Comparative Analysis of Resampling Procedures in Particle Filter Design

Isaev, A.M. Comparative Analysis of Resampling Procedures in Particle Filter Design

Abstract. Nonlinear filtering algorithms based on sequential Monte Carlo methods are considered. The scheme of an elementary particle filter is presented, and the main challenge of its implementation––the degeneracy of the algorithm––is explained. The essence of importance resampling procedures, intended to settle this challenge, is described. The resampling procedures widely used in navigation data processing and target tracking are reviewed. A criterion for determining the time point of resampling is given. The above resampling procedures are compared by analyzing the corresponding pseudocodes. As illustrative examples, two problems are considered: the estimation of a process representing the sum of a polynomial and the second integral of white noise, based on nonlinear measurements, and target tracking. For these problems, the resampling procedures are compared numerically by simulation using the method of statistical trials.

Keywords: Bayesian stochastic approach, nonlinear filtering, Monte Carlo methods, sequential importance resampling procedures, particle filter.

PDF (English)

Cite this paper

Isaev, A.M., Comparative Analysis of Resampling Procedures in Particle Filter Design. Control Sciences 3, 83–92 (2026).