Selectively iterative particle filtering and its applications for target tracking in WSNs

Particle filters (PF) have been widely used in the estimation of the state transition and observation of non-linear/non-Gaussian systems, and samples degeneracy is the main issue of particle filters. In this paper, a novel PF - selectively iterative particle filter (SIPF) is proposed for target tracking in wireless sensor networks (WSNs). There are two novel strategies in SIPF, the statistics based threshold and particles refining. The key insight of SIPF is from an experimental observation that, the more suitable divergence of particles can yield the better estimation. The performance of the proposed SIPF is tested on two theoretical models, and then it is used in a target tracking issue in WSN in the distributed model. Experimental results show that the proposed SIPF can greatly improve the accuracy of object tracking, and in theoretical models the estimation error is only about 10%, while in practical models is only about 25% compared to other existing 9 filtering methods.