acceleration.
For example, the vehicle_cmd_gate filters the acceleration command generated by the controller and was previously introducing significant delays when transitioning between a stop command where the acceleration is negative, and a move command where the acceleration is positive. Because of the jerk filter, the transition between negative and positive was not instantaneous and would take several hundreds of milliseconds.
In many vehicles, it is necessary to change gear before first starting to move the vehicle. When performed autonomously, this gear change can take some significant time. Moreover, as seen from the data recorded with manual driving, the measured gear value may be delayed.
In Autoware, the controller sends a stopping control command until the gear is changed to the drive state. This means that delays in the gear change and its reported value can greatly impact the start delay. Note that this is only an issue when the vehicle is initially in the parked gear.
The only way to reduce this delay is by tuning the vehicle to increase the gear change speed or to reduce the delay in the gear change report.
In vehicles with a brake pedal, the braking system will often be made of several moving parts which cannot move instantly. Thus, when Autoware sends brake commands to a vehicle, some delays should be expected in the actual brake applied to the wheels.
This lingering brake may prevent or delay the initial motion of the ego vehicle.
For vehicles with throttle control, one of the main cause of start delays is due to the throttle response of the vehicle. When pushing the throttle pedal, the wheels of the vehicle do not instantly start rotating. This is partly due to the inertia of the vehicle, but also to the motor which may take a significant time to start applying some torque to the wheels.
It may be possible to tune some vehicle side parameters to reduce this delay, but it is often done at the cost of reduced energy efficiency.
On the Autoware side, the only way to decrease this delay is to increase the initial throttle but this can cause uncomfortably high initial accelerations.
"},{"location":"how-to-guides/others/reducing-start-delays/#initial-acceleration-and-throttle","title":"Initial acceleration and throttle","text":"
As we just discussed, for vehicles with throttle control, an increased initial throttle value can reduce the start delay.
Since Autoware outputs an acceleration value, the conversion module raw_vehicle_cmd_converter is used to map the acceleration value from Autoware to a throttle value to be sent to the vehicle. Such mapping is usually calibrated automatically using the accel_brake_map_calibrator module, but it may produce a low initial throttle which leads to high start delays.
In order to increase the initial throttle, there are two options: increase the initial acceleration output by Autoware, or modify the acceleration to throttle mapping.
The initial acceleration output by Autoware can be tuned in the motion_velocity_smoother with parameters engage_velocity and engage_acceleration. However, the vehicle_cmd_gate applies a filter on the control command to prevent too sudden changes in jerk and acceleration, limiting the maximum allowed acceleration while the ego vehicle is stopped.
Alternatively, the mapping of acceleration can be tuned to increase the throttle corresponding to the initial acceleration. If we look at an example acceleration map, it does the following conversion: when the ego velocity is 0 (first column), acceleration values between 0.631 (first row) and 0.836 (second row) are converted to a throttle between 0% and 10%. This means that any initial acceleration bellow 0.631m/s\u00b2 will not produce any throttle. Keep in mind that after tuning the acceleration map, it may be necessary to also update the brake map.
Although CUDA installation is recommended to achieve better performance for object detection and traffic light recognition in Autoware Universe, it is possible to run these algorithms without CUDA. The following subsections briefly explain how to run each algorithm in such an environment.
"},{"location":"how-to-guides/others/running-autoware-without-cuda/#running-2d3d-object-detection-without-cuda","title":"Running 2D/3D object detection without CUDA","text":"
Autoware Universe's object detection can be run using one of five possible configurations:
lidar_centerpoint
lidar_apollo_instance_segmentation
lidar-apollo + tensorrt_yolo
lidar-centerpoint + tensorrt_yolo
euclidean_cluster
Of these five configurations, only the last one (euclidean_cluster) can be run without CUDA. For more details, refer to the euclidean_cluster module's README file.
"},{"location":"how-to-guides/others/running-autoware-without-cuda/#running-traffic-light-detection-without-cuda","title":"Running traffic light detection without CUDA","text":"
For traffic light recognition (both detection and classification), there are two modules that require CUDA:
traffic_light_ssd_fine_detector
traffic_light_classifier
To run traffic light detection without CUDA, set enable_fine_detection to false in the traffic light launch file. Doing so disables the traffic_light_ssd_fine_detector such that traffic light detection is handled by the map_based_traffic_light_detector module instead.
To run traffic light classification without CUDA, set use_gpu to false in the traffic light classifier launch file. Doing so will force the traffic_light_classifier to use a different classification algorithm that does not require CUDA or a GPU.
"},{"location":"how-to-guides/training-machine-learning-models/training-models/","title":"Training and Deploying Models","text":""},{"location":"how-to-guides/training-machine-learning-models/training-models/#training-and-deploying-models","title":"Training and Deploying Models","text":""},{"location":"how-to-guides/training-machine-learning-models/training-models/#overview","title":"Overview","text":"
The Autoware offers a comprehensive array of machine learning models, tailored for a wide range of tasks including 2D and 3D object detection, traffic light recognition and more. These models have been meticulously trained utilizing open-mmlab's extensive repositories. By leveraging the provided scripts and following the training steps, you have the capability to train these models using your own dataset, tailoring them to your specific needs.
Furthermore, you will find the essential conversion scripts to deploy your trained models into Autoware using the mmdeploy repository.
The traffic light classifier model within the Autoware has been trained using the mmlab/pretrained repository. The Autoware offers pretrained models based on EfficientNet-b1 and MobileNet-v2 architectures. To fine-tune these models, a total of 83,400 images were employed, comprising 58,600 for training, 14,800 for evaluation, and 10,000 for testing. These images represent Japanese traffic lights and were trained using TIER IV's internal dataset.
Name Input Size Test Accuracy EfficientNet-b1 128 x 128 99.76% MobileNet-v2 224 x 224 99.81%
Comprehensive training instructions for the traffic light classifier model are detailed within the readme file accompanying \"traffic_light_classifier\" package. These instructions will guide you through the process of training the model using your own dataset. To facilitate your training, we have also provided an example dataset containing three distinct classes (green, yellow, red), which you can leverage during the training process.
Detailed instructions for training the traffic light classifier model can be found here.
Autoware is scalable and can be customized to work with distributed or less powerful hardware. The minimum hardware requirements given below are just a general recommendation. However, performance will be improved with more cores, RAM and a higher-spec graphics card or GPU core.
CPU with 8 cores
16GB RAM
[Optional] NVIDIA GPU (4GB RAM)
Although GPU is not required to run basic functionality, it is mandatory to enable the following neural network related functions:
LiDAR based object detection
Camera based object detection
Traffic light detection and classification
For details of how to enable object detection and traffic light detection/classification without a GPU, refer to the Running Autoware without CUDA.
Docker can ensure that all developers in a project have a common, consistent development environment. It is recommended for beginners, casual users, people who are unfamiliar with Ubuntu.
For more information, refer to the Docker installation guide.
Source installation is for the cases where more granular control of the installation environment is needed. It is recommended for experienced users or people who want to customize their environment. Note that some problems may occur depending on your local environment.
For more information, refer to the source installation guide.
"},{"location":"installation/#installing-related-tools","title":"Installing related tools","text":"
Some other tools are required depending on the evaluation you want to do. For example, to run an end-to-end simulation you need to install an appropriate simulator.
For more information, see here.
"},{"location":"installation/#additional-settings-for-developers","title":"Additional settings for developers","text":"
There are also tools and settings for developers, such as Shells or IDEs.
For more information, see here.
"},{"location":"installation/additional-settings-for-developers/","title":"Additional settings for developers","text":""},{"location":"installation/additional-settings-for-developers/#additional-settings-for-developers","title":"Additional settings for developers","text":""},{"location":"installation/additional-settings-for-developers/#console-settings-for-ros-2","title":"Console settings for ROS 2","text":""},{"location":"installation/additional-settings-for-developers/#colorizing-logger-output","title":"Colorizing logger output","text":"
By default, ROS 2 logger doesn't colorize the output. To colorize it, write the following in your .bashrc:
export RCUTILS_COLORIZED_OUTPUT=1\n
"},{"location":"installation/additional-settings-for-developers/#customizing-the-format-of-logger-output","title":"Customizing the format of logger output","text":"
By default, ROS 2 logger doesn't output detailed information such as file name, function name, or line number. To customize it, write the following in your .bashrc:
export RCUTILS_CONSOLE_OUTPUT_FORMAT=\"[{severity} {time}] [{name}]: {message} ({function_name}() at {file_name}:{line_number})\"\n
For more options, see here.
"},{"location":"installation/additional-settings-for-developers/#network-settings-for-ros-2","title":"Network settings for ROS 2","text":"
ROS 2 employs DDS, and the configuration of ROS 2 and DDS is described separately. For ROS 2 networking concepts, refer to the official documentation.
ROS 2 multicasts data on the local network by default. Therefore, when you develop in an office, the data flows over the local network of your office. It may cause collisions of packets or increases in network traffic.
To avoid these, there are two options.
Localhost-only communication
Same domain only communication on the local network
Unless you plan to use multiple host computers on the local network, localhost-only communication is recommended. For details, refer to the sections below.
Write the following in your .bashrc: For more information, see the ROS 2 documentation.
export ROS_LOCALHOST_ONLY=1\n
If you export ROS_LOCALHOST_ONLY=1, MULTICAST must be enabled at the loopback address. To verify that MULTICAST is enabled, use the following command.
$ ip link show lo\n1: lo: <LOOPBACK,MULTICAST,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000\n
If the word MULTICAST is not printed, use the following command to enable it.
sudo ip link set lo multicast on\n
"},{"location":"installation/additional-settings-for-developers/#same-domain-only-communication-on-the-local-network","title":"Same domain only communication on the local network","text":"
ROS 2 uses ROS_DOMAIN_ID to create groups and communicate between machines in the groups. Since all ROS 2 nodes use domain ID 0 by default, it may cause unintended interference.
To avoid it, set a different domain ID for each group in your .bashrc:
# Replace X with the Domain ID you want to use\n# Domain ID should be a number in range [0, 101] (inclusive)\nexport ROS_DOMAIN_ID=X\n
Also confirm that ROS_LOCALHOST_ONLY is 0 by using the following command.
echo $ROS_LOCALHOST_ONLY # If the output is 1, localhost has priority.\n
For more information, see the ROS 2 Documentation.
Autoware uses DDS for inter-node communication. ROS 2 documentation recommends users to tune DDS to utilize its capability. Especially, receive buffer size is the critical parameter for Autoware. If the parameter is not large enough, Autoware will failed in receiving large data like point cloud or image.
Unless customized, CycloneDDS is adopted by default. For example, to execute Autoware with CycloneDDS, prepare a config file. A sample config file is given below. Save it as cyclonedds_config.xml.
This configuration is mostly taken from Eclipse Cyclone DDS:Run-time configuration documentation. You can see why each value is set as such under the documentation link.
Set the config file path and enlarge the Linux kernel maximum buffer size before launching Autoware.
For more information, Refer to ROS 2 documentation. Reading user guide for chosen DDS is helpful for more understanding.
"},{"location":"installation/additional-settings-for-developers/#tuning-dds-for-multiple-host-computers-for-advanced-users","title":"Tuning DDS for multiple host computers (for advanced users)","text":"
When Autoware runs on multiple host computers, IP Fragmentation should be taken into account. As ROS 2 documentation recommends, parameters for IP Fragmentation should be set as shown in the following example.
"},{"location":"installation/autoware/docker-installation-devel/","title":"Docker installation for development","text":""},{"location":"installation/autoware/docker-installation-devel/#docker-installation-for-development","title":"Docker installation for development","text":""},{"location":"installation/autoware/docker-installation-devel/#prerequisites","title":"Prerequisites","text":"
Git
For NVIDIA Jetson devices, install JetPack >= 5.0
"},{"location":"installation/autoware/docker-installation-devel/#how-to-set-up-a-development-environment","title":"How to set up a development environment","text":"
Clone autowarefoundation/autoware and move to the directory.
"},{"location":"installation/autoware/docker-installation-devel/#installing-dependencies-using-ansible","title":"Installing dependencies using Ansible","text":"
Be very careful with this method. Make sure you read and confirmed all the steps in the Ansible configuration before using it.
If you've manually installed the dependencies, you can skip this section.
./setup-dev-env.sh docker\n
You might need to log out and log back to make the current user able to use docker.
"},{"location":"installation/autoware/docker-installation-devel/#how-to-set-up-a-workspace","title":"How to set up a workspace","text":"
Warning
Before proceeding, confirm and agree with the NVIDIA Deep Learning Container license. By pulling and using the Autoware Universe images, you accept the terms and conditions of the license.
Create the autoware_map directory for map data later.
After that, move to the workspace in the container:
cd autoware\n
Create the src directory and clone repositories into it.
mkdir src\nvcs import src < autoware.repos\n
Update dependent ROS packages.
The dependency of Autoware may change after the Docker image was created. In that case, you need to run the following commands to update the dependency.
"},{"location":"installation/autoware/docker-installation-prebuilt/","title":"Docker installation for quick start","text":""},{"location":"installation/autoware/docker-installation-prebuilt/#docker-installation-for-quick-start","title":"Docker installation for quick start","text":""},{"location":"installation/autoware/docker-installation-prebuilt/#how-to-set-up-a-development-environment","title":"How to set up a development environment","text":"
Installing dependencies manually
Install Docker Engine
Install NVIDIA Container Toolkit
Install rocker
"},{"location":"installation/autoware/docker-installation-prebuilt/#how-to-set-up-a-workspace","title":"How to set up a workspace","text":"
Create the autoware_map directory for map data later.
Since this page explains Docker-specific information, it is recommended to see Source installation as well if you need detailed information.
Here are two ways to install Autoware by docker:
The first way is to start Autoware with prebuilt image, this is a quick start, this way you can only run Autoware simulator and not develop Autoware, it is only suitable for beginners
The second way is to start Autoware with devel image, which supports developing and running Autoware using docker
"},{"location":"installation/autoware/docker-installation/#docker-installation-for-quick-start","title":"Docker installation for quick start","text":"
docker installation for quick start
"},{"location":"installation/autoware/docker-installation/#docker-installation-for-development","title":"Docker installation for development","text":"
"},{"location":"installation/autoware/docker-installation/#cuda-error-forward-compatibility-was-attempted-on-non-supported-hw","title":"cuda error: forward compatibility was attempted on non supported hw","text":"
When starting Docker with GPU support enabled for NVIDIA graphics, you may sometimes receive the following error:
docker: Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused \"process_linux.go:449: container init caused \\\"process_linux.go:432: running prestart hook 0 caused \\\\\\\"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: cuda error: forward compatibility was attempted on non supported hw\\\\\\\\n\\\\\\\"\\\"\": unknown.\nERROR: Command return non-zero exit code (see above): 125\n
This usually indicates that a new NVIDIA graphics driver has been installed (usually via apt) but the system has not yet been restarted. A similar message may appear if the graphics driver is not available, for example because of resuming after suspend.
To fix this, restart your system after installing the new NVIDIA driver.
"},{"location":"installation/autoware/docker-installation/#docker-with-nvidia-gpu-fails-to-start-autoware-on-arm64-devices","title":"Docker with NVIDIA gpu fails to start Autoware on arm64 devices","text":"
When starting Docker with GPU support enabled for NVIDIA graphics on arm64 devices, e.g. NVIDIA jetson AGX xavier, you may receive the following error:
nvidia@xavier:~$ rocker --nvidia --x11 --user --volume $HOME/autoware -- ghcr.io/autowarefoundation/autoware-universe:humble-latest-cuda-arm64\n...\n\nCollecting staticx==0.12.3\nDownloading https://files.pythonhosted.org/packages/92/ff/d9960ea1f9db48d6044a24ee0f3d78d07bcaddf96eb0c0e8806f941fb7d3/staticx-0.12.3.tar.gz (68kB)\nComplete output from command python setup.py egg_info:\nTraceback (most recent call last):\nFile \"\", line 1, in\nFile \"/tmp/pip-install-m_nm8mya/staticx/setup.py\", line 4, in\nfrom wheel.bdist_wheel import bdist_wheel\nModuleNotFoundError: No module named 'wheel'\n\nCommand \"python setup.py egg_info\" failed with error code 1 in /tmp/pip-install-m_nm8mya/staticx/\n...\n
This error exists in current version of rocker tool, which relates to the os_detection function of rocker.
To fix this error, temporary modification of rocker source code is required, which is not recommended.
At current stage, it is recommended to run docker without NVIDIA gpu enabled for arm64 devices:
This section describes a process to run arm64 systems on amd64 systems using qemu-user-static.
Initially, your system is usually incompatible with arm64 systems. To check that:
$ docker run --rm -t arm64v8/ubuntu uname -m\nWARNING: The requested image's platform (linux/arm64/v8) does not match the detected host platform (linux/amd64) and no specific platform was requested\nstandard_init_linux.go:228: exec user process caused: exec format error\n
Installing qemu-user-static enables us to run arm64 images on amd64 systems.
$ sudo apt-get install qemu-user-static\n$ docker run --rm --privileged multiarch/qemu-user-static --reset -p yes\n$ docker run --rm -t arm64v8/ubuntu uname -m\nWARNING: The requested image's platform (linux/arm64/v8) does not match the detected host platform (linux/amd64) and no specific platform was requested\naarch64\n
To run Autoware's Docker images of arm64 architecture, add the suffix -arm64.
$ docker run --rm -it ghcr.io/autowarefoundation/autoware-universe:humble-latest-cuda-arm64\nWARNING: The requested image's platform (linux/arm64) does not match the detected host platform (linux/amd64) and no specific platform was requested\nroot@5b71391ad50f:/autoware#\n
Note: If you wish to use ROS 2 Galactic on Ubuntu 20.04, refer to installation instruction from galactic branch, but be aware that Galactic version of Autoware might not have latest features.
"},{"location":"installation/autoware/source-installation/#how-to-set-up-a-development-environment","title":"How to set up a development environment","text":"
Clone autowarefoundation/autoware and move to the directory.
If you are installing Autoware for the first time, you can automatically install the dependencies by using the provided Ansible script.
./setup-dev-env.sh\n
If you encounter any build issues, please consult the Troubleshooting section for assistance.
Info
Before installing NVIDIA libraries, please ensure that you have reviewed and agreed to the licenses.
CUDA
cuDNN
TensorRT
Note
The following items will be automatically installed. If the ansible script doesn't work or if you already have different versions of dependent libraries installed, please install the following items manually.
Install ROS 2
Install ROS 2 Dev Tools
Install the RMW Implementation
Install pacmod
Install Autoware Core dependencies
Install Autoware Universe dependencies
Install pre-commit dependencies
Install Nvidia CUDA
Install Nvidia cuDNN and TensorRT
If you didn't use ansible script you will need to download some package artifacts as explained in Manual loading of artifacts. Otherwise some packages (mostly from perception) will not be able to run as they need these artifacts for the inference.
"},{"location":"installation/autoware/source-installation/#how-to-set-up-a-workspace","title":"How to set up a workspace","text":"
Create the src directory and clone repositories into it.
Autoware uses vcstool to construct workspaces.
cd autoware\nmkdir src\nvcs import src < autoware.repos\n
Install dependent ROS packages.
Autoware requires some ROS 2 packages in addition to the core components. The tool rosdep allows an automatic search and installation of such dependencies. You might need to run rosdep update before rosdep install.
"},{"location":"installation/related-tools/","title":"Installation of related tools","text":""},{"location":"installation/related-tools/#installation-of-related-tools","title":"Installation of related tools","text":"
The Autoware perception stack uses models for inference. These models are automatically downloaded if using ansible, but they can also be downloaded manually.
"},{"location":"models/#onnx-model-files","title":"ONNX model files","text":""},{"location":"models/#download-instructions","title":"Download instructions","text":"
The ONNX model files are stored in a common location, hosted by Web.Auto
Any tool that can download files from the web (e.g. wget or curl) is the only requirement for downloading these files:
This document is created to describe and give additional information of the sensors and systems supported by Autoware.Auto software.
All equipment listed in this document has available ROS 2 drivers and has been tested by one or more of the community members on field in autonomous vehicle and robotics applications.
The listed sensors and systems are not sold, developed or given direct technical support by the Autoware community. Having said that any ROS 2 and Autoware related issue regarding the hardware usage could be asked using the community guidelines which found here.
The documents consists of the sections listed below:
Crystal Rugged solutions which is used for autonomous driving and tested by one or more community members are listed below:
Supported Products List CPU GPU RAM, Interfaces Environmental Autoware Tested (Y/N) AVC 0161-AC Intel\u00ae Xeon\u00ae Scalable Dual GPU RTX Series 2TB RAM,CAN, USB, Ethernet, Serial, Hot-Swap SSD 10-32 VoltVibration:2 G RMS 10-1000 Hz, 3 axes -
Link to company website is here.
"},{"location":"reference-hw/cameras/","title":"CAMERAs","text":""},{"location":"reference-hw/cameras/#cameras","title":"CAMERAs","text":""},{"location":"reference-hw/cameras/#tier-iv-automotive-hdr-cameras","title":"TIER IV Automotive HDR Cameras","text":"
TIER IV's Automotive HDR cameras which have ROS 2 driver and tested by TIER IV are listed below:
Supported Products List MP FPS Interface HDR LFM Trigger /Synchronization Ingress Protection ROS 2 Driver Autoware Tested (Y/N) C1 2.5 30 GMSL2 / USB3 Y (120dB) Y Y IP69K Y Y C2 5.4 30 GMSL2 / USB3 Y (120dB) Y Y IP69K Y Y C3 (to be released in 2024) 8.3 30 GMSL2 / TBD Y (120dB) Y Y IP69K Y Y
Link to ROS 2 driver: https://github.com/tier4/ros2_v4l2_camera
Link to product support site: TIER IV Edge.Auto documentation
Link to product web site: TIER IV Automotive Camera Solution
Lucid Vision cameras which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List MP FPS Interface HDR LFM Trigger /Synchronization Ingress Protection ROS 2 Driver Autoware Tested (Y/N) TRITON 054S 5.4 22 GigE Y Y Y up to IP67 Y Y TRITON 032S 3.2 35.4 GigE N/A N/A Y up to IP67 Y Y
Link to ROS 2 driver: https://gitlab.com/leo-drive/Drivers/arena_camera Link to company website: https://thinklucid.com/triton-gige-machine-vision/
The list of all drivers listed above for easy access as a table with additional information:
Type Maker Driver links License Maintainer Lidar VelodyneHesai Link Apache 2 david.wong@tier4.jpabraham.monrroy@map4.jp Lidar Velodyne Link BSD jwhitley@autonomoustuff.com Lidar Robosense Link BSD zdxiao@robosense.cn Lidar Hesai Link Apache 2 wuxiaozhou@hesaitech.com Lidar Leishen Link - - Lidar Livox Link MIT dev@livoxtech.com Lidar Ouster Link Apache 2 stevenmacenski@gmail.comtom@boxrobotics.ai Radar smartmicro Link Apache 2 opensource@smartmicro.de Camera Flir Link Apache 2 bernd.pfrommer@gmail.com Camera Lucid Vision Link - kcolak@leodrive.ai Camera Allied Vision Link Apache 2 at@email.com GNSS NovAtel Link BSD preed@swri.org GNSS SBG Systems Link MIT support@sbg-systems.com GNSS PolyExplore Link - support@polyexplore.com"},{"location":"reference-hw/imu_ahrs_gnss_ins/","title":"IMU, AHRS & GNSS/INS","text":""},{"location":"reference-hw/imu_ahrs_gnss_ins/#imu-ahrs-gnssins","title":"IMU, AHRS & GNSS/INS","text":""},{"location":"reference-hw/imu_ahrs_gnss_ins/#novatel-gnssins-sensors","title":"NovAtel GNSS/INS Sensors","text":"
NovAtel GNSS/INS sensors which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List INS Rate Roll, Pitch, Yaw Acc. GNSS ROS 2 Driver\u00a0 Autoware Tested (Y/N) PwrPak7D-E2 200 Hz R (0.013\u00b0)P (0.013\u00b0)Y (0.070\u00b0) 20 HzL1 / L2 / L5 555 Channels Y - Span CPT7 200 Hz R (0.01\u00b0)\u00a0P (0.01\u00b0)\u00a0Y (0.03\u00b0) 20 Hz L1 / L2 / L5 555 Channels Y -
Link to ROS 2 driver: https://github.com/swri-robotics/novatel_gps_driver/tree/dashing-devel
Link to company website: https://hexagonpositioning.com/
PolyExplore GNSS/INS sensors which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List INS/IMU Rate Roll, Pitch, Yaw Acc. GNSS ROS 2 Driver\u00a0 Autoware Tested (Y/N) POLYNAV 2000P 100 Hz R (0.01\u00b0)P (0.01\u00b0)Y (0.1\u00b0) L1 / L2240 Channels Y - POLYNAV 2000S 100 Hz R (0.015\u00b0)P (0.015\u00b0)Y (0.08\u00b0) L1 / L240 Channels Y -
Link to ROS 2 driver: https://github.com/polyexplore/ROS2_Driver
Link to company website: https://www.polyexplore.com/
"},{"location":"reference-hw/lidars/","title":"LIDARs","text":""},{"location":"reference-hw/lidars/#lidars","title":"LIDARs","text":""},{"location":"reference-hw/lidars/#velodyne-3d-lidar-sensors","title":"Velodyne 3D LIDAR Sensors","text":"
Velodyne Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) Alpha Prime 245m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Ultra Puck 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Puck 100m (+15\u00b0)/(-15\u00b0), (360\u00b0) Y Y Puck Hi-res 100m (+10\u00b0)/(-10\u00b0), (360\u00b0) Y Y
Link to ROS 2 drivers: https://github.com/tier4/nebula https://github.com/ros-drivers/velodyne/tree/ros2/velodyne_pointcloud https://gitlab.com/autowarefoundation/autoware.auto/AutowareAuto/-/tree/master/src/drivers/velodyne_nodes https://github.com/autowarefoundation/awf_velodyne/tree/tier4/universe
Link to company website: https://velodynelidar.com/
"},{"location":"reference-hw/lidars/#robosense-3d-lidar-sensors","title":"RoboSense 3D LIDAR Sensors","text":"
RoboSense Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) RS-Ruby 250m (+15\u00b0)/(-25\u00b0), (360) Y - RS-Ruby-Lite 230m (+15\u00b0)/(-25\u00b0), (360) Y - RS-LiDAR-32 200m (+15\u00b0)/(-25\u00b0), (360) Y - RS-LiDAR-16 150m (+15\u00b0)/(-15), (360) Y -
Link to ROS 2 driver: https://github.com/RoboSense-LiDAR/rslidar_sdk
Link to company website: https://www.robosense.ai/
"},{"location":"reference-hw/lidars/#hesai-3d-lidar-sensors","title":"HESAI 3D LIDAR Sensors","text":"
Hesai Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) Pandar 128 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y - Pandar 64 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Pandar 40P 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Pandar XT 120m (+15\u00b0)/(-16\u00b0), (360\u00b0) Y Y Pandar QT 20m (-52.1\u00b0/+52.1\u00b0)/(360\u00b0) Y Y
Link to ROS 2 drivers: https://github.com/tier4/nebula https://github.com/HesaiTechnology/HesaiLidar_General_ROS
Link to company website: https://www.hesaitech.com/en/
"},{"location":"reference-hw/lidars/#leishen-3d-lidar-sensors","title":"Leishen 3D LIDAR Sensors","text":"
Leishen Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) LS C16 150m (+15\u00b0/-15\u00b0), (360\u00b0) Y - LS C32\u00a0 150m (+15\u00b0/-15\u00b0), (360\u00b0) Y - CH 32 120m (+3.7\u00b0/-6.7\u00b0),(120\u00b0) Y - CH 128 20m (+14\u00b0/-17\u00b0)/(150\u00b0) Y -
Link to ROS 2 driver: https://github.com/leishen-lidar
Link to company website: http://www.lslidar.com/
"},{"location":"reference-hw/lidars/#livox-3d-lidar-sensors","title":"Livox 3D LIDAR Sensors","text":"
Livox Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) Horizon 260m (81.7\u00b0), (25.1\u00b0) Y Y Mid-70 90m (70.4\u00b0), (77.2\u00b0) Y - Avia 190m (70.4\u00b0), Circular Y - HAP 150m (25\u00b0), (120\u00b0) - -
Link to ROS 2 driver: https://github.com/Livox-SDK/livox_ros2_driver
Link to company website: https://www.livoxtech.com/
"},{"location":"reference-hw/lidars/#ouster-3d-lidar-sensors","title":"Ouster 3D LIDAR Sensors","text":"
Ouster Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) OS0 50m (90\u00b0), (360\u00b0) Y - OS1 120m (45\u00b0), (360\u00b0) Y - OS2 240m (22,5\u00b0), (360\u00b0) Y Y
Link to ROS 2 driver: https://github.com/ros-drivers/ros2_ouster_drivers
Smartmicro Radars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (Azimuth), (Elevation) ROS 2 Driver Autoware Tested (Y/N) Type 153 (Triple Mode Short, Medium Long) S:0.2...19 m\u00a0M:0.4...55 m L:0.8...120 m Short: (130\u00b0), (15\u00b0) Medium: (130\u00b0), (15\u00b0)Long: (100\u00b0),(15\u00b0) Y Y Type 132 ,(Dual Mode ,Medium, Long) M:0.5...64 m\u00a0\u00a0L:1...175 m Medium: (100\u00b0), (15\u00b0) Long: (32\u00b0), (15\u00b0) Y Y
Link to ROS 2 driver: https://github.com/smartmicro/smartmicro_ros2_radars
Link to company website: https://www.smartmicro.com/automotive-radar
FLIR ADK Thermal Vision cameras which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List MP FPS Interface Spectral Band FOV ROS 2 Driver Autoware Tested (Y/N) FLIR ADK 640x512 30 USB-GMSL,Ethernet 8-14 um (LWIR) 75\u02da, 50\u02da, 32\u02da, and 24\u02da - -"},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/","title":"Vehicle Drive By Wire Suppliers","text":""},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/#vehicle-drive-by-wire-suppliers","title":"Vehicle Drive By Wire Suppliers","text":""},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/#new-eagle-dbw-solutions","title":"New Eagle DBW Solutions","text":"
New Eagle DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Jeep CherokeeChrysler PacificaToyota PriusChevy BoltFord TransitRAM 1500Custom\u00a0 500W Sine Inverter2000 Watts8 Channel PDS Optional, Available Y Y
Link to company website: https://neweagle.net/autonomous-machines/
Dataspeed DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Lincoln MKZ, NautilusFord Fusion, F150, Transit Connect, RangerChrysler PacificaJeep CherokeePolaris GEM, RZR 12 Channel PDS,15 A Each at 12 V Optional, Available Y -
Link to company website: https://www.dataspeedinc.com/
Autonomous Stuff Pacmod DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Polaris GEM SeriesPolaris eLXD MY 2016+Polaris Ranger X900International ProStarLexus RX-450h MYFord RangerToyota Minivan Power distribution panel Optional, Available Y Y
Link to company website: https://autonomoustuff.com/platform/pacmod
"},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/#schaeffler-paravan-space-drive-dbw-solutions","title":"Schaeffler-Paravan Space Drive DBW Solutions","text":"
Schaeffler-Paravan Space Drive DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Custom Integration with Actuators - Optional, Available Y Y
Link to company website: https://www.schaeffler-paravan.de/en/products/space-drive-system/
PIX Moving AV solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Electric DBW Chassis and Platforms Y Y Y -
Link to company website: https://www.pixmoving.com/pixkit
Different sizes of platforms
"},{"location":"reference-hw/vehicle_platform_suppliers/#autonomoustuff-av-solutions","title":"Autonomoustuff AV Solutions","text":"
Autonomoustuff platform solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Road Vehicles, Golf Carts & Trucks Y Y Y -
Link to company website: https://autonomoustuff.com/platform
"},{"location":"reference-hw/vehicle_platform_suppliers/#navya-av-solutions","title":"NAVYA AV Solutions","text":"
NAVYA platform solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Shuttle Bus, Taxi and Tow Tractors Y Y - -
Link to company website: https://navya.tech/en
"},{"location":"reference-hw/vehicle_platform_suppliers/#zing-robotics-av-solutions","title":"ZING ROBOTICS AV Solutions","text":"
ZING Robotics platform solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Purpose built electric autonomous vehicles for aviation, military etc. Y Y - -
Link to company website: https://www.zingrobotics.com/
This page explains several documentation sites that are useful for Autoware and ROS development.
The Autoware Foundation is the official site of the Autoware Foundation. You can learn about the Autoware community here.
Autoware Documentation (this site) is the central documentation site for Autoware maintained by the Autoware community. General software-related information of Autoware is aggregated here.
Autoware Universe Documentation has READMEs and design documents of software components.
ROS Docs Guide explains the ROS 1 and ROS 2 documentation infrastructure.
Docs guide shows the list of useful documentation sites. Visit them and see if there is any information related to your problem.
Note that the documentation sites aren't always up-to-date and perfect. If you find out that some information is wrong, unclear, or missing in Autoware docs, feel free to submit a pull request following the contribution guidelines.
Warning
Since this documentation site is still under construction, there are some empty pages.
If you encounter a problem with Autoware, check existing issues and questions and search for similar issues first.
Issues
Note that Autoware has multiple repositories listed in autoware.repos. It is recommended to search across the repositories.
Questions
If no answer was found, create a new question thread here. If your question is not answered within a week, then @mention the maintainers to remind them.
Also, there are other discussion types such as feature requests or design discussions. Feel free to open or join such discussions.
If you don't know how to create a discussion, refer to GitHub Docs.
If you have a problem and you have confirmed it is a bug, find the appropriate repository and create a new issue there. If you can't determine the appropriate repository, ask the maintainers for help by creating a new discussion in the Q&A category.
Warning
Do not create issues for questions or unconfirmed bugs. If such issues are created, maintainers will transfer them to GitHub Discussions.
If you want to fix the bug by yourself, discuss the approach with maintainers and submit a pull request.
If you want to widely discuss a topic with the general Autoware and ROS community or ask a question not related to Autoware's bugs, post to the Autoware category on ROS Discourse.
Warning
Do not post questions about bugs to ROS Discourse!
Note that this may break your system and run carefully.
Run the setup script without installing CUDA-related libraries.
./setup-dev-env.sh --no-nvidia\n
Warning
Note that some components in Autoware Universe require CUDA, and only the CUDA version in the env file is supported at this time. Autoware may work with other CUDA versions, but those versions are not supported and functionality is not guaranteed.
Building Autoware requires a lot of memory, and your machine can freeze or crash if memory runs out during a build. To avoid this problem, 16-32GB of swap should be configured.
# Optional: Check the current swapfile\nfree -h\n\n# Remove the current swapfile\nsudo swapoff /swapfile\nsudo rm /swapfile\n\n# Create a new swapfile\nsudo fallocate -l 32G /swapfile\nsudo chmod 600 /swapfile\nsudo mkswap /swapfile\nsudo swapon /swapfile\n\n# Optional: Check if the change is reflected\nfree -h\n
For more detailed configuration steps, along with an explanation of swap, refer to Digital Ocean's \"How To Add Swap Space on Ubuntu 20.04\" tutorial
If there are too many CPU cores (more than 64) in your machine, it might requires larger memory. A workaround here is to limit the job number while building.
You can adjust -j4 to any number based on your system. For more details, see the manual page of GNU make.
By reducing the number of packages built in parallel, you can also reduce the amount of memory used. In the following example, the number of packages built in parallel is set to 1, and the number of jobs used by make is limited to 1.
By lowering both the number of packages built in parallel and the number of jobs used by make, you can reduce the memory usage. However, this also means that the build process takes longer.
"},{"location":"support/troubleshooting/#errors-when-using-the-latest-version-of-autoware","title":"Errors when using the latest version of Autoware","text":"
If you are working with the latest version of Autoware, issues can occur due to out-of-date software or old build files.
To resolve these types of problems, first try cleaning your build artifacts and rebuilding:
If the error is not resolved, remove src/ and update your workspace according to installation type (Docker / source).
Warning
Before removing src/, confirm that there are no modifications in your local environment that you want to keep!
If errors still persist after trying the steps above, delete the entire workspace, clone the repository once again and restart the installation process.
"},{"location":"support/troubleshooting/#errors-when-using-a-fixed-version-of-autoware","title":"Errors when using a fixed version of Autoware","text":"
In principle, errors should not occur when using a fixed version. That said, possible causes include:
ROS 2 has been updated with breaking changes.
For confirmation, check the Packaging and Release Management tag on ROS Discourse.
Your local environment is broken.
Confirm your .bashrc file, environment variables, and library versions.
In addition to the causes listed above, there are two common misunderstandings around the use of fixed versions.
You used a fixed version for autowarefoundation/autoware only. All of the repository versions in the .repos file must be specified in order to use a completely fixed version.
You didn't update the workspace after changing the branch of autowarefoundation/autoware. Changing the branch of autowarefoundation/autoware does not affect the files under src/. You have to run the vcs import command to update them.
"},{"location":"support/troubleshooting/#error-when-building-python-package","title":"Error when building python package","text":"
The error is due to the fact that for versions between 66.0.0 and 67.5.0 setuptools enforces the python packages to be PEP-440 conformant. Since version 67.5.1 setuptools has a fallback that makes it possible to work with old packages again.
The solution is to update setuptools to the newest version with the following command
If any errors occur when running Autoware with Docker or rocker, first confirm that your Docker installation is working correctly by running the following commands:
docker run --rm -it hello-world\ndocker run --rm -it ubuntu:latest\n
Next, confirm that you are able to access the base Autoware image that is stored on the GitHub Packages website
docker run --rm -it ghcr.io/autowarefoundation/autoware-universe:latest\n
"},{"location":"support/troubleshooting/#runtime-issues","title":"Runtime issues","text":""},{"location":"support/troubleshooting/#performance-related-issues","title":"Performance related issues","text":"
Symptoms:
Autoware is running slower than expected
Messages show up late in RViz2
Point clouds are lagging
Camera images are lagging behind
Point clouds or markers flicker on RViz2
When multiple subscribers use the same publishers, the message rate drops
If you have any of these symptoms, please the Performance Troubleshooting page.
"},{"location":"support/troubleshooting/#map-does-not-display-when-running-the-planning-simulator","title":"Map does not display when running the Planning Simulator","text":"
When running the Planning Simulator, the most common reason for the map not being displayed in RViz is because the map path has not been specified correctly in the launch command. You can confirm if this is the case by searching for Could not find lanelet map under {path-to-map-dir}/lanelet2_map.osm errors in the log.
Another possible reason is that map loading is taking a long time due to poor DDS performance. For this, please visit the Performance Troubleshooting page.
When multiple subscribers use the same publishers, the message rate drops
"},{"location":"support/troubleshooting/performance-troubleshooting/#diagnostic-steps","title":"Diagnostic Steps","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#check-if-multicast-is-enabled","title":"Check if multicast is enabled","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#target-symptoms","title":"Target symptoms","text":"
When multiple subscribers use the same publishers, the message rate drops
sudo ip link set multicast on {your-interface-name}\n
This way DDS will function as intended and multiple subscribers can receive data from a single publisher without any significant degradation in performance.
This is a temporary solution. And will be reverted once the computer restarts.
To make it permanent either,
Create a service to run this on startup (recommended)
OR put following lines to the ~/.bashrc file:
if [ ! -e /tmp/multicast_is_set ]; then\nsudo ip link set lo multicast on\ntouch /tmp/multicast_is_set\nfi\n
This will probably ask for password on the terminal every time you restart the computer.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-compilation-flags","title":"Check the compilation flags","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#target-symptoms_1","title":"Target symptoms","text":"
Autoware is running slower than expected
Point clouds are lagging
When multiple subscribers use the same publishers, the message rate drops even further
Check the ~/.bash_history file to see if there are any colcon build directives without -DCMAKE_BUILD_TYPE=Release or -DCMAKE_BUILD_TYPE=RelWithDebInfo flags at all.
Even if a build starts with these flags but same workspace gets compiled without these flags, it will still be a slow build in the end.
In addition, the nodes will run slow in general, especially the pointcloud_preprocessor nodes.
Remove the build, install and optionally log folders in the main autoware folder.
Compile the Autoware with either Release or RelWithDebInfo tags:
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release\n# Or build with debug flags too (comparable performance but you can debug too)\ncolcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=RelWithDebInfo\n
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-dds-settings","title":"Check the DDS settings","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#target-symptoms_2","title":"Target symptoms","text":"
Autoware is running slower than expected
Messages show up late in RViz2
Point clouds are lagging
Camera images are lagging behind
When multiple subscribers use the same publishers, the message rate drops
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-rmw-ros-middleware-implementation","title":"Check the RMW (ROS Middleware) implementation","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#diagnosis_2","title":"Diagnosis","text":"
Run following to check the middleware used:
echo $RMW_IMPLEMENTATION\n
The return line should be rmw_cyclonedds_cpp. If not, apply the solution.
If you are using a different DDS middleware, we might not have official support for it just yet.
Add export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp as a separate line in you ~/.bashrc file.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-if-the-cyclonedds-is-configured-correctly","title":"Check if the CycloneDDS is configured correctly","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#diagnosis_3","title":"Diagnosis","text":"
Run following to check the configuration .xml file of the CycloneDDS:
echo $CYCLONEDDS_URI\n
The return line should be a valid path pointing to an .xml file with CycloneDDS configuration.
Follow DDS settings:Tuning DDS documentation and make sure:
you have export CYCLONEDDS_URI=/absolute_path_to_your/cyclonedds_config.xml as a line on your ~/.bashrc file.
you have the cyclonedds_config.xml with the configuration provided in the documentation.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-linux-kernel-maximum-buffer-size","title":"Check the Linux kernel maximum buffer size","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#diagnosis_4","title":"Diagnosis","text":"
Run: sysctl net.core.rmem_max, it should return at least net.core.rmem_max = 2147483647.
This parameter specifies the maximum size of the \"receive buffer\" for each network connection, which determines the maximum amount of data that can be held in memory at any given time. By increasing the maximum buffer size, the operating system can accommodate larger bursts of data, which can help prevent network congestion and reduce packet loss, resulting in faster and more reliable data transfers.
Run: sysctl net.ipv4.ipfrag_time, it should return around: net.ipv4.ipfrag_time = 3
The \"net.ipv4.ipfrag_time\" parameter specifies the maximum time in seconds that the kernel should retain partially fragmented IP packets before discarding them. The default value for this parameter is usually set to 30 seconds, but it may vary depending on the specific operating system and configuration.
By setting this parameter to a lower value, such as 3 seconds, the kernel can free up memory resources more quickly by discarding partially fragmented packets that are no longer needed, which can help improve the overall performance and stability of the system.
Run: sysctl net.ipv4.ipfrag_high_thresh, it should return at around: net.ipv4.ipfrag_high_thresh = 134217728
The \"net.ipv4.ipfrag_high_thresh\" parameter specifies the high watermark threshold for the number of partially fragmented packets allowed in the kernel IP packet reassembly queue. When the number of partially fragmented packets in the queue exceeds this threshold, the kernel will start to drop newly arrived packets until the number of partially fragmented packets drops below the threshold.
By setting this parameter to a higher value, such as 134217728 (128 MB), the kernel can accommodate a larger number of partially fragmented packets in the queue, which can help improve the performance of network applications that transfer large amounts of data, such as file transfer protocols and multimedia streaming applications.
This will probably ask for password on the terminal every time you restart the computer.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-if-ros-localhost-only-communication-is-enabled","title":"Check if ROS localhost only communication is enabled","text":"
If you are using multi computer setup, please skip this check.
Enabling ROS localhost only communication can help improve the performance of ROS by reducing network traffic and avoiding potential conflicts with other devices on the network.
Add export $ROS_LOCALHOST_ONLY=1 as a separate line in you ~/.bashrc file.
This environment variable tells ROS to only use the loopback network interface (i.e., localhost) for communication, rather than using the network interface card (NIC) for Ethernet or Wi-Fi. This can reduce network traffic and potential conflicts with other devices on the network, resulting in better performance and stability.
Simulations provide a way of verifying Autoware's functionality before field testing with an actual vehicle. There are three main types of simulation that can be run ad hoc or via a scenario runner.
"},{"location":"tutorials/#simulation-methods","title":"Simulation methods","text":""},{"location":"tutorials/#ad-hoc-simulation","title":"Ad hoc simulation","text":"
Ad hoc simulation is a flexible method for running basic simulations on your local machine, and is the recommended method for anyone new to Autoware.
Scenario simulation uses a scenario runner to run more complex simulations based on predefined scenarios. It is often run automatically for continuous integration purposes, but can also be run on a local machine.
Planning simulation uses simple dummy data to test the Planning and Control components - specifically path generation, path following and obstacle avoidance. It verifies that a vehicle can reach a goal destination while avoiding pedestrians and surrounding cars, and is another method for verifying the validity of Lanelet2 maps. It also allows for testing of traffic light handling.
