Adaptive Bayesian Compressed Sensing based localization in wireless networks

This paper exploits the most recent developments in sparsity approximation and Compressed Sensing (CS) to efficiently perform localization in wireless networks. Based on the spatial sparsity of the mobile devices distribution, a Bayesian Compressed Sensing (BCS) scheme has been put forward to perform accurate localization. Location estimation is carried out at a network central unit (CU) thus significantly alleviating the burden of mobile devices. Since the CU can observe correlated signals from different mobile devices, the proposed method utilizes the common structure of the received measurements in order to jointly estimate the locations precisely. Moreover, when the number of mobile devices changes, we increase or decrease the measurement number adaptively depending on “error bars” along with precedent reconstruction processes. Simulation shows that the proposed method, i.e. Adaptive Multi-task BCS Localization (AMBL), results in a better accuracy in terms of mean localization error compared with traditional localization schemes.