Abstract
Since the FCC released the E-911 document and required the positioning capability in next generation mobile devices, the mobile positioning has attracted many research attentions. In order to enhance localization accuracy in line-of-sight (LOS) and non-line-of-sight (NLOS) wireless environments, this research presents a robust mobile positioning algorithm to track the position of a mobile station (MS). This thesis proposes new positioning techniques for the mobile positioning system in 3D environments. Particle filter and convex optimization are adopted to complete the estimation. For the mobile positioning, this work uses particle filter for solving the effect of non-linear channel, successfully addresses the non-linearity issue, and combats the non-ideal effect in a real channel environment. The technique improves the positioning accuracy in both indoor and outdoor conditions. Moreover, this research also proposes to improve the positioning performance by weighting the estimated distances of different particle filters from different base stations. To handle the non-light-of-sight (NLOS) problem, we propose two 3D virtual base station transformation (VBST) algorithms to estimate the position of mobile devices in 2D or 3D environments. Compare the traditional particle filter, the performance of proposed weighted VBST algorithm can improve about 40% and no weighted VBST algorithm can improve about 2% to 20% in the experiments. Then, the convex optimization is performed to obtain the optimal location among different base stations. Due to the convex optimization problem is too complicated and the mixed norm is non-differential, we also adopt subgradient method to reduce complexity of the optimization problem. The first analyzes the map factor that determines the Euclidean distance, Manhattan distance, or mixed distance, and then determines the methods to assign the map factor in different maps. Furthermore, we propose a dynamic map factor to replace the fixed map factor in order to enhance the performance. The dynamic map factor approach can improve about 5% than the fixed map factor approach. Finally, this study shows some simulations of the proposed algorithms and compares with the traditional approaches.