Multiple Model Particle Filtering for Multitarget Tracking

This paper addresses the problem of tracking multiple moving targets by recursively estimating the joint multitarget probability density (JMPD). Estimation of the JMPD is done in a Bayesian framework, providing a method of tracking multiple targets which allows nonlinear target motion, nonlinear measurement to state coupling, and non-Gaussian target state densities. We utilize a particle filter implementation which has been detailed elsewhere [1]. Real targets are poorly described by a single kinematic model. Target behavior may change dramatically – e.g. targets stop moving or begin rapid acceleration. In the literature, the Interacting Multiple Model (IMM) algorithm [4] is used to address this. The IMM uses multiple models for target behavior and adaptively determines which model(s) are the most appropriate at each time step. We demonstrate the IMM in the context of our PF based multitarget tracker in two settings. First, we consider application to targets that switch between kinematic modes. The target motion used is field data recorded during a military battle simulation and includes multiple modes of target behavior. Second, we present a nontraditional application of IMM as multiple models on the state of the filter. In the context of PF based target tracking, this technique may be viewed as a (biased) sampling scheme for particle proposal. This strategy adds robustness to the tracker as it is able to automatically detect model violations and compensate by altering the filter model. This work was supported under the USAF contract F33615-02-C1199, AFRL contract SPO900-96-D-0080 and by ARO-DARPA MURI Grant DAAD19-02-1-0262. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the United States Air Force.