Exploiting NLOS Bias Correlation in Cooperative Localization

Network localization is challenging in line-of-sight (LOS)/non-line-of-sight (NLOS) mixed environments since the statistics information of NLOS biases is generally unknown. In this paper, we investigate the cooperative localization in LOS/NLOS mixed environments with spatial correlation. A maximum-likelihood estimator (MLE) based algorithm for joint agent localization and bias estimation is proposed without knowing statistics information of NLOS biases. The non-convex MLE is relaxed into a semidefinite programming and spatial correlation constraints are used to improve the localization accuracy. Furthermore, a bias-induced optimization is implemented to improve the localization performance by identifying LOS links. Finally, numerical results validate our theoretical analysis and the performance of the proposed algorithm.