Truncated particle filter based on QMC sampling and application to target tracking

Abstract

In order to deal with the nonlinear, non-Gaussian problem, this paper proposes the truncated particle filer based on quasi-Monte Carlo (QMC) sampling. This algorithm replaces the resampling scheme with the truncating scheme which preserves the large weight particles and discards the small weight particles by the truncated threshold. Because the Monte Carlo sampling can induces the possible gaps and clusters, the new proposed algorithm utilizes the QMC method to sample particles uniformly. The results show that the algorithm improves the estimate accuracy.