Kalman filtering:  assume p(x) is a Gaussian
initial state
prediction
measurement
posterior
prediction
Key
s = x (position)
o = z (sensor)
Robot figures courtesy of Dieter Fox