Compressed Sensing Based Localization Algorithms For Unknown Target Population Environments

With the rapid development of wireless sensor networks(WSN), the WSN based positioning service has tremendous market values. Efficient, real-time and accurate positioning algorithm is the key factor to determine the potential of the positioning service. Among the traditional WSN positioning algorithms, the received signal strength(RSS) based method is a simple one, which has been highly recommended by domestic and foreign scholars. Applying compressive sensing(CS) to WSN localization can not only reduce the number of RSS measurements, but also has better accuracy and robustness. However, most of the CS algorithms need to know prior knowledge of signal sparsity level, which is not suitable for unknown target population environments. In practical applications, however, target population is often unknown. So it is urgent to find some excellent algorithms to alleviate this contradiction. Based on the theory of compressed sensing and wireless sensor networks, this thesis focuses on compressed sensing based localization algorithm for unknown target population environments.This thesis first introduces the theory of compressed sensing and wireless sensor network positioning technology briefly and then focuses on the description of the WSN positioning system model and CS reconstruction algorithms. Based on this, firstly we study the performance of the detection-based orthogonal match pursuit(DOMP) algorithm. And it is the first time to be used for WSN localization to explore the ability of target recognition for unknown target population environments. Next, we propose an improved algorithm based on greedy matching pursuit(GMP) algorithm. The performance of the proposed algorithm is studied under different measurements, SNR levels and grid numbers through extensive simulations. In addition, compared with the original GMP and OMP algorithm, the performance of the algorithm is tested in three aspects- the false alarm probability, the missing probability and the location error. The results show that the improved algorithm has a better positioning performance than that of the GMP and OMP.