Using Neural Networks for Mobile Station Location Estimation

Abstract

Positioning functions, which are automatic positioning of subscribers within cellular networks, are required for mobile communication systems of all generations. However, when new generation networks are implemented, high accuracy of determining mobile station geographical coordinates is required for development of services related to subscribers’ location. To solve the task of mobile station positioning its geographical coordinates are calculated in regard to the known coordinates of base stations. The paper proposes to use a neural network for improving the effectiveness of positioning a mobile station of a cellular communication system. Positioning methods providing usage of neural networks are based on measurements of levels for signals from base stations whose coordinates are known or all the nearest access points. After creating a software or hardware solution for the artificial neural network, one has to create a mathematical model for positioning and perform the network training procedure. The proposed localization method is based on RSSI values. The advantage of the RSSI method is that it requires no additional hardware or computing power. The disadvantage of the RSSI method is the lack of accuracy. Thus, the aim of this paper is to develop an optimized method for determining mobile station location. According to the proposed method, RSSI values ​​from several (at least three) closest base stations to a mobile station enter the neural network, after corresponding processing; the coordinates (latitude and longitude) of the mobile station appear at two outputs. The proposed neural network is a multilayer perceptron. The article presents the proposed architecture of the perceptron. The number of neurons in all the layers has been substantiated. The operation of the multilayered perceptron has been described.