search

ArduPilot / ardupilot

ArduPlane, ArduCopter, ArduRover, ArduSub source
http://ardupilot.org/
GNU General Public License v3.0
10.74k stars 17.19k forks source link

Unexplained spikes in EKF Position innovations #26271

Open EosBandi opened 7 months ago

EosBandi commented 7 months ago

Copter 4.4.4, OrangeCubePlus, Dual Here4 gps. Logs show spikes in EKF Position innovations, some are close to the error threshold. IMU0/Lane1 and IMU1/Lane2 shows frequent spikes, but IMU2/Lane3 show only a small amount, occasional and smaller spikes. No sensor affinity is set, all lanes are using the same blended GPS and first compass. Compass checked, calibrated and in sync with predicted magfield values. Magnitude of the IPN and IPE spikes are related to the actual groundspeed. No spikes on any other innovation values. Log files : https://1drv.ms/f/s!AmcHEcCG2Jqso6pDjcnW3KHbxATXgQ?e=LPDX6s

Long shoot: Here4 NEO F9P lag vs ZED F9P lag? " // F9 lag not verified yet from flight log, but likely to be at least // as good as M8 lag_sec = 0.12f; "

amilcarlucas commented 7 months ago

Nice and detailed bug report

tridge commented 6 months ago

no replay on log unfortunately the spikes are clearer in the XKF3.IPx data (which have units of meters) image

tridge commented 6 months ago

@EosBandi are you able to get a log with LOG_REPLAY=1 and LOG_DISARMED=3 ?

tridge commented 6 months ago

I suspect the issue may be some high frequency noise in the IMU data that doesn't show up in our filtered IMU logging, but would show up with replay (EKF uses unfiltered IMU data) it could be timing in the GPS data too, though I can't see it in this log

EosBandi commented 6 months ago

@EosBandi are you able to get a log with LOG_REPLAY=1 and LOG_DISARMED=3 ?

Sure, but it will take some time; the vehicles are in Bogota, the pilot is in Spain, and I'm in Hungary. I'll post the logs as soon as I manage to get the pilot and the vehicles to one place :)

bugobliterator commented 6 months ago

It seems like an issue specific to this vehicle. I did flight with Here4 and these spikes are not present. The flight was not with blended GPS though and with moving baseline enabled.

echarts (1)

Full Log: https://droneshare.cubepilot.com/#/s/log?b=1707286594399&c=71651630-c580-11ee-8113-bfa0943c0b8f&d=GPS+Yaw+Acceptance+Gate+Test2&e=2&a=cbfcd2fc-3fa9-11ed-b631-b97a8244eade

EosBandi commented 6 months ago

It seems like an issue specific to this vehicle. I did flight with Here4 and these spikes are not present. The flight was not with blended GPS though and with moving baseline enabled.

To me it seems to have the same random spikes, just a little smaller magnitudes.... image

Am I just overreacting to something normal with EKF3 ?