An Advanced Auxiliary Delay-Weight Particle Filter with Linear Computation Cost

Abstract

We investigate the problem of tracking mobile targets in wireless sensor networks. We propose an advanced auxiliary delayed-weight particle filter algorithm (ADWPF). We make a deep study on the evolvement of particles and formally define the tree-like structure relationship among particles based on observations. Most importantly, we add some auxiliary particles to these structures formed by sampled particles in order to obtain more efficient structures. Based on the newly tree-like structures formed by auxiliary particles and sampled particles, we design a well efficient delayed-weight algorithm with linear computation cost. Experiment results demonstrate that our algorithm can greatly improve the tracking accuracy of a mobile target, compared with bootstrap filter, auxiliary particle filter, and another delayed-weight particle filter.