A multiple model filter using different process noise levels

This paper derives an optimal multiple model (MM) tracking filter using classical optimization theory. Two models are used: a constant velocity (CV) model with low state process noise, and a CV model but with large state process noise. One novel feature of this filter is that it does not require the a priori knowledge of the target transition probability matrix. Simulations are performed to show the online switching capability of the new filter as well as its performance