Block Variational Bayesian Algorithm for Multiple Target Localization With Unknown and Time-Varying Transmit Powers in WSNs

Abstract

Target localization is an important research field in wireless sensor networks and the received signal strength (RSS)-based method is of particular interest. Since traditional RSS-based method is invalid for simultaneously localizing multiple targets, the compressive sensing theory has been applied to develop an effective localization framework in recent years. However, most existing works implicitly assume the transmit powers of targets are known and constant, which contradicts the practical application scenario as the power might change with time and usually difficult to be known. In this paper, we consider the challenging problem of multiple target localization with unknown and time-varying transmit powers by exploiting multiple measurement vectors (MMVs). Since the conventional MMV formulation is incapable of the time-varying environment, we develop a novel MMV framework, in which the temporal correlation of varying powers can be further exploited. In this way, the problem is transformed to reconstruct a block-sparse signal and learn its inner correlation structure. To address this, a block-sparse reconstruction algorithm is designed, based on the novel application of the variational Bayesian approximation and the expectation-maximization methodology. The proposed algorithm can jointly estimate locations and varying powers of multiple targets with high accuracy. The extensive simulation results demonstrate the superiority of the proposed method in comparison with the state-of-the-art algorithms.