"},{"location":"tutorials/#how-does-planning-simulation-work","title":"How does planning simulation work?","text":"
Generate a path to the goal destination
Control the car along the generated path
Detect and avoid any humans or other vehicles on the way to the goal destination
Digital twin simulation is a simulation type that is able to produce realistic data and simulate almost the entire system. It is also commonly referred to as end-to-end simulation.
"},{"location":"tutorials/ad-hoc-simulation/","title":"Ad hoc simulation","text":""},{"location":"tutorials/ad-hoc-simulation/#ad-hoc-simulation","title":"Ad hoc simulation","text":"
Check if you have ~/autoware_data folder and files in it.
$ cd ~/autoware_data\n$ ls -C -w 30\nimage_projection_based_fusion\nlidar_apollo_instance_segmentation\nlidar_centerpoint\ntensorrt_yolo\ntensorrt_yolox\ntraffic_light_classifier\ntraffic_light_fine_detector\ntraffic_light_ssd_fine_detector\nyabloc_pose_initializer\n
If not, please, follow Manual downloading of artifacts.
"},{"location":"tutorials/ad-hoc-simulation/planning-simulation/#2-set-an-initial-pose-for-the-ego-vehicle","title":"2. Set an initial pose for the ego vehicle","text":"
a) Click the 2D Pose estimate button in the toolbar, or hit the P key.
b) In the 3D View pane, click and hold the left-mouse button, and then drag to set the direction for the initial pose. An image representing the vehicle should now be displayed.
Warning
Remember to set the initial pose of the car in the same direction as the lane.
To confirm the direction of the lane, check the arrowheads displayed on the map.
"},{"location":"tutorials/ad-hoc-simulation/planning-simulation/#3-set-a-goal-pose-for-the-ego-vehicle","title":"3. Set a goal pose for the ego vehicle","text":"
a) Click the 2D Goal Pose button in the toolbar, or hit the G key.
b) In the 3D View pane, click and hold the left-mouse button, and then drag to set the direction for the goal pose. If done correctly, you will see a planned path from initial pose to goal pose.
"},{"location":"tutorials/ad-hoc-simulation/planning-simulation/#4-start-the-ego-vehicle","title":"4. Start the ego vehicle","text":"
Now you can start the ego vehicle driving by clicking the AUTO button on OperationMode in AutowareStatePanel. Alteratively, you can manually start the vehicle by running the following command:
source ~/autoware/install/setup.bash\nros2 service call /api/operation_mode/change_to_autonomous autoware_adapi_v1_msgs/srv/ChangeOperationMode {}\n
After that, you can see AUTONOMOUS sign on OperationMode and AUTO button is grayed out.
By default, traffic lights on the map are all treated as if they are set to green. As a result, when a path is created that passed through an intersection with a traffic light, the ego vehicle will drive through the intersection without stopping.
The following steps explain how to set and reset traffic lights in order to test how the Planning component will respond.
Go to Panels -> Add new panel, select TrafficLightPublishPanel, and then press OK.
In TrafficLightPublishPanel, set the ID and color of the traffic light.
Click the SET button.
Finally, click the PUBLISH button to send the traffic light status to the simulator. Any planned path that goes past the selected traffic light will then change accordingly.
By default, Rviz should display the ID of each traffic light on the map. You can have a closer look at the IDs by zooming in the region or by changing the View type.
In case the IDs are not displayed, try the following troubleshooting steps:
a) In the Displays panel, find the traffic_light_id topic by toggling the triangle icons next to Map > Lanelet2VectorMap > Namespaces.
b) Check the traffic_light_id checkbox.
c) Reload the topic by clicking the Map checkbox twice.
You can update the color of the traffic light by selecting the next color (in the image it is GREEN) and clicking SET button. In the image the traffic light in front of the ego vehicle changed from RED to GREEN and the vehicle restarted.
To remove a traffic light from TrafficLightPublishPanel, click the RESET button.
Check if you have ~/autoware_data folder and files in it.
$ cd ~/autoware_data\n$ ls -C -w 30\nimage_projection_based_fusion\nlidar_apollo_instance_segmentation\nlidar_centerpoint\ntensorrt_yolo\ntensorrt_yolox\ntraffic_light_classifier\ntraffic_light_fine_detector\ntraffic_light_ssd_fine_detector\nyabloc_pose_initializer\n
If not, please, follow Manual downloading of artifacts.
Sample map and rosbag: Copyright 2020 TIER IV, Inc.
Due to privacy concerns, the rosbag does not contain image data, which will cause:
Traffic light recognition functionality cannot be tested with this sample rosbag.
Object detection accuracy is decreased.
"},{"location":"tutorials/ad-hoc-simulation/rosbag-replay-simulation/#how-to-run-a-rosbag-replay-simulation","title":"How to run a rosbag replay simulation","text":"
Any kind of for-profit activity with the trial version of the MORAI SIM:Drive is strictly prohibited.
"},{"location":"tutorials/ad-hoc-simulation/digital-twin-simulation/MORAI_Sim-tutorial/#hardware-requirements","title":"Hardware requirements","text":"Minimum PC Specs OS Windows 10, Ubuntu 20.04, Ubuntu 18.04, Ubuntu 16.04 CPU Intel i5-9600KF or AMD Ryzen 5 3500X RAM DDR4 16GB GPU RTX2060 Super Required PC Specs OS Windows 10, Ubuntu 20.04, Ubuntu 18.04, Ubuntu 16.04 CPU Intel i9-9900K or AMD Ryzen 7 3700X (or higher) RAM DDR4 64GB (or higher) GPU RTX2080Ti or higher"},{"location":"tutorials/ad-hoc-simulation/digital-twin-simulation/MORAI_Sim-tutorial/#application-and-download","title":"Application and Download","text":"
Only for AWF developers, trial license for 3 months can be issued. Download the application form and send to Hyeongseok Jeon
After the trial license is issued, you can login to MORAI Sim:Drive via Launchers (Windows/Ubuntu)
CAUTION: Do not use the Launchers in the following manual
"},{"location":"tutorials/scenario-simulation/planning-simulation/random-test-simulation/","title":"Random test simulation","text":""},{"location":"tutorials/scenario-simulation/planning-simulation/random-test-simulation/#random-test-simulation","title":"Random test simulation","text":"
Note
Running the Scenario Simulator requires some additional steps on top of building and installing Autoware, so make sure that Scenario Simulator installation has been completed first before proceeding.
For more information about supported parameters, refer to the random_test_runner documentation.
"},{"location":"tutorials/scenario-simulation/planning-simulation/scenario-test-simulation/","title":"Scenario test simulation","text":""},{"location":"tutorials/scenario-simulation/planning-simulation/scenario-test-simulation/#scenario-test-simulation","title":"Scenario test simulation","text":"
Note
Running the Scenario Simulator requires some additional steps on top of building and installing Autoware, so make sure that Scenario Simulator installation has been completed first before proceeding.
Driving Log Replayer is an evaluation tool for Autoware. To get started, follow the official instruction provided by TIER IV.
"}]}
\ No newline at end of file
+{"config":{"lang":["en"],"separator":"[\\s\\-]+","pipeline":["stopWordFilter"]},"docs":[{"location":"","title":"Introduction","text":""},{"location":"#autoware-documentation","title":"Autoware Documentation","text":""},{"location":"#about-autoware","title":"About Autoware","text":"
Autoware is the world\u2019s leading open-source software project for autonomous driving. Autoware is built on Robot Operating System (ROS) and enables commercial deployment of autonomous driving in a broad range of vehicles and applications.
This page is a collection of the links to the competitions that are related to the Autoware Foundation.
Title Status Description Ongoing Autoware / TIER IV Challenge 2023 Date: May 15, 2023 - Nov. 1st, 2023 As one of the main contributors of Autoware, TIER IV has been facing many difficult challenges through development, and TIER IV would like to sponsor a challenge to solve such engineering challenges. Any researchers, students, individuals or organizations are welcome to participate and submit their solution to any of the challenges we propose. Ongoing Japan Automotive AI Challenge 2023 Registration: June 5, 2023 - July 14, 2023 Qualifiers: July 3, 2023 - Aug. 31, 2023 Finals: Nov. 12, 2023 In this competition, we focus on challenging tasks posed by autonomous driving in factory environments and aim to develop Autoware-based AD software that can overcome them. The qualifiers use the digital twin autonomous driving simulator AWSIM to complete specific tasks within a virtual environment. Teams that make it to the finals have the opportunity to run their software on actual vehicles in a test course in Japan."},{"location":"autoware-competitions/#proposing-new-competition","title":"Proposing New Competition","text":"
If you want add a new competition to this page, please propose it in a TSC meeting and get confirmation from the AWF.
Thank you for your interest in contributing! Autoware is supported by people like you, and all types and sizes of contribution are welcome.
As a contributor, here are the guidelines that we would like you to follow for Autoware and its associated repositories.
Code of Conduct
What should I know before I get started?
Autoware concepts
Contributing to open source projects
How can I get help?
How can I contribute?
Participate in discussions
Join a working group
Report bugs
Make a pull request
Like Autoware itself, these guidelines are being actively developed and suggestions for improvement are always welcome! Guideline changes can be proposed by creating a discussion in the Ideas category.
"},{"location":"contributing/#code-of-conduct","title":"Code of Conduct","text":"
To ensure the Autoware community stays open and inclusive, please follow the Code of Conduct.
If you believe that someone in the community has violated the Code of Conduct, please make a report by emailing conduct@autoware.org.
"},{"location":"contributing/#what-should-i-know-before-i-get-started","title":"What should I know before I get started?","text":""},{"location":"contributing/#autoware-concepts","title":"Autoware concepts","text":"
To gain a high-level understanding of Autoware's architecture and design, the following pages provide a brief overview:
Autoware architecture
Autoware concepts
For experienced developers, the Autoware interfaces and individual component pages should also be reviewed to understand the inputs and outputs for each component or module at a more detailed level.
"},{"location":"contributing/#contributing-to-open-source-projects","title":"Contributing to open source projects","text":"
If you are new to open source projects, we recommend reading GitHub's How to Contribute to Open Source guide for an overview of why people contribute to open source projects, what it means to contribute and much more besides.
"},{"location":"contributing/#how-can-i-get-help","title":"How can I get help?","text":"
Do not open issues for general support questions as we want to keep GitHub issues for confirmed bug reports. Instead, open a discussion in the Q&A category. For more details on the support mechanisms for Autoware, refer to the Support guidelines.
Note
Issues created for questions or unconfirmed bugs will be moved to GitHub discussions by the maintainers.
"},{"location":"contributing/#how-can-i-contribute","title":"How can I contribute?","text":""},{"location":"contributing/#discussions","title":"Discussions","text":"
You can contribute to Autoware by facilitating and participating in discussions, such as:
Proposing a new feature to enhance Autoware
Joining an existing discussion and expressing your opinion
Organizing discussions for other contributors
Answering questions and supporting other contributors
The various working groups within the Autoware Foundation are responsible for accomplishing goals set by the Technical Steering Committee. These working groups are open to everyone, and joining a particular working group will allow you to gain an understanding of current projects, see how those projects are managed within each group and to contribute to issues that will help progress a particular project.
To see the schedule for upcoming working group meetings, refer to the Autoware Foundation events calendar.
Before you report a bug, please search the issue tracker for the appropriate repository. It is possible that someone has already reported the same issue and that workarounds exist. If you can't determine the appropriate repository, ask the maintainers for help by creating a new discussion in the Q&A category.
When reporting a bug, you should provide a minimal set of instructions to reproduce the issue. Doing so allows us to quickly confirm and focus on the right problem.
If you want to fix the bug by yourself that will be appreciated, but you should discuss possible approaches with the maintainers in the issue before submitting a pull request.
Creating an issue is straightforward, but if you happen to experience any problems then create a Q&A discussion to ask for help.
You can submit pull requests for small changes such as:
Minor documentation updates
Fixing spelling mistakes
Fixing CI failures
Fixing warnings detected by compilers or analysis tools
Making small changes to a single package
If your pull request is a large change, the following process should be followed:
Create a GitHub Discussion to propose the change. Doing so allows you to get feedback from other members and the Autoware maintainers and to ensure that the proposed change is in line with Autoware's design philosophy and current development plans. If you're not sure where to have that conversation, then create a new Q&A discussion.
Create an issue following consensus in the discussions
Create a pull request to implement the changes that references the Issue created in step 2
Create documentation for the new addition (if relevant)
Examples of large changes include:
Adding a new feature to Autoware
Adding a new documentation page or section
For more information on how to submit a good pull request, have a read of the pull request guidelines and don't forget to review the required license notations!
Autoware is licensed under Apache License 2.0. Thus all contributions will be licensed as such as per clause 5 of the Apache License 2.0:
5. Submission of Contributions. Unless You explicitly state otherwise,\n any Contribution intentionally submitted for inclusion in the Work\n by You to the Licensor shall be under the terms and conditions of\n this License, without any additional terms or conditions.\n Notwithstanding the above, nothing herein shall supersede or modify\n the terms of any separate license agreement you may have executed\n with Licensor regarding such Contributions.\n
Here is an example copyright header to add to the top of a new file:
Copyright [first year of contribution] The Autoware Contributors\nSPDX-License-Identifier: Apache-2.0\n
We don't write copyright notations of each contributor here. Instead, we place them in the NOTICE file like the following.
This product includes code developed by [company name].\nCopyright [first year of contribution] [company name]\n
Let us know if your legal department has a special request for the copyright notation.
Currently, the old formats explained here are also acceptable. Those old formats can be replaced by this new format if the original authors agree. Note that we won't write their copyrights to the NOTICE file unless they agree with the new format.
Also, it is encouraged to apply Clang-Tidy to each file. For the usage, see Applying Clang-Tidy to ROS packages.
Note that not all rules are covered by Clang-Tidy.
"},{"location":"contributing/coding-guidelines/languages/cpp/#style-rules","title":"Style rules","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#include-header-files-in-the-defined-order-required-partially-automated","title":"Include header files in the defined order (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale","title":"Rationale","text":"
Due to indirect dependencies, the include system of C++ makes different behaviors if the header order is different.
To reduce unintended bugs, local header files should come first.
If there are any reasons for defining macros at a specific position, write a comment before the macro.
// Compliant\n#include \"my_header.hpp\"\n\n#include \"my_package/foo.hpp\"\n\n#include <package1/foo.hpp>\n#include <package2/bar.hpp>\n\n#include <std_msgs/msg/header.hpp>\n\n#include <iostream>\n#include <vector>\n\n// For the foo bar reason, the FOO_MACRO must be defined here.\n#define FOO_MACRO\n#include <foo/bar.hpp>\n
"},{"location":"contributing/coding-guidelines/languages/cpp/#use-lower-snake-case-for-function-names-required-partially-automated","title":"Use lower snake case for function names (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_1","title":"Rationale","text":"
It is consistent with the C++ standard library.
It is consistent with other programming languages such as Python and Rust.
"},{"location":"contributing/coding-guidelines/languages/cpp/#use-upper-camel-case-for-enum-names-required-partially-automated","title":"Use upper camel case for enum names (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_2","title":"Rationale","text":"
"},{"location":"contributing/coding-guidelines/languages/cpp/#use-lower-snake-case-for-constant-names-required-partially-automated","title":"Use lower snake case for constant names (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_3","title":"Rationale","text":"
"},{"location":"contributing/coding-guidelines/languages/cpp/#count-acronyms-and-contractions-of-compound-words-as-one-word-required-partially-automated","title":"Count acronyms and contractions of compound words as one word (required, partially automated)","text":""},{"location":"contributing/coding-guidelines/languages/cpp/#rationale_4","title":"Rationale","text":"
To clarify the boundaries of words when acronyms are consecutive.
ROS 2 logging is a powerful tool for understanding and debugging ROS nodes.
This page focuses on how to design console logging in Autoware and shows several practical examples. To comprehensively understand how ROS 2 logging works, refer to the logging documentation.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#logging-use-cases-in-autoware","title":"Logging use cases in Autoware","text":"
Developers debug code by seeing the console logs.
Vehicle operators take appropriate risk-avoiding actions depending on the console logs.
Log analysts analyze the console logs that are recorded in rosbag files.
To efficiently support these use cases, clean and highly visible logs are required. For that, several rules are defined below.
DEBUG: Use this level to show debug information for developers. Note that logs with this level is hidden by default.
INFO: Use this level to notify events (cyclic notifications during initialization, state changes, service responses, etc.) to operators.
WARN: Use this level when a node can continue working correctly, but unintended behaviors might happen.
For example, \"path optimization failed but the previous data can be used\", \"the localization score is low\", etc.
ERROR: Use this level when a node can't continue working correctly, and unintended behaviors would happen.
For example, \"path optimization failed and the path is empty\", \"the vehicle will trigger an emergency stop\", etc.
FATAL: Use this level when the entire system can't continue working correctly, and the system must be stopped.
For example, \"the vehicle control ECU doesn't respond\", \"the system storage crashed\", etc.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#filter-out-unnecessary-logs-by-setting-logging-options-required-non-automated","title":"Filter out unnecessary logs by setting logging options (required, non-automated)","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#rationale_1","title":"Rationale","text":"
Some third-party nodes such as drivers may not follow the Autoware's guidelines. If the logs are noisy, unnecessary logs should be filtered out.
Use the --log-level {level} option to change the minimum level of logs to be displayed:
<launch>\n<!-- This outputs only FATAL level logs. -->\n<node pkg=\"demo_nodes_cpp\" exec=\"talker\" ros_args=\"--log-level fatal\" />\n</launch>\n
If you want to disable only specific output targets, use the --disable-stdout-logs, --disable-rosout-logs, and/or --disable-external-lib-logs options:
<launch>\n<!-- This outputs to rosout and disk. -->\n<node pkg=\"demo_nodes_cpp\" exec=\"talker\" ros_args=\"--disable-stdout-logs\" />\n</launch>\n
<launch>\n<!-- This outputs to stdout. -->\n<node pkg=\"demo_nodes_cpp\" exec=\"talker\" ros_args=\"--disable-rosout-logs --disable-external-lib-logs\" />\n</launch>\n
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#use-throttled-logging-when-the-log-is-unnecessarily-shown-repeatedly-required-non-automated","title":"Use throttled logging when the log is unnecessarily shown repeatedly (required, non-automated)","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#rationale_2","title":"Rationale","text":"
If tons of logs are shown on the console, people miss important message.
The following cases are acceptable even if it's not throttled.
The message is really worth displaying every time.
The message level is DEBUG.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#do-not-depend-on-rclcppnode-in-core-library-classes-but-depend-only-on-rclcpplogginghpp-advisory-non-automated","title":"Do not depend on rclcpp::Node in core library classes but depend only on rclcpp/logging.hpp (advisory, non-automated)","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#rationale_3","title":"Rationale","text":"
Core library classes, which contain reusable algorithms, may also be used for non-ROS platforms. When porting libraries to other platforms, fewer dependencies are preferred.
// Compliant\n#include <rclcpp/logging.hpp>\n\nclass FooCore {\npublic:\nexplicit FooCore(const rclcpp::Logger & logger) : logger_(logger) {}\n\nvoid process() {\nRCLCPP_INFO(logger_, \"message\");\n}\n\nprivate:\nrclcpp::Logger logger_;\n};\n\n// Compliant\n// Note that logs aren't published to `/rosout` if the logger name is different from the node name.\n#include <rclcpp/logging.hpp>\n\nclass FooCore {\nvoid process() {\nRCLCPP_INFO(rclcpp::get_logger(\"foo_core_logger\"), \"message\");\n}\n};\n\n\n// Non-compliant\n#include <rclcpp/node.hpp>\n\nclass FooCore {\npublic:\nexplicit FooCore(const rclcpp::NodeOptions & node_options) : node_(\"foo_core_node\", node_options) {}\n\nvoid process() {\nRCLCPP_INFO(node_.get_logger(), \"message\");\n}\n\nprivate:\nrclcpp::Node node_;\n};\n
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#tips","title":"Tips","text":""},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#use-rqt_console-to-filter-logs","title":"Use rqt_console to filter logs","text":"
To filter logs, using rqt_console is useful:
ros2 run rqt_console rqt_console\n
For more details, refer to ROS 2 Documentation.
"},{"location":"contributing/coding-guidelines/ros-nodes/console-logging/#useful-marco-expressions","title":"Useful marco expressions","text":"
To debug program, sometimes you need to see which functions and lines of code are executed. In that case, you can use __FILE__, __LINE__ and __FUNCTION__ macro:
The commonly used coordinate systems include the world coordinate system, the vehicle coordinate system, and the sensor coordinate system.
The world coordinate system is a fixed coordinate system that defines the physical space in the environment where the vehicle is located.
The vehicle coordinate system is the vehicle's own coordinate system, which defines the vehicle's position and orientation in the world coordinate system.
The sensor coordinate system is the sensor's own coordinate system, which is used to define the sensor's position and orientation in the vehicle coordinate system.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#how-coordinates-are-used-in-autoware","title":"How coordinates are used in Autoware","text":"
In Autoware, coordinate systems are typically used to represent the position and movement of vehicles and obstacles in space. Coordinate systems are commonly used for path planning, perception and control, can help the vehicle decide how to avoid obstacles and to plan a safe and efficient path of travel.
Transformation of sensor data
In Autoware, each sensor has a unique coordinate system and their data is expressed in terms of the coordinates. In order to correlate the independent data between different sensors, we need to find the position relationship between each sensor and the vehicle body. Once the installation position of the sensor on the vehicle body is determined, it will remain fixed during running, so the offline calibration method can be used to determine the precise position of each sensor relative to the vehicle body.
ROS TF2
The TF2 system maintains a tree of coordinate transformations to represent the relationships between different coordinate systems. Each coordinate system is given a unique name and they are connected by coordinate transformations. How to use TF2, refer to the TF2 tutorial.
earth: earth coordinate system describe the position of any point on the earth in terms of geodetic longitude, latitude, and altitude. In Autoware, the earth frame is only used in the GnssInsPositionStamped message.
map: map coordinate system is used to represent the location of points on a local map. Geographical coordinate system are mapped into plane rectangular coordinate system using UTM or MGRS. The map frame`s axes point to the East, North, Up directions as explained in Coordinate Axes Conventions.
base_link: vehicle coordinate system, the origin of the coordinate system is the center of the rear axle of the vehicle.
imu, lidar, gnss, radar: these are sensor frames, transfer to vehicle coordinate system through mounting relationship.
camera_link: camera_link is ROS standard camera coordinate system .
camera_optical_link: camera_optical_link is image standard camera coordinate system.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#estimating-the-base_link-frame-by-using-the-other-sensors","title":"Estimating the base_link frame by using the other sensors","text":"
Generally we don't have the localization sensors physically at the base_link frame. So various sensors localize with respect to their own frames, let's call it sensor frame.
We introduce a new frame naming convention: x_by_y:
We cannot directly get the sensor frame. Because we would need the EKF module to estimate the base_link frame first.
Without the EKF module the best we can do is to estimate Map[map] --> sensor_by_sensor --> base_link_by_sensor using this sensor.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#example-by-the-gnssins-sensor","title":"Example by the GNSS/INS sensor","text":"
For the integrated GNSS/INS we use the following frames:
The gnss_ins_by_gnss_ins frame is obtained by the coordinates from GNSS/INS sensor. The coordinates are converted to map frame using the gnss_poser node.
Finally gnss_ins_by_gnss_ins frame represents the position of the gnss_ins estimated by the gnss_ins sensor in the map.
Then by using the static transformation between gnss_ins and the base_link frame, we can obtain the base_link_by_gnss_ins frame. Which represents the base_link estimated by the gnss_ins sensor.
"},{"location":"contributing/coding-guidelines/ros-nodes/coordinate-system/#how-they-can-be-created","title":"How they can be created","text":"
Calibration of sensor
The conversion relationship between every sensor coordinate system and base_link can be obtained through sensor calibration technology. How to calibrating your sensors refer to this link calibrating your sensors.
Localization
The relationship between the base_link coordinate system and the map coordinate system is determined by the position and orientation of the vehicle, and can be obtained from the vehicle localization result.
Geo-referencing of map data
The geo-referencing information can get the transformation relationship of earth coordinate system to local map coordinate system.
Place configuration files such as node parameters. For ROS parameters, use the extension .param.yaml. For non-ROS parameters, use the extension .yaml.
Rationale: Since ROS parameters files are type-sensitive, they should not be the target of some code formatters and linters. In order to distinguish the file type, we use different file extensions.
Place header files exposed to other packages. Do not place files directly under the include directory, but place files under the directory with the package name. This directory is used for mostly library headers. Note that many headers do not need to be placed here. It is enough to place the headers under the src directory.
Autoware use ROS 2 launch system to startup the software. Please see the official documentation to get a basic understanding about ROS 2 Launch system if you are not familiar with it.
"},{"location":"contributing/coding-guidelines/ros-nodes/launch-files/#guideline","title":"Guideline","text":""},{"location":"contributing/coding-guidelines/ros-nodes/launch-files/#the-organization-of-launch-files-in-autoware","title":"The organization of launch files in Autoware","text":"
Autoware mainly has two repositories related to launch file organization: the autoware.universe and the autoware_launch.
the autoware.universe contains the code of the main Autoware modules, and its launch directory is responsible for launching the nodes of each module. Autoware software stack is organized based on the architecture, so you may find that we try to match the launch structure similar to the architecture (splitting of files, namespace). For example, the tier4_map_launch subdirectory corresponds to the map module, so do the other tier4_*_launch subdirectories.
The autoware_launch is a repository referring to autoware.universe. The mainly purpose of introducing this repository is to provide the general entrance to start the Autoware software stacks, i.e, calling the launch file of each module.
The autoware.launch.xml is the basic launch file for road driving scenarios.
As can be seen from the content, the entire launch file is divided into several different modules, including Vehicle, System, Map, Sensing, Localization, Perception, Planning, Control, etc. By setting the launch_* argument equals to true or false , we can determine which modules to be loaded.
The logging_simulator.launch.xml is often used together with the recorded ROS bag to debug if the target module (e.g, Sensing, Localization or Perception) functions normally.
The planning_simulator.launch.xml is based on the Planning Simulator tool, mainly used for testing/validation of Planning module by simulating traffic rules, interactions with dynamic objects and control commands to the ego vehicle.
The e2e_simulator.launch.xml is the launcher for digital twin simulation environment.
"},{"location":"contributing/coding-guidelines/ros-nodes/launch-files/#add-a-new-package-in-autoware","title":"Add a new package in Autoware","text":"
If a newly created package has executable node, we expect sample launch file and configuration within the package, just like the recommended structure shown in previous directory structure page.
In order to automatically load the newly added package when starting Autoware, you need to make some necessary changes to the corresponding launch file. For example, if using ICP instead of NDT as the pointcloud registration algorithm, you can modify the autoware.universe/launch/tier4_localization_launch/launch/pose_estimator/pose_estimator.launch.xml file to load the newly added ICP package.
Another purpose of introducing the autoware_launch repository is to facilitate the parameter management of Autoware. Thinking about this situation: if we want to integrate Autoware to a specific vehicle and modify parameters, we have to fork autoware.universe which also has a lot of code other than parameters and is frequently updated by developers. By integrating these parameters in autoware_launch, we can customize the Autoware parameters just by forking autoware_launch repository. Taking the localization module as an examples:
all the \u201claunch parameters\u201d for localization component is listed in the files under autoware_launch/autoware_launch/config/localization.
the \"launch parameters\" file paths are set in the autoware_launch/autoware_launch/launch/components/tier4_localization_component.launch.xml file.
in autoware.universe/launch/tier4_localization_launch/launch, the launch files loads the \u201claunch parameters\u201d if the argument is given in the parameter configuration file. You can still use the default parameters in each packages to launch tier4_localization_launch within autoware.universe.
All the fields by default have the following units depending on their types:
type default unit distance meter (m) angle radians (rad) time second (s) speed m/s velocity m/s acceleration m/s\u00b2 angular vel. rad/s angular accel. rad/s\u00b2
If a field in a message has any of these default units, don't add any suffix or prefix denoting the type.
type non-default unit suffix distance nanometer _nm distance micrometer _um distance millimeter _mm distance kilometer _km angle degree (deg) _deg time nanosecond _ns time microsecond _us time millisecond _ms time minute _min time hour (h) _hour velocity km/h _kmph
If a unit that you'd like to use doesn't exist here, create an issue/PR to add it to this list.
"},{"location":"contributing/coding-guidelines/ros-nodes/message-guidelines/#message-field-types","title":"Message field types","text":"
For list of types supported by the ROS interfaces see here.
Also copied here for convenience:
Message Field Type C++ equivalent boolboolbyteuint8_tcharcharfloat32floatfloat64doubleint8int8_tuint8uint8_tint16int16_tuint16uint16_tint32int32_tuint32uint32_tint64int64_tuint64uint64_tstringstd::stringwstringstd::u16string"},{"location":"contributing/coding-guidelines/ros-nodes/message-guidelines/#arrays","title":"Arrays","text":"
On top of the message, briefly explain what the message contains and/or what it is used for. For an example, see sensor_msgs/msg/Imu.msg.
If necessary, add line comments before the fields that explain the context and/or meaning.
For simple fields like x, y, z, w you might not need to add comments.
Even though it is not strictly checked, try not to pass 100 characters in a line.
Example:
# Number of times the vehicle performed an emergency brake\nuint32 count_emergency_brake\n\n# Seconds passed since the last emergency brake\nuint64 duration\n
Autoware ROS nodes have declared parameters which values are provided during the node start up in the form of a parameter file. All the expected parameters with corresponding values should exist in the parameter file. Depending on the application, the parameter values might need to be modified.
Find more information on parameters from the official ROS documentation:
A ROS package which uses the declare_parameter(...) function should:
use the declare_parameter(...) with out a default value
create a parameter file
create a schema file
The rationale behind this workflow is to have a verified single source of truth to pass to the ROS node and to be used in the web documentation. The approach reduces the risk of using invalid parameter values and makes maintenance of documentation easier. This is achieved by:
declare_parameter(...) throws an exception if an expected parameter is missing in the parameter file
the schema validates the parameter file in the CI and renders a parameter table, as depicted in the graphics below
As there is no default_value provided, the function throws an exception if a parameter were to be missing in the provided *.param.yaml file. Use a type from the C++ Type column in the table below for the declare_parameter(...) function, replacing INSERT_TYPE.
ParameterType Enum C++ Type PARAMETER_BOOLboolPARAMETER_INTEGERint64_tPARAMETER_DOUBLEdoublePARAMETER_STRINGstd::stringPARAMETER_BYTE_ARRAYstd::vector<uint8_t>PARAMETER_BOOL_ARRAYstd::vector<bool>PARAMETER_INTEGER_ARRAYstd::vector<int64_t>PARAMETER_DOUBLE_ARRAYstd::vector<double>PARAMETER_STRING_ARRAYstd::vector<std::string>
The table has been derived from Parameter Type and Parameter Value.
See example: Lidar Apollo Segmentation TVM Nodes declare function
The parameter file is minimal as there is no need to provide the user with additional information, e.g., description or type. This is because the associated schema file provides the additional information. Use the template below as a starting point for a ROS node.
Note: /** is used instead of the explicit node namespace, this allows the parameter file to be passed to a ROS node which has been remapped.
To adapt the template to the ROS node, replace each INSERT_PARAMETER_..._NAME and INSERT_PARAMETER_..._VALUE for all parameters. Each declare_parameter(...) takes one parameter as input. All the parameter files should have the .param.yaml suffix so that the auto-format can be applied properly.
Autoware has the following two types of parameter files for ROS packages:
Node parameter file
Node parameter files store the default parameters provided for each package in Autoware.
For example, the parameter of behavior_path_planner
All nodes in Autoware must have a parameter file if ROS parameters are declared in the node.
For FOO_package, the parameter is expected to be stored in FOO_package/config.
The launch file for individual packages must load node parameter by default:
When a user creates a launch package for the user's vehicle, the user should copy node parameter files for the nodes that are called in the launch file as \"launch parameter files\".
Launch parameter files are then customized specifically for user's vehicle.
For example, the customized parameter of behavior_path_planner stored under autoware_launch
The examples for launch parameter files are stored under autoware_launch.
JSON Schema is used the validate the parameter file(s) ensuring that it has the correct structure and content. Using JSON Schema for this purpose is considered best practice for cloud-native development. The schema template below shall be used as a starting point when defining the schema for a ROS node.
The schema file path is INSERT_PATH_TO_PACKAGE/schema/ and the schema file name is INSERT_NODE_NAME.schema.json. To adapt the template to the ROS node, replace each INSERT_... and add all parameters 1..N.
See example: Lidar Apollo Segmentation TVM Nodes schema
Parameters have several attributes, some are required and some optional. The optional attributes are highly encouraged when applicable, as they provide useful information about a parameter and can ensure the value of the parameter is within its bounds.
a tested and verified value, see JSON Schema default
bound(s)
type dependent, e.g., integer, range and size
"},{"location":"contributing/coding-guidelines/ros-nodes/parameters/#tips-and-tricks","title":"Tips and Tricks","text":"
Using well established standards enables the use of conventional tooling. Below is an example of how to link a schema to the parameter file(s) using VS Code. This enables a developer with convenient features such as auto-complete and parameter bound validation.
In the root directory of where the project is hosted, create a .vscode folder with two files; extensions.json containing
The RedHat YAML extension enables validation of YAML files using JSON Schema and the \"yaml.schemas\" setting associates the *.schema.json file with all *.param.yaml files in the config/ folder.
ROS allows topics, parameters and nodes to be namespaced which provides the following benefits:
Multiple instances of the same node type will not cause naming clashes.
Topics published by a node can be automatically namespaced with the node's namespace providing a meaningful and easily-visible connection.
Keeps from cluttering the root namespace.
Helps to maintain separation-of-concerns.
This page focuses on how to use namespaces in Autoware and shows some useful examples. For basic information on topic namespaces, refer to this tutorial.
"},{"location":"contributing/coding-guidelines/ros-nodes/topic-namespaces/#how-topics-should-be-named-in-node","title":"How topics should be named in node","text":"
Autoware divides the node into the following functional categories, and adds the start namespace for the nodes according to the categories.
localization
perception
planning
control
sensing
vehicle
map
system
When a node is run in a namespace, all topics which that node publishes are given that same namespace. All nodes in the Autoware stack must support namespaces by avoiding practices such as publishing topics in the global namespace.
In general, topics should be namespaced based on the function of the node which produces them and not the node (or nodes) which consume them.
Classify topics as input or output topics based on they are subscribed or published by the node. In the node, input topic is named input/topic_name and output topic is named output/topic_name.
Configure the topic in the node's launch file. Take the joy_controller node as an example, in the following example, set the input and output topics and remap topics in the joy_controller.launch.xml file.
Visualization or debug purpose topics should have ~/debug/ namespace.
e.g., In node obstacle_avoidance_planner, in order to debug or visualizing topics, using topic names of type ~/debug/topic_name to publish information.
Contributions to Autoware's documentation are welcome, and the same principles described in the contribution guidelines should be followed. Small, limited changes can be made by forking this repository and submitting a pull request, but larger changes should be discussed with the community and Autoware maintainers via GitHub Discussion first.
Examples of small changes include:
Fixing spelling or grammatical mistakes
Fixing broken links
Making an addition to an existing, well-defined page, such as the Troubleshooting guide.
Examples of larger changes include:
Adding new pages with a large amount of detail, such as a tutorial
Re-organization of the existing documentation structure
You should refer to the Google developer documentation style guide as much as possible. Reading the Highlights page of that guide is recommended, but if not then the key points below should be noted.
Use standard American English spelling and punctuation.
Use sentence case for document titles and section headings.
Use descriptive link text.
Write short sentences that are easy to understand and translate.
"},{"location":"contributing/documentation-guidelines/#tips","title":"Tips","text":""},{"location":"contributing/documentation-guidelines/#how-to-preview-your-modification","title":"How to preview your modification","text":"
There are two ways to preview your modification on a documentation website.
"},{"location":"contributing/documentation-guidelines/#1-using-github-actions-workflow","title":"1. Using GitHub Actions workflow","text":"
Follow the steps below.
Create a pull request to the repository.
Add the deploy-docs label from the sidebar (See below figure).
Wait for a couple of minutes, and the github-actions bot will notify the URL for the pull request's preview.
"},{"location":"contributing/documentation-guidelines/#2-running-an-mkdocs-server-in-your-local-environment","title":"2. Running an MkDocs server in your local environment","text":"
Instead of creating a PR, you can use the mkdocs command to build Autoware's documentation websites on your local computer. Assuming that you are using Ubuntu OS, run the following to install the required libraries.
Autoware uses the fork-and-pull model. For more details about the model, refer to GitHub Docs.
The following is a general example of the pull request workflow based on the fork-and-pull model. Use this workflow as a reference when you contribute to Autoware.
Create an issue.
Discuss the approaches to the issue with maintainers.
Confirm the support guidelines before creating an issue.
Follow the discussion guidelines when you discuss with other contributors.
Create a fork repository. (for the first time only)
Write code in your fork repository according to the approach agreed upon in the issue.
Write the tests and documentation as appropriate.
Follow the coding guidelines guidelines when you write code.
Follow the Testing guidelines guidelines when you write tests.
Follow the Documentation guidelines guidelines when you write documentation.
Follow the commit guidelines when you commit your changes.
Test the code.
It is recommended that you summarize the test results, because you will need to explain the test results in the later review process.
If you are not sure what tests should be done, discuss them with maintainers.
Create a pull request.
Follow the pull request rules when you create a pull request.
Wait for the pull request to be reviewed.
The reviewers will review your code following the review guidelines.
Not only the reviewers, but also the author is encouraged to understand the review guidelines.
If CI checks have failed, fix the errors.
Address the review comments pointed out by the reviewers.
If you don't understand the meaning of a review comment, ask the reviewers until you understand it.
Fixing without understanding the reason is not recommended because the author should be responsible for the final content of their own pull request.
If you don't agree with a review comment, ask the reviewers for a rational reason.
The reviewers are obligated to make the author understand the meanings of each comment.
After you have done with the review comments, re-request a review to the reviewers and back to 6.
If there are no more new review comments, the reviewers will approve the pull request and proceed to 8.
Merge the pull request.
Anyone with write access can merge the pull request if there is no special request from maintainers.
The author is encouraged to merge the pull request to feel responsible for their own pull request.
If the author does not have write access, ask the reviewers or maintainers.
"},{"location":"contributing/pull-request-guidelines/#pull-request-rules","title":"Pull request rules","text":""},{"location":"contributing/pull-request-guidelines/#use-an-appropriate-pull-request-template-required-non-automated","title":"Use an appropriate pull request template (required, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale","title":"Rationale","text":"
The unified style of descriptions by templates can make reviews efficient.
There are two types of templates. Select one based on the following condition.
Standard change:
Complexity:
New features or significant updates.
Requires deeper understanding of the codebase.
Impact:
Affects multiple parts of the system.
Basically includes minor features, bug fixes and performance improvement.
Needs testing before merging.
Small change:
Complexity:
Documentation, simple refactoring, or style adjustments.
Easy to understand and review.
Impact:
Minimal effect on the system.
Quicker merge with less testing needed.
"},{"location":"contributing/pull-request-guidelines/#steps-to-use-an-appropriate-pull-request-template","title":"Steps to use an appropriate pull request template","text":"
Select the appropriate template, as shown in this video.
Read the selected template carefully and fill the required content.
Check the checkboxes during a review.
There are pre-review checklist and post-review checklist for the author.
"},{"location":"contributing/pull-request-guidelines/#set-appropriate-reviewers-after-creating-a-pull-request-required-partially-automated","title":"Set appropriate reviewers after creating a pull request (required, partially automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_1","title":"Rationale","text":"
Pull requests must be reviewed by appropriate reviewers to keep the quality of the codebase.
For most ROS packages, reviewers will be automatically assigned based on the maintainer information in package.xml.
If no reviewer is assigned automatically, assign reviewers manually following the instructions in GitHub Docs.
You can find the reviewers by seeing the .github/CODEOWNERS file of the repository.
If you are not sure the appropriate reviewers, ask @autoware-maintainers.
If you have no rights to assign reviewers, mention reviewers instead.
"},{"location":"contributing/pull-request-guidelines/#apply-conventional-commits-to-the-pull-request-title-required-automated","title":"Apply Conventional Commits to the pull request title (required, automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_2","title":"Rationale","text":"
Conventional Commits can generate categorized changelogs, for example using git-cliff.
For the repositories that contain code (most repositories), use the definition of conventional-commit-types for the type.
For documentation repositories such as autoware-documentation, use the following definition:
feat
Add new pages.
Add contents to the existing pages.
fix
Fix the contents in the existing pages.
refactor
Move contents to different pages.
docs
Update documentation for the documentation repository itself.
build
Update the settings of the documentation site builder.
! (breaking changes)
Remove pages.
Change the URL of pages.
perf and test are generally unused. Other types have the same meaning as the code repositories.
"},{"location":"contributing/pull-request-guidelines/#add-the-related-component-names-to-the-scope-of-conventional-commits-advisory-non-automated","title":"Add the related component names to the scope of Conventional Commits (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_3","title":"Rationale","text":"
It helps contributors find pull requests that are relevant to them.
For ROS packages, adding the package name or component name is good.
feat(trajectory_follower): add an awesome feature\nrefactor(planning, control): use common utils\n
"},{"location":"contributing/pull-request-guidelines/#keep-a-pull-request-small-advisory-non-automated","title":"Keep a pull request small (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_4","title":"Rationale","text":"
Small pull requests are easy to understand for reviewers.
Small pull requests are easy to revert for maintainers.
Avoid developing two features in one pull request.
Avoid mixing different types (feat, fix, refactor, etc.) of changes in the same commit.
"},{"location":"contributing/pull-request-guidelines/#remind-reviewers-if-there-is-no-response-for-more-than-a-week-advisory-non-automated","title":"Remind reviewers if there is no response for more than a week (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/#rationale_5","title":"Rationale","text":"
It is the author's responsibility to care about their own pull request until it is merged.
Autoware has several checks for a pull request. The results are shown at the bottom of the pull request page as below.
If the \u274c mark is shown, click the Details button and investigate the failure reason.
If the Required mark is shown, you cannot merge the pull request unless you resolve the error. If not, it is optional, but preferably it should be fixed.
The following sections explain about common CI checks in Autoware. Note that some repositories may have different settings.
The Developer Certificate of Origin (DCO) is a lightweight way for contributors to certify that they wrote or otherwise have the right to submit the code they are contributing to the project.
This workflow checks whether the pull request fulfills DCO. You need to confirm the required items and commit with git commit -s.
For more information, refer to the GitHub App page.
pre-commit is a tool to run formatters or linters when you commit.
This workflow checks whether the pull request has no error with pre-commit.
In the workflow pre-commit.ci - pr is enabled in the repository, it will automatically fix errors by pre-commit.ci as many as possible. If there are some errors remain, fix them manually.
You can run pre-commit in your local environment by the following command:
pre-commit run -a\n
Or you can install pre-commit to the repository and automatically run it before committing:
pre-commit install\n
Since it is difficult to detect errors with no false positives, some jobs are split into another config file and marked as optional. To check them, use the --config option:
pre-commit run -a --config .pre-commit-config-optional.yaml\n
This workflow detects spelling mistakes using CSpell with our dictionary file. You can submit pull requests to tier4/autoware-spell-check-dict to update the dictionary.
Since it is difficult to detect errors with no false positives, it is an optional workflow, but it is preferable to remove spelling mistakes as many as possible.
This workflow checks colcon build and colcon test for the pull request. To make the CI faster, it doesn't check all packages but only modified packages and the dependencies.
This workflow is the ARM64 version of build-and-test-differential. You need to add the ARM64 label to run this workflow.
For reference information, since ARM machines are not supported by GitHub-hosted runners, we use self-hosted runners prepared by the AWF. For the details about self-hosted runners, refer to GitHub Docs.
"},{"location":"contributing/pull-request-guidelines/commit-guidelines/#use-dash-case-for-the-separator-of-branch-names-advisory-non-automated","title":"Use dash-case for the separator of branch names (advisory, non-automated)","text":""},{"location":"contributing/pull-request-guidelines/commit-guidelines/#rationale_1","title":"Rationale","text":"
If you have already submitted a pull request, you do not have to change the branch name because you need to re-create a pull request, which is noisy and a waste of time. Be careful from the next time.
"},{"location":"contributing/pull-request-guidelines/review-tips/","title":"Review tips","text":""},{"location":"contributing/pull-request-guidelines/review-tips/#review-tips","title":"Review tips","text":""},{"location":"contributing/pull-request-guidelines/review-tips/#toggle-annotations-or-review-comments-in-the-diff-view","title":"Toggle annotations or review comments in the diff view","text":"
There might be some annotations or review comments in the diff view during your review.
