Abstract: Mobile localization in wireless sensor networks (WSNs) has attracted considerable attention in recent years. One of the most important factors affecting the accuracy of localization or tracking is non-line-of-sight (NLOS) signal propagation. The NLOS error could seriously reduce the localization accuracy. By analyzing the characteristics of the residual of NLOS distance measurements， a strict residual selection method was proposed to identify the condition of the distance measurements. In this algorithm， extend Kalman filter (EKF) linear regression model was firstly utilized to get distance residuals. Then the strict residual selection was used to filtrate the residuals. Finally the localization was finished by using the parallel variable node EKF algorithm. Simulation results show that the localization of the proposed algorithm outperforms the other algorithms compared in NLOS conditions. The proposed algorithm has better robustness and higher accuracy in different environments.

Key words: wireless sensor network, non-line-of-sight(NLOS) localization, extend Kalman filter, strict residual, linear regression model