The robotics project I'm working on suffers from some very noisy sensors. A simple 1st order low pass filter doesn't do the trick. The noise has a low frequency cyclic component as well as standard white noise, plus occasional electrical spikes. It's not great.
We tried a number of things including 2nd order filters, bandpass filtering (to try and remove the low frequncy cycle), fuzzy sets and more.
I was hoping to avoid a Kalman filter style approch (since they are painfull to formulate, and they are only meant to handle gausian noise as far as I know). H-infinity filters are another option, (which can handle non-gausian noise and don't require a complete system model) but they are still too complex for my liking.
So we settled on trying an Alpha-Beta filter since that should give a vauge idea of how the Kalman might act, and its looking promising.
I'd never heard of these before, so thank you wikipedia! I'll have to spread the good word.