Fuzzy logic based particle filter for tracking a maneuverable target

In this paper we propose a new fuzzy logic-based particle filter (FLPF) algorithm for tracking a maneuvering target. The nonlinear system which is comprised of two-input and single-output are presented by fuzzy relational equations. Each of the fuzzy relational equations is expressed in a canonical-rule based form. The dynamics of the maneuvering target are modeled by multiple switching (jump Markov) systems. We assume that the target follows one-out-of-three dynamic behavior model at any time in the observation period: constant velocity (CV) motion model, clockwise coordinated turn (CCT) model, and anticlockwise coordinated turn (ACT) model. The time-varying deviation between actual and predicted positions is inferred by a two-input single-output fuzzy relation. An example is included for visualizing the effectiveness of the proposed algorithm. For comparison purposes, we simulated a conventional sequential importance sampling (SIS) algorithm. Simulation results showed that the FLPF has better tracking performance compared to the SIS.