AP_NavEKF3: define Yaw alignment min GPS speed per vehicle #24206
Conversation
|
I wonder if we could automatically reduce the threshold based on the reported GPS accuracy, might be more trouble than its worth tho. |
| #define GPS_VEL_YAW_ALIGN_MIN_SPD 5.0F | ||
| #else | ||
| // copter, heli or undefined | ||
| #define GPS_VEL_YAW_ALIGN_MIN_SPD 2.0F |
There was a problem hiding this comment.
I'd like input from @priseborough on whether lowering this is safe for copters. GSF can be very unreliable for low speed/low accel vehicles
rmackay9
force-pushed
the
ekf3-gsf-fix
branch
from
110f7ca to
8536878
Compare
|
I've rebased this on master and this is being tested here. |
|
There is no theoretical reason why that threshold cant be lowered for copter and rover platforms. The GSF alignment is more a function of acceleration profile than it is velocity. The velocity threshold does indirectly enforce a prior speed change and therefore horizontal acceleration event which is why it was added. During the initial development of this algorithm I flew a copter with no minimum speed requirement and alignment was OK provided the GSF GPS data fusion didn't start running until alignment, but that was with a well isolated IMU and good GPs data. Has anyone else with a average HW quality copter tested this lower threshold? |
|
This is included in 4.5.2-beta1 |
This improves GSF performance for Copter and Rover 4.4 by reducing the minimum speed the vehicle must be travelling at for a GSF reset to occur.
The minimum speed requirement issue (see #24183) was introduced with PR #23515 but the minimum (5m/s) was likely designed with Planes in mind.This PR reduces the limit to 1m/s for all other vehicles (Copter, Helis, Boats, Rover, etc). While slightly arbitrary I have tested these and I could not make GSF produce an incorrect yaw even if I set the speed limit to zero.
Related discussion is here
Alternatively we could add a parameter although i wonder if anyone would ever set it.
This has been tested in SITL and on real vehicles for Copter, Rover and Boat and GSF always calculated the correct Yaw even when the limit was reduced to zero The user who reported the problem (and discovered the fix) has successfully tested 0.5m/s for Rover.
Below is a screen shot of a couple of tests on Copter and Rover. Some debug was added to show the exact GPS speed at the moment GSF was calculated. In these particular tests I was specifically trying to trigger GSF at low speeds to see if it would calculate the wrong heading (it didn't).
