Sensor Selection For Target Tracking Based On Particle Swarm Optimization

Wireless Sensor network is composed of a set of small unit which can sense and monitoring environment, and it communicates with other sensor via wireless way. Its application has been expanded from the military to many fields, such as health care, education and family. It has made great contributions to the development of society. But in WSN, the energy and communications bandwidth are limited. It has become a hot research to choose the optimal and the least amount of sensor nodes for target tracking. So the management of the sensor nodes in wireless sensor network is very meaningful.In the process of target tracking, the tracking accuracy will be very high if all of the nodes involved in the work. But it also will consume much energy in this way. For this problem, a sensor management scheme is proposed based on conditional posterior Cramer-Rao lower bounds(CPCRLB). This online sensor selection is achieved by particle filtering. And the results demonstrate the efficiency and superiority of the CPCRLB-based sensor management.Exhaustive algorithm usually be used to select sensor in WSN. The computational complexity of find an optimal subset through exhaustive search can grow exponentially with the number of sensors. In this paper, we apply the binary particle swarm optimization to the problem of selecting k sensors from a set of m sensors for the purpose of minimizing the error in parameter estimation. In addition to applying the general binary particle swarm optimization(BPSO) to the sensor selection problem, we also present a specific improvement to this population heuristic algorithm. The proposed BPSO for the sensor selection problem is computationally efficient, and its performance is verified through simulation results.In the BPSO for sensors management there is only one objective function. Usually we need to solve multiple optimization problems. In this paper, we propose multiple objective particle swarm optimization algorithm for target tracking in wireless sensor network by formulating it as a multiobjective optimization. At each time of tracking, we obtain tradeoff solutions between two conflicting objectives: minimization of the number of selected sensors and minimization of CPCRLB. Simulation results show that the sensor strategy achieves good estimation performance by significantly decreasing the number of selected sensors.