KLD-Resampling with Adjusted Variance and Gradient Data-based Particle Filter Applied to Wireless Sensor Networks

We propose a modified particle filter (PF) which adjusts variance and gradient data for Kullback-Leibler Distance (KLD)-resampling algorithm to solve the tracking target position in wireless sensor networks (WSNs). Our approach can diminish the bad effect of the received signal strength (RSS) variation by generating sample set near the high likelihood region. Finding optimal adjusted variance of this method based on the maximum of the gap mean number of particles used between proposal and KLD-resampling is presented. A number of simulations are conducted to evaluate the sample size as well as the effect of different parameters such as root mean square error (RMSE) or estimation error, mean number of particles used. Our experiments show that new method enhances the efficient number of particles used as well as estimation error compared with traditional approaches.