Are more models better?: the effect of the model transition matrix on the IMM filter

The purpose of this investigation was to determine what approach is best to deal with the question of how to choose the model transition matrix for the interactive multiple model filter. The desire is to determine an approach to choosing the model transition matrix that is "best" in some sense. In the study, many target tracking simulations were run using the interacting multiple models algorithm (IMM) with two models: constant velocity and constant acceleration. During these simulations, the transition matrix used and the target tracks used were varied to allow viewing of behavior of the IMM under different conditions. It is difficult to decide whether changing the probabilities gives better performance, since performance is not measured by one number, but by a balance of competing interests. Using a three-model IMM inputs were used in order to test which values should be on the diagonal and what effect changing the off-diagonal elements would have on the filter's behavior. The results obtained imply that having three models is not necessary except in exceptional cases. This leads one to pose the question: Are more models better?.