PX4 Dev Call: October 14, 2020

October 14, 2020

Agenda

  • Status update by component/project
  • Roadmap, and Release discussion
  • Community Q&A
  • In-Depth discussions

Join Meeting

Meeting ID: 946 175 205
Passcode: 108516

Call in using your phone, find your local number: https://zoom.us/u/aetSYKbiMF


Component update

System Architecture

The document shows that people who voted prefer the two repository names:
“PX4-Autopilot” or “PX4Autopilot”
People that have an opinion on this please prepare them for next week, we’ll cut a decision in the next dev call.

We’ve been too tolerant of flaky CI. The linked issue is probably MAVSDK crashing. But there’s multiple potential issues e.g. with timing and or lock-step. Do not tolerate CI failures!! The majority are real failures.

Goal: have a consistent [-1,1] range for all sticks. MAVLink spec MANUAL_CONTROL also has [-1000,1000] range for all sticks but QGC sends [0,1000] to work with PX4. @dagar is checking the QGC scaling and we’ll need checks to ensure the safe transition.

OS / NuttX

Either we have a variable for the absolute path that is empty on NuttX or we have an invocation command other than “source” that does the job. @david_s5 will have a look and reach out for questions and testing.

All other known issues with the new NuttX version are sorted out. Resources are tight on some older boards. @david_s5 is trying to get our required changes to upstream NuttX such that we can sty in sync with upstream.

Driver

Fix for DPS310 barometer driver on mRo boards

Commander

No updates

Estimation

This might be interfering with the current Multi-EKF efforts. We’ll discuss it in the in-depth section later.

VTOL

No updates

Fixed Wing

@Roman giving a heads up about a revision of TECS that has better performance in holding altitude during turns.

Multicopter

No updates

Avoidance

No updates

Simulation

@Roman Does JSBsim work with lockstep? Can we specify the speed factor?
@Jaeyoung-Lim Lockstep yes, speed factor is not exposed at the moment but can be added.

Shoutout: We have a direct simulation with JSBsim now and it simulates fixed wing physics much better than gazebo. There’s a lot of activity currently on that topic.

PX4 failure injection was extended to be able to disable specific e.g. gyroscopes. We’ll try to also leverage the failure injection on the JSBsim side in the future.

MAVSDK

MAVROS / DDS / ROS2

No technical updates

We are participating at the ROS World with 6x 50 minute sessions. There will be a lot of people > 2000 registrations already.

UAVCAN

Call was yesterday, you can participate looking at the calendar: https://www.dronecode.org/calendar/

Notes from call yesterday:


Reached a major milestone: Official drone namespace in the UAVCAN message specifications
Example: Battery/BMS, ESC, Servo communication


Libcanard v1 is released and will be supported by PX4

Release

Minor update 1.11.1 available with accumulated bugfixes.

Community Q&A

@camposi Emulate an Oblique filter for moving the gimbal to reduce necessary survey rows by extending the field of view taking pictures in slightly changes gimbal orientations. The idea is to have a single high-level command to make the camera cover the bigger field. This reduces the number of necessary waypoints since the constantly changing camera orientations don’t require extra in between points.

@Jaeyoung-Lim Don’t we lose straight down picture information by rolling it left and right? The frequency of pictures is increased e.g. by a factor of 3 to be able to take pictures left, middle and right.

@bresch Does the geotagging capture the camera attitude correctly? It should work since it uses the gimbal status but most mapping tools don’t take use of the tagged camera attitude.

We need to get the MAVLink spec adjustment merged first and then we can get PX4 changes and the QGC side which is also already ready in.

In-Depth discussions

  • Estimator stored mag bias variances saving working in Multi-EKF scenario

Conclusion: Move forward with the pr for one mag bias saving and skip this saving in the initial Multi-EKF implementation until we have the right selection in place.


Errata and Feedback

Let me know below if we failed to capture anything the right way, and if there are any updates (not present here), or you have feedback you would like to share with the dev-team.

I would appreciate if we had the time to discuss the following PR: