Tracking maneuvering targets using a modified Rao-Blackwellised particle filter

Abstract

Tracking single or multiple maneuvering targets is an urgent need for defense. In order to meet the military requirement, we propose a modified clustering-based Rao-Blackwellized particle filter (CBRBPF) to track single or multiple maneuvering targets with observations received by single or multiple sensors. The modified RBPF is basing on the clustering-based data association method. We partition the proposed target tracking method into two parts. In the first part, we model the similarity of the observations and design a new clustering algorithm specifically designed for data association. In the second part, we use the modified Rao-Blackwellized particle filter to estimate the states of the targets. We test the proposed method in different situations, including tracking single target with multiple sensors, tracking multiple targets with single sensor and tracking multiple targets with multiple sensors. In each situation, we also compare the proposed method with some commonly used target tracking methods. The results prove that the proposed clustering-based tracking method is more accurate and practical.