Skip to content

Commit

Permalink
Improve Alarm Docs (#1073)
Browse files Browse the repository at this point in the history
* Update alarms.rst

* Update alarms.rst

* Update alarms.rst

* Update alarms.rst

* Update alarms.rst

* Update and rename alarms.rst to alarmspart2.rst

* Update and rename alarmspart2.rst to alarms.rst

* Update index.rst

* Update and rename alarms.rst to alarmspart2.rst

* Create alarms.md

* Update alarms.md

* Update alarms.md

* Create services.md

* Update alarms.md

* Update alarms.md

* Update alarms.md

* Update alarms.md

* Update alarms.md

* Update index.rst

* Update alarms.md

* Fix docs so script will build

* Remove custom services docs page in order to avoid duplicating publicly available info about ROS

* Format markdown file to remove long line lengths

---------

Co-authored-by: Cameron Brown <[email protected]>
  • Loading branch information
MrinallU and cbrxyz authored Mar 5, 2024
1 parent 44997b7 commit e58b169
Show file tree
Hide file tree
Showing 2 changed files with 115 additions and 0 deletions.
114 changes: 114 additions & 0 deletions docs/software/alarms.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,114 @@
# Alarms in ROS

In the realm of building dependable control systems, the importance of error detection
and effective error-handling mechanisms cannot be overstated. Within this context,
MIL presents a robust solution in the form of a live alarm system. This alarm system
operates discreetly in the background of both the robot's mission and driver codebases,
ready to be activated upon the emergence of errors. Notably, the alarm code doesn't
solely serve to identify and address errors; it can also adeptly manage changes
or updates that extend beyond error scenarios.

## ROS Alarms: A Service-Oriented Architecture

The architecture of ROS alarms distinguishes itself by employing a service-oriented
model rather than the usual topic-based approach. In ROS, Services act as the
conduits for interaction between nodes, functioning in a request-response manner.
While ROS topics enable asynchronous data exchange, services facilitate nodes in
seeking specific actions or information from other nodes, awaiting a subsequent
response before proceeding. This method of waiting before proceeding is known as a
synchronous data exchange. This proves especially valuable in tasks that require
direct engagement, such as data retrieval or computations.

## Alarm System Logic

The alarm system's functionality is more intricate than that of a typical ROS
service, which usually manages operations of base types (ints, strings, etc.).
In this scenario, the alarm's service server is engineered to manage the tasks
of updating, querying, and processing an alarm object. ROS alarms encompass two
distinct types of clients: the alarm broadcaster and the alarm listener. The
broadcaster initializes and triggers alarms in response to errors or changes,
while the listener monitors the broadcaster's activity and activates designated
a callback function when alarms are raised. The callback function should handle
the error or change appropriately.

To successfully leverage alarms, the initialization of both the broadcaster and
listener is needed. The listener should be configured to execute a predefined
callback function, addressing errors or changes detected by the broadcaster.
Within your codebase, error detection and alarm-raising procedures should be
integrated. If orchestrated correctly, the callback function will be automatically
invoked, underscoring successful error mitigation.

Note that there are several special properties that can be attached to your alarm.
Here are a couple of examples:
* When you raise an alarm you can assign a severity level to the alarm [0, 5].
* You can attach multiple callback functions to the alarm.
* **This is where severity comes into play!** By specifying the required
severity level that is needed to execute the callback when initializing the
function, you can choose which callbacks are executed when the alarm is raised.
* You can also specify a range of severity levels that the alarm would need to
execute a given callback.

Here is a line-by-line breakdown of an example alarm implementation:

```python
ab = AlarmBroadcaster("test_alarm")
al = AlarmListener("test_alarm")
ab.clear_alarm()
rospy.sleep(0.1)
```
This is how you would initialize the alarm broadcaster and listener. Here
make sure to clear any previous alarm data in the broadcaster.

```python
al.add_callback(cb1)
```
Make sure to establish the callback function that should be executed once
the alarm is activated.

```python
ab.raise_alarm()
rospy.sleep(0.1)
assert al.is_raised()
assert cb1_ran
```
When the alarm is sounded via the `raise_alarm()` function, the callback will be
executed automatically.

```python
al.clear_callbacks()

al.add_callback(cb1, severity_required=2)
al.add_callback(cb2, call_when_raised=False)

rospy.loginfo("Severity Range Test 1")
ab.raise_alarm(severity=4)
rospy.sleep(0.1)
assert not cb1_ran
assert cb2_ran
cb2_ran = False

rospy.loginfo("Severity Range Test 2")
ab.raise_alarm(severity=1)
rospy.sleep(0.1)
assert cb1_ran
assert not cb2_ran
cb1_ran = False
```
Note that you can also attach some special properties to your alarm. For instance,
you can attach multiple callback functions to the alarm. You can also configure
whether the callback function should be automatically executed when the alarm is
raised or whether it should be executed manually. Finally, you can assign a
severity level to the alarm which can tell the alarm code which callback functions
should be run.

## Applications and Context

The applications of ROS alarms span various contexts, with one notable application
residing in the control of the submersible vehicle's thrust and killboard. The
thrust and killboard, responsible for the sub's electronic operations, is
integrally associated with ROS alarms. Upon the board's activation or deactivation
(hard or soft kill), alarms are invoked to apprise users of these changes. The
listener's callback function comes into play, ensuring that alarms are updated
in alignment with the board's current state. This process triggered each time
the board is deactivated, creates a system whereby users are continually informed
about the board's status changes – essentially manifesting a dynamic live alarm system.
1 change: 1 addition & 0 deletions docs/software/index.rst
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@ Various documentation related to practices followed by the MIL Software team.
asyncio
rqt
rostest
alarms
Bash Style Guide <bash_style.md>
C++ Style Guide <cpp_style.md>
Python Style Guide <python_style.md>
Expand Down

0 comments on commit e58b169

Please sign in to comment.