Likelihood adaptation of particle filter for target tracking using wireless sensor networks

The accuracy of particle filtering estimation for the tracking system is prone to be influenced by the high measurement noise (or errors) in wireless sensor networks (WSNs). We first analyze the impact of instantaneous measurement noise which is introduced into the likelihood function and biases the particle filtering estimation. Based on our analysis, we propose a likelihood adaptation method considering the prior information of measurement and introduce a belief factor θ, which is a tuning parameter for adaptation. The optimal θ is attained by deriving the minimum Kullback-Leibler divergence. We integrate our adaptation method with bootstrap particle filter for time-of-arrival based target tracking. The simulation and experiment results of demonstrate that our likelihood adaptation method has greatly improved the estimation performance of particle filter in a high noise environment.