Same as before, filter appears to be working perfectly, need to do a car test still, but the rate output has still captured some of the turns.
This time data was recorded at 1Hz. Unfortunately it is clear holding the payload enclosure with one hand whilst running across a muddy field leaves the gyro output buried in noise. However the kalman filter seems to be coping well. Test three will have to use a bike.
Same data as previous, longer test run and intervals at 1Hz this time.
Around the 75 seciond mark it can be seen that the heading is around due south. The gps heading is wandering about the +-PI boundary, and it can be seen that the filtered heading is not behaving as it should, and is trying to follow the gps. A change to the code in kalman.c is probably the best solution to this - the top of the residual vector needs to be maintained in the +-PI range so the filter is stable under these conditions.