Precise Positioning Based on NLOS Multipath Detection by Using Machine Learning in Urban Environments

Abstract

This paper describes a method of non-line-of-sight (NLOS) multipath detection by using machine learning of signal correlation value of grobal positioning system (GPS) for precise positioning in urban environments. GPS is widely used for estimating self-position in many scenes. However, in urban canyons, GPS positioning accuracy is deteriorated due to NLOS multipath signals. Because of obstacles between NLOS satellites and a GPS antenna, GPS signals omitted from GPS satellites are reflected or diffracted by obstacles. We propose the method to detect NLOS multipath signals using support vector machine (SVM) based on a machine learning of GPS signal correlation values. GPS signal correlation values can be obtained by using a software receiver. We extract the features of NLOS multipath signals from signal correlation value and create the NLOS multipath detection program by machine learning of its feature. As the result of the evaluation, the proposed method is effective to detect NLOS multipath signal.