To toggle annotations, press the A key.
Before:
After:
To toggle review comments, press the I key.
For other keyboard shortcuts, refer to GitHub Docs.
"},{"location":"contributing/pull-request-guidelines/review-tips/#view-code-in-the-web-based-visual-studio-code","title":"View code in the web-based Visual Studio Code","text":"
You can open Visual Studio Code from your browser to view code in a rich UI. To use it, press the . key on any repository or pull request.
For more detailed usage, refer to github/dev.
"},{"location":"contributing/pull-request-guidelines/review-tips/#check-out-the-branch-of-a-pull-request-quickly","title":"Check out the branch of a pull request quickly","text":"
If you want to check out the branch of a pull request, it's generally troublesome with the fork-and-pull model.
# Copy the user name and the fork URL.\ngit remote add {user-name} {fork-url}\ngit checkout {user-name}/{branch-name}\ngit remote rm {user-name} # To clean up\n
Instead, you can use GitHub CLI to simplify the steps, just run gh pr checkout {pr-number}.
You can copy the command from the top right of the pull request page.
An integration test is defined as the phase in software testing where individual software modules are combined and tested as a group. Integration tests occur after unit tests, and before validation tests.
The input to an integration test is a set of independent modules that have been unit tested. The set of modules is tested against the defined integration test plan, and the output is a set of properly integrated software modules that is ready for system testing.
"},{"location":"contributing/testing-guidelines/integration-testing/#value-of-integration-testing","title":"Value of integration testing","text":"
Integration tests determine if independently developed software modules work correctly when the modules are connected to each other. In ROS 2, the software modules are called nodes. Testing a single node is a special type of integration test that is commonly referred to as component testing.
Integration tests help to find the following types of errors:
Incompatible interactions between nodes, such as non-matching topics, different message types, or incompatible QoS settings.
Edge cases that were not touched by unit testing, such as a critical timing issue, network communication delays, disk I/O failures, and other such problems that can occur in production environments.
Issues that can occur while the system is under high CPU/memory load, such as malloc failures. This can be tested using tools like stress and udpreplay to test the performance of nodes with real data.
With ROS 2, it is possible to program complex autonomous-driving applications with a large number of nodes. Therefore, a lot of effort has been made to provide an integration-test framework that helps developers test the interaction of ROS 2 nodes.
Doing so adds smoke tests that ensure that a node can be:
Launched with a default parameter file.
Terminated with a standard SIGTERM signal.
For the full API documentation, refer to the package design page.
Note
This API is not suitable for all smoke test cases. It cannot be used when a specific file location (eg: for a map) is required to be passed to the node, or if some preparation needs to be conducted before node launch. In such cases use the manual solution from the component test section below.
"},{"location":"contributing/testing-guidelines/integration-testing/#integration-test-with-a-single-node-component-test","title":"Integration test with a single node: component test","text":"
The simplest scenario is a single node. In this case, the integration test is commonly referred to as a component test.
To add a component test to an existing node, you can follow the example of the lanelet2_map_loader in the map_loader package (added in this PR).
In package.xml, add:
<test_depend>ros_testing</test_depend>\n
In CMakeLists.txt, add or modify the BUILD_TESTING section:
Then a launch description is created to launch the node under test. Note that the test_map.osm file path is found and passed to the node, something that cannot be done with the smoke testing API:
@pytest.mark.launch_test\ndef generate_test_description():\n\n lanelet2_map_path = os.path.join(\n get_package_share_directory(\"map_loader\"), \"test/data/test_map.osm\"\n )\n\n lanelet2_map_loader = Node(\n package=\"map_loader\",\n executable=\"lanelet2_map_loader\",\n parameters=[{\"lanelet2_map_path\": lanelet2_map_path}],\n )\n\n context = {}\n\n return (\n LaunchDescription(\n [\n lanelet2_map_loader,\n # Start test after 1s - gives time for the map_loader to finish initialization\n launch.actions.TimerAction(\n period=1.0, actions=[launch_testing.actions.ReadyToTest()]\n ),\n ]\n ),\n context,\n )\n
Note
Since the node need time to process the input lanelet2 map, we use a TimerAction to delay the start of the test by 1s.
In the example above, the context is empty but it can be used to pass objects to the test cases.
You can find an example of using the context in the ROS 2 context_launch_test.py test example.
Finally, a test is executed after the node executable has been shut down (post_shutdown_test). Here we ensure that the node was launched without error and exited cleanly.
@launch_testing.post_shutdown_test()\nclass TestProcessOutput(unittest.TestCase):\n def test_exit_code(self, proc_info):\n # Check that process exits with code 0: no error\n launch_testing.asserts.assertExitCodes(proc_info)\n
"},{"location":"contributing/testing-guidelines/integration-testing/#running-the-test","title":"Running the test","text":"
Continuing the example from above, first build your package:
The simple test described in Integration test with a single node: component test can be extended in numerous directions, such as testing a node's output.
"},{"location":"contributing/testing-guidelines/integration-testing/#testing-the-output-of-a-node","title":"Testing the output of a node","text":"
To test while the node is running, create an active test by adding a subclass of Python's unittest.TestCase to *launch.test.py. Some boilerplate code is required to access output by creating a node and a subscription to a particular topic, e.g.
import unittest\n\nclass TestRunningDataPublisher(unittest.TestCase):\n\n @classmethod\n def setUpClass(cls):\n cls.context = Context()\n rclpy.init(context=cls.context)\n cls.node = rclpy.create_node(\"test_node\", context=cls.context)\n\n @classmethod\n def tearDownClass(cls):\n rclpy.shutdown(context=cls.context)\n\n def setUp(self):\n self.msgs = []\n sub = self.node.create_subscription(\n msg_type=my_msg_type,\n topic=\"/info_test\",\n callback=self._msg_received\n )\n self.addCleanup(self.node.destroy_subscription, sub)\n\n def _msg_received(self, msg):\n # Callback for ROS 2 subscriber used in the test\n self.msgs.append(msg)\n\n def get_message(self):\n startlen = len(self.msgs)\n\n executor = rclpy.executors.SingleThreadedExecutor(context=self.context)\n executor.add_node(self.node)\n\n try:\n # Try up to 60 s to receive messages\n end_time = time.time() + 60.0\n while time.time() < end_time:\n executor.spin_once(timeout_sec=0.1)\n if startlen != len(self.msgs):\n break\n\n self.assertNotEqual(startlen, len(self.msgs))\n return self.msgs[-1]\n finally:\n executor.remove_node(self.node)\n\n def test_message_content():\n msg = self.get_message()\n self.assertEqual(msg, \"Hello, world\")\n
Unit testing is the first phase of testing and is used to validate units of source code such as classes and functions. Typically, a unit of code is tested by validating its output for various inputs. Unit testing helps ensure that the code behaves as intended and prevents accidental changes of behavior.
Autoware uses the ament_cmake framework to build and run tests. The same framework is also used to analyze the test results.
ament_cmake provides several convenience functions to make it easy to register tests in a CMake-based package and to ensure that JUnit-compatible result files are generated. It currently supports a few different testing frameworks like pytest, gtest, and gmock.
In order to prevent tests running in parallel from interfering with each other when publishing and subscribing to ROS topics, it is recommended to use commands from ament_cmake_ros to run tests in isolation.
See below for an example of using ament_add_ros_isolated_gtest with colcon test. All other tests follow a similar pattern.
"},{"location":"contributing/testing-guidelines/unit-testing/#create-a-unit-test-with-gtest","title":"Create a unit test with gtest","text":"
In my_cool_pkg/test, create the gtest code file test_my_cool_pkg.cpp:
Next add an entry under BUILD_TESTING in the CMakeLists.txt to compile the test source files:
if(BUILD_TESTING)\n\nament_add_ros_isolated_gtest(test_my_cool_pkg test/test_my_cool_pkg.cpp)\ntarget_link_libraries(test_my_cool_pkg ${PROJECT_NAME})\ntarget_include_directories(test_my_cool_pkg PRIVATE src) # For private headers.\n...\nendif()\n
This automatically links the test with the default main function provided by gtest. The code under test is usually in a different CMake target (${PROJECT_NAME} in the example) and its shared object for linking needs to be added. If the test source files include private headers from the src directory, the directory needs to be added to the include path using target_include_directories() function.
To register a new gtest item, wrap the test code with the macro TEST (). TEST () is a predefined macro that helps generate the final test code, and also registers a gtest item to be available for execution. The test case name should be in CamelCase, since gtest inserts an underscore between the fixture name and the class case name when creating the test executable.
gtest/gtest.h also contains predefined macros of gtest like ASSERT_TRUE(condition), ASSERT_FALSE(condition), ASSERT_EQ(val1,val2), ASSERT_STREQ(str1,str2), EXPECT_EQ(), etc. ASSERT_* will abort the test if the condition is not satisfied, while EXPECT_* will mark the test as failed but continue on to the next test condition.
Info
More information about gtest and its features can be found in the gtest repo.
In the demo CMakeLists.txt, ament_add_ros_isolated_gtest is a predefined macro in ament_cmake_ros that helps simplify adding gtest code. Details can be viewed in ament_add_gtest.cmake.
Look in the ~/workspace/log/test_<date>/<package_name> directory for all the raw test commands, std_out, and std_err. There is also the ~/workspace/log/latest_*/ directory containing symbolic links to the most recent package-level build and test output.
To print the tests' details while the tests are being run, use the --event-handlers console_cohesion+ option to print the details directly to the console:
$ colcon test --event-handlers console_cohesion+ --packages-select my_cool_pkg\n\n...\ntest 1\n Start 1: test_my_cool_pkg\n\n1: Test command: /usr/bin/python3 \"-u\" \"~/workspace/install/share/ament_cmake_test/cmake/run_test.py\" \"~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml\" \"--package-name\" \"my_cool_pkg\" \"--output-file\" \"~/workspace/build/my_cool_pkg/ament_cmake_gtest/test_my_cool_pkg.txt\" \"--command\" \"~/workspace/build/my_cool_pkg/test_my_cool_pkg\" \"--gtest_output=xml:~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml\"\n1: Test timeout computed to be: 60\n1: -- run_test.py: invoking following command in '~/workspace/src/my_cool_pkg':\n1: - ~/workspace/build/my_cool_pkg/test_my_cool_pkg --gtest_output=xml:~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml\n1: [==========] Running 1 test from 1 test case.\n1: [----------] Global test environment set-up.\n1: [----------] 1 test from test_my_cool_pkg\n1: [ RUN ] test_my_cool_pkg.test_hello\n1: Hello World\n1: [ OK ] test_my_cool_pkg.test_hello (0 ms)\n1: [----------] 1 test from test_my_cool_pkg (0 ms total)\n1:\n1: [----------] Global test environment tear-down\n1: [==========] 1 test from 1 test case ran. (0 ms total)\n1: [ PASSED ] 1 test.\n1: -- run_test.py: return code 0\n1: -- run_test.py: inject classname prefix into gtest result file '~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml'\n1: -- run_test.py: verify result file '~/workspace/build/my_cool_pkg/test_results/my_cool_pkg/test_my_cool_pkg.gtest.xml'\n1/5 Test #1: test_my_cool_pkg ................... Passed 0.09 sec\n\n...\n\n100% tests passed, 0 tests failed out of 5\n\nLabel Time Summary:\ncopyright = 0.49 sec*proc (1 test)\ncppcheck = 0.20 sec*proc (1 test)\ngtest = 0.05 sec*proc (1 test)\nlint_cmake = 0.18 sec*proc (1 test)\nlinter = 1.34 sec*proc (4 tests)\nxmllint = 0.47 sec*proc (1 test)\n\nTotal Test time (real) = 7.91 sec\n...\n
The GNSS data is available in sensor_msgs/msg/NavSatFix message type.
But also the Applanix raw messages are also included in applanix_msgs/msg/NavigationPerformanceGsof50 and applanix_msgs/msg/NavigationSolutionGsof49 message types. In order to be able to play back these messages, you need to build and source the applanix_msgs package.
# Create a workspace and clone the repository\nmkdir -p ~/applanix_ws/src && cd \"$_\"\ngit clone https://github.com/autowarefoundation/applanix.git\ncd ..\n\n# Build the workspace\ncolcon build --symlink-install --packages-select applanix_msgs\n\n# Source the workspace\nsource ~/applanix_ws/install/setup.bash\n\n# Now you can play back the messages\n
Also make sure to source Autoware Universe workspace too.
# Install awscli\n$ sudo apt update && sudo apt install awscli -y\n\n# This will download the entire dataset to the current directory.\n# (About 10.9GB of data)\n$ aws s3 sync s3://autoware-files/collected_data/2022-08-22_leo_drive_isuzu_bags/ ./2022-08-22_leo_drive_isuzu_bags --no-sign-request\n\n# Optionally,\n# If you instead want to download a single bag file, you can get a list of the available files with following:\n$ aws s3 ls s3://autoware-files/collected_data/2022-08-22_leo_drive_isuzu_bags/ --no-sign-request\n PRE all-sensors-bag1_compressed/\n PRE all-sensors-bag2_compressed/\n PRE all-sensors-bag3_compressed/\n PRE all-sensors-bag4_compressed/\n PRE all-sensors-bag5_compressed/\n PRE all-sensors-bag6_compressed/\n PRE driving_20_kmh_2022_06_10-16_01_55_compressed/\n PRE driving_30_kmh_2022_06_10-15_47_42_compressed/\n\n# Then you can download a single bag file with the following:\naws s3 sync s3://autoware-files/collected_data/2022-08-22_leo_drive_isuzu_bags/all-sensors-bag1_compressed/ ./all-sensors-bag1_compressed --no-sign-request\n
"},{"location":"datasets/#autocoreai-lidar-ros-2-bag-file-and-pcap","title":"AutoCore.ai - lidar ROS 2 bag file and pcap","text":"
This dataset contains pcap files and ros2 bag files from Ouster OS1-64 Lidar. The pcap file and ros2 bag file is recorded in the same time with slight difference in duration.
Autoware provides the runtimes and technology components by open-source software. The runtimes are based on the Robot Operating System (ROS). The technology components are provided by contributors, which include, but are not limited to:
Sensing
Camera Component
LiDAR Component
RADAR Component
GNSS Component
Computing
Localization Component
Perception Component
Planning Component
Control Component
Logging Component
System Monitoring Component
Actuation
DBW Component
Tools
Simulator Component
Mapping Component
Remote Component
ML Component
Annotation Component
Calibration Component
"},{"location":"design/#concern-assumption-and-limitation","title":"Concern, Assumption, and Limitation","text":"
The downside of the microautonomy architecture is that the computational performance of end applications is sacrificed due to its data path overhead attributed to functional modularity. In other words, the trade-off characteristic of the microautonomy architecture exists between computational performance and functional modularity. This trade-off problem can be solved technically by introducing real-time capability. This is because autonomous driving systems are not really designed to be real-fast, that is, low-latency computing is nice-to-have but not must-have. The must-have feature for autonomous driving systems is that the latency of computing is predictable, that is, the systems are real-time. As a whole, we can compromise computational performance to an extent that is predictable enough to meet the given timing constraints of autonomous driving systems, often referred to as deadlines of computation.
The Autoware concepts page describes the design philosophy of Autoware. Readers (service providers and all Autoware users) will learn the basic concepts underlying Autoware development, such as microautonomy and the Core/Universe architecture.
The Autoware architecture page describes an overview of each module that makes up Autoware. Readers (all Autoware users) will gain a high-level picture of how each module that composes Autoware works.
The Autoware interfaces page describes in detail the interface of each module that makes up Autoware. Readers (intermediate developers) will learn how to add new functionality to Autoware and how to integrate their own modules with Autoware.
The current Autoware is defined to be a layered architecture that clarifies each module's role and simplifies the interface between them. By doing so:
Autoware's internal processing becomes more transparent.
Collaborative development is made easier because of the reduced interdependency between modules.
Users can easily replace an existing module (e.g. localization) with their own software component by simply wrapping their software to fit in with Autoware's interface.
Note that the initial focus of this architecture design was solely on driving capability, and so the following features were left as future work:
Autoware's architecture consists of the following six stacks. Each linked page contains a more detailed set of requirements and use cases specific to that stack:
A diagram showing Autoware's nodes in the default configuration can be found on the Node diagram page. Detailed documents for each node are available in the Autoware Universe docs.
Note that Autoware configurations are scalable / selectable and will vary depending on the environment and required use cases.
This document presents the design concept of the Control Component. The content is as follows:
Autoware Control Design
Outlining the policy for Autoware's control, which deals with only general information for autonomous driving systems and provides generic control commands to the vehicle.
Vehicle Adaptation Design
Describing the policy for vehicle adaptation, which utilizes adapter mechanisms to standardize the characteristics of the vehicle's drive system and integrate it with Autoware.
Control Feature Design
Demonstrating the features provided by Autoware's control.
Presenting the approach towards the functions installed in the vehicle such as ABS.
"},{"location":"design/autoware-architecture/control/#autoware-control-design","title":"Autoware Control Design","text":"
The Control Component generates the control signal to which the Vehicle Component subscribes. The generated control signals are computed based on the reference trajectories from the Planning Component.
The Control Component consists of two modules. The trajectory_follower module generates a vehicle control command to follow the reference trajectory received from the planning module. The command includes, for example, the desired steering angle and target speed. The vehicle_command_gate is responsible for filtering the control command to prevent abnormal values and then sending it to the vehicle. This gate also allows switching between multiple sources such as the MRM (minimal risk maneuver) module or some remote control module, in addition to the trajectory follower.
The Autoware control system is designed as a platform for automated driving systems that can be compatible with a diverse range of vehicles.
The control process in Autoware uses general information (such as target acceleration and deceleration) and no vehicle-specific information (such as brake pressure) is used. Hence it can be adjusted independently of the vehicle's drive interface enabling easy integration or performance tuning.
Furthermore, significant differences that affect vehicle motion constraints, such as two-wheel steering or four-wheel steering, are addressed by switching the control vehicle model, achieving control specialized for each characteristic.
Autoware's control module outputs the necessary information to control the vehicle as a substitute for a human driver. For example, the control command from the control module looks like the following:
Note that vehicle-specific values such as pedal positions and low-level information such as individual wheel rotation speeds are excluded from the command.
Autoware is designed to be an autonomous driving platform able to accommodate vehicles with various drivetrain types.
This is an explanation of how Autoware handles the standardization of systems with different vehicle drivetrain. The interfaces for vehicle drivetrain are diverse, including steering angle, steering angular velocity, steering torque, speed, accel/brake pedals, and brake pressure. To accommodate these differences, Autoware adds an adapter module between the control component and the vehicle interface. This module performs the conversion between the proprietary message types used by the vehicle (such as brake pressure) and the generic types used by Autoware (such as desired acceleration). By providing this conversion information, the differences in vehicle drivetrain can be accommodated.
If the information is not known in advance, an automatic calibration tool can be used. Calibration will occur within limited degrees of freedom, generating the information necessary for the drivetrain conversion automatically.
This configuration is summarized in the following diagram.
"},{"location":"design/autoware-architecture/control/#examples-of-several-vehicle-interfaces","title":"Examples of several vehicle interfaces","text":"
This is an example of the several drivetrain types in the vehicle interface.
Vehicle Lateral interface Longitudinal interface Note Lexus Steering angle Accel/brake pedal position Acceleration lookup table conversion for longitudinal JPN TAXI Steering angle Accel/brake pedal position Acceleration lookup table conversion for longitudinal GSM8 Steering EPS voltage Acceleration motor voltage, Deceleration brake hydraulic pressure lookup table and PID conversion for lateral and longitudinal YMC Golfcart Steering angle Velocity Logiee yaw rate Velocity F1 TENTH Steering angle Motor RPM interface code"},{"location":"design/autoware-architecture/control/#control-feature-design","title":"Control Feature Design","text":"
The following lists the features provided by Autoware's Control/Vehicle component, as well as the conditions and assumptions required to utilize them effectively.
The proper operation of the ODD is limited by factors such as whether the functions are enabled, delay time, calibration accuracy and degradation rate, and sensor accuracy.
Feature Description\u3000 Requirements/Assumptions Note \u3000Limitation for now Lateral Control Control the drivetrain system related to lateral vehicle motion Trying to increase the number of vehicle types that can be supported in the future. Only front-steering type is supported. Longitudinal Control Control the drivetrain system related to longitudinal vehicle motion Slope Compensation Supports precise vehicle motion control on slopes Gradient information can be obtained from maps or sensors attached to the chassis If gradient information is not available, the gradient is estimated from the vehicle's pitch angle. Delay Compensation Controls the drivetrain system appropriately in the presence of time delays The drivetrain delay information is provided in advance If there is no delay information, the drivetrain delay is estimated automatically (automatic calibration). However, the effect of delay cannot be completely eliminated, especially in scenarios with sudden changes in speed. Only fixed delay times can be set for longitudinal and lateral drivetrain systems separately. It does not accommodate different delay times for the accelerator and brake. Drivetrain IF Conversion (Lateral Control) Converts the drivetrain-specific information of the vehicle into the drivetrain information used by Autoware (e.g., target steering angular velocity \u2192 steering torque) The conversion information is provided in advance If there is no conversion information, the conversion map is estimated automatically (automatic calibration). The degree of freedom for conversion is limited (2D lookup table + PID FB). Drivetrain IF Conversion (Longitudinal Control) Converts the drivetrain-specific information of the vehicle into the drivetrain information used by Autoware (e.g., target acceleration \u2192 accelerator/brake pedal value) The conversion information is provided in advance If there is no conversion information, the conversion map is estimated automatically (automatic calibration). The degree of freedom for conversion is limited (2D lookup table + PID FB). Automatic Calibration Automatically estimates and applies values such as drivetrain IF conversion map and delay time. The drivetrain status can be obtained (must) Anomaly Detection Notifies when there is a discrepancy in the calibration or unexpected drivetrain behavior The drivetrain status can be obtained (must) Steering Zero Point Correction Corrects the midpoint of the steering to achieve appropriate steering control The drivetrain status can be obtained (must) Steering Deadzone Correction Corrects the deadzone of the steering to achieve appropriate steering control The steering deadzone parameter is provided in advance If the parameter is unknown, the deadzone parameter is estimated from driving information Not available now Steering Deadzone Estimation Dynamically estimates the steering deadzone from driving data Not available now Weight Compensation Performs appropriate vehicle control according to weight Weight information can be obtained from sensors If there is no weight sensor, estimate the weight from driving information. Currently not available Weight Estimation Dynamically estimates weight from driving data Currently not available
The list above does not cover wheel control systems such as ABS commonly used in vehicles. Regarding these features, the following considerations are taken into account.
"},{"location":"design/autoware-architecture/control/#integration-with-vehicle-side-functions","title":"Integration with vehicle-side functions","text":"
ABS (Anti-lock Brake System) and ESC (Electric Stability Control) are two functions that may be pre-installed on a vehicle, directly impacting its controllability. The control modules of Autoware assume that both ABS and ESC are installed on the vehicle and their absence may cause unreliable controls depending on the target ODD. For example, with low-velocity driving in a controlled environment, these functions are not necessary.
Also, note that this statement does not negate the development of ABS functionality in autonomous driving systems.
"},{"location":"design/autoware-architecture/control/#autoware-capabilities-and-vehicle-requirements","title":"Autoware Capabilities and Vehicle Requirements","text":"
As an alternative to human driving, autonomous driving systems essentially aim to handle tasks that humans can perform. This includes not only controlling the steering wheel, accel, and brake, but also automatically detecting issues such as poor brake response or a misaligned steering angle. However, this is a trade-off, as better vehicle performance will lead to superior system behavior, ultimately affecting the design of ODD.
On the other hand, for tasks that are not typically anticipated or cannot be handled by a human driver, processing in the vehicle ECU is expected. Examples of such scenarios include cases where the brake response is clearly delayed or when the vehicle rotates due to a single-side tire slipping. These tasks are typically handled by ABS or ESC.
Localization aims to estimate vehicle pose, velocity, and acceleration.
Goals:
Propose a system that can estimate vehicle pose, velocity, and acceleration for as long as possible.
Propose a system that can diagnose the stability of estimation and send a warning message to the error-monitoring system if the estimation result is unreliable.
Design a vehicle localization function that can work with various sensor configurations.
Non-goals:
This design document does not aim to develop a localization system that
is infallible in all environments
works outside of the pre-defined ODD (Operational Design Domain)
has better performance than is required for autonomous driving
This section shows example sensor configurations and their expected performances. Each sensor has its own advantages and disadvantages, but overall performance can be improved by fusing multiple sensors.
The vehicle is located in a structure-rich environment, such as an urban area
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable","title":"Situations that can make the system unstable","text":"
The vehicle is placed in a structure-less environment, such as a rural landscape, highway, or tunnel
Environmental changes have occurred since the map was created, such as snow cover or the construction/destruction of buildings.
Surrounding objects are occluded
The car is surrounded by objects undetectable by LiDAR, e.g., glass windows, reflections, or absorption (dark objects)
The environment contains laser beams at the same frequency as the car's LiDAR sensor(s)
The system can estimate the vehicle location on the point cloud map with the error of ~10cm.
The system is operable at night.
"},{"location":"design/autoware-architecture/localization/#3d-lidar-or-camera-vector-map","title":"3D-LiDAR or Camera + Vector Map","text":""},{"location":"design/autoware-architecture/localization/#expected-situation_1","title":"Expected situation","text":"
Road with clear white lines and loose curvatures, such as a highway or an ordinary local road.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_1","title":"Situations that can make the system unstable","text":"
White lines are scratchy or covered by rain or snow
Tight curvature such as intersections
Large reflection change of the road surface caused by rain or paint
The vehicle is placed in an open environment with few to no surrounding objects, such as a rural landscape.
"},{"location":"design/autoware-architecture/localization/#situation-that-can-make-the-system-unstable","title":"Situation that can make the system unstable","text":"
GNSS signals are blocked by surrounding objects, e.g., tunnels or buildings.
The vehicle is placed in an environment with rich visual features, such as an urban area.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_2","title":"Situations that can make the system unstable","text":"
The vehicle is placed in a texture-less environment.
The vehicle is surrounded by other objects.
The camera observes significant illumination changes, such as those caused by sunshine, headlights from other vehicles or when approaching the exit of a tunnel.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_3","title":"Situations that can make the system unstable","text":"
The vehicle is running on a slippery or bumpy road, which can cause incorrect observations of wheel speed.
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_4","title":"Situations that can make the system unstable","text":"
IMUs have a bias1 that is dependent on the surrounding temperature, and can cause incorrect sensor observation or odometry drift.
The vehicle is placed in an environment with low magnetic noise
"},{"location":"design/autoware-architecture/localization/#situations-that-can-make-the-system-unstable_5","title":"Situations that can make the system unstable","text":"
The vehicle is placed in an environment with high magnetic noise, such as one containing buildings or structures with reinforced steel or other materials that generate electromagnetic waves.
The car is placed in an environment with magnetic markers installed.
"},{"location":"design/autoware-architecture/localization/#situations-where-the-system-becomes-unstable","title":"Situations where the system becomes unstable","text":"
By implementing different modules, various sensor configurations and algorithms can be used.
The localization system can start pose estimation from an ambiguous initial location.
The system can produce a reliable initial location estimation.
The system can manage the state of the initial location estimation (uninitialized, initializable, or non-initializable) and can report to the error monitor.
Two architectures are defined, \"Required\" and \"Recommended\". However, the \"Required\" architecture only contains the inputs and outputs necessary to accept various localization algorithms. To improve the reusability of each module, the required components are defined in the \"Recommended\" architecture section along with a more detailed explanation.
Produces the vehicle velocity, angular velocity, acceleration, angular acceleration, and their covariances
It is possible to create a single module for both twist and acceleration or to create two separate modules - the choice of architecture is up to the developer
The twist estimator produces velocity and angular velocity from internal sensor observation
The accel estimator produces acceleration and angular acceleration from internal sensor observations
Monitors and guarantees the stability and reliability of pose estimation by fusing information obtained from multiple localization modules
Reports error status to the error monitor
"},{"location":"design/autoware-architecture/localization/#tf-tree","title":"TF tree","text":"frame meaning earth ECEF (Earth Centered Earth Fixed\uff09 map Origin of the map coordinate (ex. MGRS origin) viewer User-defined frame for rviz base_link Reference pose of the ego-vehicle (projection of the rear-axle center onto the ground surface) sensor Reference pose of each sensor
Developers can optionally add other frames such as odom or base_footprint as long as the tf structure above is maintained.
The localization module should provide pose, velocity, and acceleration for control, planning, and perception.
Latency and stagger should be sufficiently small or adjustable such that the estimated values can be used for control within the ODD (Operational Design Domain).
The localization module should produce the pose on a fixed coordinate frame.
Sensors should be independent of each other so that they can be easily replaced.
The localization module should provide a status indicating whether or not the autonomous vehicle can operate with the self-contained function or map information.
Tools or manuals should describe how to set proper parameters for the localization module
Valid calibration parameters should be provided to align different frame or pose coordinates and sensor timestamps.
To maintain sufficient pose estimation performance for safe operation, the following metrics are considered:
Safety
The distance traveled within the ODD where pose estimation met the required accuracy, divided by the overall distance traveled within the ODD, as a percentage.
The anomaly detection rate for situations where the localization module cannot estimate pose within the ODD
The accuracy of detecting when the vehicle goes outside of the ODD, as a percentage.
Computational load
Latency
"},{"location":"design/autoware-architecture/localization/#5-interface-and-data-structure","title":"5. Interface and Data Structure","text":""},{"location":"design/autoware-architecture/localization/#6-concerns-assumptions-and-limitations","title":"6. Concerns, Assumptions, and Limitations","text":""},{"location":"design/autoware-architecture/localization/#prerequisites-of-sensors-and-inputs","title":"Prerequisites of sensors and inputs","text":""},{"location":"design/autoware-architecture/localization/#sensor-prerequisites","title":"Sensor prerequisites","text":"
Input data is not defective.
Internal sensor observation such as IMU continuously keeps the proper frequency.
Input data has correct and exact time stamps.
Estimated poses can be inaccurate or unstable if the timestamps are not exact.
Sensors are correctly mounted with exact positioning and accessible from TF.
If the sensor positions are inaccurate, estimation results may be incorrect or unstable.
A sensor calibration framework is required to properly obtain the sensor positions.
Autoware relies on high-definition point cloud maps and vector maps of the driving environment to perform various tasks such as localization, route planning, traffic light detection, and predicting the trajectories of pedestrians and other vehicles.
This document describes the design of map component of Autoware, including its requirements, architecture design, features, data formats, and interface to distribute map information to the rest of autonomous driving stack.
Map should provide two types of information to the rest of the stack:
Semantic information about roads as a vector map
Geometric information about the environment as a point cloud map (optional)
A vector map contains highly accurate information about a road network, lane geometry, and traffic lights. It is required for route planning, traffic light detection, and predicting the trajectories of other vehicles and pedestrians.
A 3D point cloud map is primarily used for LiDAR-based localization and part of perception in Autoware. In order to determine the current position and orientation of the vehicle, a live scan captured from one or more LiDAR units is matched against a pre-generated 3D point cloud map. Therefore, an accurate point cloud map is crucial for good localization results. However, if the vehicle has an alternate localization method with enough accuracy, for example using camera-based localization, point cloud map may not be required to use Autoware.
In addition to above two types of maps, Autoware also requires a supplemental file for specifying the coordinate system of the map in geodetic system.
This diagram describes the high-level architecture of Map component in Autoware.
The Map component consists of the following sub-components:
Point Cloud Map Loading: Load and publish point cloud map
Vector Map Loading: Load and publish vector map
Projection Loading: Load and publish projection information for conversion between local coordinate (x, y, z) and geodetic coordinate (latitude, longitude, altitude)
"},{"location":"design/autoware-architecture/map/#4-component-interface","title":"4. Component interface","text":""},{"location":"design/autoware-architecture/map/#input-to-the-map-component","title":"Input to the map component","text":"
From file system
Point cloud map and its metadata file
Vector map
Projection information
"},{"location":"design/autoware-architecture/map/#output-from-the-map-component","title":"Output from the map component","text":"
To Sensing
Projection information: Used to convert GNSS data from geodetic coordinate system to local coordinate system
To Localization
Point cloud map: Used for LiDAR-based localization
Vector map: Used for localization methods based on road markings, etc
To Perception
Point cloud map: Used for obstacle segmentation by comparing LiDAR and point cloud map
Vector map: Used for vehicle trajectory prediction
To Planning
Vector map: Used for behavior planning
To API layer
Projection information: Used to convert localization results from local coordinate system to geodetic coordinate system
The point cloud map must be supplied as a file with the following requirements:
The point cloud map must be projected on the same coordinate defined in map_projection_loader in order to be consistent with the lanelet2 map and other packages that converts between local and geodetic coordinates. For more information, please refer to the readme of map_projection_loader.
It must be in the PCD (Point Cloud Data) file format, but can be a single PCD file or divided into multiple PCD files.
Each point in the map must contain X, Y, and Z coordinates.
An intensity or RGB value for each point may be optionally included.
It must cover the entire operational area of the vehicle. It is also recommended to include an additional buffer zone according to the detection range of sensors attached to the vehicle.
Its resolution should be at least 0.2 m to yield reliable localization results.
It can be in either local or global coordinates, but must be in global coordinates (georeferenced) to use GNSS data for localization.
For more details on divided map format, please refer to the readme of map_loader in Autoware Universe.
Note
Three global coordinate systems are currently supported by Autoware, including Military Grid Reference System (MGRS), Universal Transverse Mercator (UTM), and Japan Rectangular Coordinate System. However, MGRS is a preferred coordinate system for georeferenced maps. In a map with MGRS coordinate system, the X and Y coordinates of each point represent the point's location within the 100,000-meter square, while the Z coordinate represents the point's elevation.
The vector cloud map must be supplied as a file with the following requirements:
It must be in Lanelet2 format, with additional modifications required by Autoware.
It must contain the shape and position information of lanes, traffic lights, stop lines, crosswalks, parking spaces, and parking lots.
Except at the beginning or end of a road, each lanelet in the map must be correctly connected to its predecessor, successors, left neighbor, and right neighbor.
Each lanelet in the map must contain traffic rule information including its speed limit, right of way, traffic direction, associated traffic lights, stop lines, and traffic signs.
It must cover the entire operational area of the vehicle.
Note that the diagram is for reference. We are planning to update this diagram every release and may have old information between the releases. If you wish to check the latest node diagram use rqt_graph after launching the Autoware.
The Planning component generates the trajectory message that will be subscribed to by the Control component based on the environmental state obtained from the Localization and the Perception components.
The goal of the Planning component is to generate a trajectory (path and velocity) of the ego vehicle that is safe and well-regulated while satisfying the given mission.
Goals:
The basic functions are provided so that a simple ODD can be defined.
The functionality is modularized to accommodate the third-party components. That is, a complicated or realistic ODD needs not be defined by the basic functions provided by default.
The capability is extensible with the third-party components or the decision of human operators.
The mechanism and policy are separated to allow the system or operators to change the behavior of the ego vehicle. Ultimately speaking, the policy can be set to crash an obstacle and the mechanism always follows. Otherwise, the system is not safe from the design point of view.
Non-goals:
The Planning component is not self-contained but can be extended with third parties.
The Planning component is not aimed at the complete functionality and capability.
The Planning component is not designed to always outperform human drivers.
The Planning component is not capable of \u201cnever crashes\u201d.
This diagram describes the high-level architecture of the Planning Component.
The Planning component consists of the following sub-components:
Mission Planning: Calculates the route based on the given goal and map information.
Scenario Planning: Determines the trajectory based on the current scenario, such as Lane Driving or Parking.
Lane Driving: Calculates the trajectory for driving within constructed lanes.
Behavior Planner: Calculates suitable path based on safety considerations and traffic rules.
Motion Planner: Calculates suitable trajectory for the vehicle by taking into account safety factors, vehicle motion considerations, and instructions from the behavior planner.
Parking: Calculates the trajectory for parking in unstructured areas.
Validation: Verifies the safety of the trajectory.
Each component contains some modules that can be dynamically loaded and unloaded based on the situation. For instance, the Behavior Planning component includes modules such as lane change, intersection, and crosswalk modules.
Our planning components are built based on the microautonomy architecture with Autoware. We adopt a modular system framework where the tasks are implemented as modules that can be dynamically loaded and unloaded to achieve different features depending on the given use cases.
This section describes the inputs and outputs of the Planning Component and of its internal modules. See the Planning Component Interface page for the current implementation.
"},{"location":"design/autoware-architecture/planning/#input-to-the-planning-component","title":"Input to the planning component","text":"
From Map
Vector map: Contains all static information about the environment, including lane connection information for route planning, lane geometry for generating a reference path, and traffic rule-related information.
From Perception
Detected object information: Provides real-time information about objects that cannot be known in advance, such as pedestrians and other vehicles. The Planning Component plans maneuvers to avoid collisions with these objects.
Detected obstacle information: Supplies real-time information about the location of obstacles, which is more primitive than Detected Object and used for emergency stops and other safety measures.
Occupancy map information: Offers real-time information about the presence of pedestrians and other vehicles and occluded area information.
Traffic light recognition result: Provides the current state of each traffic light in real time. The Planning Component extracts relevant information for the planned path and determines whether to stop at intersections.
From Localization
Vehicle motion information: Includes the ego vehicle's position, velocity, acceleration, and other motion-related data.
From System
Operation mode: Indicates whether the vehicle is operating in Autonomous mode.
From Human Machine Interface (HMI)
Feature execution: Allows for executing/authorizing autonomous driving operations, such as lane changes or entering intersections, by human operators.
From API Layer
Goal: Represents the final position that the Planning Component aims to reach.
Checkpoint: Represents a midpoint along the route to the destination. This is used during route calculation.
Velocity limit: Sets the maximum speed limit for the vehicle.
"},{"location":"design/autoware-architecture/planning/#output-from-the-planning-component","title":"Output from the planning component","text":"
To Control
Trajectory: Provides a smooth sequence of pose, twist, and acceleration that the Control Component must follow. The trajectory is typically 10 seconds long with a 0.1-second resolution.
Turn Signals: Controls the vehicle's turn indicators, such as right, left, hazard, etc. based on the planned maneuvers.
To System
Diagnostics: Reports the state of the Planning Component, indicating whether the processing is running correctly and whether a safe plan is being generated.
To Human Machine Interface (HMI)
Feature execution availability: Indicates the status of operations that can be executed or are required, such as lane changes or entering intersections.
Trajectory candidate: Shows the potential trajectory that will be executed after the user's execution.
To API Layer
Planning factors: Provides information about the reasoning behind the current planning behavior. This may include the position of target objects to avoid, obstacles that led to the decision to stop, and other relevant information.
"},{"location":"design/autoware-architecture/planning/#internal-interface-in-the-planning-component","title":"Internal interface in the planning component","text":"
Mission Planning to Scenario Planning
Route: Offers guidance for the path that needs to be followed from the starting point to the destination. This path is determined based on information such as lane IDs defined on the map. At the route level, it doesn't explicitly indicate which specific lanes to take, and the route can contain multiple lanes.
Behavior Planning to Motion Planning
Path: Provides a rough position and velocity to be followed by the vehicle. These path points are usually defined with an interval of about 1 meter. Although other interval distances are possible, it may impact the precision or performance of the planning component.
Drivable area: Defines regions where the vehicle can drive, such as within lanes or physically drivable areas. It assumes that the motion planner will calculate the final trajectory within this defined area.
Scenario Planning to Validation
Trajectory: Defines the desired positions, velocities, and accelerations which the Control Component will try to follow. Trajectory points are defined at intervals of approximately 0.1 seconds based on the trajectory velocities.
Validation to Control Component
Trajectory: Same as above but with some additional safety considerations.
"},{"location":"design/autoware-architecture/planning/#how-to-add-new-modules-wip","title":"How to add new modules (WIP)","text":"
As mentioned in the goal session, this planning module is designed to be extensible by third-party components. For specific instructions on how to add new modules and expand its functionality, please refer to the provided documentation or guidelines (WIP).
"},{"location":"design/autoware-architecture/planning/#supported-functions","title":"Supported Functions","text":"Feature Description Requirements Figure Route Planning Plan route from the ego vehicle position to the destination. Reference implementation is in Mission Planner, enabled by launching the mission_planner node. - Lanelet map (driving lanelets) Path Planning from Route Plan path to be followed from the given route. Reference implementation is in Behavior Path Planner. - Lanelet map (driving lanelets) Obstacle Avoidance Plan path to avoid obstacles by steering operation. Reference implementation is in Avoidance, Obstacle Avoidance Planner. Enable flag in parameter: launch obstacle_avoidance_planner true Demonstration Video - objects information Path Smoothing Plan path to achieve smooth steering. Reference implementation is in Obstacle Avoidance Planner. Demonstration Video - Lanelet map (driving lanelet) Narrow Space Driving Plan path to drive within the drivable area. Furthermore, when it is not possible to drive within the drivable area, stop the vehicle to avoid exiting the drivable area. Reference implementation is in Obstacle Avoidance Planner. Demonstration Video - Lanelet map (high-precision lane boundaries) Lane Change Plan path for lane change to reach the destination. Reference implementation is in Lane Change. Demonstration Video - Lanelet map (driving lanelets) Pull Over Plan path for pull over to park at the road shoulder. Reference implementation is in Goal Planner. Demonstration Videos: Simple Pull Over Arc Forward Pull Over Arc Backward Pull Over - Lanelet map (shoulder lane) Pull Out Plan path for pull over to start from the road shoulder. Reference implementation is in Pull Out Module. Demonstration Video: Simple Pull Out Backward Pull Out - Lanelet map (shoulder lane) Path Shift Plan path in lateral direction in response to external instructions. Reference implementation is in Side Shift Module. - None Obstacle Stop Plan velocity to stop for an obstacle on the path. Reference implementation is in Obstacle Stop Planner, Obstacle Cruise Planner. launch obstacle_stop_planner and enable flag: TODO, launch obstacle_cruise_planner and enable flag: TODO Demonstration Video - objects information Obstacle Deceleration Plan velocity to decelerate for an obstacle located around the path. Reference implementation is in Obstacle Stop Planner, Obstacle Cruise Planner. Demonstration Video - objects information Adaptive Cruise Control Plan velocity to follow the vehicle driving in front of the ego vehicle. Reference implementation is in Obstacle Stop Planner, Obstacle Cruise Planner. - objects information Decelerate for cut-in vehicles Plan velocity to avoid a risk for cutting-in vehicle to ego lane. Reference implementation is in Obstacle Cruise Planner. - objects information Surround Check at starting Plan velocity to prevent moving when an obstacle exists around the vehicle. Reference implementation is in Surround Obstacle Checker. Enable flag in parameter: use_surround_obstacle_check true in tier4_planning_component.launch.xml Demonstration Video - objects information Curve Deceleration Plan velocity to decelerate the speed on a curve. Reference implementation is in Motion Velocity Smoother. - None Curve Deceleration for Obstacle Plan velocity to decelerate the speed on a curve for a risk of obstacle collision around the path. Reference implementation is in Obstacle Velocity Limiter. Demonstration Video - objects information - Lanelet map (static obstacle) Crosswalk Plan velocity to stop or decelerate for pedestrians approaching or walking on a crosswalk. Reference implementation is in Crosswalk Module. Demonstration Video - objects information - Lanelet map (pedestrian crossing) Intersection Oncoming Vehicle Check Plan velocity for turning right/left at intersection to avoid a risk with oncoming other vehicles. Reference implementation is in Intersection Module. Demonstration Video - objects information - Lanelet map (intersection lane and yield lane) Intersection Blind Spot Check Plan velocity for turning right/left at intersection to avoid a risk with other vehicles or motorcycles coming from behind blind spot. Reference implementation is in Blind Spot Module. Demonstration Video - objects information - Lanelet map (intersection lane) Intersection Occlusion Check Plan velocity for turning right/left at intersection to avoid a risk with the possibility of coming vehicles from occlusion area. Reference implementation is in Intersection Module. Demonstration Video - objects information - Lanelet map (intersection lane) Intersection Traffic Jam Detection Plan velocity for intersection not to enter the intersection when a vehicle is stopped ahead for a traffic jam. Reference implementation is in Intersection Module. Demonstration Video - objects information - Lanelet map (intersection lane) Traffic Light Plan velocity for intersection according to a traffic light signal. Reference implementation is in Traffic Light Module. Demonstration Video - Traffic light color information Run-out Check Plan velocity to decelerate for the possibility of nearby objects running out into the path. Reference implementation is in Run Out Module. Demonstration Video - objects information Stop Line Plan velocity to stop at a stop line. Reference implementation is in Stop Line Module. Demonstration Video - Lanelet map (stop line) Occlusion Spot Check Plan velocity to decelerate for objects running out from occlusion area, for example, from behind a large vehicle. Reference implementation is in Occlusion Spot Module. Demonstration Video - objects information - Lanelet map (private/public lane) No Stop Area Plan velocity not to stop in areas where stopping is prohibited, such as in front of the fire station entrance. Reference implementation is in No Stopping Area Module. - Lanelet map (no stopping area) Merge from Private Area to Public Road Plan velocity for entering the public road from a private driveway to avoid a risk of collision with pedestrians or other vehicles. Reference implementation is in Merge from Private Area Module. - objects information - Lanelet map (private/public lane) WIP Speed Bump Plan velocity to decelerate for speed bumps. Reference implementation is in Speed Bump Module. Demonstration Video - Lanelet map (speed bump) Detection Area Plan velocity to stop at the corresponding stop when an object exist in the designated detection area. Reference implementation is in Detection Area Module. Demonstration Video - Lanelet map (detection area) No Drivable Lane Plan velocity to stop before exiting the area designated by ODD (Operational Design Domain) or stop the vehicle if autonomous mode started in out of ODD lane. Reference implementation is in No Drivable Lane Module. - Lanelet map (no drivable lane) Collision Detection when deviating from lane Plan velocity to avoid conflict with other vehicles driving in the another lane when the ego vehicle is deviating from own lane. Reference implementation is in Out of Lane Module. - objects information - Lanelet map (driving lane) WIP Parking Plan path and velocity for given goal in parking area. Reference implementation is in Free Space Planner. Demonstration Video - objects information - Lanelet map (parking area) Autonomous Emergency Braking (AEB) Perform an emergency stop if a collision with an object ahead is anticipated. It is noted that this function is expected as a final safety layer, and this should work even in the event of failures in the Localization or Perception system. Reference implementation is in Out of Lane Module. - Primitive objects Minimum Risk Maneuver (MRM) Provide appropriate MRM (Minimum Risk Maneuver) instructions when a hazardous event occurs. For example, when a sensor trouble found, send an instruction for emergency braking, moderate stop, or pulling over to the shoulder, depending on the severity of the situation. Reference implementation is in TODO - TODO WIP Trajectory Validation Check the planned trajectory is safe. If it is unsafe, take appropriate action, such as modify the trajectory, stop sending the trajectory or report to the autonomous driving system. Reference implementation is in Planning Validator. - None Running Lane Map Generation Generate lane map from localization data recorded in manual driving. Reference implementation is in WIP - None WIP Running Lane Optimization Optimize the centerline (reference path) of the map to make it smooth considering the vehicle kinematics. Reference implementation is in Static Centerline Optimizer. - Lanelet map (driving lanes) WIP"},{"location":"design/autoware-architecture/planning/#reference-implementation","title":"Reference Implementation","text":"
The following diagram describes the reference implementation of the Planning component. By adding new modules or extending the functionalities, various ODDs can be supported.
