Learning. Cellular Networks Localization Algorithm

Cellular wireless location service is a new mobile value-added service with a good market future. Its basic principle is to implement mobile user location through estimating charactistics parameters relative to position, including time-of-arrival, time-difference-of-arrival, angle-of-arrival, ect. In cellular environments, because of the influence of several adverse factors, such as multipath, non-line-of-sight propagation, noise, interference and channel frequency characteristics, the performance of location algorithms may be significantly degraded and the position estimation of MS (Mobile Station) is inevitably biased. Non-line-of-sight (NLOS) error is a killer issue that degrades the accuracy of mobile positioning especially in complex environments, therefore how to suppress the NLOS error has been the focus of worldwide researchers.Therefore, in order to focus on the core problem of mobile location accuracy improvement, this thesis aims at learning positioning techniques in complex environments and four aspect's works were done.(1) The typical NLOS mitigation algorithm, learning positioning technique and the criterion of performance evaluation and for location algorithms are described.(2) In practice, the cellular network has some knowledge of the location of the mobile terminal. Using the prior information, we propose a non-parametric method for density estimation based on the Support Vector Machine (SVM) to locate the mobile terminals. Simulation results show this algorithm gives excellent performances at several different levels of measurement noise.(3) Learning positioning technique relies on a priori information and needs the majority of the characteristics of the position the region. Utilizing kriging, we interpolate the measurements to specific points that we are interested in to get more survey data. Simulation results indicate that the proposed algorithm provides better location accuracy even in severe NLOS conditions.(4) In order to overcome the overfitting problem caused by noises and outliers in support machine,a method for NLOS mitigation based on fuzzy least square support vector machines is proposed. Simulation results show that the proposed method is robust in NLOS environments and actually increases the accuracy of (least square support vector machines) LS-SVM.