Direct Target Tracking by Distributed Gaussian Particle Filtering for Heterogeneous Networks

Abstract

In this article, we consider the distributed direct target tracking using the received radio signal by exploiting time delay and Doppler for heterogeneous wireless sensor networks. We develop herein a distributed Gaussian particle filtering (D-GPF) algorithm for diffusion networks, along with an adaptive combiners (AC) scheme and a particle number adaptation (PNA) method. We transform the online AC optimization problem into the minimum variance unbiased estimation problem with the nonnegative constraint of the combiners imposed, and solve this constrained problem by establishing the Karush-Kuhn-Tucker (KKT) conditions. Further, we develop a variable bin-size scheme for the PNA method to improve the efficiency of particle filters in conformity with the underlying state uncertainty at each sensor of the heterogeneous networks. Simulations involving heterogeneous networks illustrate i) that the proposed AC method could improve robustness of the D-GPF against the spatial variation of Signal-to-Noise-Ratios over the network, and ii) that the proposed PNA method could achieve a tradeoff between the tracking performance and computational complexity of the D-GPF.