Note that some implementation does not adhere to the high-level architecture design and require updating.
For more details, please refer to the design documents in each package.
mission_planner: calculate route from start to goal based on the map information.
behavior_path_planner: calculates path and drivable area based on the traffic rules.
lane_following
lane_change
avoidance
pull_over
pull_out
side_shift
behavior_velocity_planner: calculates max speed based on the traffic rules.
detection_area
blind_spot
cross_walk
stop_line
traffic_light
intersection
no_stopping_area
virtual_traffic_light
occlusion_spot
run_out
no_drivable_lane
obstacle_avoidance_planner: calculate path shape under obstacle and drivable area constraints
surround_obstacle_checker: keeps the vehicle being stopped when there are obstacles around the ego-vehicle. It works only when the vehicle is stopped.
obstacle_stop_planner: When there are obstacles on or near the trajectory, it calculates the maximum velocity of the trajectory points depending on the situation: stopping, slowing down, or adaptive cruise (following the car).
stop
slow_down
adaptive_cruise
costmap_generator: generates a costmap for path generation from dynamic objects and lane information.
freespace_planner: calculates trajectory considering the feasibility (e.g. curvature) for the freespace scene. Algorithms are described here.
scenario_selector : chooses a trajectory according to the current scenario.
external_velocity_limit_selector: takes an appropriate velocity limit from multiple candidates.
motion_velocity_smoother: calculates final velocity considering velocity, acceleration, and jerk constraints.
"},{"location":"design/autoware-architecture/planning/#important-parameters","title":"Important Parameters","text":"Package Parameter Type Description obstacle_stop_plannerstop_planner.stop_position.max_longitudinal_margin double distance between the ego and the front vehicle when stopping (when cruise_planner_type:=obstacle_stop_planner) obstacle_cruise_plannercommon.safe_distance_margin double distance between the ego and the front vehicle when stopping (when cruise_planner_type:=obstacle_cruise_planner) behavior_path_planneravoidance.avoidance.lateral.lateral_collision_margin double minimum lateral margin to obstacle on avoidance behavior_path_planneravoidance.avoidance.lateral.lateral_collision_safety_buffer double additional lateral margin to obstacle if possible on avoidance obstacle_avoidance_planneroption.enable_outside_drivable_area_stop bool If set true, a stop point will be inserted before the path footprint is outside the drivable area."},{"location":"design/autoware-architecture/planning/#notation","title":"Notation","text":""},{"location":"design/autoware-architecture/planning/#1-self-crossing-road-and-overlapped","title":"[1] self-crossing road and overlapped","text":"
To support the self-crossing road and overlapped road in the opposite direction, each planning module has to meet the specifications
Currently, the supported modules are as follows.
lane_following (in behavior_path_planner)
detection_area (in behavior_velocity_planner)
stop_line (in behavior_velocity_planner)
virtual_traffic_light (in behavior_velocity_planner)
obstacle_avoidance_planner
obstacle_stop_planner
motion_velocity_smoother
"},{"location":"design/autoware-architecture/planning/#2-size-of-path-points","title":"[2] Size of Path Points","text":"
Some functions do not support paths with only one point. Therefore, each modules should generate the path with more than two path points.
Abstraction of data formats to enable usage of sensors from various vendors
Perform common/primitive sensor data processing required by each component
"},{"location":"design/autoware-architecture/sensing/#inputs","title":"Inputs","text":""},{"location":"design/autoware-architecture/sensing/#input-types","title":"Input types","text":"Sensor Data Message Type Point cloud (Lidars, depth cameras, etc.) sensor_msgs/msg/PointCloud2.msg Image (RGB, monochrome, depth, etc. cameras) sensor_msgs/msg/Image.msg Radar scan radar_msgs/msg/RadarScan.msg Radar tracks radar_msgs/msg/RadarTracks.msg GNSS-INS position sensor_msgs/msg/NavSatFix.msg GNSS-INS orientation autoware_sensing_msgs/GnssInsOrientationStamped.msg GNSS-INS velocity geometry_msgs/msg/TwistWithCovarianceStamped.msg GNSS-INS acceleration geometry_msgs/msg/AccelWithCovarianceStamped.msg Ultrasonics sensor_msgs/msg/Range.msg"},{"location":"design/autoware-architecture/sensing/#design-by-data-types","title":"Design by data-types","text":"
GNSS/INS data pre-processing design
Image pre-processing design
Point cloud pre-processing design
Radar data pre-processing design
Ultrasonics data pre-processing design
"},{"location":"design/autoware-architecture/sensing/data-types/gnss-ins-data/","title":"GNSS/INS data pre-processing design","text":""},{"location":"design/autoware-architecture/sensing/data-types/gnss-ins-data/#gnssins-data-pre-processing-design","title":"GNSS/INS data pre-processing design","text":"
In the ideal case, the driver is expected to output a point cloud with the PointXYZIRCADT point type.
name datatype derived description XFLOAT32false X position YFLOAT32false Y position ZFLOAT32false Z position I (intensity) UINT8false Measured reflectivity, intensity of the point R (return type) UINT8false Laser return type for dual return lidars C (channel) UINT16false Vertical channel id of the laser that measured the point A (azimuth) FLOAT32trueatan2(Y, X), Horizontal angle from the front of the lidar to the point D (distance) FLOAT32truehypot(X, Y, Z), Euclidean distance of the point to lidar T (time) UINT32false Nanoseconds passed since the time of the header when this point was measured
Note
A (azimuth) and D (distance) fields are derived fields. They are provided by the driver to reduce the computational load on some parts of the perception stack.
Note
If the Motion Distortion Corrector Filter won't be used, the T (time) field can be omitted, PointXYZIRCAD point type can be used.
Warning
Autoware will support conversion from PointXYZI to PointXYZIRC or PointXYZIRCAD (with channel and return is set to 0) for prototyping purposes. However, this conversion is not recommended for production use since it's not efficient.
We will use following ranges for intensity, compatible with the VLP16 User Manual:
Quoting from the VLP-16 User Manual:
For each laser measurement, a reflectivity byte is returned in addition to distance. Reflectivity byte values are segmented into two ranges, allowing software to distinguish diffuse reflectors (e.g. tree trunks, clothing) in the low range from retroreflectors (e.g. road signs, license plates) in the high range. A retroreflector reflects light back to its source with a minimum of scattering. The VLP-16 provides its own light, with negligible separation between transmitting laser and receiving detector, so retroreflecting surfaces pop with reflected IR light compared to diffuse reflectors that tend to scatter reflected energy.
Diffuse reflectors report values from 0 to 100 for reflectivities from 0% to 100%.
Retroreflectors report values from 101 to 255, where 255 represents an ideal reflection.
In a typical point cloud without retroreflectors, all intensity points will be between 0 and 100.
Retroreflective Gradient road sign, Image Source
But in a point cloud with retroreflectors, the intensity points will be between 0 and 255.
"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#intensity-mapping-for-other-lidar-brands","title":"Intensity mapping for other lidar brands","text":""},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#hesai-pandarxt16","title":"Hesai PandarXT16","text":"
Hesai Pandar XT16 User Manual
This lidar has 2 modes for reporting reflectivity:
Linear mapping
Non-linear mapping
If you are using linear mapping mode, you should map from [0, 255] to [0, 100] when constructing the point cloud.
If you are using non-linear mapping mode, you should map (hesai to autoware)
Software User Manual v2.0.0 for all Ouster sensors
In the manual it is stated:
Reflectivity [16 bit unsigned int] - sensor Signal Photons measurements are scaled based on measured range and sensor sensitivity at that range, providing an indication of target reflectivity. Calibration of this measurement has not currently been rigorously implemented, but this will be updated in a future firmware release.
So it is advised to map the 16 bit reflectivity to [0, 100] range.
I couldn't get the english user manual, link of website
In a user manual I was able to find it says:
Byte 7 represents echo strength, and the value range is 0-255. (Echo strength can reflect the energy reflection characteristics of the measured object in the actual measurement environment. Therefore, the echo strength can be used to distinguish objects with different reflection characteristics.)
So it is advised to map the [0, 255] to [0, 100] range.
Various lidars support multiple return modes. Velodyne lidars support Strongest and Last return modes.
In the PointXYZIRCT and PointXYZIRC types, R field represents return mode with an UINT8.
R (return type) Description 0 Unknown / Not Marked 1 Strongest 2 Last"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#channel","title":"Channel","text":"
The channel field is used to identify the vertical channel of the laser that measured the point. In various lidar manuals or literature, it can also be called laser id, ring, laser line.
For Velodyne VLP-16, there are 16 channels. Default order of channels in drivers are generally in firing order.
In the PointXYZIRCT and PointXYZIRC types, C field represents the vertical channel id with an UINT16.
"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#solid-state-and-petal-pattern-lidars","title":"Solid state and petal pattern lidars","text":"
Warning
This section is subject to change. Following are suggestions and open for discussion.
For solid state lidars that have lines, assign row number as the channel id.
In lidar point clouds, each point measurement can have its individual time stamp. This information can be used to eliminate the motion blur that is caused by the movement of the lidar during the scan.
The header contains a Time field. The time field has 2 components:
Field Type Description secint32 Unix time (seconds elapsed since January 1, 1970) nanosecuint32 Nanoseconds elapsed since the sec field
The header of the point cloud message is expected to have the time of the earliest point it has.
Note
The sec field is int32 in ROS 2 humble. The largest value it can represent is 2^31 seconds, it is subject to year 2038 problems. We will wait for actions on ROS 2 community side.
More info at: https://github.com/ros2/rcl_interfaces/issues/85
"},{"location":"design/autoware-architecture/sensing/data-types/point-cloud/#individual-point-time","title":"Individual point time","text":"
Each PointXYZIRCT point type has the T field for representing the nanoseconds passed since the first-shot point of the point cloud.
To calculate exact time each point was shot, the T nanoseconds are added to the header time.
Note
The T field is uint32 type. The largest value it can represent is 2^32 nanoseconds, which equates to roughly 4.29 seconds. Usual point clouds don't last more than 100ms for full cycle. So this field should be enough.
"},{"location":"design/autoware-architecture/sensing/data-types/radar-data/","title":"Radar data pre-processing design","text":""},{"location":"design/autoware-architecture/sensing/data-types/radar-data/#radar-data-pre-processing-design","title":"Radar data pre-processing design","text":"
Warning
Under Construction
"},{"location":"design/autoware-architecture/sensing/data-types/ultrasonics-data/","title":"Ultrasonics data pre-processing design","text":""},{"location":"design/autoware-architecture/sensing/data-types/ultrasonics-data/#ultrasonics-data-pre-processing-design","title":"Ultrasonics data pre-processing design","text":"
The Vehicle Interface component provides an interface between Autoware and a vehicle that passes control signals to the vehicle\u2019s drive-by-wire system and receives vehicle information that is passed back to Autoware.
The Vehicle Interface component converts Autoware commands to a vehicle-specific format and converts vehicle status in a vehicle-specific format to Autoware messages.
The interface between Autoware and the Vehicle component is abstracted and independent of hardware.
The interface is extensible such that additional vehicle-specific commands can be easily added. For example, headlight control.
Non-goals:
Accuracy of responses from the vehicle will not be defined, but example accuracy requirements from reference designs are provided as examples.
The Vehicle Interface component consists of the following components:
A Raw Vehicle Command Converter component that will pass through vehicle commands from the Control component if velocity/acceleration control is supported by the drive-by-wire system. Otherwise, the Control commands will be modified according to the control method (eg: converting a target acceleration from the Control component to a vehicle specific accel/brake pedal value through the use of an acceleration map)
A Vehicle Interface component (vehicle specific) that acts as an interface between Autoware and a vehicle to communicate control signals and to obtain information about the vehicle (steer output, tyre angle etc)
Each component contains static nodes of Autoware, while each module can be dynamically loaded and unloaded (corresponding to C++ classes). The mechanism of the Vehicle Interface component is depicted by the following figures:
The Vehicle Interface component can provide the following features in functionality and capability:
Basic functions
Converting Autoware control commands to vehicle specific command
Converting vehicle specific status information (velocity, steering) to Autoware status message
Diagnostics
List available features
Provide a warning if the Control component tries to use a feature that is not available in the Vehicle Interface component
Additional functionality and capability features may be added, depending on the vehicle hardware. Some example features are listed below:
Safety features
Disengage autonomous driving via manual intervention.
This can be done through the use of an emergency disengage button, or by a safety driver manually turning the steering wheel or pressing the brake
Optional controls
Turn indicator
Handbrake
Headlights
Hazard lights
Doors
Horn
Wipers
"},{"location":"design/autoware-architecture/vehicle/#4-interface-and-data-structure","title":"4. Interface and Data Structure","text":"
The interface of the Vehicle Interface component for other components running in the same process space to access the functionality and capability of the Vehicle Interface component is defined as follows.
From Control
Actuation Command
target acceleration, braking, and steering angle
From Planning
Vehicle Specific Commands (optional and a separate message for each type)
Shift
Door
Wiper
etc
From the vehicle
Vehicle status messages
Vehicle-specific format messages for conversion into Autoware-specific format messages
Velocity status
Steering status (optional)
Shift status (optional)
Turn signal status (optional)
Actuation status (optional)
The output interface of the Vehicle Interface component:
Vehicle control messages to the vehicle
Control signals to drive the vehicle
Depends on the vehicle type/protocol, but should include steering and velocity commands at a minimum
Vehicle status messages to Autoware
Actuation Status
Acceleration, brake, steering status
Vehicle odometry (output to Localization)
Vehicle twist information
Control mode
Information about whether the vehicle is under autonomous control or manual control
Shift status (optional)
Vehicle shift status
Turn signal status (optional)
Vehicle turn signal status
The data structure for the internal representation of semantics for the objects and trajectories used in the Vehicle Interface component is defined as follows:
"},{"location":"design/autoware-architecture/vehicle/#5-concerns-assumptions-and-limitations","title":"5. Concerns, Assumptions, and Limitations","text":"
Concerns
Architectural trade-offs and scalability
Assumptions
-
Limitations
"},{"location":"design/autoware-architecture/vehicle/#6-examples-of-accuracy-requirements-by-odd","title":"6. Examples of accuracy requirements by ODD","text":""},{"location":"design/autoware-concepts/","title":"Autoware concepts","text":""},{"location":"design/autoware-concepts/#autoware-concepts","title":"Autoware concepts","text":"
Autoware is the world\u2019s first open-source software for autonomous driving systems. Autoware provides value for both The technology developers of autonomous driving systems can create new components based on Autoware. The service operators of autonomous driving systems, on the other hand, can select appropriate technology components with Autoware. This is enabled by the microautonomy architecture that modularizes its software stack into the core and universe subsystems (modules).
Autoware uses a pipeline architecture to enable the development of autonomous driving systems. The pipeline architecture used in Autoware consists of components similar to three-layer-architecture. And they run in parallel. There are 2 main modules: the Core and the Universe. The components in these modules are designed to be extensible and reusable. And we call it microautonomy architecture.
The Core module contains basic runtimes and technology components that satisfy the basic functionality and capability of sensing, computing, and actuation required for autonomous driving systems. AWF develops and maintains the Core module with their architects and leading members through their working groups. Anyone can contribute to the Core but the PR(Pull Request) acceptance criteria is more strict compared to the Universe.
The Universe modules are extensions to the Core module that can be provided by the technology developers to enhance the functionality and capability of sensing, computing, and actuation. AWF provides the base Universe module to extend from. A key feature of the microautonomy architecture is that the Universe modules can be contributed to by any organization and individual. That is, you can even create your Universe and make it available for the Autoware community and ecosystem. AWF is responsible for quality control of the Universe modules through their development process. As a result, there are multiple types of the Universe modules - some are verified and validated by AWF and others are not. It is up to the users of Autoware which Universe modules are selected and integrated to build their end applications.
The interface design is the most essential piece of the microautonomy architecture, which is classified into internal and external interfaces. The component interface is designed for the components in a Universe module to communicate with those in other modules, including the Core module, within Autoware internally. The AD(Autonomous Driving) API, on the other hand, is designed for the applications of Autoware to access the technology components in the Core and Universe modules of Autoware externally. Designing solid interfaces, the microautonomy architecture is made possible with AWF's partners, and at the same time is made feasible for the partners.
A grand challenge of the microautonomy architecture is to achieve real-time capability, which guarantees all the technology components activated in the system to predictably meet timing constraints (given deadlines). In general, it is difficult, if not impossible, to tightly estimate the worst-case execution times (WCETs) of components.
In addition, it is also difficult, if not impossible, to tightly estimate the end-to-end latency of components connected by a DAG. Autonomous driving systems based on the microautonomy architecture, therefore, must be designed to be fail-safe but not never-fail. We accept that the timing constraints may be violated (the given deadlines may be missed) as far as the overrun is taken into account. The overrun handlers are two-fold: (i) platform-defined and (ii) user-defined. The platform-defined handler is implemented as part of the platform by default, while the user-defined handler can overwrite it or add a new handler to the system. This is what we call \u201cfail-safe\u201d on a timely basis.
"},{"location":"design/autoware-concepts/#requirements-and-roadmap","title":"Requirements and roadmap","text":"
Goals:
All open-source
Use case driven
Real-time (predictable) framework with overrun handling
Code quality
"},{"location":"design/autoware-concepts/difference-from-ai-and-auto/","title":"How is Autoware Core/Universe different from Autoware.AI and Autoware.Auto?","text":""},{"location":"design/autoware-concepts/difference-from-ai-and-auto/#how-is-autoware-coreuniverse-different-from-autowareai-and-autowareauto","title":"How is Autoware Core/Universe different from Autoware.AI and Autoware.Auto?","text":"
Autoware is the world's first \"all-in-one\" open-source software for self-driving vehicles. Since it was first released in 2015, there have been multiple releases made with differing underlying concepts, each one aimed at improving the software.
Autoware.AI is the first distribution of Autoware that was released based on ROS 1. The repository contains a variety of packages covering different aspects of autonomous driving technologies - sensing, actuation, localization, mapping, perception and planning.
While it was successful in attracting many developers and contributions, it was difficult to improve Autoware.AI's capabilities for a number of reasons:
A lack of concrete architecture design leading to a lot of built-up technical debt, such as tight coupling between modules and unclear module responsibility.
Differing coding standards for each package, with very low test coverage.
Furthermore, there was no clear definition of the conditions under which an Autoware-enabled autonomous vehicle could operate, nor of the use cases or situations supported (eg: the ability to overtake a stationary vehicle).
From the lessons learned from Autoware.AI development, a different development process was taken for Autoware.Auto to develop a ROS 2 version of Autoware.
Warning
Autoware.AI is currently in maintenance mode and will reach end-of-life at the end of 2022.
Autoware.Auto is the second distribution of Autoware that was released based on ROS 2. As part of the transition to ROS 2, it was decided to avoid simply porting Autoware.AI from ROS 1 to ROS 2. Instead, the codebase was rewritten from scratch with proper engineering practices, including defining target use cases and ODDs (eg: Autonomous Valet Parking [AVP], Cargo Delivery, etc.), designing a proper architecture, writing design documents and test code.
Autoware.Auto development seemed to work fine initially, but after completing the AVP and and Cargo Delivery ODD projects, we started to see the following issues:
The barrier to new engineers was too high.
A lot of work was required to merge new features into Autoware.Auto, and so it was difficult for researchers and students to contribute to development.
As a consequence, most Autoware.Auto developers were from companies in the Autoware Foundation and so there were very few people who were able to add state-of-the-art features from research papers.
Making large-scale architecture changes was too difficult.
To try out experimental architecture, there was a very large overhead involved in keeping the main branch stable whilst also making sure that every change satisfied the continuous integration requirements.
In order to address the issues with Autoware.Auto development, the Autoware Foundation decided to create a new architecture called Autoware Core/Universe.
Autoware Core carries over the original policy of Autoware.Auto to be a stable and well-tested codebase. Alongside Autoware Core is a new concept called Autoware Universe, which acts as an extension of Autoware Core with the following benefits:
Users can easily replace a Core component with a Universe equivalent in order to use more advanced features, such as a new Localization or Perception algorithm.
Code quality requirements for Universe are more relaxed to make it easier for new developers, students and researchers to contribute, but will still be stricter than the requirements for Autoware.AI.
Any advanced features added to Universe that are useful to the wider Autoware community will be reviewed and considered for potential inclusion in the main Autoware Core codebase.
This way, the primary requirement of having a stable and safe autonomous driving system can be achieved, whilst simultaneously enabling access to state-of-the-art features created by third-party contributors. For more details about the design of Autoware Core/Universe, refer to the Autoware concepts documentation page.
Autoware defines three categories of interfaces. The first one is Autoware AD API for operating the vehicle from outside the autonomous driving system such as the Fleet Management System (FMS) and Human Machine Interface (HMI) for operators or passengers. The second one is Autoware component interface for components to communicate with each other. The last one is the local interface used inside the component.
The components of Autoware are connected via the component interface. Each component uses the interface to provide functionality and to access other components. AD API implementation is also a component. Since the functional elements required for AD API are defined as the component interface, other components do not need to consider AD API directly. Tools for evaluation and debugging, such as simulators, access both AD API and the component interface.
The component interface has a hierarchical specification. The top-level architecture consists of some components. Each component has some options of the next-level architecture. Developers select one of them when implementing the component. The simplest next-level architecture is monolithic. This is an all-in-one and black box implementation, and is suitable for small group development, prototyping, and very complex functions. Others are arbitrary architecture consists of sub-components and have advantages for large group development. A sub-component can be combined with others that adopt the same architecture. Third parties can define and publish their own architecture and interface for open source development. It is desirable to propose them for standardization if they are sufficiently evaluated.
As shown in the table below, interfaces are classified into four communication methods to define their behavior. Function Call is a request-response communication and is used for processing that requires immediate results. The others are publish-subscribe communication. Notification is used to process data that changes with some event, typically a callback. Streams handle continuously changing data. Reliable Stream expects all data to arrive without loss, Realtime Stream expects the latest data to arrive with low delay.
These methods are provided as services or topics of ROS since Autoware is developed using ROS and mainly communicates with its packages. On the other hand, FMS and HMI are often implemented without ROS, Autoware is also expected to communicate with applications that do not use ROS. It is wasteful for each of these applications to have an adapter for Autoware, and a more suitable means of communication is required. HTTP and MQTT are suggested as additional options because these protocols are widely used and can substitute the behavior of services and topics. In that case, text formats such as JSON where field names are repeated in an array of objects, are inefficient and it is necessary to consider the serialization.
The name of the interface must be /<component name>/api/<interface name>, where <component name> is the name of the component. For an AD API component, omit this part and start with /api. The <interface name> is an arbitrary string separated by slashes. Note that this rule causes a restriction that the namespace api must not be used as a name other than AD API and the component interface.
The following are examples of correct interface names for AD API and the component interface:
/api/autoware/state
/api/autoware/engage
/planning/api/route/set
/vehicle/api/status
The following are examples of incorrect interface names for AD API and the component interface:
It is recommended to log the interface for analysis of vehicle behavior. If logging is needed, rosbag is available for topics, and use logger in rclcpp or rclpy for services. Typically, create a wrapper for service and client classes that logs when a service is called.
For each API, consider the restrictions such as following and describe them if necessary.
Services:
response time
pre-condition
post-condition
execution order
concurrent execution
Topics:
recommended delay range
maximum delay
recommended frequency range
minimum frequency
default frequency
"},{"location":"design/autoware-interfaces/#data-structure","title":"Data structure","text":""},{"location":"design/autoware-interfaces/#data-type-definition","title":"Data type definition","text":"
Do not share the types in AD API unless they are obviously the same to avoid changes in one API affecting another. Also, implementation-dependent types, including the component interface, should not be used in AD API for the same reason. Use the type in AD API in implementation, or create the same type and copy the data to convert the type.
"},{"location":"design/autoware-interfaces/#constants-and-enumeration","title":"Constants and enumeration","text":"
Since ROS don't support enumeration, use constants instead. The default value of type such as zero and empty string should not be used to detect that a variable is unassigned. Alternatively, assign it a dedicated name to indicate that it is undefined. If one type has multiple enumerations, comment on the correspondence between constants and variables. Do not use enumeration values directly, as assignments are subject to change when the version is updated.
Clarify what the timestamp indicates. for example, send time, measurement time, update time, etc. Consider having multiple timestamps if necessary. Use std_msgs/msg/Header when using ROS transform. Also consider whether the header is common to all data, independent for each data, or additional timestamp is required.
The interfaces whose communication method is Function Call use a common response status to unify the error format. These interfaces should include a variable of ResponseStatus with the name status in the response. See autoware_adapi_v1_msgs/msg/ResponseStatus for details.
"},{"location":"design/autoware-interfaces/#concerns-assumptions-and-limitations","title":"Concerns, assumptions and limitations","text":"
The applications use the version information provided by AD API to check compatibility. Unknown versions are also treated as available as long as the major versions match (excluding major version 0). Compatibility between AD API and the component interface is assumed to be maintained by the version management system.
If an unintended behavior of AD API is detected, the application should take appropriate action. Autoware tries to keep working as long as possible, but it is not guaranteed to be safe. Safety should be considered for the entire system, including the applications.
"},{"location":"design/autoware-interfaces/ad-api/","title":"Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/#autoware-ad-api","title":"Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/#overview","title":"Overview","text":"
Autoware AD API is the interface for operating the vehicle from outside the autonomous driving system. See here for the overall interface design of Autoware.
The user stories are service scenarios that AD API assumes. AD API is designed based on these scenarios. Each scenario is realized by a combination of use cases described later. If there are scenarios that cannot be covered, please discuss adding a user story.
Use cases are partial scenarios derived from the user story and generically designed. Service providers can combine these use cases to define user stories and check if AD API can be applied to their own scenarios.
Request to cooperate (RTC) is a feature that enables a human operator to support the decision in autonomous driving mode. Autoware usually drives the vehicle using its own decisions, but the operator may prefer to make their decisions in experiments and complex situations.
The planning component manages each situation that requires decision as a scene. Each scene has an ID that doesn't change until the scene is completed or canceled. The operator can override the decision of the target scene using this ID. In practice, the user interface application can hides the specification of the ID and provides an abstracted interface to the operator.
For example, in the situation in the diagram below, vehicle is expected to make two lane changes and turning left at the intersection. Therefore the planning component generates three scene instances for each required action, and each scene instance will wait for the decision to be made, in this case \"changing or keeping lane\" and \"turning left or waiting at the intersection\". Here Autoware decides not to change lanes a second time due to the obstacle, so the vehicle will stop there. However, operator could overwrite that decision through RTC function and force the lane change so that vehicle could reach to it's goal. Using RTC, the operator can override these decisions to continue driving the vehicle to the goal.
Modules that support RTC have the operator decision and cooperation policy in addition to the module decision as shown below. These modules use the merged decision that is determined by these values when planning vehicle behavior. See decisions section for details of these values. The cooperation policy is used when there is no operator decision and has a default value set by the system settings. If the module supports RTC, these information are available in velocity factors or steering factors as cooperation status.
This is an example sequence that overrides the scene decision to force a lane change. It is for the second scene in the diagram in the architecture section. Here let's assume the cooperation policy is set to optional, see the decisions section described later for details.
A planning module creates a scene instance with unique ID when approaching a place where a lane change is needed.
The scene instance generates the module decision from the current situation. In this case, the module decision is not to do a lane change due to the obstacle.
The scene instance generates the merged decision. At this point, there is no operator decision yet, so it is based on the module decision.
The scene instance plans the vehicle to keep the lane according to the merged decision.
The scene instance sends a cooperation status.
The operator receives the cooperation status.
The operator sends a cooperation command to override the module decision and to do a lane change.
The scene instance receives the cooperation command and update the operator decision.
The scene instance updates the module decision from the current situation.
The scene instance updates the merged decision. It is based on the operator decision received.
The scene instance plans the vehicle to change the lane according to the merged decision.
The merged decision is determined by the module decision, operator decision, and cooperation policy, each of which takes the value shown in the table below.
The meanings of these values are as follows. Note that the cooperation policy is common per module, so changing it will affect all scenes in the same module.
Value Description deactivate An operator/module decision to plan vehicle behavior with priority on safety. activate An operator/module decision to plan vehicle behavior with priority on driving. autonomous An operator decision that follows the module decision. none An initial value for operator decision, indicating that there is no operator decision yet. required A policy that requires the operator decision to continue driving. optional A policy that does not require the operator decision to continue driving.
The following flow is how the merged decision is determined.
This is an example of cooperation for lane change module. The behaviors by the combination of decisions are as follows.
Operator decision Policy Module decision Description deactivate - - The operator instructs to keep lane regardless the module decision. So the vehicle keeps the lane by the operator decision. activate - - The operator instructs to change lane regardless the module decision. So the vehicle changes the lane by the operator decision. autonomous - deactivate The operator instructs to follow the module decision. So the vehicle keeps the lane by the module decision. autonomous - activate The operator instructs to follow the module decision. So the vehicle changes the lane by the module decision. none required - The required policy is used because no operator instruction. So the vehicle keeps the lane by the cooperation policy. none optional deactivate The optional policy is used because no operator instruction. So the vehicle keeps the lane by the module decision. none optional activate The optional policy is used because no operator instruction. So the vehicle change the lane by the module decision."},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/","title":"Fail-safe","text":""},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/#fail-safe","title":"Fail-safe","text":""},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/#related-api","title":"Related API","text":"
This API manages the behavior related to the abnormality of the vehicle. It provides the state of Request to Intervene (RTI), Minimal Risk Maneuver (MRM) and Minimal Risk Condition (MRC). As shown below, Autoware has the gate to switch between the command during normal operation and the command during abnormal operation. For safety, Autoware switches the operation to MRM when an abnormality is detected. Since the required behavior differs depending on the situation, MRM is implemented in various places as a specific mode in a normal module or as an independent module. The fail-safe module selects the behavior of MRM according to the abnormality and switches the gate output to that command.
The MRM state indicates whether MRM is operating. This state also provides success or failure. Generally, MRM will switch to another behavior if it fails.
State Description NONE MRM is not operating. OPERATING MRM is operating because an abnormality has been detected. SUCCEEDED MRM succeeded. The vehicle is in a safe condition. FAILED MRM failed. The vehicle is still in an unsafe condition."},{"location":"design/autoware-interfaces/ad-api/features/fail-safe/#behavior","title":"Behavior","text":"
There is a dependency between MRM behaviors. For example, it switches from a comfortable stop to a emergency stop, but not the other way around. This is service dependent. Autoware supports the following transitions by default.
State Description NONE MRM is not operating or is operating but no special behavior is required. COMFORTABLE_STOP The vehicle will stop quickly with a comfortable deceleration. EMERGENCY_STOP The vehicle will stop immediately with as much deceleration as possible."},{"location":"design/autoware-interfaces/ad-api/features/interface/","title":"Interface","text":""},{"location":"design/autoware-interfaces/ad-api/features/interface/#interface","title":"Interface","text":""},{"location":"design/autoware-interfaces/ad-api/features/interface/#related-api","title":"Related API","text":"
This API provides the interface version of the set of AD APIs. It follows Semantic Versioning in order to provide an intuitive understanding of the changes between versions.
This API manages the initialization of localization. Autoware requires a global pose as the initial guess for localization.
"},{"location":"design/autoware-interfaces/ad-api/features/localization/#states","title":"States","text":"State Description UNINITIALIZED Localization is not initialized. Waiting for a global pose as the initial guess. INITIALIZING Localization is initializing. INITIALIZED Localization is initialized. Initialization can be requested again if necessary."},{"location":"design/autoware-interfaces/ad-api/features/motion/","title":"Motion","text":""},{"location":"design/autoware-interfaces/ad-api/features/motion/#motion","title":"Motion","text":""},{"location":"design/autoware-interfaces/ad-api/features/motion/#related-api","title":"Related API","text":"
This API manages the current behavior of the vehicle. Applications can notify the vehicle behavior to the people around and visualize it for operator and passengers.
The motion state manages the stop and start of the vehicle. Once the vehicle has stopped, the state will be STOPPED. After this, when the vehicle tries to start (is still stopped), the state will be STARTING. In this state, calling the start API changes the state to MOVING and the vehicle starts. This mechanism can add processing such as announcements before the vehicle starts. Depending on the configuration, the state may transition directly from STOPPED to MOVING.
State Description STOPPED The vehicle is stopped. STARTING The vehicle is stopped, but is trying to start. MOVING The vehicle is moving. BRAKING (T.B.D.) The vehicle is decelerating strongly."},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/","title":"Operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#operation-mode","title":"Operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#related-api","title":"Related API","text":"
As shown below, Autoware assumes that the vehicle interface has two modes, Autoware control and direct control. In direct control mode, the vehicle is operated using devices such as steering and pedals. If the vehicle does not support direct control mode, it is always treated as Autoware control mode. Autoware control mode has four operation modes.
Mode Description Stop Keep the vehicle stopped. Autonomous Autonomously control the vehicle. Local Manually control the vehicle from nearby with some device such as a joystick. Remote Manually control the vehicle from a web application on the cloud. "},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#states","title":"States","text":""},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#autoware-control-flag","title":"Autoware control flag","text":"
The flag is_autoware_control_enabled indicates if the vehicle is controlled by Autoware. The enable and disable APIs can be used if the control can be switched by software. These APIs will always fail if the vehicle does not support mode switching or is switched by hardware.
"},{"location":"design/autoware-interfaces/ad-api/features/operation_mode/#operation-mode-and-change-flags","title":"Operation mode and change flags","text":"
The state operation_mode indicates what command is used when Autoware control is enabled. The flags change_to_* can be used to check if it is possible to transition to each mode.
Since Autoware may not be able to guarantee safety, such as switching to autonomous mode during overspeed. There is the flag is_in_transition for this situation and it will be true when changing modes. The operator who changed the mode should ensure safety while this flag is true. The flag will be false when the mode change is complete.
This API manages the planned behavior of the vehicle. Applications can notify the vehicle behavior to the people around and visualize it for operator and passengers.
The velocity factors is an array of information on the behavior that the vehicle stops or slows down. Each factor has a behavior type which is described below. Some behavior types have sequence and details as additional information.
Behavior Description surrounding-obstacle There are obstacles immediately around the vehicle. route-obstacle There are obstacles along the route ahead. intersection There are obstacles in other lanes in the path. crosswalk There are obstacles on the crosswalk. rear-check There are obstacles behind that would be in a human driver's blind spot. user-defined-attention-area There are obstacles in the predefined attention area. no-stopping-area There is not enough space beyond the no stopping area. stop-sign A stop by a stop sign. traffic-signal A stop by a traffic signal. v2x-gate-area A stop by a gate area. It has enter and leave as sequences and v2x type as details. merge A stop before merging lanes. sidewalk A stop before crossing the sidewalk. lane-change A lane change. avoidance A path change to avoid an obstacle in the current lane. emergency-operation A stop by emergency instruction from the operator.
Each factor also provides status, poses in the base link frame, and distance from that pose. As the vehicle approaches the stop position, this factor appears with a status of APPROACHING. And when the vehicle reaches that position and stops, the status will be STOPPED. The pose indicates the stop position, or the base link if the stop position cannot be calculated.
The steering factors is an array of information on the maneuver that requires use of turn indicators, such as turning left or right. Each factor has a behavior type which is described below and steering direction. Some behavior types have sequence and details as additional information.
Behavior Description intersection A turning left or right at an intersection. lane-change A lane change. avoidance A path change to avoid an obstacle. It has a sequence of change and return. start-planner T.B.D. goal-planner T.B.D. emergency-operation A path change by emergency instruction from the operator.
Each factor also provides status, poses in the base link frame, and distances from that poses. As the vehicle approaches the position to start steering, this factor appears with a status of APPROACHING. And when the vehicle reaches that position, the status will be TURNING. The poses indicate the start and end position of the section where the status is TURNING.
In cases such as lane change and avoidance, the vehicle will start steering at any position in the range depending on the situation. For these types, the section where the status is TURNING will be updated dynamically and the poses will follow that.
This API manages destination and waypoints. Note that waypoints are not like stops and just points passing through. In other words, Autoware does not support the route with multiple stops, the application needs to split it up and switch them. There are two ways to set the route. The one is a generic method that uses pose, another is a map-dependent.
"},{"location":"design/autoware-interfaces/ad-api/features/routing/#states","title":"States","text":"State Description UNSET The route is not set. Waiting for a route request. SET The route is set. ARRIVED The vehicle has arrived at the destination. CHANGING Trying to change the route. Not implemented yet."},{"location":"design/autoware-interfaces/ad-api/features/routing/#goal-modification","title":"Goal modification","text":"
Autoware tries to look for an alternate goal when goal is unreachable (e.g., when there is an obstacle on the given goal). When setting a route from the API, applications can choose whether they allow Autoware to adjust goal pose in such situation. When set false, Autoware may get stuck until the given goal becomes reachable.
This feature is available if the vehicle provides a software interface for the doors. It can be used to create user interfaces for passengers or to control sequences at bus stops.
Each door in a vehicle is assigned an array index. This assignment is vehicle dependent. The layout API returns this information. The description field is a string to display in the user interface, etc. This is an arbitrary string and is not recommended to use for processing in applications. Use the roles field to know doors for getting on and off. Below is an example of the information returned by the layout API.
Index Description Roles 0 front right - 1 front left GET_ON 2 rear right GET_OFF 3 rear left GET_ON, GET_OFF"},{"location":"design/autoware-interfaces/ad-api/features/vehicle-doors/#status","title":"Status","text":"
The status API provides an array of door status. This array order is consistent with the layout API.
Use the command API to control doors. Unlike the status and layout APIs, array index do not correspond to doors. The command has a field to specify the target door index.
This is an estimate of the vehicle kinematics. The vehicle position is necessary for applications to schedule dispatches. Also, using velocity and acceleration, applications can find vehicles that need operator assistance, such as stuck or brake suddenly.
This is the status provided by the vehicle. The indicators and steering are mainly used for visualization and remote control. The remaining energy can be also used for vehicle scheduling.
The vehicle dimensions are used to know the actual distance between the vehicle and objects because the vehicle position in kinematics is the coordinates of the base link. This is necessary for visualization when supporting vehicles remotely.
"},{"location":"design/autoware-interfaces/ad-api/list/","title":"List of Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/list/#list-of-autoware-ad-api","title":"List of Autoware AD API","text":"
Get the MRM state. For details, see the fail-safe.
"},{"location":"design/autoware-interfaces/ad-api/list/api/fail_safe/mrm_state/#message","title":"Message","text":"Name Type Description state uint16 The state of MRM operation. behavior uint16 The currently selected behavior of MRM."},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/","title":"/api/interface/version","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/#apiinterfaceversion","title":"/api/interface/version","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_version_msgs/srv/InterfaceVersion
"},{"location":"design/autoware-interfaces/ad-api/list/api/interface/version/#response","title":"Response","text":"Name Type Description major uint16 major version minor uint16 minor version patch uint16 patch version"},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/","title":"/api/localization/initialization_state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/#apilocalizationinitialization_state","title":"/api/localization/initialization_state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/LocalizationInitializationState
Get the initialization state of localization. For details, see the localization.
"},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialization_state/#message","title":"Message","text":"Name Type Description state uint16 A value of the localization initialization state."},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/","title":"/api/localization/initialize","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#apilocalizationinitialize","title":"/api/localization/initialize","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/InitializeLocalization
Request to initialize localization. For details, see the localization.
"},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#request","title":"Request","text":"Name Type Description pose geometry_msgs/msg/PoseWithCovarianceStamped[<=1] A global pose as the initial guess. If omitted, the GNSS pose will be used."},{"location":"design/autoware-interfaces/ad-api/list/api/localization/initialize/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/","title":"/api/motion/accept_start","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/#apimotionaccept_start","title":"/api/motion/accept_start","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/#status","title":"Status","text":"
"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/accept_start/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/","title":"/api/motion/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/#apimotionstate","title":"/api/motion/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/#status","title":"Status","text":"
Get the motion state. For details, see the motion state.
"},{"location":"design/autoware-interfaces/ad-api/list/api/motion/state/#message","title":"Message","text":"Name Type Description state uint16 A value of the motion state."},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/","title":"/api/operation_mode/change_to_autonomous","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/#apioperation_modechange_to_autonomous","title":"/api/operation_mode/change_to_autonomous","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_autonomous/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/","title":"/api/operation_mode/change_to_local","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/#apioperation_modechange_to_local","title":"/api/operation_mode/change_to_local","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_local/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/","title":"/api/operation_mode/change_to_remote","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/#apioperation_modechange_to_remote","title":"/api/operation_mode/change_to_remote","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_remote/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/","title":"/api/operation_mode/change_to_stop","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/#apioperation_modechange_to_stop","title":"/api/operation_mode/change_to_stop","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/change_to_stop/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/","title":"/api/operation_mode/disable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/#apioperation_modedisable_autoware_control","title":"/api/operation_mode/disable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/disable_autoware_control/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/","title":"/api/operation_mode/enable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/#apioperation_modeenable_autoware_control","title":"/api/operation_mode/enable_autoware_control","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/ChangeOperationMode
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/enable_autoware_control/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/","title":"/api/operation_mode/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/#apioperation_modestate","title":"/api/operation_mode/state","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/OperationModeState
Get the operation mode state. For details, see the operation mode.
"},{"location":"design/autoware-interfaces/ad-api/list/api/operation_mode/state/#message","title":"Message","text":"Name Type Description mode uint8 The selected command for Autoware control. is_autoware_control_enabled bool True if vehicle control by Autoware is enabled. is_in_transition bool True if the operation mode is in transition. is_stop_mode_available bool True if the operation mode can be changed to stop. is_autonomous_mode_available bool True if the operation mode can be changed to autonomous. is_local_mode_available bool True if the operation mode can be changed to local. is_remote_mode_available bool True if the operation mode can be changed to remote."},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/","title":"/api/perception/objects","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/#apiperceptionobjects","title":"/api/perception/objects","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/#status","title":"Status","text":"
Latest Version: v1.0.0
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/DynamicObjectArray
Get the recognized objects array with label, shape, current position and predicted path For details, see the perception.
