Kalman Filter Window Depth Settings

Filter depth has very little effect on the Kalman filter. Therefore for most cases, you can set the filter depth fairly low.

The two analyses below show a simulated x-axis movement with the raw unfiltered data compared to the exponential moving average filter and the kalman filter. Noise is simulated and includes an accelerometer disturbance with a subsequent increase in noise.

*note that each line in the chart is offset for clarity

Analysis 1 (aretas.locationfilter.minwindowsize=20):

The green line is the Kalman filtered x-axis data. You can see the green line is fairly smooth compared to the unfiltered red line. The Exponential Moving Average filter also performs fairly well.

Analysis 2 (aretas.locationfilter.minwindowsize=1):

The green line is again the Kalman filtered data and is very smooth compared to the raw unfiltered data. The exponential moving average filter does not perform as well in this configuration.

Conclusion: Results with lower min filter window settings when using Kalman filtering are indistinguishable from higher settings, however lower settings have much improved latency.