Skip to content

v1.4.3

Latest

Choose a tag to compare

@github-actions github-actions released this 14 Nov 16:54
· 1 commit to main since this release
fd9b2ec

[1.4.3] - 2026-11-14 - Baro Alt Improvements

Added

  • baro_alt now considers the roll/pitch uncertainty of the supplied
    attitude instead of a fixed constant uncertainty.
  • baro_alt takes a body-frame accelerometer bias vector from outside
    and subtracts it before rotating the sample.
    So baro_alt can benefit from an up-to-date accelerometer bias
    estimation from ins.c.
  • Frozen-IMU veto for the stillness detectors improved: a sensor that repeats
    itself bit for bit on all three axes is a frozen sensor or bus, not a
    standstill, so it no longer arms ZUPT/ZARU.
    Noise-free synthetic IMU streams now need a small dither to arm the
    detectors.
  • ins_init() and ahrs_init() return -1 if any float of their
    configuration is NaN or Inf, and ins_set_world_model() ignores a
    non-finite vector.

Changed

  • The 3x3 rotation helpers in ins.c are unrolled instead of going
    through the generic matmul/sgemm_.
  • The layout of baro_alt_t, nav_suite_t, ins_t and ahrs_t changes, code
    built against the 1.4.2 headers has to be recompiled.

Fixed

  • The GNSS fusion rate limit (gnss.min_delay_ms) dropped fixes of a
    receiver running exactly at the limit: a fix stamped 99.9 ms after the
    last one was skipped and the next one came 200 ms later.
  • The health check of INS did not look at the bias states. A non-finite
    accelerometer, gyro or magnetometer bias now shuts the filter down.
  • The magnetometer heading of the INS bootstrap ignored a hard-iron
    bias carried over from the previous instance and started the yaw off
    by the error that bias had already removed.
  • baro_alt_predict_step(), baro_alt_correct_step(),
    baro_alt_update(), local_gnss_alt_update(), ahrs_predict_step()
    and ahrs_correct_step() dereferenced a NULL instance. They now return
    without effect.