Particle filter based average consensus target tracking with preservation of network connectivity

In this paper a mobile sensor network is tasked with cooperatively tracking a moving target. Each sensor runs a distributed particle filter to estimate the states of the target. Since some sensors may experience occasional failures or do not trust their own observations, a general case is considered, where only a portion of the sensors have the observations of the target. An average consensus algorithm is developed to allow each sensor to dynamically track the global mean of all local estimates about the target from the network by using communication with neighboring sensors only. Due to the limited communication capability on each sensor, the consensus algorithm and the cooperative objective of tracking the target must be accomplished while ensuring that the overall network remains connected. Based on the estimated global mean, a potential field based distributed motion control law is then designed to drive the sensors to track the target cooperatively, while preserving the network connectivity and avoiding collision during the mission operation. Simulation results verify the effectiveness of the proposed approach.