Research On Multi-target Tracking In Sensor Network Based On Swarm Intelligence And Game Theory

Wireless sensor network is a network consist of a large number of wireless sensor nodes withbuilt-in various wireless sensors. This network is used to detect, record, analyze and transmit thestatus data of specific object covered by the network, and all data are finally sent to the receiver viathe network by wireless communication.Target tracking is the process of that people use a variety of observational and computationalmeans to locate the moving target and track its trajectory. Wireless sensor network target trackingcan be divided into single-target tracking and multi-target tracking classified by the number oftargets. Multi-target tracking in wireless sensor network is a process that collecting the data oftargets from the sensor nodes and calculating this data to get the trajectory. In the process ofmulti-target tracking in wireless sensor network, how to improve the efficiency of networkself-organization and the accuracy of target tracking, are challenges that multi-target tracking inwireless sensor network should face.This paper introduces and compares several sensor network self-organization methods orientedto multi-target tracking, and then this paper analyzes the Particle Swarm intelligence Optimization(PSO) in the sensor network self-organization oriented to multi-target tracking. By analyzing thedefect of PSO algorithm with a central node (CPSO), this paper put forward a distributed PSOalgorithm (DPSO). This algorithm uses particles to represent candidate network self-organizationplans, it improves the self-organization of the network by calculating the best position of particles.Followed by a simulation, the result proves that DPSO performs better in sensor networkself-organization method oriented to multi-target tracking. This paper also analyzes the filteringalgorithm used in the multi-target tracking in wireless sensor network, and then puts forward theminimized-maximum-error filter algorithm based on game equilibrium. This algorithm reduces themaximum error value by considering the worst situation in target tracking and achieving thezero-sum game equilibrium. Followed by a simulation, the result proves that the new algorithmperforms better than standard kalman fliter algorithm, it can improve the accuracy of multi-targettracking in sensor network.