"},{"location":"design/autoware-interfaces/ad-api/list/api/perception/objects/#message","title":"Message","text":"Name Type Description objects.id unique_identifier_msgs/msg/UUID The UUID of each object objects.existence_probability float64 The probability of the object exits objects.classification autoware_adapi_v1_msgs/msg/ObjectClassification[] The type of the object recognized and the confidence level objects.kinematics autoware_adapi_v1_msgs/msg/DynamicObjectKinematics Consist of the object pose, twist, acceleration and the predicted_paths objects.shape shape_msgs/msg/SolidPrimitive escribe the shape of the object with dimension, and polygon"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/","title":"/api/planning/steering_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/#apiplanningsteering_factors","title":"/api/planning/steering_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/#status","title":"Status","text":"
Latest Version: not released
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/SteeringFactorArray
Get the steering factors, sorted in ascending order of distance. For details, see the planning factors.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/steering_factors/#message","title":"Message","text":"Name Type Description factors.pose geometry_msgs/msg/Pose[2] The base link pose related to the steering factor. factors.distance float32[2] The distance from the base link to the above pose. factors.direction uint16 The direction of the steering factor. factors.status uint16 The status of the steering factor. factors.behavior string The behavior type of the steering factor. factors.sequence string The sequence type of the steering factor. factors.detail string The additional information of the steering factor. factors.cooperation autoware_adapi_v1_msgs/msg/CooperationStatus[<=1] The cooperation status if the module supports."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/","title":"/api/planning/velocity_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/#apiplanningvelocity_factors","title":"/api/planning/velocity_factors","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/#status","title":"Status","text":"
Latest Version: not released
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/VelocityFactorArray
Get the velocity factors, sorted in ascending order of distance. For details, see the planning factors.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/velocity_factors/#message","title":"Message","text":"Name Type Description factors.pose geometry_msgs/msg/Pose The base link pose related to the velocity factor. factors.distance float32 The distance from the base link to the above pose. factors.status uint16 The status of the velocity factor. factors.behavior string The behavior type of the velocity factor. factors.sequence string The sequence type of the velocity factor. factors.detail string The additional information of the velocity factor. factors.cooperation autoware_adapi_v1_msgs/msg/CooperationStatus[<=1] The cooperation status if the module supports."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/","title":"/api/planning/cooperation/get_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/#apiplanningcooperationget_policies","title":"/api/planning/cooperation/get_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/GetCooperationPolicies
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/get_policies/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status policies.behavior string The type of the target behavior. policies.sequence string The type of the target sequence. policies.policy uint8 The type of the cooporation policy."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/","title":"/api/planning/cooperation/set_commands","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#apiplanningcooperationset_commands","title":"/api/planning/cooperation/set_commands","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/SetCooperationCommands
Set the operator's decision for cooperation. For details, see the cooperation.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#request","title":"Request","text":"Name Type Description commands.uuid unique_identifier_msgs/msg/UUID The ID in the cooperation status. commands.cooperator autoware_adapi_v1_msgs/msg/CooperationDecision The operator's decision."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_commands/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/","title":"/api/planning/cooperation/set_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#apiplanningcooperationset_policies","title":"/api/planning/cooperation/set_policies","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/SetCooperationPolicies
Set the default decision that is used instead when the operator's decision is undecided. For details, see the cooperation.
"},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#request","title":"Request","text":"Name Type Description policies.behavior string The type of the target behavior. policies.sequence string The type of the target sequence. policies.policy uint8 The type of the cooporation policy."},{"location":"design/autoware-interfaces/ad-api/list/api/planning/cooperation/set_policies/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/","title":"/api/routing/clear_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/#apiroutingclear_route","title":"/api/routing/clear_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/#status","title":"Status","text":"
"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/clear_route/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/","title":"/api/routing/route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/#apiroutingroute","title":"/api/routing/route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/#status","title":"Status","text":"
Get the route with the waypoint segments in lanelet format. It is empty if route is not set.
"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/route/#message","title":"Message","text":"Name Type Description header std_msgs/msg/Header header for pose transformation data autoware_adapi_v1_msgs/msg/RouteData[<=1] The route in lanelet format"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/set_route/","title":"/api/routing/set_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/set_route/#apiroutingset_route","title":"/api/routing/set_route","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/routing/set_route/#status","title":"Status","text":"
Get the route state. For details, see the routing.
"},{"location":"design/autoware-interfaces/ad-api/list/api/routing/state/#message","title":"Message","text":"Name Type Description state uint16 A value of the route state."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/","title":"/api/vehicle/dimensions","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/#apivehicledimensions","title":"/api/vehicle/dimensions","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/GetVehicleDimensions
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/dimensions/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status dimensions autoware_adapi_v1_msgs/msg/VehicleDimensions vehicle dimensions"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/","title":"/api/vehicle/kinematics","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/#apivehiclekinematics","title":"/api/vehicle/kinematics","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/#status","title":"Status","text":"
Latest Version: not released
Method: realtime stream
Data Type: autoware_adapi_v1_msgs/msg/VehicleKinematics
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/kinematics/#message","title":"Message","text":"Name Type Description geographic_pose geographic_msgs/msg/GeoPointStamped The longitude and latitude of the vehicle. If the map uses local coordinates, it will not be available. pose geometry_msgs/msg/PoseWithCovarianceStamped The pose with covariance from the base link. twist geometry_msgs/msg/TwistWithCovarianceStamped Vehicle current twist with covariance. accel geometry_msgs/msg/AccelWithCovarianceStamped Vehicle current acceleration with covariance."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/","title":"/api/vehicle/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/#apivehiclestatus","title":"/api/vehicle/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/#status","title":"Status","text":"
Latest Version: not released
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/VehicleStatus
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/status/#message","title":"Message","text":"Name Type Description gear autoware_adapi_v1_msgs/msg/Gear Gear status. turn_indicators autoware_adapi_v1_msgs/msg/TurnIndicators Turn indicators status, only either left or right will be enabled. hazard_lights autoware_adapi_v1_msgs/msg/HazardLights Hazard lights status. steering_tire_angle float64 Vehicle current tire angle in radian. energy_percentage float32 Battery percentage or fuel percentage, it will depends on the vehicle."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/","title":"/api/vehicle/doors/command","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#apivehicledoorscommand","title":"/api/vehicle/doors/command","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/SetDoorCommand
Set the door command. This API is only available if the vehicle supports software door control.
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#request","title":"Request","text":"Name Type Description doors.index uint32 The index of the target door. doors.command uint8 The command for the target door."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/command/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/","title":"/api/vehicle/doors/layout","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/#apivehicledoorslayout","title":"/api/vehicle/doors/layout","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/#status","title":"Status","text":"
Latest Version: not released
Method: function call
Data Type: autoware_adapi_v1_msgs/srv/GetDoorLayout
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/layout/#response","title":"Response","text":"Name Type Description status autoware_adapi_v1_msgs/msg/ResponseStatus response status doors.roles uint8[] The roles of the door in the service the vehicle provides. doors.description string The description of the door for display in the interface."},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/","title":"/api/vehicle/doors/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/#apivehicledoorsstatus","title":"/api/vehicle/doors/status","text":""},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/#status","title":"Status","text":"
Latest Version: not released
Method: notification
Data Type: autoware_adapi_v1_msgs/msg/DoorStatusArray
"},{"location":"design/autoware-interfaces/ad-api/list/api/vehicle/doors/status/#message","title":"Message","text":"Name Type Description doors.status uint8 current door status"},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/#user-story-of-bus-service","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/#overview","title":"Overview","text":"
This user story is a bus service that goes around the designated stops.
"},{"location":"design/autoware-interfaces/ad-api/stories/bus-service/#scenario","title":"Scenario","text":"Step Operation Use Case 1 Startup the autonomous driving system. Launch and terminate 2 Drive the vehicle from the garage to the waiting position. Change the operation mode 3 Enable autonomous control. Change the operation mode 4 Drive the vehicle to the next bus stop. Drive to the designated position 5 Get on and off the vehicle. Get on and get off 6 Return to step 4 unless it's the last bus stop. 7 Drive the vehicle to the waiting position. Drive to the designated position 8 Drive the vehicle from the waiting position to the garage. Change the operation mode 9 Shutdown the autonomous driving system. Launch and terminate"},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/#user-story-of-bus-service","title":"User story of bus service","text":""},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/#overview","title":"Overview","text":"
This user story is a taxi service that picks up passengers and drives them to their destination.
"},{"location":"design/autoware-interfaces/ad-api/stories/taxi-service/#scenario","title":"Scenario","text":"Step Operation Use Case 1 Startup the autonomous driving system. Launch and terminate 2 Drive the vehicle from the garage to the waiting position. Change the operation mode 3 Enable autonomous control. Change the operation mode 4 Drive the vehicle to the position to pick up. Drive to the designated position 5 Get on the vehicle. Get on and get off 6 Drive the vehicle to the destination. Drive to the designated position 7 Get off the vehicle. Get on and get off 8 Drive the vehicle to the waiting position. Drive to the designated position 9 Return to step 4 if there is another request. 10 Drive the vehicle from the waiting position to the garage. Change the operation mode 11 Shutdown the autonomous driving system. Launch and terminate"},{"location":"design/autoware-interfaces/ad-api/types/","title":"Types of Autoware AD API","text":""},{"location":"design/autoware-interfaces/ad-api/types/#types-of-autoware-ad-api","title":"Types of Autoware AD API","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationCommand/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationDecision
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationCommand/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationDecision/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationDecision/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationPolicy/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationPolicy/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationStatus/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationDecision
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/CooperationStatus/#this-type-is-used-by","title":"This type is used by","text":"
uint8 OPEN = 1\nuint8 CLOSE = 2\n\nuint32 index\nuint8 command\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorCommand/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorCommand/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorLayout/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorLayout/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatus/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatus/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatusArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DoorStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DoorStatusArray/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObject/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DynamicObject
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectArray/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectKinematics/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DynamicObjectPath
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectKinematics/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectPath/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/DynamicObjectPath/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/HazardLights/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/HazardLights/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/LocalizationInitializationState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/LocalizationInitializationState/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MotionState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MotionState/#this-type-is-used-by","title":"This type is used by","text":"
builtin_interfaces/Time stamp\n\n# For common use\nuint16 UNKNOWN = 0\n\n# For state\nuint16 NORMAL = 1\nuint16 MRM_OPERATING = 2\nuint16 MRM_SUCCEEDED = 3\nuint16 MRM_FAILED = 4\n\n# For behavior\nuint16 NONE = 1\nuint16 EMERGENCY_STOP = 2\nuint16 COMFORTABLE_STOP = 3\n\nuint16 state\nuint16 behavior\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MrmState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/MrmState/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ObjectClassification/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ObjectClassification/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/OperationModeState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/OperationModeState/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ResponseStatus/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/ResponseStatus/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/Route/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/RouteData
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/Route/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteData/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/RouteSegment
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteData/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteOption/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteOption/#this-type-is-used-by","title":"This type is used by","text":"
int64 id\nstring type # The same id may be used for each type.\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RoutePrimitive/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RoutePrimitive/#this-type-is-used-by","title":"This type is used by","text":"
autoware_adapi_v1_msgs/RoutePrimitive preferred\nautoware_adapi_v1_msgs/RoutePrimitive[] alternatives # Does not include the preferred primitive.\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteSegment/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/RoutePrimitive
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteSegment/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteState/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/RouteState/#this-type-is-used-by","title":"This type is used by","text":"
# constants for common use\nuint16 UNKNOWN = 0\n\n# constants for direction\nuint16 LEFT = 1\nuint16 RIGHT = 2\nuint16 STRAIGHT = 3\n\n# constants for status\nuint16 APPROACHING = 1\nuint16 TURNING = 3\n\n# variables\ngeometry_msgs/Pose[2] pose\nfloat32[2] distance\nuint16 direction\nuint16 status\nstring behavior\nstring sequence\nstring detail\nautoware_adapi_v1_msgs/CooperationStatus[<=1] cooperation\n\n\n\n# deprecated constants for type\nuint16 INTERSECTION = 1\nuint16 LANE_CHANGE = 2\nuint16 AVOIDANCE_PATH_CHANGE = 3\nuint16 AVOIDANCE_PATH_RETURN = 4\nuint16 STATION = 5\nuint16 PULL_OUT = 6 # Deprecated. Use START_PLANNER.\nuint16 START_PLANNER = 6\nuint16 PULL_OVER = 7 # Deprecated. Use GOAL_PLANNER.\nuint16 GOAL_PLANNER = 7\nuint16 EMERGENCY_OPERATION = 8\n\n# deprecated constants for status\nuint16 TRYING = 2\n\n# deprecated variables\nuint16 type\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactor/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactor/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactorArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/SteeringFactor
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/SteeringFactorArray/#this-type-is-used-by","title":"This type is used by","text":"
# constants\nuint8 UNKNOWN = 0\nuint8 DISABLE = 1\nuint8 LEFT = 2\nuint8 RIGHT = 3\n\nuint8 status\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/TurnIndicators/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/TurnIndicators/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleDimensions/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleDimensions/#this-type-is-used-by","title":"This type is used by","text":"
# Geographic point, using the WGS 84 reference ellipsoid.\n# This data will be invalid If Autoware does not provide projection information between geographic coordinates and local coordinates.\ngeographic_msgs/GeoPointStamped geographic_pose\n\n# Local coordinate from the autoware\ngeometry_msgs/PoseWithCovarianceStamped pose\ngeometry_msgs/TwistWithCovarianceStamped twist\ngeometry_msgs/AccelWithCovarianceStamped accel\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleKinematics/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleKinematics/#this-type-is-used-by","title":"This type is used by","text":"
builtin_interfaces/Time stamp\nautoware_adapi_v1_msgs/Gear gear\nautoware_adapi_v1_msgs/TurnIndicators turn_indicators\nautoware_adapi_v1_msgs/HazardLights hazard_lights\nfloat64 steering_tire_angle\nfloat32 energy_percentage # Battery percentage or fuel percentage, it will depends on the vehicle.\n
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleStatus/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/Gear
autoware_adapi_v1_msgs/msg/HazardLights
autoware_adapi_v1_msgs/msg/TurnIndicators
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VehicleStatus/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactor/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactor/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactorArray/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/VelocityFactor
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/msg/VelocityFactorArray/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/AcceptStart/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/AcceptStart/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ChangeOperationMode/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ChangeOperationMode/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ClearRoute/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/ClearRoute/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetCooperationPolicies/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationPolicy
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetCooperationPolicies/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetDoorLayout/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DoorLayout
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetDoorLayout/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetVehicleDimensions/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
autoware_adapi_v1_msgs/msg/VehicleDimensions
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/GetVehicleDimensions/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/InitializeLocalization/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/InitializeLocalization/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationCommands/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationCommand
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationCommands/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationPolicies/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/CooperationPolicy
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetCooperationPolicies/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetDoorCommand/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/DoorCommand
autoware_adapi_v1_msgs/msg/ResponseStatus
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetDoorCommand/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoute/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
autoware_adapi_v1_msgs/msg/RouteOption
autoware_adapi_v1_msgs/msg/RouteSegment
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoute/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoutePoints/#this-type-uses","title":"This type uses","text":"
autoware_adapi_v1_msgs/msg/ResponseStatus
autoware_adapi_v1_msgs/msg/RouteOption
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_v1_msgs/srv/SetRoutePoints/#this-type-is-used-by","title":"This type is used by","text":"
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_version_msgs/srv/InterfaceVersion/#this-type-uses","title":"This type uses","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/types/autoware_adapi_version_msgs/srv/InterfaceVersion/#this-type-is-used-by","title":"This type is used by","text":"
None
"},{"location":"design/autoware-interfaces/ad-api/use-cases/change-operation-mode/","title":"Change the operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/change-operation-mode/#change-the-operation-mode","title":"Change the operation mode","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/change-operation-mode/#related-api","title":"Related API","text":"
"},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/","title":"Drive to the designated position","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/#drive-to-the-designated-position","title":"Drive to the designated position","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/#related-api","title":"Related API","text":"
Driving
Route
"},{"location":"design/autoware-interfaces/ad-api/use-cases/drive-designated-position/#sequence","title":"Sequence","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/","title":"Get on and get off","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/#get-on-and-get-off","title":"Get on and get off","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/#related-api","title":"Related API","text":"
Door
"},{"location":"design/autoware-interfaces/ad-api/use-cases/get-on-off/#sequence","title":"Sequence","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/initialize-pose/","title":"Initialize the pose","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/initialize-pose/#initialize-the-pose","title":"Initialize the pose","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/initialize-pose/#related-api","title":"Related API","text":"
"},{"location":"design/autoware-interfaces/ad-api/use-cases/launch-terminate/","title":"Launch and terminate","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/launch-terminate/#launch-and-terminate","title":"Launch and terminate","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/launch-terminate/#related-api","title":"Related API","text":"
AD API provides current vehicle status for remote monitoring, visualization for passengers, etc. Use the API below depending on the data you want to monitor.
The vehicle status provides basic information such as kinematics, indicators, and dimensions. This allows a remote operator to know the position and velocity of the vehicle. For applications such as FMS, it can help find vehicles that need assistance, such as vehicles that are stuck or brake suddenly. It is also possible to determine the actual distance to an object from the vehicle dimensions.
The planning factors provides the planning status of the vehicle. HMI can use this to warn of sudden movements of the vehicle, and to share the stop reason with passengers for comfortable driving.
The perception provides the objects detected by Autoware. HMI can use this to visualize objects around the vehicle.
"},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/","title":"Vehicle operation","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/#vehicle-operation","title":"Vehicle operation","text":""},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/#request-to-intervene","title":"Request to intervene","text":"
Request to intervene (RTI) is a feature that requires the operator to switch to manual driving mode. It is also called Take Over Request (TOR). Interfaces for RTI are currently being discussed. For now assume that manual driving is requested if the MRM state is not NORMAL. See fail-safe for details.
"},{"location":"design/autoware-interfaces/ad-api/use-cases/vehicle-operation/#request-to-cooperate","title":"Request to cooperate","text":"
Request to cooperate (RTC) is a feature that the operator supports the decision in autonomous driving mode. Autoware usually drives the vehicle using its own decisions, but the operator may prefer to make their own decisions in complex situations. Since RTC only overrides the decision and does not need to change operation mode, the vehicle can continue autonomous driving, unlike RTC. See cooperation for details.
Actuation status of the ego vehicle for acceleration, steering, and brake.
TODO This represents the reported physical efforts exerted by the vehicle actuators. Published by the Vehicle Interface.
ActuationStatus (github discussion).
builtin_interfaces::msg::Time stamp
float32 acceleration
float32 steering
"},{"location":"design/autoware-interfaces/components/control/#output","title":"Output","text":""},{"location":"design/autoware-interfaces/components/control/#vehicle-control-command","title":"Vehicle Control Command","text":"
A motion signal to drive the vehicle, achieved by the low-level controller in the vehicle layer. Used by the Vehicle Interface.
Current velocity of the ego vehicle, published by the vehicle interface.
autoware_auto_vehicle_msgs/msg/VelocityReport
std_msgs/msg/Header header;
float longitudinal_velocity;
float lateral_velocity;
float heading_rate;
Before the velocity input localization interface, module vehicle_velocity_converter converts message type autoware_auto_vehicle_msgs/msg/VelocityReport to geometry_msgs/msg/TwistWithCovarianceStamped.
Current pose, velocity and acceleration of ego, calculated from localization interface.
Note: Kinematic state contains pose, velocity and acceleration. In the future, pose, velocity and acceleration will not be used as output for localization.
Autoware relies on high-definition point cloud maps and vector maps of the driving environment to perform various tasks. Before launching Autoware, you need to load the pre-created map files.
It loads a Lanelet2 file and publishes the map data as autoware_auto_mapping_msgs/msg/HADMapBin message. The lan/lon coordinates are projected onto the MGRS coordinates.
This page provides specific specifications about the Interface of the Planning Component. Please refer to the planning architecture design document for high-level concepts and data flow.
TODO: The detailed definitions (meanings of elements included in each topic) are not described yet, need to be updated.
"},{"location":"design/autoware-interfaces/components/planning/#input","title":"Input","text":""},{"location":"design/autoware-interfaces/components/planning/#from-map-component","title":"From Map Component","text":"Name Topic Type Description Vector Map /map/vector_map autoware_auto_mapping_msgs/msg/HADMapBin Map of the environment where the planning takes place."},{"location":"design/autoware-interfaces/components/planning/#from-localization-component","title":"From Localization Component","text":"Name Topic Type Description Vehicle Kinematic State /localization/kinematic_state nav_msgs/msg/Odometry Current position, orientation and velocity of ego. Vehicle Acceleration /localization/acceleration geometry_msgs/msg/AccelWithCovarianceStamped Current acceleration of ego.
TODO: acceleration information should be merged into the kinematic state.
"},{"location":"design/autoware-interfaces/components/planning/#from-perception-component","title":"From Perception Component","text":"Name Topic Type Description Objects /perception/object_recognition/objects autoware_auto_perception_msgs/msg/PredictedObjects Set of perceived objects around ego that need to be avoided or followed when planning a trajectory. This contains semantics information such as a object class (e.g. vehicle, pedestrian, etc) or a shape of the objects. Obstacles /perception/obstacle_segmentation/pointcloud sensor_msgs/msg/PointCloud2 Set of perceived obstacles around ego that need to be avoided or followed when planning a trajectory. This only contains a primitive information of the obstacle. No shape nor velocity information. Occupancy Grid Map /perception/occupancy_grid_map/map nav_msgs/msg/OccupancyGrid Contains the presence of obstacles and blind spot information (represented as UNKNOWN). Traffic Signal /perception/traffic_light_recognition/traffic_signals autoware_auto_perception_msgs/msg/TrafficSignalArray Contains the traffic signal information such as a color (green, yellow, read) and an arrow (right, left, straight).
TODO: The type of the Obstacles information should not depend on the specific sensor message type (now PointCloud). It needs to be fixed.
"},{"location":"design/autoware-interfaces/components/planning/#from-api","title":"From API","text":"Name Topic Type Description Max Velocity /planning/scenario_planning/max_velocity_default autoware_adapi_v1_msgs/srv/SetRoutePoints Indicate the maximum value of the vehicle speed plan Operation Mode /system/operation_mode/state autoware_adapi_v1_msgs/msg/OperationModeState Indicates the current operation mode (automatic/manual, etc.). Route Set /planning/mission_planning/set_route autoware_adapi_v1_msgs/srv/SetRoute Indicates to set the route when the vehicle is stopped. Route Points Set /planning/mission_planning/set_route_points autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to set the route with points when the vehicle is stopped. Route Change /planning/mission_planning/change_route autoware_adapi_v1_msgs/srv/SetRoute Indicates to change the route when the vehicle is moving. Route Points Change /planning/mission_planning/change_route_points autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to change the route with points when the vehicle is moving. Route Clear /planning/mission_planning/clear_route autoware_adapi_v1_msgs/srv/ClearRoute Indicates to clear the route information. MRM Route Set Points /planning/mission_planning/mission_planner/srv/set_mrm_route autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to set the emergency route. MRM Route Clear /planning/mission_planning/mission_planner/srv/clear_mrm_route autoware_adapi_v1_msgs/srv/SetRoutePoints Indicates to clear the emergency route."},{"location":"design/autoware-interfaces/components/planning/#output","title":"Output","text":""},{"location":"design/autoware-interfaces/components/planning/#to-control","title":"To Control","text":"Name Topic Type Description Trajectory /planning/trajectory autoware_auto_planning_msgs/msg/Trajectory A sequence of space and velocity and acceleration points to be followed by the controller. Turn Indicator /planning/turn_indicators_cmd autoware_auto_vehicle_msgs/msg/TurnIndicatorsCommand Turn indicator signal to be followed by the vehicle. Hazard Light /planning/hazard_lights_cmd autoware_auto_vehicle_msgs/msg/HazardLightsCommand Hazard light signal to be followed by the vehicle."},{"location":"design/autoware-interfaces/components/planning/#to-system","title":"To System","text":"Name Topic Type Description Diagnostics /planning/hazard_lights_cmd diagnostic_msgs/msg/DiagnosticArray Diagnostic status of the Planning component reported to the System component."},{"location":"design/autoware-interfaces/components/planning/#to-api","title":"To API","text":"Name Topic Type Description Path Candidate /planning/path_candidate/* autoware_auto_planning_msgs/msg/Path The path Autoware is about to take. Users can interrupt the operation based on the path candidate information. Steering Factor /planning/steering_factor/* autoware_adapi_v1_msgs/msg/SteeringFactorArray Information about the steering maneuvers performed by Autoware (e.g., steering to the right for a right turn, etc.) Velocity Factor /planning/velocity_factors/* autoware_adapi_v1_msgs/msg/VelocityFactorArray Information about the velocity maneuvers performed by Autoware (e.g., stop for an obstacle, etc.)"},{"location":"design/autoware-interfaces/components/planning/#planning-internal-interface","title":"Planning internal interface","text":"
This section explains the communication between the different planning modules shown in the Planning Architecture Design.
"},{"location":"design/autoware-interfaces/components/planning/#from-mission-planning-to-scenario-planning","title":"From Mission Planning to Scenario Planning","text":"Name Topic Type Description Route /planning/mission_planning/route autoware_planning_msgs/msg/LaneletRoute A sequence of lane IDs on a Lanelet map, from the starting point to the destination."},{"location":"design/autoware-interfaces/components/planning/#from-behavior-planning-to-motion-planning","title":"From Behavior Planning to Motion Planning","text":"Name Topic Type Description Path /planning/scenario_planning/lane_driving/behavior_planning/path autoware_auto_planning_msgs/msg/Path A sequence of approximate vehicle positions for driving, along with information on the maximum speed and the drivable areas. Modules receiving this message are expected to make changes to the path within the constraints of the drivable areas and the maximum speed, generating the desired final trajectory."},{"location":"design/autoware-interfaces/components/planning/#from-scenario-planning-to-validation","title":"From Scenario Planning to Validation","text":"Name Topic Type Description Trajectory /planning/scenario_planning/trajectory autoware_auto_planning_msgs/msg/Trajectory A sequence of precise vehicle positions, speeds, and accelerations required for driving. It is expected that the vehicle will follow this trajectory."},{"location":"design/autoware-interfaces/components/sensing/","title":"Sensing","text":""},{"location":"design/autoware-interfaces/components/sensing/#sensing","title":"Sensing","text":"
The Vehicle Interface receives the Vehicle Signal Commands and Vehicle Control Commands and publishes the vehicle status. It also communicates with vehicle by the vehicle-specific protocol.
The Gate switches multiple Vehicle Control Commands. These signals include autonomous diving command, joystick, remote control, and emergency operation, etc. The Adapter converts generalized control command (target steering, steering rate, velocity, acceleration, jerk) into vehicle-specific control values (steering-torque, wheel-torque, voltage, pressure, accel pedal position, etc).
Commands for various elements of the vehicle unrelated to motion. Published by the Planning module.
"},{"location":"design/autoware-interfaces/components/vehicle-interface/#outputs","title":"Outputs","text":""},{"location":"design/autoware-interfaces/components/vehicle-interface/#vehicle-signal-reports","title":"Vehicle Signal Reports","text":"
Reports for various elements of the vehicle unrelated to motion. Published by the Vehicle Interface.
Actuation status of the ego vehicle for acceleration, steering, and brake. This represents the reported physical efforts exerted by the vehicle actuators. Published by the Vehicle Interface.
Actuation command sent to the ego vehicle. This represents the requested physical efforts to be exerted by the vehicle actuators. Published by the Vehicle Interface as generated by the adapter.
Defining temporal performance metrics on components
An example procedure for adding and evaluating a new node
TODO: Write the following contents.
Create an Autoware package
etc.
"},{"location":"how-to-guides/integrating-autoware/overview/","title":"Overview","text":""},{"location":"how-to-guides/integrating-autoware/overview/#overview","title":"Overview","text":""},{"location":"how-to-guides/integrating-autoware/overview/#requirement-prepare-your-real-vehicle-hardware","title":"Requirement: prepare your real vehicle hardware","text":"
Prerequisites for the vehicle:
An onboard computer that satisfies the Autoware installation prerequisites
The following devices attached
Drive-by-wire interface
LiDAR
Optional: Inertial measurement unit
Optional: Camera
Optional: GNSS
"},{"location":"how-to-guides/integrating-autoware/overview/#1-creating-your-autoware-meta-repository","title":"1. Creating your Autoware meta-repository","text":"
Create your Autoware meta-repository. One easy way is to fork autowarefoundation/autoware and clone it. For how to fork a repository, refer to GitHub Docs.
If you set up multiple types of vehicles, adding a suffix like \"autoware.vehicle_A\" or \"autoware.vehicle_B\" is recommended.
"},{"location":"how-to-guides/integrating-autoware/overview/#2-creating-the-your-vehicle-and-sensor-description","title":"2. Creating the your vehicle and sensor description","text":"
Next, you need to create description packages that define the vehicle and sensor configuration of your vehicle.
Create the following two packages:
YOUR_VEHICLE_launch (see here for example)
YOUR_SENSOR_KIT_launch (see here for example)
Once created, you need to update the autoware.repos file of your cloned Autoware repository to refer to these two description packages.
Create a launch file (refer to the sample vehicle launch package for example). If you have multiple vehicles with the same hardware setup, you can specify vehicle_id to distinguish them.
"},{"location":"how-to-guides/integrating-autoware/overview/#adapt-your_sensor_kit_description-for-autoware-launching-system","title":"Adapt YOUR_SENSOR_KIT_description for autoware launching system","text":""},{"location":"how-to-guides/integrating-autoware/overview/#at-your_sensor_kit_description","title":"At YOUR_SENSOR_KIT_description","text":"
Define URDF and extrinsic parameters for all the sensors here (refer to the sample sensor kit description package for example). Note that you need to calibrate extrinsic parameters for all the sensors beforehand.
Create launch/sensing.launch.xml that launches the interfaces of all the sensors on the vehicle. (refer to the sample sensor kit launch package for example).
Note
At this point, you are now able to run Autoware's Planning Simulator to do a basic test of your vehicle and sensing packages. To do so, you need to build and install Autoware using your cloned repository. Follow the steps for either Docker or source installation (starting from the dependency installation step) and then run the following command:
"},{"location":"how-to-guides/integrating-autoware/overview/#3-create-a-vehicle_interface-package","title":"3. Create a vehicle_interface package","text":"
You need to create an interface package for your vehicle. The package is expected to provide the following two functions.
Receive command messages from vehicle_cmd_gate and drive the vehicle accordingly
Send vehicle status information to Autoware
You can find detailed information about the requirements of the vehicle_interface package in the Vehicle Interface design documentation. You can also refer to TIER IV's pacmod_interface repository as an example of a vehicle interface package.
You need both a pointcloud map and a vector map in order to use Autoware. For more information on map design, please click here.
"},{"location":"how-to-guides/integrating-autoware/overview/#create-a-pointcloud-map","title":"Create a pointcloud map","text":"
Use third-party tools such as a LiDAR-based SLAM (Simultaneous Localization And Mapping) package to create a pointcloud map in the .pcd format. For more information, please click here.
Click the 2D Nav Goal button in the toolbar, or hit the G key
In the 3D View pane, click and hold the left mouse button, and then drag to set the direction for the goal pose. If successful, you will see the calculated planning path on RViz.
You can also engage via RViz with \"AutowareStatePanel\". The panel can be found in Panels > Add New Panel > tier4_state_rviz_plugin > AutowareStatePanel.
Now the vehicle should drive along the calculated path!
"},{"location":"how-to-guides/integrating-autoware/overview/#6-tune-parameters-for-your-vehicle-environment","title":"6. Tune parameters for your vehicle & environment","text":"
You may need to tune your parameters depending on the domain in which you will operate your vehicle.
Autoware requires a pointcloud map and a vector map for the vehicle's operating environment. (Check the map design documentation page for the detailed specification).
This page explains how users can create maps that can be used for Autoware.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/#creating-a-point-cloud-map","title":"Creating a point cloud map","text":"
Traditionally, a Mobile Mapping System (MMS) is used in order to create highly accurate large-scale point cloud maps. However, since a MMS requires high-end sensors for precise positioning, its operational cost can be very expensive and may not be suitable for a relatively small driving environment. Alternatively, a Simultaneous Localization And Mapping (SLAM) algorithm can be used to create a point cloud map from recorded LiDAR scans. Some of the useful open-source SLAM implementations are listed in this page.
If you prefer proprietary software that is easy to use, you can try a fully automatic mapping tool from MAP IV, Inc., MapIV Engine. They currently provide a trial license for Autoware users free of charge.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/#creating-a-vector-map","title":"Creating a vector map","text":"
The easiest way to create an Autoware-compatible vector map is to use Vector Map Builder, a free web-based tool provided by TIER IV, Inc.. Vector Map Builder allows you to create lanes and add additional regulatory elements such as stop signs or traffic lights using a point cloud map as a reference.
For open-source software options, MapToolbox is a plugin for Unity specifically designed to create Lanelet2 maps for Autoware. Although JOSM is another open-source tool that can be used to create Lanelet2 maps, be aware that a number of modifications must be done manually to make the map compatible with Autoware. This process can be tedious and time-consuming, so the use of JOSM is not recommended.
If it is not possible to create HD maps yourself, you can use a mapping service from the following Autoware-compatible map providers instead:
MAP IV, Inc.
AISAN TECHNOLOGY CO., LTD.
TomTom
The table below shows each company's mapping technology and the types of HD maps they support.
Company Mapping technology Available maps MAP IV, Inc. SLAM Point cloud and vector maps AISAN TECHNOLOGY CO., LTD. MMS Point cloud and vector maps TomTom MMS Vector map*
Note
Maps provided by TomTom use their proprietary AutoStream format, not Lanelet2. The open-source AutoStreamForAutoware tool can be used to convert an AutoStream map to a Lanelet2 map. However, the converter is still in its early stages and has some known limitations.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/converting-utm-to-mgrs-map/","title":"Converting UTM maps to MGRS map format","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/converting-utm-to-mgrs-map/#converting-utm-maps-to-mgrs-map-format","title":"Converting UTM maps to MGRS map format","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/converting-utm-to-mgrs-map/#overview","title":"Overview","text":"
If you want to use MGRS (Military Grid Reference System) format in Autoware, you need to convert UTM (Universal Transverse Mercator) map to MGRS format. In order to do that, we will use UTM to MGRS pointcloud converter ROS 2 package provided by Leo Drive.
After the installation of converter tool, we need to define northing, easting and ellipsoid height of local UTM map origin in pc_utm_to_mgrs_converter.param.yaml. For example, you can use latitude, longitude and altitude values in the navsatfix message from your GNSS/INS sensor.
After that, you need to convert latitude and longitude values to northing and easting values. You can use any converter on the internet for converting latitude longitude values to UTM. (i.e., UTMconverter)
Now, we are ready to update pc_utm_to_mgrs_converter.param.yaml, example for our navsatfix message:
/**:\n ros__parameters:\n # Northing of local origin\n- Northing: 4520550.0\n+ Northing: 4542871.33\n\n # Easting of local origin\n- Easting: 698891.0\n+ Easting: 658659.84\n\n # Elipsoid Height of local origin\n- ElipsoidHeight: 47.62\n+ ElipsoidHeight: 74.28\n
Lastly, we will update input and pointcloud the map path in pc_utm_to_mgrs_converter.launch.xml:
The conversion process will be started, you should see Saved <YOUR-MAP-POINTS-SIZE> data points saved to <YOUR-OUTPUT-MAP-PATH> message on your terminal. So, MGRS format pointcloud map saved on your output map directory.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/","title":"Creating a vector map","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/#creating-a-vector-map","title":"Creating a vector map","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/#overview","title":"Overview","text":"
In this section, we will explain how to create Lanelet2 maps with TIER IV's vector map builder tool.
If you want to look at another method, MapToolbox is a plugin for Unity specifically designed to create Lanelet2 maps for Autoware. We didn't recommend JOSM since it needs modifications for Autoware usage. So, this process can be tedious and time-consuming.
You need a TIER IV account for using vector map builder tool. So, if you have not before, please create a TIER IV account in order to use vector map builder tool. For more information about this tool, please check the official guide.
You can follow these pages for creating a Lanelet2 map and its regulatory elements.
Behavior velocity planner's crosswalk module plans velocity to stop or decelerate for pedestrians approaching or walking on a crosswalk. In order to operate that, we will add crosswalk attribute to our lanelet2 map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/crosswalk/#creating-a-crosswalk-attribute","title":"Creating a crosswalk attribute","text":"
In order to create a crosswalk on your map, please follow these steps:
Click Abstraction button on top panel.
Select Crosswalk from the panel.
Click and draw crosswalk on your pointcloud map.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/crosswalk/#testing-created-crosswalk-with-planning-simulator","title":"Testing created crosswalk with planning simulator","text":"
After the completing of creating the map, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Click 2D Pose Estimate button on rviz or press P and give a pose for initialization.
Click 2D Goal Pose button on rviz or press G and give a pose for goal point.
We need to add pedestrians to crosswalk, so activate interactive pedestrians from Tool Properties panel on rviz.
After that, please press Shift, then click right click button for inserting pedestrians.
You can control inserted pedestrian via dragging right click.
Crosswalk markers on rviz:
Crosswalk test on the created map.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/","title":"Creating a Lanelet","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#creating-a-lanelet","title":"Creating a Lanelet","text":"
At this page, we will explain how to create a simple lanelet on your pointcloud map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#creating-a-lanelet2","title":"Creating a Lanelet2","text":"
Firstly, we need to import our pointcloud map to vector map builder tool:
Please click File.
Then, click Import PCD.
Click Browse and select your .pcd file.
You will display the point cloud on your vector map builder tool after the upload is complete:
Uploaded pointcloud map file on vector map builder
Now, we are ready to create lanelet2 map on our pointcloud map:
Please click Create.
Then, click Create Lanelet2Maps.
Please fill your map name
Please fill your MGRS zone. (At tutorial_vehicle, MGRS grid zone: 35T - MGRS 100,000-meter square: PF)
Click Create.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#creating-a-simple-lanelet","title":"Creating a simple lanelet","text":"
In order to create a simple lanelet on your map, please follow these steps:
CLick Lanelet2Maps on the bar
Enable Lanelet mode via selecting Lanelet.
Then, you can click the pointcloud map to create lanelet.
If your lanelet is finished, you can disable Lanelet.
If you want to change your lanelet width, click lanelet --> Change Lanelet Width, then you can enter the lanelet width.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#join-two-lanelets","title":"Join two lanelets","text":"
In order to join two lanelets, please follow these steps:
Please create two distinct lanelet.
Select a Lanelet, then press Shift and select other lanelet.
Now, you can see Join Lanelets button, just press it.
In order to add (join) two or more lanelets to another lanelet, please follow these steps:
Create multiple lanelets.
You can join the first two lanelets like the steps before.
Please check end points ids of first lanelet.
Then you need to change these ids with third lanelet's start point. (Please change with selecting linestring of lanelet)
You will see two next lanes of the first lanelet will be appeared.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#change-speed-limit-of-lanelet","title":"Change Speed Limit Of Lanelet","text":"
In order to change the speed limit of lanelet, please follow these steps:
Select the lanelet where the speed limit will be changed
Set speed limit on the right panel.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/lanelet2/#test-lanelets-with-planning-simulator","title":"Test lanelets with planning simulator","text":"
After the completing of creating lanelets, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Click 2D Pose Estimate button on rviz or press P and give a pose for initialization.
Click 2D Goal Pose button on rviz or press G and give a pose for goal point.
Testing our created vector map with planning simulator"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/","title":"Stop Line","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/#stop-line","title":"Stop Line","text":"
Behavior velocity planner's stop line module plans velocity to stop right before stop lines and restart driving after stopped. In order to operate that, we will add stop line attribute to our lanelet2 map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/#creating-a-stop-line-regulatory-element","title":"Creating a stop line regulatory element","text":"
In order to create a stop line on your pointcloud map, please follow these steps:
Please select lanelet which stop line to be added.
Click Abstraction button on top panel.
Select Stop Line from the panel.
Click on the desired area for inserting stop line.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/stop-line/#testing-created-the-stop-line-element-with-planning-simulator","title":"Testing created the stop line element with planning simulator","text":"
After the completing of creating the map, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Behavior velocity planner's traffic light module plans velocity according to the traffic light status. In order to operate that, we will add stop line attribute to our lanelet2 map.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/traffic-light/#creating-a-traffic-light-regulatory-element","title":"Creating a traffic light regulatory element","text":"
In order to create a traffic light on your pointcloud map, please follow these steps:
Please select lanelet which traffic light to be added.
Click Abstraction button on top panel.
Select Traffic Light from the panel.
Click on the desired area for inserting traffic light.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/creating-vector-map/traffic-light/#testing-created-the-traffic-light-element-with-planning-simulator","title":"Testing created the traffic light element with planning simulator","text":"
After the completing of creating the map, we need to save it. To that please click File --> Export Lanelet2Maps then download.
After the download is finished, we need to put lanelet2 map and pointcloud map on the same location. The directory structure should be like this:
Click 2D Pose Estimate button on rviz or press P and give a pose for initialization.
Click Panels -> Add new panel, select TrafficLightPublishPanel, and then press OK.
In TrafficLightPublishPanel, set the ID and color of the traffic light.
Then, Click SET and PUBLISH button.
Click 2D Goal Pose button on rviz or press G and give a pose for goal point.
You can see the traffic light marker on the rviz screen if you set the traffic light color as RED.
Traffic Light markers on rviz:
Traffic light test on the created map.
Video Demonstration:
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/","title":"Available Open Source SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/#available-open-source-slam","title":"Available Open Source SLAM","text":"
This page provides the list of available open source Simultaneous Localization And Mapping (SLAM) implementation that can be used to generate a point cloud (.pcd) map file.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/#selecting-which-implementation-to-use","title":"Selecting which implementation to use","text":"
Lidar odometry drifts accumulatively as time goes by and there is solutions to solve that problem such as graph optimization, loop closure and using gps sensor to decrease accumulative drift error. Because of that, a SLAM algorithm should have loop closure feature, graph optimization and should use gps sensor. Additionally, some of the algorithms are using IMU sensor to add another factor to graph for decreasing drift error. While some of the algorithms requires 9-axis IMU sensor strictly, some of them requires only 6-axis IMU sensor or not even using the IMU sensor. Before choosing an algorithm to create maps for Autoware please consider these factors depends on your sensor setup or expected quality of generated map.
Commonly used open-source SLAM implementations are lidarslam-ros2 (LiDAR, IMU*) and LIO-SAM (LiDAR, IMU, GNSS). The required sensor data for each algorithm is specified in the parentheses, where an asterisk (*) indicates that such sensor data is optional. For supported LiDAR models, please check the GitHub repository of each algorithm. While these ROS 2-based SLAM implementations can be easily installed and used directly on the same machine that runs Autoware, it is important to note that they may not be as well-tested or as mature as ROS 1-based alternatives.
The notable open-source SLAM implementations that are based on ROS 1 include hdl-graph-slam (LiDAR, IMU*, GNSS*), LeGO-LOAM (LiDAR, IMU*), LeGO-LOAM-BOR (LiDAR), and LIO-SAM (LiDAR, IMU, GNSS).
