The Sensor Selection Based On PCRLB And Convex Optimization

Wireless sensor networks(Wireless Sensor Networks, WSN) have attracted much attention in recent years. One of the most important areas where the advantages of Sensor networks can be exploited is for tracking mobile targets Due to such large networks has their inherent limitations, the sensors are generally resource constrained, and hence their service life and channel between the center and the fusion center is limited. To this end, we should develop some effective strategies to prolong the networks’ life.In order to overcome the network’s shortcomings, This paper propose a sensor selection method based on quantized observations. When the nodes receive the raw measurements of the target state, they don’t transmit the measurements but compress it firstly, with this method the channel’s bandwidth is saved. In order to solve the sensor selection problem, This paper propose a indication function for the sensor management strategy-the Posterior Cramer-Law Lower Bounds. This bound provides the performance limit for Bayesian estimation problem. In this paper, we develop a modified PCRLB which take the measurements in account. The new bound is a real-time management performance evaluation than the traditional PCRLB.We can use the PCRLB as the objective function of the sensor selection, along with the selected number of the sensors, this sensor selection problem can convert to a convex optimization problem. For this optimization problem, we develop a interior-point method to solve convex problem.In order to show the superiority of the algorithm, we provide the detail process for solving this problem and compare this algorithm with the enumeration method and the approximate dynamic programming method and give the simulation results to show these methods. then discuss the influence of the uncertainty of the target motion, the number of the selected sensors and the quantization bits to the object tracking performance.