Ardupilot EU Dev Call 2024-08-07

Attendees: ~14ppl

Merged.

Merged.

Randy: No GCSs rely on the bug.
Peter: We should be using the method to get the value, not multiply manually.
Andrew: Let’s use the AMSL accessor in the future.
Merged for now.

A: Would require testing.
Requires a switch to operate the engine.
Seems like a reasonable approach.
Missed an ICE_START_CHAN instance in a parameter file.
The hidden parameter technique is a valid approach.
Marked as needs testing, SITL will have to do, unfortunately.
Would be nice to prevent arming when there are old, non-existent parameters in parameters default file.

A: Needs flight testing.
Better also use the Filter Review Tool.

Andy: Tested on this new vesrion.
A: Marked for backport.
Merged.

Andy: The changes are included in #27335. Let’s merge that instead.
It’s a wonder how this was working before.
A: Closed.

A: Would be nice to tell between forward and rear transitions.
Apply the minimum it only in auto throttle modes.
Use does_auto_throttle to differentiate.
Also convert thrust_type into a getter.

A: Merged.
Peter: The failing test is known flapping.

Peter: It now catches invalid numerical values.
A: We’re better off rejecting invalid values in the future.
Randy: Would be nice to have a pre-takeoff mission checker.
Approved.
A: Merged.

A: Would be nice to have definite values in the enum members, to ensure that logs will be consistent.
Merged.

A: Leonard mentions a bug.
R: But a fix has been merged already. Leonard probably approves.
A: save_tuning_gains() is redundant. The current gains will be saved on landing anyway.
P: Maybe it’s in preparation of the QuickTune integration.
A: Soft disarming needs to happen before saving the gains. Otherwise only the first 30 will be saved.
Andy, R: Could be a separate PR.
A: Rescheduling for Tuesday.

R: Approved.
A: Merged.

Andy: It has been tested.
A: Approved and Merged.

P: Still in testing.
A: Approved on CI pass.

A: Approved and waiting for testing.

P: Highlights how https://githubcom/ArduPilot/ardupilot/pull/27745 leads into a paradox.
A: Will defer for next week, perhaps it’s too hard to solve.

A: Approved and merged.

Andy: A partner had an issue with the recommended 1/4.
A: Okay, let’s give up on this prearm check.
Would be nice to know how much gain/phase/noise margin we need in each control dimension.
Leonard: SYSID gives the system response. It doesn’t characterize noise effects.
And then SYDID would characterize hover only, while the issues will appear at the edges of the flight envelope.
A: Flight testing with an increasing parameter (gain/attenuation/bandwidth) would be a fine approach for estimating those margins, no?
L: Gain and phase margins are easier to measure, but you might still have no oscillation past the crossover, until something introduces the oscillation.
A: Would be nice to have some way to warn users about blatantly wrong configurations. But perhaps it’s impossible to find one rule for all cases.
A,L: (Discussion on the potential existence of a valid/reasonable and at the same time useful checking threshold.)
A: Merged.

A: Ideally we would want these parameters to appear only for the Septentrio backend.
(Bonus coding session included in the dev call.)
Chiara to test Andrew’s commit.