Commit Graph
453 Commits
Author SHA1 Message Date
bresch dae246d7e2 ekf2: do not continuously use mag decl fusion when GNSS fusion is active
This prevents over-constraining the heading from mag fusion. An
incorrect mag yaw rotation can be absorbed as a declination error.
2024-04-05 22:26:03 -04:00
bresch 2f51db7284 ekf2: limit mag heading fusion to prevent heading overconfidence 2024-04-05 22:26:03 -04:00
bresch 51883fe5d4 ekf2: integrate mag heading into mag 3D 2024-04-05 22:26:03 -04:00
bresch 0c9e4013d2 ekf2: improve tilt leveling speed
Starting with no yaw uncertainty makes the tilt more observable when
using fake position fusion during the quasi-stationary alignment phase.
2024-04-05 22:26:03 -04:00
bresch 12fefbcfbd ekf2: use global definition of quaternion error 2024-04-05 22:26:03 -04:00
murata,katsutoshi 749f88b62b ekf2: gps control lazily check yaw_failure() only after in_air 2024-03-26 19:50:57 -04:00
bresch 6d819343aa ekf2: fix direct state measurement update for suboptimal K case
The duration of a unit test had to be increased because the incorrect
covariance matrix update, was making the unit test passing faster
(over-optimistic variance).
2024-03-20 14:13:49 +01:00
muramura 32aa3263a6 EKF: Change a typo 2024-03-18 10:29:44 +01:00
bresch 7c6ecd95a8 mc_wind_estimator_tuning: optionally use GNSS velocity
Sometimes GNSS is logged but not used
2024-03-14 21:06:24 +01:00
bresch b2f1122372 ekf2: remove old yaw 321 and 312 derivations 2024-03-14 11:33:45 -04:00
bresch e3f67d5c1a ekf2: new yaw derivation
Instead of euler angles, compute measurement jacobian using a small
global perturbation around the vertical axis
2024-03-14 11:33:45 -04:00
bresch 0639f5370c ekf2: fix mag and wind covariance prediction 2024-03-07 11:06:31 -05:00
bresch 421f13e4b5 ekf2: fix joseph covariance update for Schmidt-Kalman filter
If part of the Kalman gain is zeroed, the first step of the joseph
update does not produce a symmetrical matrix.
2024-03-07 15:11:47 +01:00
bresch 6f9a378247 yaw_est: force set gyro bias when at rest
The gyro bias estimate from EFK2 is really good when at rest and should
be used by the yaw estimator to prevent heading drifts due to poor
heading observability.
2024-03-05 14:00:06 -05:00
bresch b80f15f7b5 ekf2-mag_auto: always use mag 3D after takeoff 2024-02-27 12:33:43 -05:00
bresch 051baec9c4 ekf2: allow wind dead-reckoning after manual position reset
Reset velocity using airspeed and start navigating
2024-02-27 09:34:05 -05:00
Daniel Agar 8001132d33 ekf2: ZeroGyroUpdate move to fuseDirectStateMeasurement 2024-02-26 12:32:59 -05:00
bresch d501d8e1d4 ekf2: use Joseph stabilized update in direct state observations 2024-02-26 12:32:59 -05:00
bresch 9d9766c6cf ekf2: use Joseph stabilized covariance update 2024-02-26 12:32:59 -05:00
bresch 37a40d3fc2 baro static pressure compensation tuning: remove dependency to baro bias
`estimator_baro_bias` requires to have GNSS and baro hgt active and GNSS as the reference. This is quite restrictive. Instead, we can simply use a high-passed version of the baro error.
2024-02-23 10:07:13 +01:00
Daniel Agar 506c60c471 ekf2: declination fusion don't use uninitialized parameter (EKF2_MAG_DECL) 2024-02-21 09:45:44 -05:00
Daniel Agar 8243b4f474 ekf2: move vel/pos reset helpers 2024-02-20 13:16:24 -05:00
Daniel Agar 22b957696d ekf2: velocity/position fusion helper minor consistency cleanup 2024-02-20 13:16:24 -05:00
Daniel Agar c338891677 ekf2: split vel_pos_fusion.cpp 2024-02-20 13:16:24 -05:00
Daniel Agar c4c41c49e5 ekf2: move fuseVelPosHeight() -> fuseDirectStateMeasurement()
- don't bother keeping bad_vel_{N,E,D} and bad_pos_{N,E,D} fault status bits
2024-02-20 13:16:24 -05:00
Daniel Agar 021dd0d0af ekf2: fix EV height bias predict call
- needs to be called every iteration
2024-02-20 11:47:53 -05:00
bresch c221da27a7 ekf2: set attitude validity flag using centralized function 2024-02-20 11:33:30 -05:00
Daniel Agar 1032dd3470 ekf: fix measurementUpdate comment typo 2024-02-19 09:41:49 +01:00
Daniel Agar 8dc3975456 ekf2: only populate gnss pos aid src status if ref initialized
- this is a minor logging improvement when plotting the position from the beginning of the log (often a replay session)
2024-02-15 13:13:10 -05:00
muramura f636414ca7 tuning_tools: Change 1G to a more accurate value 2024-02-09 10:26:57 -05:00
bresch 17d55dddd6 ekf2-drag: do not generate Kalman gain to save flash 2024-02-06 12:16:33 +01:00
bresch 1efb08375a ekf2: let drag fusion affect the complete state vector
This improves tilt estimation and can extend the inertial dead-reckoning
validity period
2024-02-06 12:16:33 +01:00
bresch 9c02e384e6 ekf2-agp: follow measurement reset 2024-01-30 11:23:55 -05:00
bresch 5d9081b0dd ekf2-agp: ensure logging of AGP aid_src topic 2024-01-30 11:23:55 -05:00
bresch 4268759d4a ekf2-agp: reset to measurement on fusion timeout 2024-01-30 11:23:55 -05:00
bresch 80f20e619c ekf2: zvup sequential fusion 2024-01-29 12:12:37 -05:00
Daniel Agarandbresch 51155f7a29 ekf2: treat inhibited states as "consider states" and remove conservative covariance matrix fixes (#22597)
When a state stops to be estimated it becomes a "consider state". Its value and variance are frozen but its covariance with other states continue to evolve normally.

