PX4 Dev Call: March 30, 2022

rroche · March 23, 2022, 5:00pm

March 30, 2022

Community Q&A : Maximum 3 - 5 min limit for each question. If the discussion isn’t finished, ask the Questioner to create a Github Issue about it and continue discussion there.
Limit discussions to maximum 5 min, then it should be continued in In-Depth discussions.
Always consider the perspective of community members who may not have a deep understanding on certain concepts

Timeline:

Since we are back in person, we are pushing for increased content diversity, and we welcome any of the below:

Safy: What’s the prefered way to do what a goto command in QGC does but through mavros and offboard mode.
The command for goto see https://mavlink.io/en/messages/common.html#MAV_CMD_DO_REPOSITION is exposed in mavros through the command plugin. It’s best to use this command for waypoints because it results in a jerk optimized trajectory to the goal point. When arriving at the destination the vehicle can be switched to offboard mode but then the mavros algorithm needs to take care of the speed and smoothness of further movements.

Follow-up question about how the offabord interface works and if it could be used. It’s documented here: https://docs.px4.io/master/en/computer_vision/obstacle_avoidance.html#obstacle-avoidance-interface
@junwoo0914 had an issue with the build because of changes in MAVLink:
https://github.com/PX4/PX4-Autopilot/pull/19375/files
It’s probably related to the submodule not being in sync with breaking changes.

Updates from the team

@bresch implemented a script that helps to tune the drag coefficients which are dependent on the vehicle setup and are used by the estimator to calculate a wind estimate: https://github.com/Auterion/Flight_Control_Prototyping_Scripts/pull/17
Compared to before where the coefficients had to be tuned manually mostly based on a replayable log and with manual trial and error the new scripts works on any log that was captured in no wind with a high speed envelope and will spit out optimized parameters.
@david_s5 has a fix for one of the release blockers where the NuttX version was not reported correctly: https://github.com/PX4/PX4-Autopilot/pull/19383/
Some of the tags in the history didn’t meet the expected format so now it will use the last suitable tag for the version report.
Thomas Debrunner: Rebased the pr for using the SIH (simulation in hardware) module in SITL (simulation in the loop).
https://github.com/PX4/PX4-Autopilot/pull/19028
The motivation is to have a more robust and hopefully less resource-intensive way for running SITL tests locally and also in CI. The naming is confusing right now because SIH is not purely in hardware anymore so we should revise the naming. The global goal is to unify the simulators and have a core implementation in PX4 allowing optional extra models and sensors driven externally. One of the reoccurring problems is to make sure there is an exclusive source is driving the time for lockstep.

Trying jerk-limited trajectory controller for the orbit angle control, still reviewing and tidying PR

@tstastny : The code style is repeatedly under discussion because we seem to have found a not completely consistent way of writing new code but a lot of style rules are not enforced and an inconsistent across the codebase. We came up with a suggestion to employ the Google style with as few exceptions as possible to account for not changing every line and our needs:
Suggested exceptions:
- Tabs instead of space to not change every line.
- Column width because we have wide monitors and don’t want to waste all the space.
- Function names in lower-case starting camel case funtionName() to be able to distinguish them easily from ClassNames. make format needs to work just like before. There should not be anything hindering further work while transitioning.
We could use clang-format and clang-tidy to enforce the style, there are also integrations with IDEs like VS code. We should transition gradually and not change everything at once.
@junwoo0914 : It’s currently not efficient for developers to navigate QGC. He’s suggesting keyboard hotkeys. There’s further discussion about how the menu could be improved such that the UX for developers but also users require less clicks.

Parameter Name limit increasing
- Update: In discussion with @bkueng it was found that this is hard because the entire parameter system has to be changed at the same time on all sides. We can still do it but MAVLink has to be used differently and the existing parameter system has to be dealt with in some backward compatible way.
Flight Mode dynamic
- Update: Some comments were added to the document. Any further readers, commenters and feedback are appreciated.

If you have any feedback or corrections please comment on this post!

Updates on each sectors of PX4 System

mwbb · March 30, 2022, 3:36pm

Thanks for organizing this meeting @junwoo0914