Mobile localization based on improved non-line-of-sight classification

One of the main problems facing accurate localization using Time-of-Arrival (TOA) measurements in wireless communication systems is Non-Line-of-Sight (NLOS) propagation. Classification is among the important NLOS mitigation approaches whereby the system attempts to identify and localize with the LOS base stations (BSs) only. In this paper, the impact of the number of NLOS BSs and their geometrical distribution on the localization process is investigated. With the use of a residual test (RT) classification algorithm, it is shown that, under certain conditions, the system that localizes with all BSs (including both LOS and NLOS) can benefit from error cancellation, and may perform better than the system that implements “pure” classification. Based on this observation, it is concluded that BSs identified as NLOS should not always be dropped, and a modified classification algorithm is proposed. The new algorithm drops BSs identified as NLOS only when there are enough geometrically “well distributed” BSs to localize with. Simulation results show that the proposed algorithm can outperform other conventional classification schemes.