PX4 Maintainers Call: March 28, 2023

:calendar: March 28, 2023

The maintainers meeting is a meeting for the developer team to coordinate on pressing issues and to plan the development of the PX4 Autopilot project, the community is welcome to join and listen, but won’t be able to speak unless specific access is granted ahead of time.

:selfie: Meeting Link

:notebook_with_decorative_cover: Agenda

  • Release 1.14.0 beta
  • Testing for 1.14.0 beta release

:memo: Meeting Notes

v1.14.0 beta release

Development Priorities:

  1. Random 100% thrust - alex & Matthias
  2. Preflight failure with Optical Flow on ground - alex & daniel
  3. Failsafe flight task triggered after startup - Matthias
  4. MicroDDS Client Agnet IP via parameter - Beniamino (Tested & Only a review is needed)
  5. ADS1115 battery sensor issue user report. What should we do?

Documentation Priorities:

  1. Docs sync - Hamish
  2. MicroRTPS cleanup - Beniamino

Questions

  1. Heightrate setpoint for FW offboard: @jaeyounglim, is this a release blocker? If not, could we push this to the next minor release since it’s WIP?
  2. Multi EKF Sensor priority parameter: @Alex Klimaj is this a release blocker?
  3. As I pointed out few messages up, the v1.14.0-beta2 tag was already released, but could we hold tagging until we get the outstanding issues resolved & then release the official beta? (It will have to be beta3, or 2.0.0-beta1): Release v1.14.0-beta2 · PX4/PX4-Autopilot · GitHub

TODOs

  1. (from last call) DDS testing from users not so familiar with ROS2 & asking for actual use cases. Identify use cases of the DDS: Organize another call to discuss this thoroughly (controller interface, offboard control kinda use cases, etc) > Thursday 5pm CET. (Make DDS convention consistent)
  2. (from last call) Docs sync meeting → And support tiers.

Discussions

  1. Offboard control for fixed wing is completely broken now - Jay
    • Trajectory setpoint is not working
    • Idea: Test this in the CI? - Daniel
  2. TECS refactor changed landing / etc FW behavior - Jay

Release plan for the 1.14.0 beta

  • We will be releasing beta further
  • Once all the non-trivial issues in project board is resolved, we can call it the ‘release candidate’
  • Keep on releasing betas (flashable via QGC), and constantly remind in Discord
  • Once we have in Release Candidate, maybe do social media post (less noise) & encourage even more testing.

Critical bug fixes we need to call it a release candidate

Testing for 1.14.0 beta :test_tube:

Add the items as the ‘testing matrix’ tryout from the components here: PX4 Release procedure by junwoo091400 · Pull Request #2331 · PX4/PX4-user_guide · GitHub

QGC coordination

  • Release coming up for QGC as well, but the support for PX4 v1.14 is not relevant now.
  • Parameter description/tree structure could be improved btw.
  • Note: XRCE DDS parameters are at the end of the parameters group in QGC

v1.14 or 2.0? :pirate_flag:

I know we are bringing this up & down again, but should we release* v1.14.0, or v2.0.0*?

We did discuss that we should get v1.14.0 and then on releases afterwards stick to a Semantic Versioning, but since the need to bump major version is constantly emerging, I think we should discuss about this

  • Pro 2.0.0: Control allocation is a big change for a user. People get confused a lot, and having major version bump can indicate a big change.

Breaking change: If we had a mixer file, it wouldn’t work anymore.
Do we want to call it 2.0? It’s a big shift for the perspective of users :thinking:

We should probably keep it low-profile, and not say it’s a MASSIVE change (may give false interpretation) to the users - Daniel

If we do 2.0: We probably should also explain semantic versioning change to the community, as that is the actual change that is coming.

E.g. if QGC and PX4 are also consistently progressing, we should consider that too.

Action Item: Ramon will discuss with marketing team to figure out what effort is needed for the 2.0.0.

ROSCON

  • Since the dev summit 2023 is next to ROSCON, we could flush out the ROS support better and make PX4 more ROS native, and have synergy :thinking:
  • E.g. ‘barallel roll’ flight mode implementation in ROS

100% Thrust

  • Trigger case: Take off, land then hold throttle low, then the hover thrust estimate goes super low & acceleration setpoint saturates (velocity integral goes high)
  • Estimate of the hover thrust is valid, although it shouldn’t be valid

Side discussion: Land detector is also faulty. It works somehow to multirotor, but this hover thrust estimator problem is also related to land detector. So long term maybe we should figure out where the land detector is screwing up.