Most of these algorithms already have a built-in loop-closure and pose graph optimization. However, if the built-in, automatic loop-closure fails or does not work correctly, you can use Interactive SLAM to adjust and optimize a pose graph manually.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/#list-of-third-party-slam-implementations","title":"List of Third Party SLAM Implementations","text":"Package Name Explanation Repository Link Loop Closure Sensors ROS Version Dependencies FAST-LIO-LC A computationally efficient and robust LiDAR-inertial odometry package with loop closure module and graph optimization https://github.com/yanliang-wang/FAST_LIO_LC ✓ LidarIMUGPS [Optional] ROS 1 ROS MelodicPCL >= 1.8Eigen >= 3.3.4GTSAM >= 4.0.0 FAST_LIO_SLAM FAST_LIO_SLAM is the integration of FAST_LIO and SC-PGO which is scan context based loop detection and GTSAM based pose-graph optimization https://github.com/gisbi-kim/FAST_LIO_SLAM ✓ LidarIMUGPS [Optional] ROS 1 PCL >= 1.8Eigen >= 3.3.4 FD-SLAM FD_SLAM is Feature&Distribution-based 3D LiDAR SLAM method based on Surface Representation Refinement. In this algorithm novel feature-based Lidar odometry used for fast scan-matching, and used a proposed UGICP method for keyframe matching https://github.com/SLAMWang/FD-SLAM ✓ LidarIMU [Optional]GPS ROS 1 PCLg2oSuitesparse hdl_graph_slam An open source ROS package for real-time 6DOF SLAM using a 3D LIDAR. It is based on 3D Graph SLAM with NDT scan matching-based odometry estimation and loop detection. It also supports several graph constraints, such as GPS, IMU acceleration (gravity vector), IMU orientation (magnetic sensor), and floor plane (detected in a point cloud) https://github.com/koide3/hdl_graph_slam ✓ LidarIMU [Optional]GPS [Optional] ROS 1 PCLg2oOpenMP IA-LIO-SAM IA_LIO_SLAM is created for data acquisition in unstructured environment and it is a framework for Intensity and Ambient Enhanced Lidar Inertial Odometry via Smoothing and Mapping that achieves highly accurate robot trajectories and mapping https://github.com/minwoo0611/IA_LIO_SAM ✓ LidarIMUGPS ROS 1 GTSAM ISCLOAM ISCLOAM presents a robust loop closure detection approach by integrating both geometry and intensity information https://github.com/wh200720041/iscloam ✓ Lidar ROS 1 Ubuntu 18.04ROS MelodicCeresPCLGTSAMOpenCV LeGO-LOAM-BOR LeGO-LOAM-BOR is improved version of the LeGO-LOAM by improving quality of the code, making it more readable and consistent. Also, performance is improved by converting processes to multi-threaded approach https://github.com/facontidavide/LeGO-LOAM-BOR ✓ LidarIMU ROS 1 ROS MelodicPCLGTSAM LIO_SAM A framework that achieves highly accurate, real-time mobile robot trajectory estimation and map-building. It formulates lidar-inertial odometry atop a factor graph, allowing a multitude of relative and absolute measurements, including loop closures, to be incorporated from different sources as factors into the system https://github.com/TixiaoShan/LIO-SAM ✓ LidarIMUGPS [Optional] ROS 1ROS 2 PCLGTSAM Optimized-SC-F-LOAM An improved version of F-LOAM and uses an adaptive threshold to further judge the loop closure detection results and reducing false loop closure detections. Also it uses feature point-based matching to calculate the constraints between a pair of loop closure frame point clouds and decreases time consumption of constructing loop frame constraints https://github.com/SlamCabbage/Optimized-SC-F-LOAM ✓ Lidar ROS 1 PCLGTSAMCeres SC-A-LOAM A real-time LiDAR SLAM package that integrates A-LOAM and ScanContext. https://github.com/gisbi-kim/SC-A-LOAM ✓ Lidar ROS 1 GTSAM >= 4.0 SC-LeGO-LOAM SC-LeGO-LOAM integrated LeGO-LOAM for lidar odometry and 2 different loop closure methods: ScanContext and Radius search based loop closure. While ScanContext is correcting large drifts, radius search based method is good for fine-stitching https://github.com/irapkaist/SC-LeGO-LOAM ✓ LidarIMU ROS 1 PCLGTSAM"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/","title":"FAST_LIO_LC","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/#fast_lio_lc","title":"FAST_LIO_LC","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/#what-is-fast_lio_lc","title":"What is FAST_LIO_LC?","text":"
A computationally efficient and robust LiDAR-inertial odometry package with loop closure module and graph optimization.
mkdir -p ~/ws_fastlio_lc/src\n cd ~/ws_fastlio_lc/src\n git clone https://github.com/gisbi-kim/FAST_LIO_SLAM.git\n git clone https://github.com/Livox-SDK/livox_ros_driver\n cd ..\n catkin_make\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-lc/#2-set-parameters","title":"2) Set parameters","text":"
After downloading the repository, change topic and sensor settings on the config file (workspace/src/FAST_LIO_LC/FAST_LIO/config/ouster64_mulran.yaml) with the lidar topic name in your bag file.
For imu-lidar compatibility, extrinsic matrices from calibration must be changed.
To enable auto-save, pcd_save_enable must be 1 from the launch file (workspace/src/FAST_LIO_LC/FAST_LIO/launch/mapping_ouster64_mulran.launch).
# open new terminal: run FAST-LIO\nroslaunch fast_lio mapping_ouster64.launch\n\n# open the other terminal tab: run SC-PGO\nroslaunch aloam_velodyne fastlio_ouster64.launch\n\n# play bag file in the other terminal\nrosbag play RECORDED_BAG.bag --clock\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/","title":"FAST_LIO_SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/#fast_lio_slam","title":"FAST_LIO_SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/#what-is-fast_lio_slam","title":"What is FAST_LIO_SLAM?","text":"
FAST_LIO_SLAM is the integration of FAST_LIO and SC-PGO which is scan context based loop detection and GTSAM based pose-graph optimization.
mkdir -p ~/catkin_fastlio_slam/src\n cd ~/catkin_fastlio_slam/src\n git clone https://github.com/gisbi-kim/FAST_LIO_SLAM.git\n git clone https://github.com/Livox-SDK/livox_ros_driver\n cd ..\n catkin_make\n source devel/setup.bash\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fast-lio-slam/#2-set-parameters","title":"2) Set parameters","text":"
Set imu and lidar topic on Fast_LIO/config/ouster64.yaml
# terminal 1: run FAST-LIO2\nroslaunch fast_lio mapping_ouster64.launch\n\n # open the other terminal tab: run SC-PGO\ncd ~/catkin_fastlio_slam\n source devel/setup.bash\n roslaunch aloam_velodyne fastlio_ouster64.launch\n\n # play bag file in the other terminal\nrosbag play xxx.bag -- clock --pause\n
You may have an interest in this version of FAST-LIO + Loop closure, implemented by yanliang-wang
Maintainer: Giseop Kim (paulgkim@kaist.ac.kr)
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/","title":"FD-SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/#fd-slam","title":"FD-SLAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/#what-is-fd-slam","title":"What is FD-SLAM?","text":"
FD_SLAM is Feature&Distribution-based 3D LiDAR SLAM method based on Surface Representation Refinement. In this algorithm novel feature-based Lidar odometry used for fast scan-matching, and used a proposed UGICP method for keyframe matching.
/hdl_graph_slam/dump (hdl_graph_slam/DumpGraph)\n- save all the internal data (point clouds, floor coeffs, odoms, and pose graph) to a directory.\n\n/hdl_graph_slam/save_map (hdl_graph_slam/SaveMap)\n- save the generated map as a PCD file.\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/fd-slam/#3-set-parameters","title":"3) Set parameters","text":"
All the configurable parameters are listed in launch/****.launch as ros params.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/hdl-graph-slam/","title":"hdl_graph_slam","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/hdl-graph-slam/#hdl_graph_slam","title":"hdl_graph_slam","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/hdl-graph-slam/#what-is-hdl_graph_slam","title":"What is hdl_graph_slam?","text":"
An open source ROS package for real-time 6DOF SLAM using a 3D LIDAR. It is based on 3D Graph SLAM with NDT scan matching-based odometry estimation and loop detection. It also supports several graph constraints, such as GPS, IMU acceleration (gravity vector), IMU orientation (magnetic sensor), and floor plane (detected in a point cloud).
Kenji Koide, Jun Miura, and Emanuele Menegatti, A Portable 3D LIDAR-based System for Long-term and Wide-area People Behavior Measurement, Advanced Robotic Systems, 2019 [link].
[Active Intelligent Systems Laboratory, Toyohashi University of Technology, Japan] [Mobile Robotics Research Team, National Institute of Advanced Industrial Science and Technology (AIST), Japan]
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/ia-lio-slam/","title":"IA-LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/ia-lio-slam/#ia-lio-sam","title":"IA-LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/ia-lio-slam/#what-is-ia-lio-sam","title":"What is IA-LIO-SAM?","text":"
IA_LIO_SLAM is created for data acquisition in unstructured environment and it is a framework for Intensity and Ambient Enhanced Lidar Inertial Odometry via Smoothing and Mapping that achieves highly accurate robot trajectories and mapping.
# open new terminal: run IA_LIO\n source devel/setup.bash\n roslaunch lio_sam mapping_ouster64.launch\n\n # play bag file in the other terminal\n rosbag play RECORDED_BAG.bag --clock\n
Thank you for citing IA-LIO-SAM(./config/doc/KRS-2021-17.pdf) if you use any of this code.
Part of the code is adapted from LIO-SAM (IROS-2020).
@inproceedings{legoloam2018shan,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Shan, Tixiao and Englot, Brendan},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
IA-LIO-SAM is based on LIO-SAM (T. Shan, B. Englot, D. Meyers, W. Wang, C. Ratti, and D. Rus. LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping).
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/","title":"ISCLOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#iscloam","title":"ISCLOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#what-is-iscloam","title":"What is ISCLOAM?","text":"
ISCLOAM presents a robust loop closure detection approach by integrating both geometry and intensity information.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#build-and-run","title":"Build and Run","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#1-clone-repository","title":"1. Clone repository","text":"
cd ~/catkin_ws/src\ngit clone https://github.com/wh200720041/iscloam.git\ncd ..\ncatkin_make -j1\nsource ~/catkin_ws/devel/setup.bash\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#2-set-parameter","title":"2. Set Parameter","text":"
Change the bag location and sensor parameters on launch files.
if you would like to generate the map of environment at the same time, you can run
roslaunch iscloam iscloam_mapping.launch\n
Note that the global map can be very large, so it may takes a while to perform global optimization, some lag is expected between trajectory and map since they are running in separate thread. More CPU usage will happen when loop closure is identified.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/iscloam/#ground-truth-comparison","title":"Ground Truth Comparison","text":"
If you use this work for your research, you may want to cite the paper below, your citation will be appreciated
@inproceedings{wang2020intensity,\n author={H. {Wang} and C. {Wang} and L. {Xie}},\n booktitle={2020 IEEE International Conference on Robotics and Automation (ICRA)},\n title={Intensity Scan Context: Coding Intensity and Geometry Relations for Loop Closure Detection},\n year={2020},\n volume={},\n number={},\n pages={2095-2101},\n doi={10.1109/ICRA40945.2020.9196764}\n}\n
Thanks for A-LOAM and LOAM(J. Zhang and S. Singh. LOAM: Lidar Odometry and Mapping in Real-time) and LOAM_NOTED.
Author: Wang Han, Nanyang Technological University, Singapore
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/","title":"LeGO-LOAM-BOR","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/#lego-loam-bor","title":"LeGO-LOAM-BOR","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/#what-is-lego-loam-bor","title":"What is LeGO-LOAM-BOR?","text":"
LeGO-LOAM-BOR is improved version of the LeGO-LOAM by improving quality of the code, making it more readable and consistent. Also, performance is improved by converting processes to multi-threaded approach.
cd ~/catkin_ws/src\ngit clone https://github.com/facontidavide/LeGO-LOAM-BOR.git\ncd ..\ncatkin_make\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lego-loam-bor/#2-set-parameters","title":"2) Set parameters","text":"
Thank you for citing our LeGO-LOAM paper if you use any of this code:
@inproceedings{legoloam2018,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Tixiao Shan and Brendan Englot},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/","title":"LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#lio-sam","title":"LIO-SAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#what-is-lio-sam","title":"What is LIO-SAM?","text":"
A framework that achieves highly accurate, real-time mobile robot trajectory estimation and map-building. It formulates lidar-inertial odometry atop a factor graph, allowing a multitude of relative and absolute measurements, including loop closures, to be incorporated from different sources as factors into the system.
*Robosense lidars aren't supported officially, but their Helios series can be used as Velodyne lidars.
The system architecture of LIO-SAM method described in the following diagram, please look at the official repository for getting more information.
System Architecture of LIO-SAM
We are using Robosense Helios 5515 and CLAP B7 sensor on tutorial_vehicle, so we will use these sensors for running LIO-SAM.
Additionally, LIO-SAM tested with Applanix POS LVX and Hesai Pandar XT32 sensor setup. Some additional information according to the sensors will be provided in this page.
After the building of LIO-SAM, we need to record ROS 2 Bag file with including necessary topics for LIO-SAM. The necessary topics are described in the config file on LIO-SAM.
ROS 2 Bag example for LIO-SAM with Robosense Helios and CLAP B7
We will update sensor settings also. Since Robosense Lidars aren't officially supported, we will set our 32-channel Robosense Helios 5515 lidar as Velodyne:
The mapping direction is towards to the going direction in the real world. If LiDAR sensor is backwards, according to the direction you are moving, then you need to change the extrinsicRot too. Unless the IMU tries to go in the wrong direction, and it may occur problems.
For example, in our Applanix POS LVX and Hesai Pandar XT32 setup, IMU direction was towards to the going direction and LiDAR direction has 180 degree difference in Z-axis according to the IMU direction. In other words, they were facing back to each other. The tool may need a transformation for IMU for that.
In that situation, the calibration parameters changed as this:
If you are set configurations and create bag file for LIO-SAM, you can launch LIO-SAM with:
ros2 launch lio_sam run.launch.py\n
The rviz2 screen will be open, then you can play your bag file:
ros2 bag play <YOUR-BAG-FILE>\n
If the mapping process is finished, you can save map with calling this service:
ros2 service call /lio_sam/save_map lio_sam/srv/SaveMap \"{resolution: 0.2, destination: <YOUR-MAP-DIRECTORY>}\"\n
Here is the video for demonstration of LIO-SAM mapping in our campus environment:
The output map format is local UTM, we will change local UTM map to MGRS format for tutorial_vehicle. Also, if you want change UTM to MGRS for autoware, please follow convert-utm-to-mgrs-map page.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#example-result","title":"Example Result","text":"Sample Map Output for our Campus Environment"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/lio-sam/#paper","title":"Paper","text":"
Thank you for citing LIO-SAM (IROS-2020) if you use any of this code.
@inproceedings{liosam2020shan,\n title={LIO-SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping},\n author={Shan, Tixiao and Englot, Brendan and Meyers, Drew and Wang, Wei and Ratti, Carlo and Rus Daniela},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={5135-5142},\n year={2020},\n organization={IEEE}\n}\n
Part of the code is adapted from LeGO-LOAM.
@inproceedings{legoloam2018shan,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Shan, Tixiao and Englot, Brendan},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
LIO-SAM is based on LOAM (J. Zhang and S. Singh. LOAM: Lidar Odometry and Mapping in Real-time).
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/","title":"Optimized-SC-F-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#optimized-sc-f-loam","title":"Optimized-SC-F-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#what-is-optimized-sc-f-loam","title":"What is Optimized-SC-F-LOAM?","text":"
An improved version of F-LOAM and uses an adaptive threshold to further judge the loop closure detection results and reducing false loop closure detections. Also it uses feature point-based matching to calculate the constraints between a pair of loop closure frame point clouds and decreases time consumption of constructing loop frame constraints.
In this folder, Ground Truth information, optimized pose information, F-LOAM pose information and time information are stored
mkdir -p ~/message/Scans\n\nChange line 383 in the laserLoopOptimizationNode.cpp to your own \"message\" folder path\n
(Do not forget to rebuild your package)
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#3-set-parameters","title":"3) Set parameters","text":"
Set LIDAR topic and LIDAR properties on 'sc_f_loam_mapping.launch'
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#example-result","title":"Example Result","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#results-on-kitti-sequence-00-and-sequence-05","title":"Results on KITTI Sequence 00 and Sequence 05","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/optimized-sc-f-loam/#comparison-of-trajectories-on-kitti-dataset","title":"Comparison of trajectories on KITTI dataset","text":"
Test on KITTI sequence You can download the sequence 00 and 05 datasets from the KITTI official website and convert them into bag files using the kitti2bag open source method.
00: 2011_10_03_drive_0027 000000 004540
05: 2011_09_30_drive_0018 000000 002760
See the link: https://github.com/ethz-asl/kitti_to_rosbag
Thanks for SC-A-LOAM(Scan context: Egocentric spatial descriptor for place recognition within 3d point cloud map) and F-LOAM(F-LOAM : Fast LiDAR Odometry and Mapping).
@misc{https://doi.org/10.48550/arxiv.2204.04932,\n doi = {10.48550/ARXIV.2204.04932},\n\n url = {https://arxiv.org/abs/2204.04932},\n\n author = {Liao, Lizhou and Fu, Chunyun and Feng, Binbin and Su, Tian},\n\n keywords = {Robotics (cs.RO), FOS: Computer and information sciences, FOS: Computer and information sciences},\n\n title = {Optimized SC-F-LOAM: Optimized Fast LiDAR Odometry and Mapping Using Scan Context},\n\n publisher = {arXiv},\n\n year = {2022},\n\n copyright = {arXiv.org perpetual, non-exclusive license}\n}\n
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/","title":"SC-A-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/#sc-a-loam","title":"SC-A-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/#what-is-sc-a-loam","title":"What is SC-A-LOAM?","text":"
A real-time LiDAR SLAM package that integrates A-LOAM and ScanContext.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-a-loam/#4-saving-as-pcd-file","title":"4) Saving as PCD file","text":"
"},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-lego-loam/","title":"SC-LeGO-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-lego-loam/#sc-lego-loam","title":"SC-LeGO-LOAM","text":""},{"location":"how-to-guides/integrating-autoware/creating-maps/open-source-slam/sc-lego-loam/#what-is-sc-lego-loam","title":"What is SC-LeGO-LOAM?","text":"
SC-LeGO-LOAM integrated LeGO-LOAM for lidar odometry and 2 different loop closure methods: ScanContext and Radius search based loop closure. While ScanContext is correcting large drifts, radius search based method is good for fine-stitching.
@INPROCEEDINGS { gkim-2018-iros,\n author = {Kim, Giseop and Kim, Ayoung},\n title = { Scan Context: Egocentric Spatial Descriptor for Place Recognition within {3D} Point Cloud Map },\n booktitle = { Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems },\n year = { 2018 },\n month = { Oct. },\n address = { Madrid }\n}\n
and
@inproceedings{legoloam2018,\n title={LeGO-LOAM: Lightweight and Ground-Optimized Lidar Odometry and Mapping on Variable Terrain},\n author={Shan, Tixiao and Englot, Brendan},\n booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},\n pages={4758-4765},\n year={2018},\n organization={IEEE}\n}\n
In some cases, for example, when your created point cloud map is either too dense or too large (i.e., exceeding 300 MB), you may want to downsample it for improved computational and memory efficiency. Also, you can consider using dynamic map loading with partial loading, please check map_loader package for more information.
At tutorial_vehicle implementation we will use the whole map, so we will downsample it with using CloudCompare.
Please check the official page for installing options.
"},{"location":"how-to-guides/integrating-autoware/creating-maps/pointcloud-map-downsampling/#downsampling-a-pointcloud-map","title":"Downsampling a pointcloud map","text":"
There are three subsampling methods on CloudCompare, we are using Space method for subsampling, but you can use other methods if you want.
Please open CloudCompare and drag your pointcloud to here, then you can select your pointcloud map by just clicking on the map at the DB tree panel.
Then you can click subsample button on the top panel.
CloudCompare
Please select on your subsample method, we will use space for tutorial_vehicle.
Then you can select options. For example, we need to determine minimum space between points. (Please be careful in this section, subsampling is depending on your map size, computer performance, etc.) We will set this value 0.2 for tutorial_vehicle's map.
Pointcloud subsampling
After the subsampling process is finished, you should select pointcloud on the DB Tree panel as well.
Select your downsampled pointcloud
Now, you can save your downsampled pointcloud with ctrl + s or you can click save button from File bar. Then, this pointcloud can be used by autoware.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/","title":"Calibrating your sensors","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#calibrating-your-sensors","title":"Calibrating your sensors","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#overview","title":"Overview","text":"
Autoware expects to have multiple sensors attached to the vehicle as input to perception, localization, and planning stack. These sensors must be calibrated correctly and their positions must be defined using either urdf files (as in sample_sensor_kit) or as tf launch files.
Navigation2 provides a good tutorial for camera internal calibration.
AutoCore provides a light-weight tool.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#lidar-lidar-calibration","title":"Lidar-lidar calibration","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/calibrating-sensors/#lidar-lidar-calibration-tool-from-autocore","title":"Lidar-Lidar Calibration tool from Autocore","text":"
LL-Calib on GitHub, provided by AutoCore, is a lightweight toolkit for online/offline 3D LiDAR to LiDAR calibration. It's based on local mapping and \"GICP\" method to derive the relation between main and sub lidar. Information on how to use the tool, troubleshooting tips and example rosbags can be found at the above link.
Developed by MathWorks, The Lidar Camera Calibrator app enables you to interactively estimate the rigid transformation between a lidar sensor and a camera.
SensorsCalibration toolbox v0.1: One more open source method for Lidar-camera calibration. This is a project for LiDAR to camera calibration,including automatic calibration and manual calibration
Developed by AutoCore, an easy-to-use lightweight toolkit for Lidar-camera-calibration is proposed. Only in three steps, a fully automatic calibration will be done.
Developed by APRIL Lab at Zhejiang University in China, the LI-Calib calibration tool is a toolkit for calibrating the 6DoF rigid transformation and the time offset between a 3D LiDAR and an IMU, based on continuous-time batch optimization. IMU-based cost and LiDAR point-to-surfel (surfel = surface element) distance are minimized jointly, which renders the calibration problem well-constrained in general scenarios.
AutoCore has forked the original LI-Calib tool and overwritten the Lidar input for more general usage. Information on how to use the tool, troubleshooting tips and example rosbags can be found at the LI-Calib fork on GitHub.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/","title":"Creating vehicle and sensor description","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/#creating-vehicle-and-sensor-description","title":"Creating vehicle and sensor description","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/#introduction","title":"Introduction","text":"
Resolves the positions of sensors with base_link as the parent frame and defines the positions and orientations based on sensors_calibration.yaml in individual_params.
In Autoware, <YOUR_SENSOR_KIT_description>/config/sensors_calibration.yaml is not used.
A sensor_kit refers to a subset that includes multiple sensors, and sensor_kit_base_link is the name of its frame. The positions and orientations within the kit are defined in sensor_kit.xacro.
Resolves the positions of sensors with sensor_kit_base_link as the parent and defines the positions and orientations based on sensor_kit_calibration.yaml in individual_params.
In Autoware, <YOUR_SENSOR_KIT_description>/config/sensor_kit_calibration.yaml is not used.
The individual_parameter is where parameters referenced by sensors.xacro and sensor_kit.xacro are stored. As the name imply, it is intended to manage parameters for multiple individual instances.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-and-sensor-description/creating-vehicle-and-sensor-description/#introduction-to-various-parameters","title":"Introduction to Various Parameters","text":"
vehicle_interface.launch.xml is the launch file related to the drive system. Please modify it according to the configuration of your vehicle's drive system.
If you are operating multiple vehicles, use the vehicle_id to switch to the corresponding configuration for each vehicle.
pointcloud_preprocessor.launch.py is the launch file to convert the raw sensor data. For more information, please click here.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/","title":"Creating a vehicle interface for an Ackermann kinematic model","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#creating-a-vehicle-interface-for-an-ackermann-kinematic-model","title":"Creating a vehicle interface for an Ackermann kinematic model","text":"
This page introduces a module vehicle interface and explains how to implement it.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#what-is-a-vehicle-interface","title":"What is a vehicle interface","text":"
Vehicle interface is an interface that connects the control commands and your vehicle's control device. Autoware publishes control commands such as:
Velocity control
Steering control
Car light commands
Then, the vehicle interface converts these commands into actuation such like:
Motor and brake activation
Steering wheel operation
Lighting control
So think of the vehicle interface as a module that runs the vehicle's control device to realize the input commands provided by Autoware.
An example of inputs and outputs for vehicle interface
This page shows you a brief explanation how to implement your vehicle interface, but you can see further information of vehicle interface in the \"design\" page.
Note that there is no package named \"vehicle interface\" prepared in Autoware. It is a necessary package to actuate your vehicle, but you have to create one by yourself since it is very specific to your vehicle's control device.
For example, if you are using a by-wire kit PACMod, a vehicle interface named pacmod_interface published by TIER IV, Inc. is available. However, if you have constructed something original and haven't found an open source vehicle interface applicable, you have to implement your own vehicle interface from scratch.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#how-to-implement-a-vehicle-interface","title":"How to implement a vehicle interface","text":"
The following instructions describe how to create a vehicle interface.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#1-create-a-directory-for-vehicle-interface","title":"1. Create a directory for vehicle interface","text":"
It is recommended to create your vehicle interface at <your-autoware-dir>/src/vehicle/external
cd <your-autoware-dir>/src/vehicle/external\n
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#2-install-or-implement-your-own-vehicle-interface","title":"2. Install or implement your own vehicle interface","text":"
If there is an already complete vehicle interface package (like pacmod_interface), you can install it to your environment. If not, you have to implement your own vehicle interface by yourself. Let's create a new package by ros2 pkg create. The following example will show you how to create a vehicle interface package named my_vehicle_interface.
Then, you should write your implementation of vehicle interface in my_vehicle_interface/src. Again, since this implementation is so specific to the control device of your vehicle, it is beyond the scope of this document to describe how to implement your vehicle interface in detail. Here are some factors that might be considered.
Subscription of control command topics from Autoware
Communication between the vehicle interface and your vehicle's control device
Modification of control values if needed
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#3-prepare-a-launch-file","title":"3. Prepare a launch file","text":"
After you implement your vehicle interface or you want to debug it by launching it, create a launch file of your vehicle interface, and include it to vehicle_interface.launch.xml.
Do not get confused. First, you need to create a launch file for your own vehicle interface module (like my_vehicle_interface.launch.xml) and then include that to vehicle_interface.launch.xml which exists in another directory. Here are the details.
Add a launch directory in the my_vehicle_interface directory, and create a launch file of your own vehicle interface in it. Take a look at Creating a launch file in the ROS 2 documentation.
Next, go to <your-autoware-dir>/src/vehicle, copy the directory /sample_vehicle_launch/, and paste it to the same place (which means it should be lined up with external and sample_vehicle_launch).
You have to rename each \"sample_vehicle\" to something else. For example, if you want to rename \"sample_vehicle\" to \"my_vehicle_name\", you need to change the following. Note that it is restricted to keep the \"_launch\" and \"_description\" part.
After you rename your directories, rename each \"sample_vehicle\" to \"my_vehicle_name\" in the source code.
my_vehicle_name_description/CMakeLists.txt
my_vehicle_name_description/package.xml
my_vehicle_name_description/urdf/vehicle.xacro (there are two parts)
my_vehicle_name_launch/CMakeLists.txt
my_vehicle_name_launch/package.xml
README.md
Include your launch file to my_vehicle_name_launch/my_vehicle_name_launch/launch/vehicle_interface.launch.xml by opening it and add the include terms like below.
Finally, your directory structure may look like below. Most of the files are omitted for clarity, but the files shown here needs modification as said in the previous and current process.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#4-build-the-vehicle-interface-package-and-the-launch-package","title":"4. Build the vehicle interface package and the launch package","text":"
Build three packages my_vehicle_interface, my_vehicle_name_launch and my_vehicle_name_description by colcon build, or you can just build the entire Autoware if you have done other things.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#5-when-you-launch-autoware","title":"5. When you launch Autoware","text":"
Finally, you are done implementing your vehicle interface module! Be careful that you need to launch Autoware with the proper vehicle_model option like the example below. This example is launching planning simulator.
You can subdivide your vehicle interface into smaller packages if you want. Then your directory structure may look like below (not the only way though). Do not forget to launch all packages in my_vehicle_interface.launch.xml.
If you are using a vehicle interface and launch package from a open git repository, or created your own as a git repository, it is highly recommended to add those repositories to your autoware.repos file which is located to directly under your autoware folder like the example below. You can specify the branch or commit hash by the version tag.
autoware.repos
# vehicle (this section should be somewhere in autoware.repos and add the below)\nvehicle/my_vehicle_name_launch:\ntype: git\nurl: https://github.com/<repository-name-A>/my_vehicle_name_launch.git\nversion: main\nvehicle/external/my_vehicle_interface:\ntype: git\nurl: https://github.com/<repository-name-B>/my_vehicle_interface.git\nversion: main\n
Then you can import your entire environment easily to another local device by using the vcs import command. (See the source installation guide)
Autoware now supports control inputs for vehicles based on an Ackermann kinematic model. This section introduces you a brief concept of Ackermann kinematic model and explains how Autoware controls it.
If your vehicle does not suit the Ackermann kinematic model, you have to modified the control commands. Another document gives you an example how to convert your Ackermann kinematic model control inputs into a differential drive model.
The basic style of Ackermann kinematic model has four wheels with an Ackermann link on the front, and it is powered by the rear wheels. The key point of Ackermann kinematic model is that the axes of all wheels intersect at a same point, which means all wheels will trace a circular trajectory with a different radii but a common center point (See the figure below). Therefore, this model has a great advantage that it minimizes the slippage of the wheels, and prevent tires to get worn soon.
In general, Ackermann kinematic model accepts the longitudinal speed \\(v\\) and the steering angle \\(\\phi\\) as inputs. In autoware, \\(\\phi\\) is positive if it is steered counter clockwise, so the steering angle in the figure below is actually negative.
The basic style of an Ackermann kinematic model. The left figure shows a vehicle facing straight forward, while the right figure shows a vehicle steering to the right."},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/creating-a-vehicle-interface-for-an-ackermann-kinematic-model/#control","title":"Control","text":"
Autoware publishes a ROS 2 topic named control_cmd from several types of publishers. A control_cmd topic is a AckermannControlCommand type message that contains
See the AckermannLateralCommand.idl and LongitudinalCommand.idl for details.
The vehicle interface should realize these control commands through your vehicle's control device.
Moreover, Autoware also provides brake commands, light commands, and more (see vehicle interface design), so the vehicle interface module should be applicable to these commands as long as there are devices available to handle them.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/","title":"Customizing for differential drive vehicle","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#customizing-for-differential-drive-vehicle","title":"Customizing for differential drive vehicle","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#1-introduction","title":"1. Introduction","text":"
Currently, Autoware assumes that vehicles use an Ackermann kinematic model with Ackermann steering. Thus, Autoware adopts the Ackermann command format for the Control module's output (see the AckermannDrive ROS message definition for an overview of Ackermann commands, and the AckermannControlCommands struct used in Autoware for more details).
However, it is possible to integrate Autoware with a vehicle that follows a differential drive kinematic model, as commonly used by small mobile robots.
One simple way of using Autoware with a differential drive vehicle is to create a vehicle_interface package that translates Ackermann commands to differential drive commands. Here are two points that you need to consider:
Create vehicle_interface package for differential drive vehicle
Set an appropriate wheel_base
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#21-create-a-vehicle_interface-package-for-differential-drive-vehicle","title":"2.1 Create a vehicle_interface package for differential drive vehicle","text":"
An Ackermann command in Autoware consists of two main control inputs:
steering angle (\\(\\omega\\))
velocity (\\(v\\))
Conversely, a typical differential drive command consists of the following inputs:
left wheel velocity (\\(v_l\\))
right wheel velocity (\\(v_r\\))
So, one way in which an Ackermann command can be converted to a differential drive command is by using the following equations:
\\[ v_l = v - \\frac{l\\omega}{2}, v_r = v + \\frac{l\\omega}{2} \\]
where \\(l\\) denotes wheel tread.
For information about other factors that need to be considered when creating a vehicle_interface package, refer to the vehicle_interface component page.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#22-set-an-appropriate-wheel_base","title":"2.2 Set an appropriate wheel_base","text":"
A differential drive robot does not necessarily have front and rear wheels, which means that the wheelbase (the horizontal distance between the axles of the front and rear wheels) cannot be defined. However, Autoware expects wheel_base to be set in vehicle_info.param.yaml with some value. Thus, you need to set a pseudo value for wheel_base.
The appropriate pseudo value for wheel_base depends on the size of your vehicle. Setting it to be the same value as wheel_tread is one possible choice.
Warning
If the wheel_base value is set too small then the vehicle may behave unexpectedly. For example, the vehicle may drive beyond the bounds of a calculated path.
Conversely, if wheel_base is set too large, the vehicle's range of motion will be restricted. The reason being that Autoware's Planning module will calculate an overly conservative trajectory based on the assumed vehicle length.
"},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#3-known-issues","title":"3. Known issues","text":""},{"location":"how-to-guides/integrating-autoware/creating-vehicle-interface-package/customizing-for-differential-drive-model/#motion-model-incompatibility","title":"Motion model incompatibility","text":"
Since Autoware assumes that vehicles use a steering system, it is not possible to take advantage of the flexibility of a differential drive system's motion model.
For example, when planning a parking maneuver with the freespace_planner module, Autoware may drive the differential drive vehicle forward and backward, even if the vehicle can be parked with a simpler trajectory that uses pure rotational movement.
"},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/","title":"Creating Autoware meta-repository","text":""},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#creating-autoware-meta-repository","title":"Creating Autoware meta-repository","text":""},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#what-is-meta-repository","title":"What is Meta-repository?","text":"
A meta-repository is a repository that manages multiple repositories, and Autoware is one of them. It serves as a centralized control point for referencing, configuring, and versioning other repositories.
By using Ansible and VCS, you can automatically set up your Autoware. autoware.repos file manages the configuration of multiple repositories.
Note: VCS stands for Version Control System, such as Git or Subversion.
"},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#how-to-create-and-customize-your-autoware-meta-repository","title":"How to create and customize your autoware meta-repository","text":""},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#1-create-autoware-repository","title":"1. Create autoware repository","text":"
If you want to integrate Autoware into your vehicle, the first step is to create an Autoware meta-repository.
One easy way is to fork autowarefoundation/autoware and clone it. For how to fork a repository, refer to GitHub Docs.
If you set up multiple types of vehicles, adding a suffix like autoware.vehicle_A or autoware.vehicle_B is recommended
"},{"location":"how-to-guides/integrating-autoware/creating-your-autoware-meta-repository/creating-autoware-meta-repository/#2-customize-your-autowarerepos-for-your-environment","title":"2. Customize your autoware.repos for your environment","text":"
You need to customize autoware.repos for your own vehicle's Autoware.
For example, if you want to customize the parameters in your individual_params or autoware_launch package to fit your vehicle, you can modify the configuration of each package and use them accordingly.
Please edit the parameters in Autoware's autoware_individual_params and autoware_launch packages to match your vehicle's specific requirements, as these packages provide sample parameters and may not be tailored to your vehicle by default.
If you want to fork autoware_individual_params and make modifications, it would be as follows:
Example: If you fork individual_params and rename autoware_individual_params.vehicle_A:
Please refer to the following documentation link for instructions on how to create and customize each vehicle_interface:
creating-vehicle-and-sensor-description
creating-vehicle-interface-package
customizing-for-differential-drive-model
Please remember to add all your custom packages, such as interfaces and descriptions, to your autoware.repos to ensure that your packages are properly included and managed within the Autoware repository.
It is possible to specify which components to launch using command-line arguments. For example, if you don't need to launch perception, planning, and control for localization debug, you can launch the following:
The basic command-line options are documented in autoware.launch.xml.
There are options available to switch between different methods for some component. For example, by specifying pose_source/twist_source or perception_mode, you can switch localization and perception methods, respectively. These options allow you to choose the desired algorithms or sensor configurations for the respective functionalities.
For options on eagleye component, please refer to the sub-pages.
Click the 2D Nav Goal button in the toolbar, or hit the G key
In the 3D View pane, click and hold the left mouse button, and then drag to set the direction for the goal pose. If successful, you will see the calculated planning path on RViz.
You can also engage via RViz with \"AutowareStatePanel\". The panel can be found in Panels > Add New Panel > tier4_state_rviz_plugin > AutowareStatePanel.
Once the route is computed, the \"AUTO\" button becomes active. Pressing the AUTO button engages the autonomous driving mode.
Now the vehicle should drive along the calculated path!
During the autonomous driving, the StatePanel appears as shown in the image below. Pressing the \"STOP\" button allows you to stop the vehicle.
By specifying the perception_mode, users can switch between different sensor configurations for perception. This allows you to choose the specific sensor setup that you want to use for the perception tasks.
Current localization launcher implemented by TIER IV supports multiple localization methods, both pose estimators and twist estimators. tier4_localization_component.launch.xml has two arguments to select which estimators to launch:
pose_source : an argument to select pose_estimator, currently supporting ndt (default), yabloc, and eagleye
twist_source : an argument to select twist_estimator, currently supporting gyro_odom (default), and eagleye
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/#ndt-scan-matcher-a-lidar-and-pointcloud-map-based-pose-estimator-default","title":"NDT scan matcher: a LiDAR and pointcloud map based pose estimator (default)","text":"
By default, Autoware launches ndt_scan_matcher for pose estimator. In order to launch this explicitly, you need to specify as follows:
Note that currently twist_source is set to Gyro Odometer as default, so you can skip this argument.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/#yabloc-a-camera-and-vector-map-based-pose-estimator","title":"YabLoc: a camera and vector map based pose estimator","text":"
You can use YabLoc as a camera-based localization method. For more details on YabLoc, please refer to the README of YabLoc in autoware.universe.
To use YabLoc as a pose_estimator, add pose_source:=yabloc when launching Autoware. By default, the pose_source is set to ndt. By specifying this command-line argument, YabLoc nodes will be automatically launched while the NDT nodes will not be started.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/#eagleye-a-gnss-imu-wheel-odometry-based-pose-and-twist-estimator","title":"Eagleye: a GNSS & IMU & wheel odometry based pose and twist estimator","text":"
You can use Eagleye as a GNSS & IMU & wheel odometry-based localization method. For more details on Eagleye, please refer to the Eagleye.
Eagleye has a function for position estimation and twist estimation, namely pose_estimator and twist_estimator, respectively. When running Eagleye in twist_estimator mode with other pose_estimator such as ndt_scan_matcher, Eagleye is still helpful since it can improve scan matching by providing accurate twists using GNSS doppler.
You can use Eagleye by specifying the pose_source and twist_source accordingly through command-line arguments.
Example of using Eagleye as the pose twist estimator:
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/","title":"Eagleye","text":""},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#using-eagleye-with-autoware","title":"Using Eagleye with Autoware","text":"
This page will show you how to set up Eagleye in order to use it with Autoware. For the details of the integration proposal, please refer to this discussion.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#what-is-eagleye","title":"What is Eagleye?","text":"
Eagleye is an open-source GNSS/IMU-based localizer initially developed by MAP IV. Inc. It provides a cost-effective alternative to LiDAR and point cloud-based localization by using low-cost GNSS and IMU sensors to provide vehicle position, orientation, and altitude information.
Eagleye can be utilized in the Autoware localization stack in two ways:
Feed only twist into the EKF localizer.
Feed both twist and pose from Eagleye into the EKF localizer (twist can also be used with regular gyro_odometry).
Note: RTK positioning is required when using Eagleye as the pose estimator. On the other hand, it is not mandatory when using it as the twist estimator.
Eagleye requires latitude/longitude height and doppler velocity generated by the GNSS receiver. Your GNSS ROS driver must publish the following messages:
sensor_msgs/msg/NavSatFix: This message contains latitude, longitude, and height information.
geometry_msgs/msg/TwistWithCovarianceStamped: This message contains gnss doppler velocity information.
Eagleye has been tested with the following example GNSS ROS drivers: ublox_gps and septentrio_gnss_driver. The settings needed for each of these drivers are as follows:
GNSS ROS drivers modification ublox_gps No additional settings are required. It publishes sensor_msgs/msg/NavSatFix and geometry_msgs/msg/TwistWithCovarianceStamped required by Eagleye with default settings. septentrio_gnss_driver Set publish.navsatfix and publish.twist in the config file gnss.yaml to true"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#parameter-modifications-for-integration-into-your-vehicle","title":"Parameter Modifications for Integration into Your Vehicle","text":""},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#topic-name-topic-type","title":"topic name & topic type","text":"
The users must correctly specify input topics for GNSS latitude, longitude, and height , GNSS doppler speed , IMU , and vehicle speed in the eagleye_config.yaml.
Also, the frequency of GNSS and IMU must be set in eagleye_config.yaml
common:\nimu_rate: 50\ngnss_rate: 5\n
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#conversion-from-fix-to-pose","title":"Conversion from fix to pose","text":"
The parameters for converting sensor_msgs/msg/NavSatFix to geometry_msgs/msg/PoseWithCovarianceStamped is listed in fix2pose.yaml. If you use a different geoid or projection type, change these parameters.
The other parameters are described here. Basically, these do not need to be changed .
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#notes-on-initialization","title":"Notes on initialization","text":"
Eagleye requires an initialization process for proper operation. Without initialization, the output for twist will be in the raw value, and the pose data will not be available.
The first step is static initialization, which involves allowing the Eagleye to remain stationary for approximately 5 seconds after startup to estimate the yaw-rate offset.
The next step is dynamic initialization, which involves running the Eagleye in a straight line for approximately 30 seconds. This process estimates the scale factor of wheel speed and azimuth angle.
Once dynamic initialization is complete, the Eagleye will be able to provide corrected twist and pose data.
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#how-to-check-the-progress-of-initialization","title":"How to check the progress of initialization","text":"
TODO
"},{"location":"how-to-guides/integrating-autoware/launch-autoware/localization-methods/eagleye-guide/#note-on-georeferenced-maps","title":"Note on georeferenced maps","text":"
Note that the output position might not appear to be in the point cloud maps if you are using maps that are not properly georeferenced. In the case of a single GNSS antenna, initial position estimation (dynamic initialization) can take several seconds to complete after starting to run in an environment where GNSS positioning is available.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/","title":"Evaluating the controller performance","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#evaluating-the-controller-performance","title":"Evaluating the controller performance","text":"
This page shows how to use control_performance_analysis package to evaluate the controllers.
control_performance_analysis is the package to analyze the tracking performance of a control module and monitor the driving status of the vehicle.
If you need more detailed information about package, refer to the control_performance_analysis.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#how-to-use","title":"How to use","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#before-driving","title":"Before Driving","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#1-firstly-you-need-to-launch-autoware-you-can-also-use-this-tool-with-real-vehicle-driving","title":"1. Firstly you need to launch Autoware. You can also use this tool with real vehicle driving","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#2-initialize-the-vehicle-and-send-goal-position-to-create-route","title":"2. Initialize the vehicle and send goal position to create route","text":"
If you have any problem with launching Autoware, please see the tutorials page.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#3-launch-the-control_performance_analysis-package","title":"3. Launch the control_performance_analysis package","text":"
After this command, you should be able to see the driving monitor and error variables in topics.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#4-run-the-plotjuggler-in-sourced-terminal","title":"4. Run the PlotJuggler in sourced terminal","text":"
source ~/autoware/install/setup.bash\n
ros2 run plotjuggler plotjuggler\n
If you do not have PlotJuggler in your computer, please refer here for installation guideline.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#5-increase-the-buffer-size-maximum-is-100-and-import-the-layout-from-autowareuniversecontrolcontrol_performance_analysisconfigcontroller_monitorxml","title":"5. Increase the buffer size (maximum is 100), and import the layout from /autoware.universe/control/control_performance_analysis/config/controller_monitor.xml","text":"
After import the layout, please specify the topics that are listed below.
/localization/kinematic_state
/vehicle/status/steering_status
/control_performance/driving_status
/control_performance/performance_vars
Please mark the If present, use the timestamp in the field [header.stamp] box, then select the OK.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#6-now-you-can-start-to-driving-you-should-see-all-the-performance-and-driving-variables-in-plotjuggler","title":"6. Now, you can start to driving. You should see all the performance and driving variables in PlotJuggler","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#after-driving","title":"After Driving","text":""},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-controller-performance/#1-you-can-export-the-statistical-output-and-all-data-to-compare-and-later-usage","title":"1. You can export the statistical output and all data to compare and later usage","text":"
With statistical data, you can export the all statistical values like (min, max, average) to compare the controllers.
You can also export all data to later use. To investigate them again, after launch PlotJuggler, import the .cvs file from data section.
You can plot the vehicle position. Select the two curve (keeping CTRL key pressed) and Drag & Drop them using the RIGHT Mouse button. Please visit the Help -> Cheatsheet in PlotJuggler to see more tips about it.
If you see too much noised curve in plots, you can adjust the odom_interval and low_pass_filter_gain from here to avoid noised data.
Autoware should be real-time system when integrated to a service. Therefore, the response time of each callback should be as small as possible. If Autoware appears to be slow, it is imperative to conduct performance measurements and implement improvements based on the analysis. However, Autoware is a complex software system comprising numerous ROS 2 nodes, potentially complicating the process of identifying bottlenecks. To address this challenge, we will discuss methods for conducting detailed performance measurements for Autoware and provide case studies. It is worth noting that multiple factors can contribute to poor performance, such as scheduling and memory allocation in the OS layer, but our focus in this page will be on user code bottlenecks. The outline of this section is as follows:
Improvement is impossible without precise measurements. To measure the performance of the application code, it is essential to eliminate any external influences. Such influences include interference from the operating system and CPU frequency fluctuations. Scheduling effects also occur when core resources are shared by multiple threads. This section outlines a technique for accurately measuring the performance of the application code for a specific node. Though this section only discusses the case of Linux on Intel CPUs, similar considerations should be made in other environments.
To eliminate the influence of scheduling, the node being measured should operate independently, using the same logic as when the entire Autoware system is running. To accomplish this, record all input topics of the node to be measured while the whole Autoware system is running. To achieve this objective, a tool called ros2_single_node_replayer has been prepared.
