synthetic data and notebook

[jerabaul29]

The signal processing algorithms can be really tricky to get right; I think the best / only way to make sure they work correctly would be to use some synthetic data in and compare with the output; for example generate a synthetic signal that corresponds to a sensor that:

• goes up and down following a given sinusoidal (possibly with random amplitude and frequency varying slowly in time)
• tilts in X and Y directions following a given sinusoidal (possibly with random amplitude and frequency varying slowly in time)
• has a given N(0, sigma_channel) known noise on each channel independently

this can be sampled at any frequency, filtered by the Kalman filter, and compared to the true value.

Doing so would be the "ultimate" test that all is correct.

[gauteh]

There is a python interface here: https://github.com/gauteh/ahrs-fusion/blob/main/python/tests/test_nxp.py which should make it easy to test the Kalman filter.