Using the oscillation observed under proportional control, a simple ode based model can be tested.
It can be seen that rogallo yaw behaviour is quite complex, however, there may have been significant wind disturbance during the test. The second fit (yellow) uses a yaw_rate^2 damping term, wheras the first (purple) uses an aerodynamic damping term proportional to yaw_rate . It can be seen that yaw_rate^2 gives better results, so this will be used in the kalman filtering. Interestingly, the servo gain factor, which describes how a given servo displacement effects yaw acceleration (simple linear approx) has to be significantly larger (2.4) in the second fit than in the first. The nonlinear damping significantly decreases the frequency of oscillation. The simcode is very simple, based on the code on the MiHAB glider page
Work is now underway on fine tuning the Rogallo control loop.
Heading offset data recording at 5 samples per second:
A Fourier transform, the peak is at 0.17857 Hz or 5.6 seconds period
These graphs show heading recorded at 1 second intervals, some nice oscillation is seen, which will allow the control response to be studied. The first graph shows a flight when the wind had dropped, so it would appear that the wind is driving the oscillation in the second graph.