Details on how to use the tool can be found in the README. This tool records the input topics of a specific node during the entire Autoware operation and replays it in a single node with the same logic. The tool relies on the ros2 bag record command, and the recording of service/action is not supported as of ROS 2 Humble, so nodes that use service/action as their main logic may not work well.
Isolated cores running the node to be measured must meet the following conditions.
Fix CPU frequency and disable turbo boost
Minimize timer interruptions
Offload RCU (Read Copy Update) callback
Isolate the paired core if hyper-threading enabled
To fulfill these conditions on Linux, a custom kernel build with the following kernel configurations is required. You can find many resources to instruct you on how to build a custom Linux kernel (like this one). Note that even if full tickless is enabled, timer interrupts are generated for scheduling if more than two tasks exist in one core.
# Enable CONFIG_NO_HZ_FULL\n-> General setup\n-> Timers subsystem\n-> Timer tick handling (Full dynticks system (tickless))\n(X) Full dynticks system (tickless)\n\n# Allows RCU callback processing to be offloaded from selected CPUs\n# (CONFIG_RCU_NOCB_CPU=y)\n-> General setup\n-> RCU Subsystem\n-*- Offload RCU callback processing from boot-selected CPUs\n
Additionally, the kernel boot parameters need to be set as follows.
In the above configuration, for example, the node to be measured is assumed to run on core 2, and core 8, which is a hyper-threading pair, is also being isolated. Appropriate decisions on which cores to run the measurement target and which nodes to isolate need to be made based on the cache and core layout of the measurement machine. You can easily check if it is properly configured by running cat /proc/softirqs. Since intel_pstate=disable is specified in the kernel boot parameter, userspace can be specified in the scaling governor.
cat /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor // ondemand\nsudo sh -c \"echo userspace > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor\"\n
This allows you to freely set the desired frequency within a defined range.
sudo sh -c \"echo <freq(kz)> > /sys/devices/system/cpu/cpu2/cpufreq/scaling_setspeed\"\n
Turbo Boost needs to be switched off on Intel CPUs, which is often overlooked.
sudo sh -c \"echo 0 > /sys/devices/system/cpu/cpufreq/boost\"\n
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-real-time-performance/#run-single-node-separately","title":"Run single node separately","text":"
Following the instructions in the ros2_single_node_replayer README, start the node and play the dedicated rosbag created by the tool. Before playing the rosbag, appropriately set the CPU affinity of the thread on which the node runs, so it is placed on the isolated core prepared.
taskset --cpu-list -p <target cpu> <pid>\n
To avoid interference in the last level cache, minimize the number of other applications running during the measurement.
"},{"location":"how-to-guides/integrating-autoware/tuning-parameters-and-performance/evaluating-real-time-performance/#measurement-and-visualization","title":"Measurement and visualization","text":"
To visualize the performance of the measurement target, embed code for logging timestamps and performance counter values in the target source code. To achieve this objective, a tool called pmu_analyzer has been prepared.
Details on how to use the tool can be found in the README. This tool can measure the turnaround time of any section in the source code, as well as various performance counters.
In this section, we will present several case studies that demonstrate the performance improvements. These examples not only showcase our commitment to enhancing the system's efficiency but also serve as a valuable resource for developers who may face similar challenges in their own projects. The performance improvements discussed here span various components of the Autoware system, including sensing modules and planning modules. There are tendencies for each component regarding which points are becoming bottlenecks. By examining the methods, techniques, and tools employed in these case studies, readers can gain a better understanding of the practical aspects of optimizing complex software systems like Autoware.
First, we will explain the procedure for performance improvement, taking the node ring_outlier_filter as an example. Refer to the Pull Request for details.
The following figure is a time-series plot of the turnaround time of the main processing part of ring_outlier_filter, analyzed as described in the \"Performance Measurement\" section above.
The horizontal axis indicates the number of callbacks called (i.e., callback index), and the vertical axis indicates the turnaround time.
When analyzing the performance of the sensing module from the viewpoint of performance counter, pay attention to instructions, LLC-load-misses, LLC-store-misses, cache-misses, and minor-faults.
Analysis of the performance counter shows that the largest fluctuations come from minor-faults (i.e., soft page faults), the second largest from LLC-store-misses and LLC-load-misses (i.e., cache misses in the last level cache), and the slowest fluctuations come from instructions (i.e., message data size fluctuations). For example, when we plot minor-faults on the horizontal axis and turnaround time on the vertical axis, we can see the following dominant proportional relationship.
To achieve zero soft page faults, heap allocations must only be made from areas that have been first touched in advance. We have developed a library called heaphook to avoid soft page faults while running Autoware callback. If you are interested, refer to the GitHub discussion and the issue.
To reduce LLC misses, it is necessary to reduce the working set and to use cache-efficient access patterns.
In the sensing component, which handles large message data such as LiDAR point cloud data, minimizing copying is important. A callback that takes sensor data message types as input and output should be written in an in-place algorithm as much as possible. This means that in the following pseudocode, when generating output_msg from input_msg, it is crucial to avoid using buffers as much as possible to reduce the number of memory copies.
To improve cache efficiency, implement an in-place style as much as possible, instead of touching memory areas sporadically. In ROS applications using PCL, the code shown below is often seen.
void callback(const sensor_msgs::PointCloud2ConstPtr &input_msg) {\npcl::PointCloud<PointT>::Ptr input_pcl(new pcl::PointCloud<PointT>);\npcl::fromROSMsg(*input_msg, *input_pcl);\n\n// Algorithm is described for point cloud type of pcl\npcl::PointCloud<PointT>::Ptr output_pcl(new pcl::PointCloud<PointT>);\nfill_pcl(*input_pcl, *output_pcl);\n\nauto output_msg = allocate_msg<sensor_msgs::PointCloud2>(output_size);\npcl::toROSMsg(*output_pcl, *output_msg);\npublish(std::move(output_msg));\n}\n
To use the PCL library, fromROSMsg() and toROSMsg() are used to perform message type conversion at the beginning and end of the callback. This is a wasteful copying process and should be avoided. We should eliminate unnecessary type conversions by removing dependencies on PCL (e.g., https://github.com/tier4/velodyne_vls/pull/39). For large message types such as map data, there should be only one instance in the entire system in terms of physical memory.
First, we will pick up detection_area module in behavior_velocity_planner node, which tends to have long turnaround time. We have followed the performance analysis steps above to obtain the following graph. Axises are the same as the graphs in the sensing case study.
Using pmu_analyzer tool to further identify the bottleneck, we have found that the following multiple loops were taking up a lot of processing time:
for ( area : detection_areas )\nfor ( point : point_clouds )\nif ( boost::geometry::within(point, area) )\n// do something with O(1)\n
It checks whether each point cloud is contained in each detection area. Let N be the size of point_clouds and M be the size of detection_areas, then the computational complexity of this program is O(N^2 * M), since the complexity of within is O(N). Here, given that most of the point clouds are located far away from a certain detection area, a certain optimization can be achieved. First, calculate the minimum enclosing circle that completely covers the detection area, and then check whether the points are contained in that circle. Most of the point clouds can be quickly ruled out by this method, we don\u2019t have to call the within function in most cases. Below is the pseudocode after optimization.
for ( area : detection_areas )\ncircle = calc_minimum_enclosing_circle(area)\nfor ( point : point_clouds )\nif ( point is in circle )\nif ( boost::geometry::within(point, area) )\n// do something with O(1)\n
By using O(N) algorithm for minimum enclosing circle, the computational complexity of this program is reduced to almost O(N * (N + M)) (note that the exact computational complexity does not really change). If you are interested, refer to the Pull Request.
Similar to this example, in the planning component, we take into consideration thousands to tens of thousands of point clouds, thousands of points in a path representing our own route, and polygons representing obstacles and detection areas in the surroundings, and we repeatedly create paths based on them. Therefore, we access the contents of the point clouds and paths multiple times using for-loops. In most cases, the bottleneck lies in these naive for-loops. Here, understanding Big O notation and reducing the order of computational complexity directly leads to performance improvements.
"},{"location":"how-to-guides/others/add-a-custom-ros-message/","title":"Add a custom ROS message","text":""},{"location":"how-to-guides/others/add-a-custom-ros-message/#add-a-custom-ros-message","title":"Add a custom ROS message","text":""},{"location":"how-to-guides/others/add-a-custom-ros-message/#overview","title":"Overview","text":"
During the Autoware development, you will probably need to define your own messages. Read the following instructions before adding a custom message.
Message in autoware_msgs define interfaces of Autoware Core.
If a contributor wishes to make changes or add new messages to autoware_msgs, they should first create a new discussion post under the Design category.
Any other minor or proposal messages used for internal communication within a component(such as planning) should be defined in another repository.
tier4_autoware_msgs is an example of that.
The following is a simple tutorial of adding a message package to autoware_msgs. For the general ROS 2 tutorial, see Create custom msg and srv files.
"},{"location":"how-to-guides/others/add-a-custom-ros-message/#how-to-create-custom-message","title":"How to create custom message","text":"
Make sure you are in the Autoware workspace, and then run the following command to create a new package. As an example, let's create a package to define sensor messages.
Create a package
cd ./src/core/autoware_msgs\nros2 pkg create --build-type ament_cmake autoware_sensing_msgs\n
Create custom messages
You should create .msg files and place them in the msg directory.
NOTE: The initial letters of the .msg and .srv files must be capitalized.
As an example, let's make .msg files GnssInsOrientation.msg and GnssInsOrientationStamped.msg to define GNSS/INS orientation messages:
We need to replace <buildtool_depend>ament_cmake</buildtool_depend> with <buildtool_depend>ament_cmake_auto</buildtool_depend> in the package.xml file.
Build the custom message package
You can build the package in the root of your workspace, for example by running the following command:
Now the GnssInsOrientationStamped message will be discoverable by other packages in Autoware.
"},{"location":"how-to-guides/others/add-a-custom-ros-message/#how-to-use-custom-messages-in-autoware","title":"How to use custom messages in Autoware","text":"
You can use the custom messages in Autoware by following these steps:
Add dependency in package.xml.
For example, <depend>autoware_sensing_msgs</depend>.
Include the .hpp file of the relevant message in the code.
For example, #include <autoware_sensing_msgs/msg/gnss_ins_orientation_stamped.hpp>.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/","title":"Advanced usage of colcon","text":""},{"location":"how-to-guides/others/advanced-usage-of-colcon/#advanced-usage-of-colcon","title":"Advanced usage of colcon","text":"
This page shows some advanced and useful usage of colcon. If you need more detailed information, refer to the colcon documentation.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#common-mistakes","title":"Common mistakes","text":""},{"location":"how-to-guides/others/advanced-usage-of-colcon/#do-not-run-from-other-than-the-workspace-root","title":"Do not run from other than the workspace root","text":"
It is important that you always run colcon build from the workspace root because colcon builds only under the current directory. If you have mistakenly built in a wrong directory, run rm -rf build/ install/ log/ to clean the generated files.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#do-not-unnecessarily-overlay-workspaces","title":"Do not unnecessarily overlay workspaces","text":"
colcon overlays workspaces if you have sourced the setup.bash of other workspaces before building a workspace. You should take care of this especially when you have multiple workspaces.
Run echo $COLCON_PREFIX_PATH to check whether workspaces are overlaid. If you find some workspaces are unnecessarily overlaid, remove all built files, restart the terminal to clean environment variables, and re-build the workspace.
For more details about workspace overlaying, refer to the ROS 2 documentation.
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#cleaning-up-the-build-artifacts","title":"Cleaning up the build artifacts","text":"
colcon sometimes causes errors of because of the old cache. To remove the cache and rebuild the workspace, run the following command:
rm -rf build/ install/\n
In case you know what packages to remove:
rm -rf {build,install}/{package_a,package_b}\n
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#selecting-packages-to-build","title":"Selecting packages to build","text":"
"},{"location":"how-to-guides/others/advanced-usage-of-colcon/#changing-the-default-configuration-of-colcon","title":"Changing the default configuration of colcon","text":"
Create $COLCON_HOME/defaults.yaml to change the default configuration.
mkdir -p ~/.colcon\ncat << EOS > ~/.colcon/defaults.yaml\n{\n\"build\": {\n\"symlink-install\": true\n}\n}\n
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/","title":"An example procedure for adding and evaluating a new node","text":""},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#an-example-procedure-for-adding-and-evaluating-a-new-node","title":"An example procedure for adding and evaluating a new node","text":""},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#overview","title":"Overview","text":"
This page provides a guide for evaluating Autoware when a new node is implemented, especially about developing a novel localization node.
The workflow involves initial testing and rosbag recording using a real vehicle or AWSIM, implementing the new node, subsequent testing using the recorded rosbag, and finally evaluating with a real vehicle or AWSIM.
It is assumed that the method intended for addition has already been verified well with public datasets and so on.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#1-running-autoware-in-its-standard-configuration","title":"1. Running Autoware in its standard configuration","text":"
First of all, it is important to be able to run the standard Autoware to establish a basis for performance and behavior comparison.
Autoware constantly incorporates new features. It is crucial to initially confirm that it operates as expected with the current version, which helps in problem troubleshooting.
In this context, AWSIM is presumed. Therefore, AWSIM simulator can be useful. If you are using actual hardware, please refer to the How-to guides.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#2-recording-a-rosbag-using-autoware","title":"2. Recording a rosbag using Autoware","text":"
Before developing a new node, it is recommended to record a rosbag in order to evaluate. If you need a new sensor, you should add it to your vehicle or AWSIM.
In this case, it is recommended to save all topics regardless of whether they are necessary or not. For example, in Localization, since the initial position estimation service is triggered by the input to rviz and the GNSS topic, the initial position estimation does not start when playing back data unless those topics are saved.
Consider the use of the mcap format if data capacity becomes a concern.
It is worth noting that using ros2 bag record increases computational load and might affect performance. After data recording, verifying the smooth flow of sensor data and unchanged time series is advised. This verification can be accomplished, for example, by inspecting the image data with rqt_image_view during ros2 bag play.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#3-developing-the-new-node","title":"3. Developing the new node","text":"
When developing a new node, it could be beneficial to reference a package that is similar to the one you intend to create.
It is advisable to thoroughly read the Design page, contemplate the addition or replacement of nodes in Autoware, and then implement your solution.
For example, a node doing NDT, a LiDAR-based localization method, is ndt_scan_matcher. If you want to replace this with a different approach, implement a node which produces the same topics and provides the same services.
ndt_scan_matcher is launched as pose_estimator, so it is necessary to replace the launch file as well.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#4-evaluating-by-a-rosbag-based-simulator","title":"4. Evaluating by a rosbag-based simulator","text":"
Once the new node is implemented, it is time to evaluate it. logging_simulator is a tool of how to evaluate the new node using the rosbag captured in step 2.
When you run the logging_simulator, you can set planning:=false or control:=false to disable the launch of specific component nodes.
After launching logging_simulator, the rosbag file obtained in step 2 should be replayed using ros2 bag play <rosbag_file>.
If you remap the topics related to the localization that you want to verify this time, Autoware will use the data it is calculating this time instead of the data it recorded. Also, using the --topics option of ros2 bag play, you can publish only specific topics in rosbag.
There is ros2bag_extensions available to filter the rosbag file and create a new rosbag file that contains only the topics you need.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#5-evaluating-in-a-realtime-environment","title":"5. Evaluating in a realtime environment","text":"
Once you have sufficiently verified the behavior in the logging_simulator, let's run it as Autoware with new nodes added in the realtime environment.
To debug Autoware, the method described at debug-autoware is useful.
For reproducibility, you may want to fix the GoalPose. In such cases, consider using the tier4_automatic_goal_rviz_plugin.
"},{"location":"how-to-guides/others/an-example-procedure-for-adding-and-evaluating-a-new-node/#6-sharing-the-results","title":"6. Sharing the results","text":"
If your implementation works successfully, please consider a pull request to Autoware.
It is also a good idea to start by presenting your ideas in Discussion at Show and tell.
For localization, YabLoc's Proposal may provide valuable insights.
"},{"location":"how-to-guides/others/applying-clang-tidy-to-ros-packages/","title":"Applying Clang-Tidy to ROS packages","text":""},{"location":"how-to-guides/others/applying-clang-tidy-to-ros-packages/#applying-clang-tidy-to-ros-packages","title":"Applying Clang-Tidy to ROS packages","text":"
The essential thing for debug is to print the program information clearly, which can quickly judge the program operation and locate the problem. Autoware uses ROS 2 logging tool to print debug messages, how to design console logging refer to tutorial Console logging.
"},{"location":"how-to-guides/others/debug-autoware/#using-ros-tools-debug-autoware","title":"Using ROS tools debug Autoware","text":""},{"location":"how-to-guides/others/debug-autoware/#using-command-line-tools","title":"Using command line tools","text":"
ROS 2 includes a suite of command-line tools for introspecting a ROS 2 system. The main entry point for the tools is the command ros2, which itself has various sub-commands for introspecting and working with nodes, topics, services, and more. How to use the ROS 2 command line tool refer to tutorial CLI tools.
Rviz2 is a port of Rviz to ROS 2. It provides a graphical interface for users to view their robot, sensor data, maps, and more. You can run Rviz2 tool easily by:
rviz2\n
When Autoware launch the simulators, the Rviz2 tool is opened by default to visualize the autopilot graphic information.
RQt is a graphical user interface framework that implements various tools and interfaces in the form of plugins. You can run any RQt tools/plugins easily by:
rqt\n
This GUI allows you to choose any available plugins on your system. You can also run plugins in standalone windows. For example, RQt Console:
ros2_graph can be used to generate mermaid description of ROS 2 graphs to add on your markdown files.
It can also be used as a colorful alternative to rqt_graph even though it would require some tool to render the generated mermaid diagram.
It can be installed with:
pip install ros2-graph\n
Then you can generate a mermaid description of the graph with:
ros2_graph your_node\n\n# or like with an output file\nros2_graph /turtlesim -o turtle_diagram.md\n\n# or multiple nodes\nros2_graph /turtlesim /teleop_turtle\n
When your ROS 2 setup is not running as expected, you can check its settings with the ros2doctor tool.
ros2doctor checks all aspects of ROS 2, including platform, version, network, environment, running systems and more, and warns you about possible errors and reasons for issues.
It's as simple as just running ros2 doctor in your terminal.
It has the ability to list \"Subscribers without publishers\" for all topics in the system.
And this information can help you find if a necessary node isn't running.
For more details, see the following official documentation for Using ros2doctor to identify issues.
"},{"location":"how-to-guides/others/debug-autoware/#using-a-debugger-with-breakpoints","title":"Using a debugger with breakpoints","text":"
Many IDE(e.g. Visual Studio Code, CLion) supports debugging C/C++ executable with GBD on linux platform. The following lists some references for using the debugger:
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/","title":"Defining temporal performance metrics on components","text":""},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#defining-temporal-performance-metrics-on-components","title":"Defining temporal performance metrics on components","text":""},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#motivation-to-defining-temporal-performance-metrics","title":"Motivation to defining temporal performance metrics","text":""},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#objective-of-the-page","title":"Objective of the page","text":"
This page introduces policies to define metrics to evaluate temporal performance on components of Autoware. The term \"temporal performance\" is often used throughout the page in order to distinguish between functional performance, which referred to as accuracy as well, and time-related performance.
It is expected that most algorithms employed for Autoware are executed with as high frequency and short response time as possible. In order to achieve safe autonomous driving, one of the desired outcomes is no time gap between perceived and actual situation. The time gap is commonly referred to as delay. If the delay is significant, the system may determine trajectory and maneuver based on outdated situation. Consequently, if the actual situation differs from the perceived one due to the delay, the system may make unexpected decisions.
As mentioned above, this page presents the policies to define metrics. Besides, the page contains lists of sample metrics that are crucial for the main functionalities of Autoware: Localization, Perception, Planning, and Control.
Note
Other functionalities, such as system components for diagnosis, are excluded currently. However they will be taken into account in the near future.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#contribution-of-the-temporal-performance-metrics","title":"Contribution of the temporal performance metrics","text":"
Temporal performance metrics are important for evaluating Autoware. These metrics are particularly useful for assessing delays caused by new algorithms and logic. They can be employed when comparing the temporal performance of software on a desktop computer with that on a vehicle during the vehicle integration phase.
In addition, these metrics are useful for designers and evaluators of middleware, operating systems, and computers. They are selected based on user and product requirements. One of these requirements is to provide sufficient temporal performance for executing Autoware. \"Sufficient temporal performance\" is defined as a temporal performance requirement, but it can be challenging to define the requirement because it varies depending on the product type, Operational Design Domain (ODD), and other factors. Then, this page specifically focuses on temporal performance metrics rather than requirements.
Temporal performance metrics are important for evaluating the reliability of Autoware. However, ensuring the reliability of Autoware requires consideration of not only temporal performance metrics but also other metrics.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#tools-for-evaluating-the-metrics","title":"Tools for evaluating the metrics","text":"
There are several tools available for evaluating Autoware according to the metrics listed in the page. For example, both CARET and ros2_tracing are recommended options when evaluating Autoware on Linux and ROS 2. If you want to measure the metrics with either of these tools, refer to the corresponding user guide for instructions. It's important to note that if you import Autoware to a platform other than Linux and ROS 2, you will need to choose a supported tool for evaluation.
Note
TIER IV plans to measure Autoware, which is running according to the tutorial, and provide a performance evaluation report periodically. An example of such a report can be found here, although it may not include all of the metrics listed.
The page does not aim to provide instructions on how to use these tools or measure the metrics. Its primary focus is on the metrics themselves, as they are more important than the specific tools used. These metrics retain their relevance regardless of the employed platform.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#policies-to-define-temporal-performance-metrics","title":"Policies to define temporal performance metrics","text":"
As mentioned above, the configuration of Autoware varies by the product type, ODD, and other factors. The variety of configurations makes it difficult to define the uniform metrics for evaluating Autoware. However, the policies used to define them are basically reused even when the configuration changes. Each of temporal performance metrics is categorized into two types: execution frequency and response time. Although there are many types of metrics, such as communication latency, the only two types are considered for simplicity. Execution frequency is observed using rate of Inter-Process Communication (IPC) messages. You will find an enormous number of messages in Autoware, but you don't have to take care of all. Some messages might be critical to functionality and they should be chosen for evaluation. Response time is duration elapsed through a series of processing. A series of processing is referred to as a path. Response time is calculated from timestamps of start and end of a path. Although many paths can be defined in Autoware, you have to choose significant paths.
As a hint, here are some characteristics of message and path in order to choose metrics.
Messages and paths on boundaries where observed values from sensors are consumed
Messages and paths on boundaries of functions, e.g., a boundary of perception and planning
Messages and paths on boundaries where timer-based frequency is switched
Messages and paths on boundaries where two different messages are synchronized and merged
Messages that must be transmitted at expected frequency, e.g., vehicle command messages
Those hints would be helpful for most configurations but there may be exclusions. Defining metrics precisely requires an understanding of configuration.
In addition, it is recommended that metrics be determined incrementally from the architectural level to the detailed design and implementation level. Mixing metrics at different levels of granularity can be confusing.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#list-of-sample-metrics","title":"List of sample metrics","text":"
This section demonstrates how to define metrics according to the policies explained and has lists of the metrics for Autoware launched according to the tutorial. The section is divided into multiple subsections, each containing a model diagram and an accompanying list that explains the important temporal performance metrics. Each model is equipped with checkpoints that serve as indicators for these metrics.
The first subsection presents the top-level temporal performance metrics, which are depicted in the abstract structure of Autoware as a whole. The detailed metrics are not included in the model as they would add complexity to it. Instead, the subsequent section introduces the detailed metrics. The detailed metrics are subject to more frequent updates compared to the top-level ones, which is another reason for categorizing them separately.
Each list includes a column for the reference value. The reference value represents the observed value of each metric when Autoware is running according to the tutorial. It is important to note that the reference value is not a required value, meaning that Autoware does not necessarily fail in the tutorial execution if certain metrics do not fulfill the reference value.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#top-level-temporal-performance-metrics-for-autoware","title":"Top-level temporal performance metrics for Autoware","text":"
The diagram below introduces the model for top-level temporal performance metrics.
The following three policies assist in selecting the top-level performance metrics:
Splitting Autoware based on components that consume observed values, such as sensor data, and considering the processing frequency and response time around these components
Dividing Autoware based on the entry point of Planning and Control and considering the processing frequency and response time around these components
Showing the minimum metrics for the Vehicle Interface, as they may vary depending on the target vehicle
Additionally, it is assumed that algorithms are implemented as multiple nodes and function as a pipeline processing system.
ID Representation in the model Metric meaning Related functionality Reference value Reason to choose it as a metric Note AWOV-001 Message rate from CPA #9 to CPA #18 Update rate of result from Prediction to Planning. Perception 10 Hz Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. AWOV-002 Response time from CPA #0 to CPA #20 via CPA #18 Response time in main body of Perception. Perception N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is used if delay compensation is disabled in Tracking. AWOV-003 Response time from CPA #7 to CPA #20 Response time from Tracking output of Tracking to its data consumption in Planning. Perception N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is used if delay compensation is enabled in Tracking. AWOV-004 Response time from CPA #0 to CPA #6 Duration to process pointcloud data in Sensing and Detection. Perception N/A Tracking relies on detection to provide real-time and up-to-date sensed data for accurate tracking. The metric is used if delay compensation is enabled in Tracking. AWOV-005 Message rate from CPA #4 to CPA #5 Update rate of Detection result received by Tracking. Perception 10 Hz Tracking relies on detection to provide real-time and up-to-date sensed data for accurate tracking. AWOV-006 Response time from CPA #0 to CPA #14 Response time from output of observed data from LiDARs to its consumption in EKF Localizer via NDT Scan Matcher. Localization N/A EKF Localizer relies on fresh and up-to-date observed data from sensors for accurate estimation of self pose. AWOV-007 Message rate from CPA #11 to CPA #13 Update rate of pose estimated by NDT Scan Matcher. Localization 10 Hz EKF Localizer relies on fresh and up-to-date observed data from sensors for accurate estimation of self pose. AWOV-008 Message rate from CPA #15 to CPA #12 Update rate of feed backed pose estimated by EKF Localizer. Localization 50 Hz NDT Scan Matcher relies on receiving estimated pose from EKF Localizer smoothly for linear interpolation. AWOV-009 Message rate from CPA #17 to CPA #19 Update rate of Localization result received by Planning. Localization 50 Hz Planning relies on Localization to update the estimated pose frequently. AWOV-010 Response time from CPA #20 to CPA #23 Processing time from beginning of Planning to consumption of Trajectory message in Control. Planning N/A A vehicle relies on Planning to update trajectory within a short time frame to achieve safe driving behavior. AWOV-011 Message rate from CPA #21 to CPA #22 Update rate of Trajectory message from Planning. Planning 10 Hz A vehicle relies on Planning to update trajectory frequently to achieve safe driving behavior. AWOV-012 Message rate from CPA #24 to CPA #25 Update rate of Control command. Control 33 Hz Control stability and comfort relies on sampling frequency of Control. AWOV-013 Message rate between CPA #26 and Vehicle Communication rate between Autoware and Vehicle. Vehicle Interface N/A A vehicle requires Autoware to communicate with each other at predetermined frequency. Temporal performance requirement varies depending on vehicle type.
Note
There is an assumption that each of sensors, such as LiDARs and cameras, outputs a set of pointcloud with a timestamp. CPA #0 is observed with the timestamp. If the sensors are not configured to output the timestamp, the time when Autoware receives the pointcloud is used instead. That is represented by CPA #1 in the model. The detailed metrics employs the idea as well.
"},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#detailed-temporal-performance-metrics-for-perception","title":"Detailed temporal performance metrics for Perception","text":"
The diagram below introduces the model for temporal performance metrics for Perception.
The following two policies assist in selecting the performance metrics:
Regarding the frequency and response time at which Recognition results from Object Recognition and Traffic Light Recognition are consumed in Planning
Splitting Perception component on merging points of data from multiple processing paths and considering the frequency and response time around that point
The following list shows the temporal performance metrics for Perception.
ID Representation in the model Metric meaning Related functionality Reference value Reason to choose it as a metric Note APER-001 Message rate from CPP #2 to CPP #26 Update rate of Traffic Light Recognition. Traffic Light Recognition 10 Hz Planning relies on fresh and up-to-date perceived data from Traffic Light Recognition for making precise decisions. APER-002 Response time from CPP #0 to CPP #30 Response time from camera input to consumption of the result in Planning. Traffic Light Recognition N/A Planning relies on fresh and up-to-date perceived data from Traffic Light Recognition for making precise decisions. APER-003 Message rate from CPP #25 to CPP #28 Update rate of result from Prediction (Object Recognition) to Planning. Object Recognition 10 Hz Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is same as AWOV-001. APER-004 Response time from CPP #6 to CPP #30 Response time from Tracking output of Tracking to its data consumption in Planning. Object Recognition N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is same as AWOV-002 and used if delay compensation is disabled in Tracking. APER-005 Response time from CPP #23 to CPP #30 Response time from Tracking output of Tracking to its data consumption in Planning. Object Recognition N/A Planning relies on fresh and up-to-date perceived data from Perception for creating accurate trajectory. The metric is same as AWOV-003 and used if delay compensation is enabled in Tracking. APER-006 Response time from CPP #6 to CPP #21 Duration to process pointcloud data in Sensing and Detection. Object Recognition N/A Tracking relies on Detection to provide real-time and up-to-date perceived data. The metrics is same as AWOV-004 and used if delay compensation is enabled in Tracking. APER-007 Message rate from CPP #20 to CPP #21 Update rate of Detection result received by Tracking. Object Recognition 10 Hz Tracking relies on detection to provide real-time and up-to-date sensed data for accurate tracking. The metric is same as AWOV-005 APER-008 Message rate from CPP #14 to CPP #19 Update rate of data sent from Sensor Fusion. Object Recognition 10 Hz Association Merger relies on the data to be updated at expected frequency for data synchronization. APER-009 Message rate from CPP #16 to CPP #19 Update rate of data sent from Detection by Tracker. Object Recognition 10 Hz Association Merger relies on the data to be updated at expected frequency for data synchronization. APER-010 Message rate from CPP #18 to CPP #19 Update rate of data sent from Validation Object Recognition. 10 Hz Association Merger relies on the data to be updated at expected frequency for data synchronization. APER-011 Response time from CPP #6 to CPP #19 via CPP #14 Response time to consume data sent from Sensor Fusion after LiDARs output pointcloud. Object Recognition N/A Association Merger relies on fresh and up-to-date data for data synchronization. APER-012 Response time from CPP #6 to CPP #19 via CPP #16 Response time to consume data sent from Detection by Tracker after LiDARs output pointcloud. Object Recognition N/A Association Merger relies on fresh and up-to-date data for data synchronization. APER-013 Response time from CPP #6 to CPP #19 via CPP #18 Response time to consume data sent from Validator after LiDARs output pointcloud. Object Recognition N/A Association Merger relies on fresh and up-to-date data for data synchronization. APER-014 Message rate from CPP #10 to CPP #13 Update rate of data sent from Clustering. Object Recognition 10 Hz Sensor Fusion relies on the data to be updated at expected frequency for data synchronization. APER-015 Message rate from CPP #5 to CPP #13 Update rate of data sent from Camera-based Object detection. Object Recognition 10 Hz Sensor Fusion relies on the data to be updated at expected frequency for data synchronization. APER-016 Response time from CPP #6 to CPP #13 Response time to consume data sent from Clustering after LiDARs output pointcloud. Object Recognition N/A Sensor Fusion relies on fresh and up-to-date data for data synchronization. APER-017 Response time from CPP #3 to CPP #13 Response time to consume data sent from Camera-based Object detection after Cameras output images. Object Recognition N/A Sensor Fusion relies on fresh and up-to-date data for data synchronization. APER-018 Message rate from CPP #10 to CPP #17 Update rate of data sent from Clustering. Object Recognition 10 Hz Validator relies on the data to be updated at expected frequency for data synchronization. It seems similar to APER-014, but the topic message is different. APER-019 Message rate from CPP #12 to CPP #17 Update rate of data sent from DNN-based Object Recognition. Object Recognition 10 Hz Validator relies on the data to be updated at expected frequency for data synchronization. APER-020 Response time from CPP #6 to CPP #17 via CPP #10 Response time to consume data sent from Clustering after LiDARs output pointcloud. Object Recognition N/A Validator relies on fresh and update-date data for data synchronization. It seems similar to APER-015, but the topic message is different. APER-021 Response time from CPP #6 to CPP #17 via CPP #12 Response time to consume data sent from DNN-based Object Recognition after LiDARs output pointcloud. Object Recognition N/A Validator relies on fresh and update-date data for data synchronization."},{"location":"how-to-guides/others/defining-temporal-performance-metrics/#detailed-temporal-performance-metrics-for-paths-between-obstacle-segmentation-and-planning","title":"Detailed temporal performance metrics for Paths between Obstacle segmentation and Planning","text":"
Obstacle segmentation, which is a crucial part of Perception, transmits data to Planning. The figure below illustrates the model that takes into account performance metrics related to Obstacle segmentation and Planning.
Note
Both the Obstacle grid map and Obstacle segmentation transmit data to multiple sub-components of Planning. However, not all of these sub-components are described in the model. This is because our primary focus is on the paths from LiDAR to Planning via Obstacle segmentation.
The following list shows the temporal performance metrics around Obstacle segmentation and Planning.
ID Representation in the model Metric meaning Related functionality Reference value Reason to choose it as a metric Note OSEG-001 Message rate from CPS #4 to CPS #7 Update rate of Occupancy grid map received by Planning (behavior_path_planner) Obstacle segmentation 10 Hz Planning relies on Occupancy grid map to be updated frequently and smoothly for creating accurate trajectory. OSEG-002 Response time from CPS #0 to CPS #9 via CPS #7 Response time to consume Occupancy grid map after LiDARs output sensing data. Obstacle segmentation N/A Planning relies on fresh and up-to-date perceived data from Occupancy grid map for creating accurate trajectory.. OSEG-003 Message rate from CPS #6 to CPS #11 Update rate of obstacle segmentation received by Planning (behavior_velocity_planner). Obstacle segmentation 10 Hz Planning relies on Obstacle segmentation to be updated frequently and smoothly for creating accurate trajectory. OSEG-004 Response time from CPS #0 to CPS #13 via CPS #11 Response time to consume Obstacle segmentation after LiDARs output sensing data. Obstacle segmentation N/A Planning relies on fresh and up-to-date perceived data from Obstacle segmentation for creating accurate trajectory.."},{"location":"how-to-guides/others/determining-component-dependencies/","title":"Determining component dependencies","text":""},{"location":"how-to-guides/others/determining-component-dependencies/#determining-component-dependencies","title":"Determining component dependencies","text":"
For any developers who wish to try and deploy Autoware as a microservices architecture, it is necessary to understand the software dependencies, communication, and implemented features of each ROS package/node.
As an example, the commands necessary to determine the dependencies for the Perception component are shown below.
Autoware manages dependent package versions in autoware.repos. For example, let's say you make a branch in autoware.universe and add new features. Suppose you update other dependencies with vcs pull after cutting a branch from autoware.universe. Then the version of autoware.universe you are developing and other dependencies will become inconsistent, and the entire Autoware build will fail. We recommend saving the dependent package versions by executing the following command when starting the development.
vcs export src --exact > my_autoware.repos\n
"},{"location":"how-to-guides/others/reducing-start-delays/","title":"Reducing start delays on real vehicles","text":""},{"location":"how-to-guides/others/reducing-start-delays/#reducing-start-delays-on-real-vehicles","title":"Reducing start delays on real vehicles","text":"
In simulation, the ego vehicle reacts nearly instantly to the control commands generated by Autoware. However, with a real vehicle, some delays occur that may make ego feel less responsive.
This page presents start delays experienced when using Autoware on a real vehicle. We define the start delay as the time between (a) when Autoware decides to make the ego vehicle start and (b) when the vehicle actually starts moving. More precisely:
(a) is the time when the speed or acceleration command output by Autoware switches to a non-zero value.
(b) is the time when the measured velocity of the ego vehicle switches to a positive value.
"},{"location":"how-to-guides/others/reducing-start-delays/#start-delay-with-manual-driving","title":"Start delay with manual driving","text":"
First, let us look at the start delay when a human is driving.
The following figure shows the start delay when a human driver switches the gear from parked to drive and instantly releases the brake to push the throttle pedal and make the velocity of the vehicle increase.
There are multiple things to note from this figure.
Brake (red): despite the driver instantly releasing the brake pedal, we see that the measured brake takes around 150ms to go from 100% to 0%.
Gear (orange): the driver switches gear before releasing the brake pedal, but the gear is measured to switch after the brake is released.
Throttle (green) and velocity (blue): the driver pushes the throttle pedal and the vehicle is measured to start moving around 500ms later.
To guarantee passenger comfort, some Autoware modules implement filters on the jerk of the vehicle, preventing sudden changes in acceleration.
For example, the vehicle_cmd_gate filters the acceleration command generated by the controller and was previously introducing significant delays when transitioning between a stop command where the acceleration is negative, and a move command where the acceleration is positive. Because of the jerk filter, the transition between negative and positive was not instantaneous and would take several hundreds of milliseconds.
In many vehicles, it is necessary to change gear before first starting to move the vehicle. When performed autonomously, this gear change can take some significant time. Moreover, as seen from the data recorded with manual driving, the measured gear value may be delayed.
In Autoware, the controller sends a stopping control command until the gear is changed to the drive state. This means that delays in the gear change and its reported value can greatly impact the start delay. Note that this is only an issue when the vehicle is initially in the parked gear.
The only way to reduce this delay is by tuning the vehicle to increase the gear change speed or to reduce the delay in the gear change report.
In vehicles with a brake pedal, the braking system will often be made of several moving parts which cannot move instantly. Thus, when Autoware sends brake commands to a vehicle, some delays should be expected in the actual brake applied to the wheels.
This lingering brake may prevent or delay the initial motion of the ego vehicle.
For vehicles with throttle control, one of the main cause of start delays is due to the throttle response of the vehicle. When pushing the throttle pedal, the wheels of the vehicle do not instantly start rotating. This is partly due to the inertia of the vehicle, but also to the motor which may take a significant time to start applying some torque to the wheels.
It may be possible to tune some vehicle side parameters to reduce this delay, but it is often done at the cost of reduced energy efficiency.
On the Autoware side, the only way to decrease this delay is to increase the initial throttle but this can cause uncomfortably high initial accelerations.
"},{"location":"how-to-guides/others/reducing-start-delays/#initial-acceleration-and-throttle","title":"Initial acceleration and throttle","text":"
As we just discussed, for vehicles with throttle control, an increased initial throttle value can reduce the start delay.
Since Autoware outputs an acceleration value, the conversion module raw_vehicle_cmd_converter is used to map the acceleration value from Autoware to a throttle value to be sent to the vehicle. Such mapping is usually calibrated automatically using the accel_brake_map_calibrator module, but it may produce a low initial throttle which leads to high start delays.
In order to increase the initial throttle, there are two options: increase the initial acceleration output by Autoware, or modify the acceleration to throttle mapping.
The initial acceleration output by Autoware can be tuned in the motion_velocity_smoother with parameters engage_velocity and engage_acceleration. However, the vehicle_cmd_gate applies a filter on the control command to prevent too sudden changes in jerk and acceleration, limiting the maximum allowed acceleration while the ego vehicle is stopped.
Alternatively, the mapping of acceleration can be tuned to increase the throttle corresponding to the initial acceleration. If we look at an example acceleration map, it does the following conversion: when the ego velocity is 0 (first column), acceleration values between 0.631 (first row) and 0.836 (second row) are converted to a throttle between 0% and 10%. This means that any initial acceleration bellow 0.631m/s\u00b2 will not produce any throttle. Keep in mind that after tuning the acceleration map, it may be necessary to also update the brake map.
Although CUDA installation is recommended to achieve better performance for object detection and traffic light recognition in Autoware Universe, it is possible to run these algorithms without CUDA. The following subsections briefly explain how to run each algorithm in such an environment.
"},{"location":"how-to-guides/others/running-autoware-without-cuda/#running-2d3d-object-detection-without-cuda","title":"Running 2D/3D object detection without CUDA","text":"
Autoware Universe's object detection can be run using one of five possible configurations:
lidar_centerpoint
lidar_apollo_instance_segmentation
lidar-apollo + tensorrt_yolo
lidar-centerpoint + tensorrt_yolo
euclidean_cluster
Of these five configurations, only the last one (euclidean_cluster) can be run without CUDA. For more details, refer to the euclidean_cluster module's README file.
"},{"location":"how-to-guides/others/running-autoware-without-cuda/#running-traffic-light-detection-without-cuda","title":"Running traffic light detection without CUDA","text":"
For traffic light recognition (both detection and classification), there are two modules that require CUDA:
traffic_light_ssd_fine_detector
traffic_light_classifier
To run traffic light detection without CUDA, set enable_fine_detection to false in the traffic light launch file. Doing so disables the traffic_light_ssd_fine_detector such that traffic light detection is handled by the map_based_traffic_light_detector module instead.
To run traffic light classification without CUDA, set use_gpu to false in the traffic light classifier launch file. Doing so will force the traffic_light_classifier to use a different classification algorithm that does not require CUDA or a GPU.
"},{"location":"how-to-guides/training-machine-learning-models/training-models/","title":"Training and Deploying Models","text":""},{"location":"how-to-guides/training-machine-learning-models/training-models/#training-and-deploying-models","title":"Training and Deploying Models","text":""},{"location":"how-to-guides/training-machine-learning-models/training-models/#overview","title":"Overview","text":"
The Autoware offers a comprehensive array of machine learning models, tailored for a wide range of tasks including 2D and 3D object detection, traffic light recognition and more. These models have been meticulously trained utilizing open-mmlab's extensive repositories. By leveraging the provided scripts and following the training steps, you have the capability to train these models using your own dataset, tailoring them to your specific needs.
Furthermore, you will find the essential conversion scripts to deploy your trained models into Autoware using the mmdeploy repository.
The traffic light classifier model within the Autoware has been trained using the mmlab/pretrained repository. The Autoware offers pretrained models based on EfficientNet-b1 and MobileNet-v2 architectures. To fine-tune these models, a total of 83,400 images were employed, comprising 58,600 for training, 14,800 for evaluation, and 10,000 for testing. These images represent Japanese traffic lights and were trained using TIER IV's internal dataset.
Name Input Size Test Accuracy EfficientNet-b1 128 x 128 99.76% MobileNet-v2 224 x 224 99.81%
Comprehensive training instructions for the traffic light classifier model are detailed within the readme file accompanying \"traffic_light_classifier\" package. These instructions will guide you through the process of training the model using your own dataset. To facilitate your training, we have also provided an example dataset containing three distinct classes (green, yellow, red), which you can leverage during the training process.
Detailed instructions for training the traffic light classifier model can be found here.
Autoware is scalable and can be customized to work with distributed or less powerful hardware. The minimum hardware requirements given below are just a general recommendation. However, performance will be improved with more cores, RAM and a higher-spec graphics card or GPU core.
CPU with 8 cores
16GB RAM
[Optional] NVIDIA GPU (4GB RAM)
Although GPU is not required to run basic functionality, it is mandatory to enable the following neural network related functions:
LiDAR based object detection
Camera based object detection
Traffic light detection and classification
For details of how to enable object detection and traffic light detection/classification without a GPU, refer to the Running Autoware without CUDA.
Docker can ensure that all developers in a project have a common, consistent development environment. It is recommended for beginners, casual users, people who are unfamiliar with Ubuntu.
For more information, refer to the Docker installation guide.
Source installation is for the cases where more granular control of the installation environment is needed. It is recommended for experienced users or people who want to customize their environment. Note that some problems may occur depending on your local environment.
For more information, refer to the source installation guide.
"},{"location":"installation/#installing-related-tools","title":"Installing related tools","text":"
Some other tools are required depending on the evaluation you want to do. For example, to run an end-to-end simulation you need to install an appropriate simulator.
For more information, see here.
"},{"location":"installation/#additional-settings-for-developers","title":"Additional settings for developers","text":"
There are also tools and settings for developers, such as Shells or IDEs.
For more information, see here.
"},{"location":"installation/additional-settings-for-developers/","title":"Additional settings for developers","text":""},{"location":"installation/additional-settings-for-developers/#additional-settings-for-developers","title":"Additional settings for developers","text":""},{"location":"installation/additional-settings-for-developers/#console-settings-for-ros-2","title":"Console settings for ROS 2","text":""},{"location":"installation/additional-settings-for-developers/#colorizing-logger-output","title":"Colorizing logger output","text":"
By default, ROS 2 logger doesn't colorize the output. To colorize it, write the following in your .bashrc:
export RCUTILS_COLORIZED_OUTPUT=1\n
"},{"location":"installation/additional-settings-for-developers/#customizing-the-format-of-logger-output","title":"Customizing the format of logger output","text":"
By default, ROS 2 logger doesn't output detailed information such as file name, function name, or line number. To customize it, write the following in your .bashrc:
export RCUTILS_CONSOLE_OUTPUT_FORMAT=\"[{severity} {time}] [{name}]: {message} ({function_name}() at {file_name}:{line_number})\"\n
For more options, see here.
