RBPF-based tracker with adaptive parameter model for maneuvering tracking

Abstract

A new time-varying parameter-based model for target tracking in two dimensions is developed. The target maneuvers are modeled by incorporating tangential- and normal-tack acceleration inputs, the former is characterized by a time-correlated stochastic process and the latter is parameterized by heading-rate, therefore unlike the usual evolution equation with fixed-structure, the proposed model describes complex maneuvering a parameterized formula. By adopting a modified Rao-Blackwellised scheme to evolve the unknown parameter - heading online, the approach implements a tracking by using a varying-structure model. Revealing by a number of Monte-Carlo simulations, the proposed tracker exhibits considerably well computational accuracy compared with a standard interacting multiple model(IMM) estimator, meanwhile the time consumption is more less than that of IMM.