Ardupilot EU Dev Call 2024-08-14

Attendees (max): 11

UTC0701:

Andrew: Existing tool is quite old.
Randy: Inspired by CUAV.
A: CUAV is suffering bitrot, nice to see it revived.
R: This iteration doesn’t have automatic search, but has better manual search.
Right now it’s offline, but I can also envision a live version.
I would prefer for it to be under tools, but I couldn’t place it in the correct menu.
A: Usually non-interactive tools are stand-alone (e.g. MAVExplorer.py)
But do whatever works for you now.
Try out your edits inside the __main__ function, for faster testing.
GhatGPT might also help.
It would be nice to have some way to autotest the tool, to prevent further bitrot.
You can drag a folder or file onto the MAVExplorer window.
Remember to use multiproc.freeze_support() to support windows systems.
There is MAVproxyWinBuild to create .exe files for Windows. But also the CI does that too.

UTC0719:

A: Amilcar is working on fleshing out the mavftp library.
Will it work for both fetching files and parametes?
Amilcar: Yes.
Tested on a Cube Black.
There is a bug: subsequent calls (sessions) seem to get mixed up.
Possibly it’s on Ardupilot side.
A: Ardupilot ought to be doing multi-session.
Am: MAVProxy also has the same bug.
A: Probably reproducing in SITL with MAVProxy would be the way to go about fixing it.
Remember to fixup the commits.
Discord thread created to discuss: “code_review/mav ftp library”

UTC0729:

A: There is a back-transition state for SLT. Automatic ones go through position_control_state
Remove active back for now, since it’s not used.
Talk to Henry about Altitude.cpp:222. It’s not clear what is the requirement/use case.

UTC0740:

G: Oscillations are gone.
Biggest change is that TAKEOFF mode does all of the climb in FlightStage::TAKEOFF
A: Could Peter test a compass-less TAKEOFF mode takeoffs? To ensure that yaw hold works properly.
Also George to test a compass-less takeoff in SITL, by launching with CMAC_South and `COMPASS_USE*=0.
It would be a lot better if the throttle limits were defined and set in only one place.
G: Emphasis on the pitch throttle constrain, for Paul to review.

UTC0815

A: Please test.
Remember to clear all flash errors before running the test.

UTC0819

A: Left a comment about code duplication.

UTC0823:

Andy: Would prefer not to place it on the custom build server.
A: It’s Randy’s choice.
I would prefer it to be in the build server, as it’s a lot of changes.
Andy: It’s many, small changes, it’s going to be a lot of guard defines.
A: user_rate_loop_gyro_samples check needs to allow correct operation even with the feature disabled.
If we enable this feature in all copters for 4.6, people are going to be trying it. And this makes me a bit nervous, for such a new feature.
Sid: I am concerned about the CPU load mostly.
Andy: The downside is that it will get very little test, if it’s on the build server.
A: This feature costs 1k of flash for plane, this is quaint.
I don’t like how this feature adds a lot of code for Plane, which can’t use it.
Andy: It was impossible for me to encapsulate the changes in a separate place.
A: Why is it needed to smooth out the IMU dt? Don’t we take it into account in the rate controller?
Andy: This feature really needs a nice sample throughput.
A: But this can negatively affect bus bandwidth for vehicles that don’t use it.
Andy: We have been doing something similar for years, in the same file. I don’t think it changes much in practice.
A: Setting the minimum scheduler period to 0 is more concerning. It can block all lower priority tasks.
S: It doesn’t even affect only low-power boards. I see H7 CPU struggling with some triple IMU IOs.
A: The sensor FIFO is currently what’s used to acquire the gyro signal. You can’t expect the FIFO to always contain the same amount of samples. And that means that you don’t get strict update sensor rates either anyway.
Andy: What’s the way forward?
A: Make it less scary for non-using users.
Andy: I’ll try to split it into more digestible, piece-meal PRs.

UTC0924:

A: A dead-reckoning of 4 that is actually a dead-reckoning fix is dangerous if it goes into the EKF with an assumed good accuracy.
Peter: A user mentioned the GCS warning and this triggered the PR.
A: Let’s try get a flight log out of this unit.

UTC0931:

A: Take the armed state into account, so that you avoid accidentally sending a full reverse command.

UTC0938:

Merged!

UTC0939:

Merged!

UTC0939:

Merged!

UTC0941:

Chiara: I can’t set the parameter from Mission Planner.
I’m working on the Septentrio simulator.
A: You can try to use the actual GNSS RX in SITL, and then use the debugger to step through the code.
Peter can help, since he’s the only one who can find their Septentrio.

2 Likes