Add a bitmask parameter to control bias learning for individual axes. This is achieved by setting the disabled states to zero together with their corresponding covariances.
Minor cleanup of the covariance prediction comments.
Removal of unnecessary variable copy operations.
Replace index operations to initialise covariance to zero with the more efficient memset.
* mc_pos_control: ignore a NaN/inf position setpoint
This is a hotfix that prevents the position controller from trying to
do velocity control with a position setpoint that isn't valid in the
first place.
This is only a workaround, ideally the controls later should not scale
down throttle to the minimum just because the position setpoint is far
away if they still have valid setpoint in z.
* mc_pos_control: use PX4_ISFINITE and not isfinite
- Use the course over ground as loiter exit criteria (better wind behavior)
- Compute the tangent origin coordinate explicitly instead of using immediate position
IO needs atomic updates of a few critical status flags, but doing these always atomic (with interrupts disabled) might cause a too high interrupts disabled interval. In order to avoid this only operations that change the state of the target variable are done with interrupts disabled, while operations without an effect on the target variable are not executed.
This enables to use the RAM normally consumed by the log buffer to be used for calibration and other memory-intense tasks.
These run typically only disarmed when logging is not enabled.
Initialisation changes to address valgrind errors
Change to default GPS and Airspeed time delay (these are overwritten by ekf2_main parameter settings)
Increase sensitivity of bad accelerometer checks
Squashed commits:
[290660d] voted_sensors_update: revert the new_accel_data check (and others)
The check removed the ability to detect sensor timeouts.
[c8dc7ad] sensors: publish changes to sensor selection
[dd90dec] sensors: ensure all sensor selections published first time
Mohammed Kabir