Distributed Sensor Selection Based On The Conditional PCRLB

Wireless sensor network (WSN) is a synthesis of modern sensor technology,communication technology and microelectronic technology. With the development andprogress of sensor hardware in wireless communication and network technology, variousapplications of wireless sensor network (WSN) has been widely applied in military and civilfield. Because of the limitations of physical resource (energy, frequency, bandwidth, etc.), theimplementation of effective sensor nodes management is of great significance to prolong thelife of the sensor during target tracking.Wireless sensor network structure includes centralized wireless sensor network anddistributed wireless sensor network. In centralized wireless sensor network, a sensormanagement scheme is proposed based on conditional posterior Cramer-Rao lowerbounds(CPCRLB). The particle filtering is employed to estimate the target state and CPCRLBapproximately. The online sensor selection is achieved by particle filtering. The sensorselection schemes based on CPCRLB, mutual information and nearest neighbor method arecompared by simulations. The simulation results demonstrate the efficiency and superiority ofthe CPCRLB-based sensor management.Distributed network structure has lower energy consumption, higher robustness andflexibility than centralized network structure. The paper proposes a distributed sensorselection based on CPCRLB. The CPCRLB on the mean-squared error of target state estimateis derived based on RSS measurements on decentralized architectures, and the distributedCPCRLB is used as criteria to select the optimal sensor nodes to target tracking. The iterationcalculation method of distributed CPCRLB that based on the consensus algorithm is given，and the distributed particle filtering is used to estimate the CPCRLB approximately. Theonline selection of distributed sensor nodes is achieved by the distributed particle filtering.In order to realize distributed CPCRLB sensor node algorithm and verify its validity inthe distributed network structure, iterative algorithm is presented in the paper. And the sensorselection schemes based on distributed CPCRLB, nearest neighbor algorithm and randomalgorithm are compared by simulations for target tracking. The simulation results show the validity of distributed sensor selection. Decentralized CPCRLB is superior to otheralgorithms.