Consensus-based distributed particle filters in sensor networks

This paper considers the problem of distributed particle filtering using consensus algorithms. The monitored environment may possess nonlinear dynamics, nonlinear measurements, and non-Gaussian process and observation noises. It considers the scenario in which a set of sensor nodes make multiple, noisy measurements of the monitored system. The goal of the proposed approach is to perform an on-line, distributed estimation of the current state at multiple sensor nodes. In this new proposed algorithm, average consensus filters are well organized to do distributed computation and information consensus in distributed particle filtering. Furthermore, sensors' energy consumption concerns are considered partially here. In order to achieve almost full environment information, sensors are assumed to have different sensing models, but same dimensions. As a case study, the application of the proposed algorithm to state estimation of an unmanned air vehicle is considered here. Simulation results show the good efficiency of the algorithm in the nonlinear state estimation.