 - removes conservative accel bias variance limiting
 - force symmetry is skipped after fusion of NED vel/pos (a direct measurement)

---------

Co-authored-by: bresch <brescianimathieu@gmail.com>
2024-01-24 14:14:09 -05:00
Mathieu Brescianiandbresch cb396a6339 ekf2: fix computation of tilt and yaw variances add them to logging
Co-authored-by: bresch <bresch@users.noreply.github.com>
2024-01-24 12:14:15 -05:00
Daniel Agar bf7da6430d ekf2: consolidate LPOS & GPOS accuracy methods 2024-01-22 12:34:08 -05:00
bresch 7c7a3c117a ekf2-gravity: nomalize gravity fusion and proper sequential fusion 2024-01-18 20:39:16 -05:00
bresch d624fbba07 ekf2-grav: lower gate to reject real acceleration more effectively 2024-01-18 20:39:16 -05:00
bresch c28972d15e ekf2-grav: only use filtered accel norm to start/stop the fusion
Using the raw data makes the swith too sensitive to noise
2024-01-18 20:39:16 -05:00
Daniel Agarandbresch ed0d26de8a ekf2: improve attitude estimation without horizontal aiding
- fake_pos only if at rest or tilt variances becomes large
 - fake pos: don't run when grav fusion is enabled
 - gravity fusion enabled by default
 - gravity: only fuse when accel norm and lpf norm are consistent

Co-authored-by: bresch <brescianimathieu@gmail.com>
2024-01-17 10:27:25 -05:00
Daniel Agar 64f28c4c07 ekf2: delete unused gps error norm field 2024-01-12 10:14:32 -05:00
Daniel Agar 8a031677d5 ekf2: verbose debug print status include state variances 2024-01-11 11:53:20 -05:00
Daniel Agar 8bcba6128a ekf2: fix resetGlobalPosToExternalObservation whitespace 2024-01-11 10:59:56 -05:00
Roman Bapst 603c3f6636 added support to reset vehicle position based on external position (#22444)
via command MAV_CMD_EXTERNAL_POSITION_ESTIMATE

Signed-off-by: RomanBapst <bapstroman@gmail.com>
2024-01-11 13:09:22 +01:00
Daniel Agarandbresch d45c3d3407 ekf2: verbose (DEBUG_BUILD) print status with Matrix improved print
* ekf2: verbose print status
 * matrix/Matrix improve print output
 * bold diagonal elements, print ring buffer entry size
 * print in scientific notation when >= 10 to respect max size

---------

Co-authored-by: bresch <brescianimathieu@gmail.com>
2024-01-10 09:46:04 -05:00
bresch 36eb319834 ekf2-yaw_est: split imu and velocity updates 2023-12-18 18:11:20 +01:00
bresch a653073d4f ekf2: perform GNSS checks at delayed-time horizon
- never fuse a measurement that is not passing the checks
- cleanup and simplify GNSS vel/pos control logic
2023-12-18 18:11:20 +01:00