@Barbara Anderson yeah you've already identified the exact problem — COG heading is basically useless below about 0.5-1 m/s because the positional noise starts to dominate over the actual displacement vector. At 0.3 m/s you're moving maybe 1.5-2.5cm between fixes even at 18Hz, and the 1.5m accuracy spec on the MultiNav means that positional jitter is orders of magnitude larger than your actual movement per sample. You'll get heading estimates that are all over the place.
The good news is the RFOXiA Sensors Module has a magnetometer and a full IMU (accel + gyro) built in, so the natural fix here is sensor fusion — use the GNSS for position and velocity reference, and the magnetometer/gyro for heading at low speeds. The 18Hz GNSS rate is actually really helpful for this because it gives you a tighter velocity estimate when you're above the noise floor, which you can use to periodically correct gyro drift. Below that threshold you lean on the mag and integrate the gyro.
If you're building a proper control loop, a complementary filter or a basic EKF fusing those two data sources will handle the speed transition cleanly. The hard part is magnetometer calibration in the field — if there's any motor or battery interference near the sensor you'll need to account for that.
