Filter ekf variances for failsafe and vibration checks #25621
Conversation
|
One important fix but otherwise this looks pretty good to me. We should test the impact of the filtering on the EKF failsafe and vibration failsafe. Knowing how much further or higher the vehicle flies by the slight delay in triggering of the failsafe would be good to know. |
andyp1per
force-pushed
the
pr-ekf-failsafe
branch
from
8de345f to
aff706a
Compare
|
I did a test flight of this over the weekend, but my GPS was performing quite poorly for some reason, so it actually EKF failsafed during the flight, but I don't think that's to do with this PR, but I don't think that flight was really a valid test of this PR. I do think that it would be better if this type of thing was in vehicle or the EKF rather than Copter, though, since EKF changes really should be applied to all vehicles. This won't affect the frequent lane switching though, will it? |
|
waiting for test results! |
andyp1per
force-pushed
the
pr-ekf-failsafe
branch
from
aff706a to
1f20d81
Compare
| vel_var = vel_variance_filt.apply(vel_var, dt); | ||
| height_var = hgt_variance_filt.apply(height_var, dt); | ||
|
|
||
| last_ekf_check_us = now_us; |
There was a problem hiding this comment.
nitpick: could move this up below the calculation of dt just to keep it close to related calculations.
rmackay9
dismissed
tridge’s stale review
Tridge's requests have been addressed.
|
I've tested the vibration failsafe with this PR and it seems to act the same as now. I also think that if FS_EKF_FILT is set to 0 then it should act exactly like master right? so we have that easy backout if required. Basically I'm happy with this going in. Perhaps we should consider doing the same for Rover but that's not a blocker for this PR. |
Yes, right |
add a parameter to control EKF failsafe filtering
andyp1per
force-pushed
the
pr-ekf-failsafe
branch
from
1f20d81 to
bcbcc89
Compare
|
Ideally this would go into 4.5.1 - not sure what you think @rmackay9 |
|
@andyp1per, sure, we're still pretty early in the beta.. |
From a discussion with @priseborough