Learning how to drive.

This article in Bayesian integration in sensorimotor learning tells us that it looks like the central nervous system's idea of learning is refining a probability. So it is all the more interesting that FASTSLAM uses a similar approach for rovers in unknown environment.