Attendees (unique): 7
UTC0802
Peter: AI reports a lot of bugs with the hwdef. It would be nice if we could have some Python automation to catch these earlier, instead of the end of the week.
Andrew: nARMED isn’t really something people care about. In total we could approve the PR with those small bugs existing.
P: No, let’s ask the author to have another look and address them.
UTC0811
Amilcar: I’m looking at apdef and notch filters. My Claude ran out, can I ask for an extended plan?
A: We have open ChatGPT seats. Claude, we don’t have an open seat right now. But you can certainly open a request.
Am: I can certainly start with ChatGPT.
UTC0818
A: This was necessary to tell SIM to enable momentum drag in tailsitters.
- It’s also very hard to load SITL frame parameters via JSON in Mission Planner.
- This enables changing airframe parameters in-flight as well.
Approved!
UTC0824
Randy: I think this one is correct.
Peter: It would be nice if we could do away with the rc methods do construct the angular targets.
R: I’ll keep working on it.
UTC0834
G: Reinstating support for my last_letter simulator.
- You are supposed to compile the binary and have it in your path.
- You can modify the models installed in
~/last_letter_models - Then ArduPilot will find it and launch it.
A: perror doesn’t exit, just prints a message. If you want to exit, use exit.
- Also a SITL reboot will bug out.
- Let’s create a new .yml test and a new container.
P: Maybe let’s create a separate Docker, containing the required binaries.
UTC0847
P: Let’s ask Jake if he wants it in now.
A: Happy for it to go in.
UTC0852
P: We were adding in timestamps in the Microstrain driver to interpolate in case of very slow position upate rates.
Merged!
UTC0854
Merged!
UTC0854
A: I thought we agreed there would be a parameter, and the crux of the point is the default value?
P: I don’t think we can resolve this now.
UTC0859
A: Approved!
UTC0901
P: I think FAILED is the correct return result, compared to DENIED.
R: I’ll look into it in the next few days.
UTC0908
P: In the past we could not afford the buffer to write to two files.
A: LOG_DISARMED=3 and rebooting prior to arming would help here.
Rishabh: This workflow doesn’t always help.
A: Writing two logs at once, at different pages, might cause problems and severe slowdowns.
- Also a lot of messages don’t re-emit unless they change, so rotating a log won’t result in something usable.
Ri: I want this for a production system, where you want to do a lot of stock flights, while monitoring a new feature.
A: You want to tell the DAL to re-initialize so that the rotated log gets all the initialization messages.
- It would be nice if we can have replay-only-when-armed. We’d need to dump the EKF state in the log upon arming.
Ri: The EKF things to log would be just too many.
A: They can fit in a large struct. The downside is that we’d have to update this big structure every time we update the relevant EKF code, which would be a lot more often than what we do now.
- If we save the covariance matrices data and keep the initialization values. Then we could omit the pre-arm data and continue the log upon arming.
Ri: Yes, perhaps that would work.
A: Another thing we could do would be to change the EKF sched. rate when disarmed.
G: I expect that EKF would blow up by intergrating noise over a very long period of time.
A: It’s worth trying.
Lupus: You could detect motion and increase the loop rate temporarily.
A: What if you took advantage of the EKF reset AUX feature?
UTC0929
A: I’ll leave this to Pete.
UTC0932
Merged!
A: Let’s test by faking a sinusoidal temperature signal.
UTC0943
P: I had issues with TRAINING mode.
G: I’ll add one more exception, sorry about that.
UCT0947
https://github.com/ArduPilot/mavlink/pull/2555
Merged!