Residual-Feedback Particle Filter for Maneuvering Target Tracking

Abstract

In this paper, we propose a residual-feedback particle filter (RFPF) for maneuvering target tracking, whose key idea is to adjust the process noise and particle number in a real-time manner according to the measurement residual. Simulations were conducted on a typical maneuvering motion and the results indicate that the proposed RFPF shows similar performance with the multiple model particle filter (MMPF) but requires no knowledge of acceleration, uses only one state model and reduces computational complexity.