Improved Resampling Particle Filter for Maneuvering Point Target Tracking

Abstract

The sample impoverishment during particles resampling is a challenging problem in using particle filter to track point target. The reason is that the impoverishment will lead to the divergence of particles and failure of tracking. To solve the problem, a new resampling algorithm is presented in this study. The idea is that random particle samples are drawn from the neighborhoods of previous samples with high weights according to Gaussian distribution instead of simple duplication. Therefore, during the resampling, the effect of sample impoverishment is reduced and diversity of particle samples is enriched because of the samples expansion while the low weights samples are discarded. To illustrate proposed method, the resampling algorithm with simple duplication and the resampling algorithm with random drawing samples based on Gaussian distribution are compared. Simulation results show that the tracking failure is reduced in the quantitative criteria-RMSE. © 2013 Asian Network for Scientific Information.