Multiple model algorithm based on particle filters for ground target tracking

Abstract

In this paper a novel multiple model particle filter algorithm for tracking ground targets on constrained paths is developed The algorithm is designed to let the different modes be represented by constrained likelihood models, whereas the state dynamics are the same for all models. The mixing procedure is performed over the likelihood models and the mixing parameters are calculated in a standard interacting multiple model (IMM) manner. The performance of the developed estimator is compared with several other multiple model particle filters in a Monte Carlo simulation study. A ground target scenario consisting of road networks is used to evaluate the behaviour of the tracking filters and to illustrate the selection of design parameters.