Wireless Sensor Network Localization Based On BP Neural Network And Compressed Sensing

Wireless sensor networks(WSNs) are widely used in the field of agriculture, environmental monitoring, smart home and target tracking by deploying low-cost sensors in large numbers. It is therefore very important to develop accurate and efficient techniques for localizing wireless sensor network(WSN) nodes to perform some supervisory work based on the information gathered by them. This paper focuses on the WSN location based on BP neural network and Compressed Sensing(CS):(1) It does research on the key technologies of WSN, and focuses on the Range-based and Range-free localization algorithms. TDOA and RSSI algorithms are mainly used in this paper. It also studies the computation methods of node localization and the evaluation indices of localization performance.(2) Considering the non-line-of-sight(NLOS) error existing in the localization of WSN, together with the strong anti-noise ability, good data approximation and flexible parallel data processing ability of BP neural network, this paper proposed the method of using BP neural network to optimize the location of WSN nodes. The paper analyzes the source of error in NLOS environment. In this algorithm, it optimizes the traditional BP neural network on the structure and algorithm to improve the convergence speed, and it uses reliable nodes to train the neural network, and finally it applies the trained neural network to fulfill the position of unknown nodes. The simulation results show that this algorithm can suppress the NLOS error efficiently and the localization precision is better than the traditional Taylor and Chan algorithm.(3) Considering the shortcomings like high power consumption and complexity existing in WSN localization, a new algorithm is proposed. It takes advantage of Compressed Sensing(CS) and BP neural network. In CS, the specific localization of the unknown nodes in the grid is determined. Then, BP neural network is used to correct error. Finally, the three-sided measurement is used to compute the fine localization. This algorithm uses CS to reduce the power consumption and uses the BP neural network to improve the localization precision.