Skip to content

Latest commit

 

History

History
128 lines (95 loc) · 6.06 KB

installation.md

File metadata and controls

128 lines (95 loc) · 6.06 KB
Raw

Installation

The easiest way to get started is to install the PX4 dev environment, specifically the ubuntu_sim_ros_melodic.sh script. It installs many of the required dependencies, including ROS, Gazebo, and MAVROS:

wget -q -O - https://raw.githubusercontent.com/PX4/Devguide/master/build_scripts/ubuntu_sim_ros_melodic.sh | bash

Watch the output of the script closely as you may have to run it twice to make sure everything installs/builds correctly.

Install Gazebo with physics and sensor lockstep

The default ROS distribution comes with Gazebo 9.0.0 installed. However, to use physics and sensor lockstep we need at least version 9.16.0 which can be installed by using the official Ubuntu packages:

sudo sh -c 'echo "deb http://packages.osrfoundation.org/gazebo/ubuntu-stable `lsb_release -cs` main" > /etc/apt/sources.list.d/gazebo-stable.list'
wget https://packages.osrfoundation.org/gazebo.key -O - | sudo apt-key add -
sudo apt update
sudo apt install gazebo9
sudo apt upgrade  # this is needed to update libignition!
gazebo --version  # should output version 9.16.0 (or higher)

Install the PX4 autopilot

To install and build PX4, issue the following commands:

git clone -b v1.10.2 https://github.com/PX4/PX4-Autopilot.git px4
cd px4
echo "export PX4_ROOT=$(realpath .)" >> ~/.bashrc  # or .zshrc
sudo apt install libgstreamer-plugins-base1.0-dev # libgstreamer is a dependency!
sed -i 's/TRUE/true/' Tools/sitl_gazebo/include/gazebo_opticalflow_plugin.h # use boolean instead of macro
make px4_sitl gazebo  # Gazebo should open with a 3DR Iris quadcopter

Install selected ROS packages with Python 3 support

Since we are using Python 3 (specifically version 3.6.9), we need to build the following packages with Python 3 support:

The vision_opencv package is required for converting images using cv_bridge, whereas geometry2 and orocos_kinematics_dynamics are required for frame transformations with tf2.

To clone the repositories into the workspace, we use wstool:

cd ~/catkin_ws/src
git clone https://github.com/fabianschilling/vswarm.git
wstool init  # should already be initialized!
wstool merge vswarm/vswarm.rosinstall
wstool update

Update dependencies using rosdep:

rosdep install --from-paths ~/catkin_ws/ --ignore-src -r -y

Tell ROS we'll be using Python 3:

echo "export ROS_PYTHON_VERSION=3" >> "$HOME"/.bashrc
sed -i 's/python/python3/' ~/catkin_ws/src/gazebo_ros_pkgs/gazebo_ros/scripts/spawn_model

Update Python dependencies:

cd ~/catkin_ws/src/
sudo apt install python3-pip  # required for pip installations
sudo apt install python3-sip-dev  # PyKDL requirement
python3 -m pip install --user -r vswarm/requirements.txt
# This is a bad hack, but it works :) (build requirement for python_orocos_kdl)
ln -sf /usr/lib/python2.7/dist-packages/sipconfig{,_nd}.py ~/.local/lib/python3.6/site-packages/
ln -sf ${PX4_ROOT}/ROMFS ~/.ros/  # Let PX4 find its config (even when working dir is ~/.ros)

Before building the above packages, invoke catkin with a configuration that contains the following additional CMake arguments:

cd ~/catkin_ws
catkin config --append-args \
    -DPYTHON_EXECUTABLE=/usr/bin/python3 \
    -DPYTHON_INCLUDE_DIR=/usr/include/python3.6m \
    -DPYTHON_LIBRARY=/usr/lib/x86_64-linux-gnu/libpython3.6m.so
source ~/.bashrc  # Make sure ROS_PYTHON_VERSION=3 is set before you invoke catkin build
catkin build  # You can (probably) safely ignore the warnings

Finally, automatically source the vswarm-specific environment and define some convenience aliases:

# The following line sets an empty PYTHONPATH at the beginning of the ~/.bashrc
# This stops nasty shufflying of order in the PYTHONPATH
# We always want our local catkin workspace to overshadow the global one
sed -i '1s/^/export PYTONPATH=\n/' ~/.bashrc
echo "source ~/catkin_ws/src/vswarm/util/setup.zsh" >> ~/.bashrc
echo "alias vswarm='rosrun vswarm swarm_control.py'" >> ~/.bashrc
echo "alias rqt_reconfigure='rosrun rqt_reconfigure rqt_reconfigure'" >> ~/.bashrc

Download the detection model

The vswarm drone detection model (~30MB) can be downloaded here: [Google Drive] [SWITCHdrive].

The model is a single-class version of YOLOv3-tiny adapted from the ultralytics/yolov3 repository. The .cfg file follows the Darknet config format and the .pt file is a serialized Python dictionary with a single model key that contains a PyTorch state_dict.

To make the ROS detection node aware of the detection model you downloaded, set the following environment variable:

echo "export DETECTION_CHECKPOINT=/path/to/yolov3_tiny.pt" >> ~/.bashrc

Useful tools

The following are not strictly required but are either nice to have or in some cases needed to enable some of the functionality of this repository:

  • Shell: zsh (with oh-my-zsh ) as a more powerful and user-friendly alternative to bash
  • Terminal multiplexer: tmux (with the tmuxp session manager).