Instead of having a single height source fused into the EKF and the
other ones "waiting" for a failure or the primary sensor, fuse all
sources in EKF2 at the same time. To prevent the sources from fighting against each
other, the "primary" source is set as reference and the other ones are
running a bias estimator in order to make all the secondary height
sources converge to the primary one.
If the reference isn't available, another one is automatically selected
from a priority list. This secondary reference keeps its current bias
estimate but stops updating it in order to be the new reference as close
as possible to the primary one.
- replace float32[21] URT covariances with smaller dedicated position/velocity/orientation variances (the crossterms are unused, awkward, and relatively costly)
- these are easier to casually inspect and more representative of what's actually being used currently and reduces the size of vehicle_odometry_s quite a bit
- ekf2: add new helper to get roll/pitch/yaw covariances
- mavlink: receiver ODOMETRY handle more frame types for both pose (MAV_FRAME_LOCAL_NED, MAV_FRAME_LOCAL_ENU, MAV_FRAME_LOCAL_FRD, MAV_FRAME_LOCAL_FLU) and velocity (MAV_FRAME_LOCAL_NED, MAV_FRAME_LOCAL_ENU, MAV_FRAME_LOCAL_FRD, MAV_FRAME_LOCAL_FLU, MAV_FRAME_BODY_FRD)
- mavlink: delete unused ATT_POS_MOCAP stream (this is just a passthrough)
Co-authored-by: Mathieu Bresciani <brescianimathieu@gmail.com>
- refactor updateQuaternion() to compute the yaw jacobian directly (respecting the rotation sequence determination)
- fuseHeading()/fuseYaw321()/fuseYaw312() helpers are eliminated and now mag heading fusion and EV yaw fusion compute the innovation in place
- clear up logic for performing zero innovation heading fusion when quaternion variance exceeds threshold (no more _is_yaw_fusion_inhibited flag manipulation)
- when at rest continue fusing last static heading with very low variance even if other heading sources are active
split the fusion process into:
1. updateAirspeed: computes innov, innov_var, obs_var, ...
2. fuseAirspeed: uses data computed in 1. to generate K, H and fuse the
observation
- all sources of optical flow publish sensor_optical_flow
- sensor_optical_flow is aggregated by the sensors module, aligned with integrated gyro, and published as vehicle_optical_flow
Co-authored-by: alexklimaj <alex@arkelectron.com>
Calculating K(HP) takes less operations than (KH)P because K and H are
vectors.
Without considering the sparsity optimization:
- KH (24*24 operations) is then a 24x24 matrix an it takes
24^3 operations to multiply it with P. Total: 14400 op
- HP (24*(24+24-1) operations) is a row vector
and it takes 24 operations to left-multiply it by K. Total:1152 op
- the estimated mag bias was requiring > 30 seconds of continuous 3d
mag fusion to be reported stable (and saved back to mag cal), this
restores the original intent requiring 30 seconds of accumulated valid
3d mag fusion
- Also use the delayed current data instead of newest that might not be
available (gps buffer is sometimes empty if the dt between samples is
larger than the delayed horizon).
- Separate "baro fault" from "baro intermittent": intermittent is a
temporary failure and should prevent from switching to baro right now,
but "fault" means that it should never be used anymore
- In case of height timeout, check for metrics but not for consistency
as the filter is likely to have diverged already.
- remove class _mag_sample_delayed
- update mag fusion call graph to use mag sample delayed functionally
- Ekf::resetMagHeading()
- use low pass mag directly, but check if valid (magnitude)
- MAG_FUSE_TYPE_INDOOR treat like auto if heading required
- introduces new parameter EKF2_PREDICT_US to configure the filter
update period in microseconds
- actual filter update period will be an integer multiple of IMU
bresch