Multiple model filters for systems with possibly erroneous measurements

In this paper a new method to deal with possibly erroneous measurements is presented. In practical tracking systems it is common use to either use a certain measurement or to not use it, i.e. make an instantaneous hard decision w.r.t. the correctness of the measurement. We propose a soft decision making strategy and show by means of a realistic example that this performs better than a scheme in which a hard decision is made. The soft decision strategy that is proposed can be used in any multiple model filter. We have incorporated it in the so called MMMH filter (Boers ansd Driessen 2001). There is however no limitation to incorporate it in the same way into, for example, the IMM filter (Mazor et al. 1998; Blom 1990). A long range search radar tracking system is used as an example.