Robust Student’s T Distribution Based PHD/CPHD Filter for Multiple Targets Tracking Using Variational Bayesian Approach

Abstract

Measurement-outliers caused by non-linear observation model or random disturbance will lead to the accuracy decline of a target tracking filter. This paper proposes a robust probability hypothesis density (PHD) filter to handle the measurement-outlier problem based on Student’s T Kalman (TK) filtering technique and Variational Bayesian (VB) method. First, the non-standard measurement noise is considered to follow the Student’s T distribution. Second, the TK filtering technique is employed to update the target states. Third, the posterior likelihood is updated by the VB approach. Simulation results show that the proposed method can reduce the optimal subpattern assignment (OSPA) error in the non-standard observation scenarios with measurement-outliers, compared with other typical multiple target tracking filters.