Multitarget tracking with Interacting Population-based MCMC-PF

In this paper we address the problem of tracking multiple targets based on raw measurements by means of Particle filtering. This strategy leads to a high computational complexity as the number of targets increases, so that an efficient implementation of the tracker is necessary. We propose a new multitarget Particle Filter (PF) that solves such challenging problem. We call our filter Interacting Population-based MCMC-PF (IP-MCMC-PF) since our approach is based on parallel usage of multiple population-based Metropolis-Hastings (M-H) samplers. Furthermore, to improve the chains mixing properties, we exploit genetic alike moves performing interaction between the Markov Chain Monte Carlo (MCMC) chains. Simulation analyses verify a dramatic reduction in terms of computational time for a given track accuracy, and an increased robustness w.r.t. conventional MCMC based PF.