A Bayesian Compressed Sensing Approach to Robust Object Localization in Wireless Sensor Networks

The recent growing interest for location-based services (LBSs) has created a demand on more accurate and robust object localization approaches. In this paper, the Bayesian compressed sensing (BCS) is employed to localize a single or multiple objects in a wireless sensor network (WSN). This is motivated by the advantages of BCS such as closer to l₀-norm, and better performance of reconstruction in case of noisy measurements. Due to the spatial sparsity of number of girds containing an object (comparing with the total number of grids in the region of interest), the localization problem can be transferred into recovering a spare index vector, which is reformulated as a Bayesian estimation problem according to the BCS theory. The proposed method is relieved from the requirement on accurate prior position knowledge of beacon nodes. Besides, by building location fingerprinting based on both line-of-sight (LOS) and non-line-of-sight (NLOS) measurements, the proposed method is robust and applicable to mixed LOS/NLOS environment. Finally, simulation examples are included to demonstrate the superiority of the proposed method.