Performance of Sampling/Resampling-based Particle Filters Applied to Non-Linear Problems

Abstract

In this work, we propose a wireless body area sensor network (WBASN) to monitor patient position. Localization and tracking are enhanced by improving the effect of the received signal strength (RSS) variation. First, we propose a modified particle filter (PF) that adjusts resampling parameters for the Kullback-Leibler distance (KLD)-resampling algorithm to ameliorate the effect of RSS variation by generating a sample set near the high-likelihood region. The key issue of this method is to use a resampling parameter lower bound for reducing both the root mean square error (RMSE) and the mean number of particles used. To determine this lower bound, an optimal algorithm is proposed based on the maximum RMSE between the proposed algorithm and the KLD-resampling algorithm or based on the maximum mean number of particles used of these algorithms. Finally, PFs based on KLD-sampling and KLD-resampling are proposed to minimize the efficient number of particles and to reduce the estimation error compared to traditional algorithms.