"},{"location":"installation/additional-settings-for-developers/#network-settings-for-ros-2","title":"Network settings for ROS 2","text":"
ROS 2 employs DDS, and the configuration of ROS 2 and DDS is described separately. For ROS 2 networking concepts, refer to the official documentation.
ROS 2 multicasts data on the local network by default. Therefore, when you develop in an office, the data flows over the local network of your office. It may cause collisions of packets or increases in network traffic.
To avoid these, there are two options.
Localhost-only communication
Same domain only communication on the local network
Unless you plan to use multiple host computers on the local network, localhost-only communication is recommended. For details, refer to the sections below.
Write the following in your .bashrc: For more information, see the ROS 2 documentation.
export ROS_LOCALHOST_ONLY=1\n
If you export ROS_LOCALHOST_ONLY=1, MULTICAST must be enabled at the loopback address. To verify that MULTICAST is enabled, use the following command.
$ ip link show lo\n1: lo: <LOOPBACK,MULTICAST,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000\n
If the word MULTICAST is not printed, use the following command to enable it.
sudo ip link set lo multicast on\n
"},{"location":"installation/additional-settings-for-developers/#same-domain-only-communication-on-the-local-network","title":"Same domain only communication on the local network","text":"
ROS 2 uses ROS_DOMAIN_ID to create groups and communicate between machines in the groups. Since all ROS 2 nodes use domain ID 0 by default, it may cause unintended interference.
To avoid it, set a different domain ID for each group in your .bashrc:
# Replace X with the Domain ID you want to use\n# Domain ID should be a number in range [0, 101] (inclusive)\nexport ROS_DOMAIN_ID=X\n
Also confirm that ROS_LOCALHOST_ONLY is 0 by using the following command.
echo $ROS_LOCALHOST_ONLY # If the output is 1, localhost has priority.\n
For more information, see the ROS 2 Documentation.
Autoware uses DDS for inter-node communication. ROS 2 documentation recommends users to tune DDS to utilize its capability. Especially, receive buffer size is the critical parameter for Autoware. If the parameter is not large enough, Autoware will failed in receiving large data like point cloud or image.
Unless customized, CycloneDDS is adopted by default. For example, to execute Autoware with CycloneDDS, prepare a config file. A sample config file is given below. Save it as cyclonedds_config.xml.
This configuration is mostly taken from Eclipse Cyclone DDS:Run-time configuration documentation. You can see why each value is set as such under the documentation link.
Set the config file path and enlarge the Linux kernel maximum buffer size before launching Autoware.
For more information, Refer to ROS 2 documentation. Reading user guide for chosen DDS is helpful for more understanding.
"},{"location":"installation/additional-settings-for-developers/#tuning-dds-for-multiple-host-computers-for-advanced-users","title":"Tuning DDS for multiple host computers (for advanced users)","text":"
When Autoware runs on multiple host computers, IP Fragmentation should be taken into account. As ROS 2 documentation recommends, parameters for IP Fragmentation should be set as shown in the following example.
"},{"location":"installation/autoware/docker-installation-devel/","title":"Docker installation for development","text":""},{"location":"installation/autoware/docker-installation-devel/#docker-installation-for-development","title":"Docker installation for development","text":""},{"location":"installation/autoware/docker-installation-devel/#prerequisites","title":"Prerequisites","text":"
Git
For NVIDIA Jetson devices, install JetPack >= 5.0
"},{"location":"installation/autoware/docker-installation-devel/#how-to-set-up-a-development-environment","title":"How to set up a development environment","text":"
Clone autowarefoundation/autoware and move to the directory.
"},{"location":"installation/autoware/docker-installation-devel/#installing-dependencies-using-ansible","title":"Installing dependencies using Ansible","text":"
Be very careful with this method. Make sure you read and confirmed all the steps in the Ansible configuration before using it.
If you've manually installed the dependencies, you can skip this section.
./setup-dev-env.sh docker\n
You might need to log out and log back to make the current user able to use docker.
"},{"location":"installation/autoware/docker-installation-devel/#how-to-set-up-a-workspace","title":"How to set up a workspace","text":"
Warning
Before proceeding, confirm and agree with the NVIDIA Deep Learning Container license. By pulling and using the Autoware Universe images, you accept the terms and conditions of the license.
Create the autoware_map directory for map data later.
After that, move to the workspace in the container:
cd autoware\n
Create the src directory and clone repositories into it.
mkdir src\nvcs import src < autoware.repos\n
Update dependent ROS packages.
The dependency of Autoware may change after the Docker image was created. In that case, you need to run the following commands to update the dependency.
"},{"location":"installation/autoware/docker-installation-prebuilt/","title":"Docker installation for quick start","text":""},{"location":"installation/autoware/docker-installation-prebuilt/#docker-installation-for-quick-start","title":"Docker installation for quick start","text":""},{"location":"installation/autoware/docker-installation-prebuilt/#how-to-set-up-a-development-environment","title":"How to set up a development environment","text":"
Installing dependencies manually
Install Docker Engine
Install NVIDIA Container Toolkit
Install rocker
"},{"location":"installation/autoware/docker-installation-prebuilt/#how-to-set-up-a-workspace","title":"How to set up a workspace","text":"
Create the autoware_map directory for map data later.
Since this page explains Docker-specific information, it is recommended to see Source installation as well if you need detailed information.
Here are two ways to install Autoware by docker:
The first way is to start Autoware with prebuilt image, this is a quick start, this way you can only run Autoware simulator and not develop Autoware, it is only suitable for beginners
The second way is to start Autoware with devel image, which supports developing and running Autoware using docker
"},{"location":"installation/autoware/docker-installation/#docker-installation-for-quick-start","title":"Docker installation for quick start","text":"
docker installation for quick start
"},{"location":"installation/autoware/docker-installation/#docker-installation-for-development","title":"Docker installation for development","text":"
"},{"location":"installation/autoware/docker-installation/#cuda-error-forward-compatibility-was-attempted-on-non-supported-hw","title":"cuda error: forward compatibility was attempted on non supported hw","text":"
When starting Docker with GPU support enabled for NVIDIA graphics, you may sometimes receive the following error:
docker: Error response from daemon: OCI runtime create failed: container_linux.go:349: starting container process caused \"process_linux.go:449: container init caused \\\"process_linux.go:432: running prestart hook 0 caused \\\\\\\"error running hook: exit status 1, stdout: , stderr: nvidia-container-cli: initialization error: cuda error: forward compatibility was attempted on non supported hw\\\\\\\\n\\\\\\\"\\\"\": unknown.\nERROR: Command return non-zero exit code (see above): 125\n
This usually indicates that a new NVIDIA graphics driver has been installed (usually via apt) but the system has not yet been restarted. A similar message may appear if the graphics driver is not available, for example because of resuming after suspend.
To fix this, restart your system after installing the new NVIDIA driver.
"},{"location":"installation/autoware/docker-installation/#docker-with-nvidia-gpu-fails-to-start-autoware-on-arm64-devices","title":"Docker with NVIDIA gpu fails to start Autoware on arm64 devices","text":"
When starting Docker with GPU support enabled for NVIDIA graphics on arm64 devices, e.g. NVIDIA jetson AGX xavier, you may receive the following error:
nvidia@xavier:~$ rocker --nvidia --x11 --user --volume $HOME/autoware -- ghcr.io/autowarefoundation/autoware-universe:humble-latest-cuda-arm64\n...\n\nCollecting staticx==0.12.3\nDownloading https://files.pythonhosted.org/packages/92/ff/d9960ea1f9db48d6044a24ee0f3d78d07bcaddf96eb0c0e8806f941fb7d3/staticx-0.12.3.tar.gz (68kB)\nComplete output from command python setup.py egg_info:\nTraceback (most recent call last):\nFile \"\", line 1, in\nFile \"/tmp/pip-install-m_nm8mya/staticx/setup.py\", line 4, in\nfrom wheel.bdist_wheel import bdist_wheel\nModuleNotFoundError: No module named 'wheel'\n\nCommand \"python setup.py egg_info\" failed with error code 1 in /tmp/pip-install-m_nm8mya/staticx/\n...\n
This error exists in current version of rocker tool, which relates to the os_detection function of rocker.
To fix this error, temporary modification of rocker source code is required, which is not recommended.
At current stage, it is recommended to run docker without NVIDIA gpu enabled for arm64 devices:
This section describes a process to run arm64 systems on amd64 systems using qemu-user-static.
Initially, your system is usually incompatible with arm64 systems. To check that:
$ docker run --rm -t arm64v8/ubuntu uname -m\nWARNING: The requested image's platform (linux/arm64/v8) does not match the detected host platform (linux/amd64) and no specific platform was requested\nstandard_init_linux.go:228: exec user process caused: exec format error\n
Installing qemu-user-static enables us to run arm64 images on amd64 systems.
$ sudo apt-get install qemu-user-static\n$ docker run --rm --privileged multiarch/qemu-user-static --reset -p yes\n$ docker run --rm -t arm64v8/ubuntu uname -m\nWARNING: The requested image's platform (linux/arm64/v8) does not match the detected host platform (linux/amd64) and no specific platform was requested\naarch64\n
To run Autoware's Docker images of arm64 architecture, add the suffix -arm64.
$ docker run --rm -it ghcr.io/autowarefoundation/autoware-universe:humble-latest-cuda-arm64\nWARNING: The requested image's platform (linux/arm64) does not match the detected host platform (linux/amd64) and no specific platform was requested\nroot@5b71391ad50f:/autoware#\n
Note: If you wish to use ROS 2 Galactic on Ubuntu 20.04, refer to installation instruction from galactic branch, but be aware that Galactic version of Autoware might not have latest features.
"},{"location":"installation/autoware/source-installation/#how-to-set-up-a-development-environment","title":"How to set up a development environment","text":"
Clone autowarefoundation/autoware and move to the directory.
If you are installing Autoware for the first time, you can automatically install the dependencies by using the provided Ansible script.
./setup-dev-env.sh\n
If you encounter any build issues, please consult the Troubleshooting section for assistance.
Info
Before installing NVIDIA libraries, please ensure that you have reviewed and agreed to the licenses.
CUDA
cuDNN
TensorRT
Note
The following items will be automatically installed. If the ansible script doesn't work or if you already have different versions of dependent libraries installed, please install the following items manually.
Install ROS 2
Install ROS 2 Dev Tools
Install the RMW Implementation
Install pacmod
Install Autoware Core dependencies
Install Autoware Universe dependencies
Install pre-commit dependencies
Install Nvidia CUDA
Install Nvidia cuDNN and TensorRT
If you didn't use ansible script you will need to download some package artifacts as explained in Manual loading of artifacts. Otherwise some packages (mostly from perception) will not be able to run as they need these artifacts for the inference.
"},{"location":"installation/autoware/source-installation/#how-to-set-up-a-workspace","title":"How to set up a workspace","text":"
Create the src directory and clone repositories into it.
Autoware uses vcstool to construct workspaces.
cd autoware\nmkdir src\nvcs import src < autoware.repos\n
Install dependent ROS packages.
Autoware requires some ROS 2 packages in addition to the core components. The tool rosdep allows an automatic search and installation of such dependencies. You might need to run rosdep update before rosdep install.
"},{"location":"installation/related-tools/","title":"Installation of related tools","text":""},{"location":"installation/related-tools/#installation-of-related-tools","title":"Installation of related tools","text":"
The Autoware perception stack uses models for inference. These models are automatically downloaded if using ansible, but they can also be downloaded manually.
"},{"location":"models/#onnx-model-files","title":"ONNX model files","text":""},{"location":"models/#download-instructions","title":"Download instructions","text":"
The ONNX model files are stored in a common location, hosted by Web.Auto
Any tool that can download files from the web (e.g. wget or curl) is the only requirement for downloading these files:
This document is created to describe and give additional information of the sensors and systems supported by Autoware.Auto software.
All equipment listed in this document has available ROS 2 drivers and has been tested by one or more of the community members on field in autonomous vehicle and robotics applications.
The listed sensors and systems are not sold, developed or given direct technical support by the Autoware community. Having said that any ROS 2 and Autoware related issue regarding the hardware usage could be asked using the community guidelines which found here.
The documents consists of the sections listed below:
Crystal Rugged solutions which is used for autonomous driving and tested by one or more community members are listed below:
Supported Products List CPU GPU RAM, Interfaces Environmental Autoware Tested (Y/N) AVC 0161-AC Intel\u00ae Xeon\u00ae Scalable Dual GPU RTX Series 2TB RAM,CAN, USB, Ethernet, Serial, Hot-Swap SSD 10-32 VoltVibration:2 G RMS 10-1000 Hz, 3 axes -
Link to company website is here.
"},{"location":"reference-hw/cameras/","title":"CAMERAs","text":""},{"location":"reference-hw/cameras/#cameras","title":"CAMERAs","text":""},{"location":"reference-hw/cameras/#tier-iv-automotive-hdr-cameras","title":"TIER IV Automotive HDR Cameras","text":"
TIER IV's Automotive HDR cameras which have ROS 2 driver and tested by TIER IV are listed below:
Supported Products List MP FPS Interface HDR LFM Trigger /Synchronization Ingress Protection ROS 2 Driver Autoware Tested (Y/N) C1 2.5 30 GMSL2 / USB3 Y (120dB) Y Y IP69K Y Y C2 5.4 30 GMSL2 / USB3 Y (120dB) Y Y IP69K Y Y C3 (to be released in 2024) 8.3 30 GMSL2 / TBD Y (120dB) Y Y IP69K Y Y
Link to ROS 2 driver: https://github.com/tier4/ros2_v4l2_camera
Link to product support site: TIER IV Edge.Auto documentation
Link to product web site: TIER IV Automotive Camera Solution
Lucid Vision cameras which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List MP FPS Interface HDR LFM Trigger /Synchronization Ingress Protection ROS 2 Driver Autoware Tested (Y/N) TRITON 054S 5.4 22 GigE Y Y Y up to IP67 Y Y TRITON 032S 3.2 35.4 GigE N/A N/A Y up to IP67 Y Y
Link to ROS 2 driver: https://gitlab.com/leo-drive/Drivers/arena_camera Link to company website: https://thinklucid.com/triton-gige-machine-vision/
The list of all drivers listed above for easy access as a table with additional information:
Type Maker Driver links License Maintainer Lidar VelodyneHesai Link Apache 2 david.wong@tier4.jpabraham.monrroy@map4.jp Lidar Velodyne Link BSD jwhitley@autonomoustuff.com Lidar Robosense Link BSD zdxiao@robosense.cn Lidar Hesai Link Apache 2 wuxiaozhou@hesaitech.com Lidar Leishen Link - - Lidar Livox Link MIT dev@livoxtech.com Lidar Ouster Link Apache 2 stevenmacenski@gmail.comtom@boxrobotics.ai Radar smartmicro Link Apache 2 opensource@smartmicro.de Camera Flir Link Apache 2 bernd.pfrommer@gmail.com Camera Lucid Vision Link - kcolak@leodrive.ai Camera Allied Vision Link Apache 2 at@email.com GNSS NovAtel Link BSD preed@swri.org GNSS SBG Systems Link MIT support@sbg-systems.com GNSS PolyExplore Link - support@polyexplore.com"},{"location":"reference-hw/imu_ahrs_gnss_ins/","title":"IMU, AHRS & GNSS/INS","text":""},{"location":"reference-hw/imu_ahrs_gnss_ins/#imu-ahrs-gnssins","title":"IMU, AHRS & GNSS/INS","text":""},{"location":"reference-hw/imu_ahrs_gnss_ins/#novatel-gnssins-sensors","title":"NovAtel GNSS/INS Sensors","text":"
NovAtel GNSS/INS sensors which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List INS Rate Roll, Pitch, Yaw Acc. GNSS ROS 2 Driver\u00a0 Autoware Tested (Y/N) PwrPak7D-E2 200 Hz R (0.013\u00b0)P (0.013\u00b0)Y (0.070\u00b0) 20 HzL1 / L2 / L5 555 Channels Y - Span CPT7 200 Hz R (0.01\u00b0)\u00a0P (0.01\u00b0)\u00a0Y (0.03\u00b0) 20 Hz L1 / L2 / L5 555 Channels Y -
Link to ROS 2 driver: https://github.com/swri-robotics/novatel_gps_driver/tree/dashing-devel
Link to company website: https://hexagonpositioning.com/
PolyExplore GNSS/INS sensors which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List INS/IMU Rate Roll, Pitch, Yaw Acc. GNSS ROS 2 Driver\u00a0 Autoware Tested (Y/N) POLYNAV 2000P 100 Hz R (0.01\u00b0)P (0.01\u00b0)Y (0.1\u00b0) L1 / L2240 Channels Y - POLYNAV 2000S 100 Hz R (0.015\u00b0)P (0.015\u00b0)Y (0.08\u00b0) L1 / L240 Channels Y -
Link to ROS 2 driver: https://github.com/polyexplore/ROS2_Driver
Link to company website: https://www.polyexplore.com/
"},{"location":"reference-hw/lidars/","title":"LIDARs","text":""},{"location":"reference-hw/lidars/#lidars","title":"LIDARs","text":""},{"location":"reference-hw/lidars/#velodyne-3d-lidar-sensors","title":"Velodyne 3D LIDAR Sensors","text":"
Velodyne Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) Alpha Prime 245m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Ultra Puck 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Puck 100m (+15\u00b0)/(-15\u00b0), (360\u00b0) Y Y Puck Hi-res 100m (+10\u00b0)/(-10\u00b0), (360\u00b0) Y Y
Link to ROS 2 drivers: https://github.com/tier4/nebula https://github.com/ros-drivers/velodyne/tree/ros2/velodyne_pointcloud https://gitlab.com/autowarefoundation/autoware.auto/AutowareAuto/-/tree/master/src/drivers/velodyne_nodes https://github.com/autowarefoundation/awf_velodyne/tree/tier4/universe
Link to company website: https://velodynelidar.com/
"},{"location":"reference-hw/lidars/#robosense-3d-lidar-sensors","title":"RoboSense 3D LIDAR Sensors","text":"
RoboSense Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) RS-Ruby 250m (+15\u00b0)/(-25\u00b0), (360) Y - RS-Ruby-Lite 230m (+15\u00b0)/(-25\u00b0), (360) Y - RS-LiDAR-32 200m (+15\u00b0)/(-25\u00b0), (360) Y - RS-LiDAR-16 150m (+15\u00b0)/(-15), (360) Y -
Link to ROS 2 driver: https://github.com/RoboSense-LiDAR/rslidar_sdk
Link to company website: https://www.robosense.ai/
"},{"location":"reference-hw/lidars/#hesai-3d-lidar-sensors","title":"HESAI 3D LIDAR Sensors","text":"
Hesai Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) Pandar 128 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y - Pandar 64 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Pandar 40P 200m (+15\u00b0)/(-25\u00b0), (360\u00b0) Y Y Pandar XT 120m (+15\u00b0)/(-16\u00b0), (360\u00b0) Y Y Pandar QT 20m (-52.1\u00b0/+52.1\u00b0)/(360\u00b0) Y Y
Link to ROS 2 drivers: https://github.com/tier4/nebula https://github.com/HesaiTechnology/HesaiLidar_General_ROS
Link to company website: https://www.hesaitech.com/en/
"},{"location":"reference-hw/lidars/#leishen-3d-lidar-sensors","title":"Leishen 3D LIDAR Sensors","text":"
Leishen Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) LS C16 150m (+15\u00b0/-15\u00b0), (360\u00b0) Y - LS C32\u00a0 150m (+15\u00b0/-15\u00b0), (360\u00b0) Y - CH 32 120m (+3.7\u00b0/-6.7\u00b0),(120\u00b0) Y - CH 128 20m (+14\u00b0/-17\u00b0)/(150\u00b0) Y -
Link to ROS 2 driver: https://github.com/leishen-lidar
Link to company website: http://www.lslidar.com/
"},{"location":"reference-hw/lidars/#livox-3d-lidar-sensors","title":"Livox 3D LIDAR Sensors","text":"
Livox Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) Horizon 260m (81.7\u00b0), (25.1\u00b0) Y Y Mid-70 90m (70.4\u00b0), (77.2\u00b0) Y - Avia 190m (70.4\u00b0), Circular Y - HAP 150m (25\u00b0), (120\u00b0) - -
Link to ROS 2 driver: https://github.com/Livox-SDK/livox_ros2_driver
Link to company website: https://www.livoxtech.com/
"},{"location":"reference-hw/lidars/#ouster-3d-lidar-sensors","title":"Ouster 3D LIDAR Sensors","text":"
Ouster Lidars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (V), (H) ROS 2 Driver Autoware Tested (Y/N) OS0 50m (90\u00b0), (360\u00b0) Y - OS1 120m (45\u00b0), (360\u00b0) Y - OS2 240m (22,5\u00b0), (360\u00b0) Y Y
Link to ROS 2 driver: https://github.com/ros-drivers/ros2_ouster_drivers
Smartmicro Radars which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List Range FOV (Azimuth), (Elevation) ROS 2 Driver Autoware Tested (Y/N) Type 153 (Triple Mode Short, Medium Long) S:0.2...19 m\u00a0M:0.4...55 m L:0.8...120 m Short: (130\u00b0), (15\u00b0) Medium: (130\u00b0), (15\u00b0)Long: (100\u00b0),(15\u00b0) Y Y Type 132 ,(Dual Mode ,Medium, Long) M:0.5...64 m\u00a0\u00a0L:1...175 m Medium: (100\u00b0), (15\u00b0) Long: (32\u00b0), (15\u00b0) Y Y
Link to ROS 2 driver: https://github.com/smartmicro/smartmicro_ros2_radars
Link to company website: https://www.smartmicro.com/automotive-radar
FLIR ADK Thermal Vision cameras which has ROS 2 driver and tested by one or more community members are listed below:
Supported Products List MP FPS Interface Spectral Band FOV ROS 2 Driver Autoware Tested (Y/N) FLIR ADK 640x512 30 USB-GMSL,Ethernet 8-14 um (LWIR) 75\u02da, 50\u02da, 32\u02da, and 24\u02da - -"},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/","title":"Vehicle Drive By Wire Suppliers","text":""},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/#vehicle-drive-by-wire-suppliers","title":"Vehicle Drive By Wire Suppliers","text":""},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/#new-eagle-dbw-solutions","title":"New Eagle DBW Solutions","text":"
New Eagle DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Jeep CherokeeChrysler PacificaToyota PriusChevy BoltFord TransitRAM 1500Custom\u00a0 500W Sine Inverter2000 Watts8 Channel PDS Optional, Available Y Y
Link to company website: https://neweagle.net/autonomous-machines/
Dataspeed DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Lincoln MKZ, NautilusFord Fusion, F150, Transit Connect, RangerChrysler PacificaJeep CherokeePolaris GEM, RZR 12 Channel PDS,15 A Each at 12 V Optional, Available Y -
Link to company website: https://www.dataspeedinc.com/
Autonomous Stuff Pacmod DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Polaris GEM SeriesPolaris eLXD MY 2016+Polaris Ranger X900International ProStarLexus RX-450h MYFord RangerToyota Minivan Power distribution panel Optional, Available Y Y
Link to company website: https://autonomoustuff.com/platform/pacmod
"},{"location":"reference-hw/vehicle_drive_by_wire_suppliers/#schaeffler-paravan-space-drive-dbw-solutions","title":"Schaeffler-Paravan Space Drive DBW Solutions","text":"
Schaeffler-Paravan Space Drive DBW Controllers which is used for autonomous driving and tested by one or more community members are listed below:
Supported Vehicles Power Remote Control ROS 2 Support Autoware Tested (Y/N) Custom Integration with Actuators - Optional, Available Y Y
Link to company website: https://www.schaeffler-paravan.de/en/products/space-drive-system/
PIX Moving AV solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Electric DBW Chassis and Platforms Y Y Y -
Link to company website: https://www.pixmoving.com/pixkit
Different sizes of platforms
"},{"location":"reference-hw/vehicle_platform_suppliers/#autonomoustuff-av-solutions","title":"Autonomoustuff AV Solutions","text":"
Autonomoustuff platform solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Road Vehicles, Golf Carts & Trucks Y Y Y -
Link to company website: https://autonomoustuff.com/platform
"},{"location":"reference-hw/vehicle_platform_suppliers/#navya-av-solutions","title":"NAVYA AV Solutions","text":"
NAVYA platform solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Shuttle Bus, Taxi and Tow Tractors Y Y - -
Link to company website: https://navya.tech/en
"},{"location":"reference-hw/vehicle_platform_suppliers/#zing-robotics-av-solutions","title":"ZING ROBOTICS AV Solutions","text":"
ZING Robotics platform solutions which is used for autonomous development and tested by one or more community members are listed below:
Vehicle Types Sensors Integrated Autoware Installed ROS 2 Support Autoware Tested (Y/N) Purpose built electric autonomous vehicles for aviation, military etc. Y Y - -
Link to company website: https://www.zingrobotics.com/
This page explains several documentation sites that are useful for Autoware and ROS development.
The Autoware Foundation is the official site of the Autoware Foundation. You can learn about the Autoware community here.
Autoware Documentation (this site) is the central documentation site for Autoware maintained by the Autoware community. General software-related information of Autoware is aggregated here.
Autoware Universe Documentation has READMEs and design documents of software components.
ROS Docs Guide explains the ROS 1 and ROS 2 documentation infrastructure.
Docs guide shows the list of useful documentation sites. Visit them and see if there is any information related to your problem.
Note that the documentation sites aren't always up-to-date and perfect. If you find out that some information is wrong, unclear, or missing in Autoware docs, feel free to submit a pull request following the contribution guidelines.
Warning
Since this documentation site is still under construction, there are some empty pages.
If you encounter a problem with Autoware, check existing issues and questions and search for similar issues first.
Issues
Note that Autoware has multiple repositories listed in autoware.repos. It is recommended to search across the repositories.
Questions
If no answer was found, create a new question thread here. If your question is not answered within a week, then @mention the maintainers to remind them.
Also, there are other discussion types such as feature requests or design discussions. Feel free to open or join such discussions.
If you don't know how to create a discussion, refer to GitHub Docs.
If you have a problem and you have confirmed it is a bug, find the appropriate repository and create a new issue there. If you can't determine the appropriate repository, ask the maintainers for help by creating a new discussion in the Q&A category.
Warning
Do not create issues for questions or unconfirmed bugs. If such issues are created, maintainers will transfer them to GitHub Discussions.
If you want to fix the bug by yourself, discuss the approach with maintainers and submit a pull request.
If you want to widely discuss a topic with the general Autoware and ROS community or ask a question not related to Autoware's bugs, post to the Autoware category on ROS Discourse.
Warning
Do not post questions about bugs to ROS Discourse!
Note that this may break your system and run carefully.
Run the setup script without installing CUDA-related libraries.
./setup-dev-env.sh --no-nvidia\n
Warning
Note that some components in Autoware Universe require CUDA, and only the CUDA version in the env file is supported at this time. Autoware may work with other CUDA versions, but those versions are not supported and functionality is not guaranteed.
Building Autoware requires a lot of memory, and your machine can freeze or crash if memory runs out during a build. To avoid this problem, 16-32GB of swap should be configured.
# Optional: Check the current swapfile\nfree -h\n\n# Remove the current swapfile\nsudo swapoff /swapfile\nsudo rm /swapfile\n\n# Create a new swapfile\nsudo fallocate -l 32G /swapfile\nsudo chmod 600 /swapfile\nsudo mkswap /swapfile\nsudo swapon /swapfile\n\n# Optional: Check if the change is reflected\nfree -h\n
For more detailed configuration steps, along with an explanation of swap, refer to Digital Ocean's \"How To Add Swap Space on Ubuntu 20.04\" tutorial
If there are too many CPU cores (more than 64) in your machine, it might requires larger memory. A workaround here is to limit the job number while building.
You can adjust -j4 to any number based on your system. For more details, see the manual page of GNU make.
By reducing the number of packages built in parallel, you can also reduce the amount of memory used. In the following example, the number of packages built in parallel is set to 1, and the number of jobs used by make is limited to 1.
By lowering both the number of packages built in parallel and the number of jobs used by make, you can reduce the memory usage. However, this also means that the build process takes longer.
"},{"location":"support/troubleshooting/#errors-when-using-the-latest-version-of-autoware","title":"Errors when using the latest version of Autoware","text":"
If you are working with the latest version of Autoware, issues can occur due to out-of-date software or old build files.
To resolve these types of problems, first try cleaning your build artifacts and rebuilding:
If the error is not resolved, remove src/ and update your workspace according to installation type (Docker / source).
Warning
Before removing src/, confirm that there are no modifications in your local environment that you want to keep!
If errors still persist after trying the steps above, delete the entire workspace, clone the repository once again and restart the installation process.
"},{"location":"support/troubleshooting/#errors-when-using-a-fixed-version-of-autoware","title":"Errors when using a fixed version of Autoware","text":"
In principle, errors should not occur when using a fixed version. That said, possible causes include:
ROS 2 has been updated with breaking changes.
For confirmation, check the Packaging and Release Management tag on ROS Discourse.
Your local environment is broken.
Confirm your .bashrc file, environment variables, and library versions.
In addition to the causes listed above, there are two common misunderstandings around the use of fixed versions.
You used a fixed version for autowarefoundation/autoware only. All of the repository versions in the .repos file must be specified in order to use a completely fixed version.
You didn't update the workspace after changing the branch of autowarefoundation/autoware. Changing the branch of autowarefoundation/autoware does not affect the files under src/. You have to run the vcs import command to update them.
"},{"location":"support/troubleshooting/#error-when-building-python-package","title":"Error when building python package","text":"
The error is due to the fact that for versions between 66.0.0 and 67.5.0 setuptools enforces the python packages to be PEP-440 conformant. Since version 67.5.1 setuptools has a fallback that makes it possible to work with old packages again.
The solution is to update setuptools to the newest version with the following command
If any errors occur when running Autoware with Docker or rocker, first confirm that your Docker installation is working correctly by running the following commands:
docker run --rm -it hello-world\ndocker run --rm -it ubuntu:latest\n
Next, confirm that you are able to access the base Autoware image that is stored on the GitHub Packages website
docker run --rm -it ghcr.io/autowarefoundation/autoware-universe:latest\n
"},{"location":"support/troubleshooting/#runtime-issues","title":"Runtime issues","text":""},{"location":"support/troubleshooting/#performance-related-issues","title":"Performance related issues","text":"
Symptoms:
Autoware is running slower than expected
Messages show up late in RViz2
Point clouds are lagging
Camera images are lagging behind
Point clouds or markers flicker on RViz2
When multiple subscribers use the same publishers, the message rate drops
If you have any of these symptoms, please the Performance Troubleshooting page.
"},{"location":"support/troubleshooting/#map-does-not-display-when-running-the-planning-simulator","title":"Map does not display when running the Planning Simulator","text":"
When running the Planning Simulator, the most common reason for the map not being displayed in RViz is because the map path has not been specified correctly in the launch command. You can confirm if this is the case by searching for Could not find lanelet map under {path-to-map-dir}/lanelet2_map.osm errors in the log.
Another possible reason is that map loading is taking a long time due to poor DDS performance. For this, please visit the Performance Troubleshooting page.
When multiple subscribers use the same publishers, the message rate drops
"},{"location":"support/troubleshooting/performance-troubleshooting/#diagnostic-steps","title":"Diagnostic Steps","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#check-if-multicast-is-enabled","title":"Check if multicast is enabled","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#target-symptoms","title":"Target symptoms","text":"
When multiple subscribers use the same publishers, the message rate drops
sudo ip link set multicast on {your-interface-name}\n
This way DDS will function as intended and multiple subscribers can receive data from a single publisher without any significant degradation in performance.
This is a temporary solution. And will be reverted once the computer restarts.
To make it permanent either,
Create a service to run this on startup (recommended)
OR put following lines to the ~/.bashrc file:
if [ ! -e /tmp/multicast_is_set ]; then\nsudo ip link set lo multicast on\ntouch /tmp/multicast_is_set\nfi\n
This will probably ask for password on the terminal every time you restart the computer.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-compilation-flags","title":"Check the compilation flags","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#target-symptoms_1","title":"Target symptoms","text":"
Autoware is running slower than expected
Point clouds are lagging
When multiple subscribers use the same publishers, the message rate drops even further
Check the ~/.bash_history file to see if there are any colcon build directives without -DCMAKE_BUILD_TYPE=Release or -DCMAKE_BUILD_TYPE=RelWithDebInfo flags at all.
Even if a build starts with these flags but same workspace gets compiled without these flags, it will still be a slow build in the end.
In addition, the nodes will run slow in general, especially the pointcloud_preprocessor nodes.
Remove the build, install and optionally log folders in the main autoware folder.
Compile the Autoware with either Release or RelWithDebInfo tags:
colcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=Release\n# Or build with debug flags too (comparable performance but you can debug too)\ncolcon build --symlink-install --cmake-args -DCMAKE_BUILD_TYPE=RelWithDebInfo\n
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-dds-settings","title":"Check the DDS settings","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#target-symptoms_2","title":"Target symptoms","text":"
Autoware is running slower than expected
Messages show up late in RViz2
Point clouds are lagging
Camera images are lagging behind
When multiple subscribers use the same publishers, the message rate drops
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-rmw-ros-middleware-implementation","title":"Check the RMW (ROS Middleware) implementation","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#diagnosis_2","title":"Diagnosis","text":"
Run following to check the middleware used:
echo $RMW_IMPLEMENTATION\n
The return line should be rmw_cyclonedds_cpp. If not, apply the solution.
If you are using a different DDS middleware, we might not have official support for it just yet.
Add export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp as a separate line in you ~/.bashrc file.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-if-the-cyclonedds-is-configured-correctly","title":"Check if the CycloneDDS is configured correctly","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#diagnosis_3","title":"Diagnosis","text":"
Run following to check the configuration .xml file of the CycloneDDS:
echo $CYCLONEDDS_URI\n
The return line should be a valid path pointing to an .xml file with CycloneDDS configuration.
Follow DDS settings:Tuning DDS documentation and make sure:
you have export CYCLONEDDS_URI=/absolute_path_to_your/cyclonedds_config.xml as a line on your ~/.bashrc file.
you have the cyclonedds_config.xml with the configuration provided in the documentation.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-the-linux-kernel-maximum-buffer-size","title":"Check the Linux kernel maximum buffer size","text":""},{"location":"support/troubleshooting/performance-troubleshooting/#diagnosis_4","title":"Diagnosis","text":"
Run: sysctl net.core.rmem_max, it should return at least net.core.rmem_max = 2147483647.
This parameter specifies the maximum size of the \"receive buffer\" for each network connection, which determines the maximum amount of data that can be held in memory at any given time. By increasing the maximum buffer size, the operating system can accommodate larger bursts of data, which can help prevent network congestion and reduce packet loss, resulting in faster and more reliable data transfers.
Run: sysctl net.ipv4.ipfrag_time, it should return around: net.ipv4.ipfrag_time = 3
The \"net.ipv4.ipfrag_time\" parameter specifies the maximum time in seconds that the kernel should retain partially fragmented IP packets before discarding them. The default value for this parameter is usually set to 30 seconds, but it may vary depending on the specific operating system and configuration.
By setting this parameter to a lower value, such as 3 seconds, the kernel can free up memory resources more quickly by discarding partially fragmented packets that are no longer needed, which can help improve the overall performance and stability of the system.
Run: sysctl net.ipv4.ipfrag_high_thresh, it should return at around: net.ipv4.ipfrag_high_thresh = 134217728
The \"net.ipv4.ipfrag_high_thresh\" parameter specifies the high watermark threshold for the number of partially fragmented packets allowed in the kernel IP packet reassembly queue. When the number of partially fragmented packets in the queue exceeds this threshold, the kernel will start to drop newly arrived packets until the number of partially fragmented packets drops below the threshold.
By setting this parameter to a higher value, such as 134217728 (128 MB), the kernel can accommodate a larger number of partially fragmented packets in the queue, which can help improve the performance of network applications that transfer large amounts of data, such as file transfer protocols and multimedia streaming applications.
This will probably ask for password on the terminal every time you restart the computer.
"},{"location":"support/troubleshooting/performance-troubleshooting/#check-if-ros-localhost-only-communication-is-enabled","title":"Check if ROS localhost only communication is enabled","text":"
If you are using multi computer setup, please skip this check.
Enabling ROS localhost only communication can help improve the performance of ROS by reducing network traffic and avoiding potential conflicts with other devices on the network.
Add export $ROS_LOCALHOST_ONLY=1 as a separate line in you ~/.bashrc file.
This environment variable tells ROS to only use the loopback network interface (i.e., localhost) for communication, rather than using the network interface card (NIC) for Ethernet or Wi-Fi. This can reduce network traffic and potential conflicts with other devices on the network, resulting in better performance and stability.
Simulations provide a way of verifying Autoware's functionality before field testing with an actual vehicle. There are three main types of simulation that can be run ad hoc or via a scenario runner.
"},{"location":"tutorials/#simulation-methods","title":"Simulation methods","text":""},{"location":"tutorials/#ad-hoc-simulation","title":"Ad hoc simulation","text":"
Ad hoc simulation is a flexible method for running basic simulations on your local machine, and is the recommended method for anyone new to Autoware.
Scenario simulation uses a scenario runner to run more complex simulations based on predefined scenarios. It is often run automatically for continuous integration purposes, but can also be run on a local machine.
Planning simulation uses simple dummy data to test the Planning and Control components - specifically path generation, path following and obstacle avoidance. It verifies that a vehicle can reach a goal destination while avoiding pedestrians and surrounding cars, and is another method for verifying the validity of Lanelet2 maps. It also allows for testing of traffic light handling.
"},{"location":"tutorials/#how-does-planning-simulation-work","title":"How does planning simulation work?","text":"
Generate a path to the goal destination
Control the car along the generated path
Detect and avoid any humans or other vehicles on the way to the goal destination
Digital twin simulation is a simulation type that is able to produce realistic data and simulate almost the entire system. It is also commonly referred to as end-to-end simulation.
"},{"location":"tutorials/ad-hoc-simulation/","title":"Ad hoc simulation","text":""},{"location":"tutorials/ad-hoc-simulation/#ad-hoc-simulation","title":"Ad hoc simulation","text":"
Check if you have ~/autoware_data folder and files in it.
$ cd ~/autoware_data\n$ ls -C -w 30\nimage_projection_based_fusion\nlidar_apollo_instance_segmentation\nlidar_centerpoint\ntensorrt_yolo\ntensorrt_yolox\ntraffic_light_classifier\ntraffic_light_fine_detector\ntraffic_light_ssd_fine_detector\nyabloc_pose_initializer\n
If not, please, follow Manual downloading of artifacts.
"},{"location":"tutorials/ad-hoc-simulation/planning-simulation/#2-set-an-initial-pose-for-the-ego-vehicle","title":"2. Set an initial pose for the ego vehicle","text":"
a) Click the 2D Pose estimate button in the toolbar, or hit the P key.
b) In the 3D View pane, click and hold the left-mouse button, and then drag to set the direction for the initial pose. An image representing the vehicle should now be displayed.
Warning
Remember to set the initial pose of the car in the same direction as the lane.
To confirm the direction of the lane, check the arrowheads displayed on the map.
"},{"location":"tutorials/ad-hoc-simulation/planning-simulation/#3-set-a-goal-pose-for-the-ego-vehicle","title":"3. Set a goal pose for the ego vehicle","text":"
a) Click the 2D Goal Pose button in the toolbar, or hit the G key.
b) In the 3D View pane, click and hold the left-mouse button, and then drag to set the direction for the goal pose. If done correctly, you will see a planned path from initial pose to goal pose.
"},{"location":"tutorials/ad-hoc-simulation/planning-simulation/#4-start-the-ego-vehicle","title":"4. Start the ego vehicle","text":"
Now you can start the ego vehicle driving by clicking the AUTO button on OperationMode in AutowareStatePanel. Alteratively, you can manually start the vehicle by running the following command:
source ~/autoware/install/setup.bash\nros2 service call /api/operation_mode/change_to_autonomous autoware_adapi_v1_msgs/srv/ChangeOperationMode {}\n
After that, you can see AUTONOMOUS sign on OperationMode and AUTO button is grayed out.
By default, traffic lights on the map are all treated as if they are set to green. As a result, when a path is created that passed through an intersection with a traffic light, the ego vehicle will drive through the intersection without stopping.
The following steps explain how to set and reset traffic lights in order to test how the Planning component will respond.
Go to Panels -> Add new panel, select TrafficLightPublishPanel, and then press OK.
In TrafficLightPublishPanel, set the ID and color of the traffic light.
Click the SET button.
Finally, click the PUBLISH button to send the traffic light status to the simulator. Any planned path that goes past the selected traffic light will then change accordingly.
By default, Rviz should display the ID of each traffic light on the map. You can have a closer look at the IDs by zooming in the region or by changing the View type.
In case the IDs are not displayed, try the following troubleshooting steps:
a) In the Displays panel, find the traffic_light_id topic by toggling the triangle icons next to Map > Lanelet2VectorMap > Namespaces.
b) Check the traffic_light_id checkbox.
c) Reload the topic by clicking the Map checkbox twice.
You can update the color of the traffic light by selecting the next color (in the image it is GREEN) and clicking SET button. In the image the traffic light in front of the ego vehicle changed from RED to GREEN and the vehicle restarted.
To remove a traffic light from TrafficLightPublishPanel, click the RESET button.
Check if you have ~/autoware_data folder and files in it.
$ cd ~/autoware_data\n$ ls -C -w 30\nimage_projection_based_fusion\nlidar_apollo_instance_segmentation\nlidar_centerpoint\ntensorrt_yolo\ntensorrt_yolox\ntraffic_light_classifier\ntraffic_light_fine_detector\ntraffic_light_ssd_fine_detector\nyabloc_pose_initializer\n
If not, please, follow Manual downloading of artifacts.
Sample map and rosbag: Copyright 2020 TIER IV, Inc.
Due to privacy concerns, the rosbag does not contain image data, which will cause:
Traffic light recognition functionality cannot be tested with this sample rosbag.
Object detection accuracy is decreased.
"},{"location":"tutorials/ad-hoc-simulation/rosbag-replay-simulation/#how-to-run-a-rosbag-replay-simulation","title":"How to run a rosbag replay simulation","text":"
Any kind of for-profit activity with the trial version of the MORAI SIM:Drive is strictly prohibited.
"},{"location":"tutorials/ad-hoc-simulation/digital-twin-simulation/MORAI_Sim-tutorial/#hardware-requirements","title":"Hardware requirements","text":"Minimum PC Specs OS Windows 10, Ubuntu 20.04, Ubuntu 18.04, Ubuntu 16.04 CPU Intel i5-9600KF or AMD Ryzen 5 3500X RAM DDR4 16GB GPU RTX2060 Super Required PC Specs OS Windows 10, Ubuntu 20.04, Ubuntu 18.04, Ubuntu 16.04 CPU Intel i9-9900K or AMD Ryzen 7 3700X (or higher) RAM DDR4 64GB (or higher) GPU RTX2080Ti or higher"},{"location":"tutorials/ad-hoc-simulation/digital-twin-simulation/MORAI_Sim-tutorial/#application-and-download","title":"Application and Download","text":"
Only for AWF developers, trial license for 3 months can be issued. Download the application form and send to Hyeongseok Jeon
After the trial license is issued, you can login to MORAI Sim:Drive via Launchers (Windows/Ubuntu)
CAUTION: Do not use the Launchers in the following manual
"},{"location":"tutorials/scenario-simulation/planning-simulation/random-test-simulation/","title":"Random test simulation","text":""},{"location":"tutorials/scenario-simulation/planning-simulation/random-test-simulation/#random-test-simulation","title":"Random test simulation","text":"
Note
Running the Scenario Simulator requires some additional steps on top of building and installing Autoware, so make sure that Scenario Simulator installation has been completed first before proceeding.
For more information about supported parameters, refer to the random_test_runner documentation.
"},{"location":"tutorials/scenario-simulation/planning-simulation/scenario-test-simulation/","title":"Scenario test simulation","text":""},{"location":"tutorials/scenario-simulation/planning-simulation/scenario-test-simulation/#scenario-test-simulation","title":"Scenario test simulation","text":"
Note
Running the Scenario Simulator requires some additional steps on top of building and installing Autoware, so make sure that Scenario Simulator installation has been completed first before proceeding.
+ 
+ 
+