Algorithm for improving the positioning performance of GNSS using machine learning in multi-path environment

Abstract

The Global Navigation Satellite System (GNSS) is the most representative positioning system for providing user location information. Along with the fourth industrial revolution, this satellite navigation system is becoming increasingly important for civilian, military, traffic, security, and financial activities. Moreover, it is essential in many convergence technologies such as emergency rescue and safety services, and for autonomous vehicles. However, the GNSS has the problem of performance degradation in multipath environments, like deep urban environments, and a fundamental solution has not yet been found. The reason for the degradation of the positioning performance of the GNSS in the urban environment is that the first arrival path (FAP) delay of a multipath signal cannot be accurately estimated at the signal tracking stage of the GNSS receiver

so the pseudorange measured is larger than the actual pseudorange. In this paper, we propose an algorithm to find the FAP delay at the signal tracking stage using machine learning in order to improve positioning performance in multi-path environments. In addition, we propose an algorithm to replace cases where machine learning is not available at the signal tracking stage by applying machine learning at the position calculation stage of the GNSS receiver. Then, performance of the algorithms applied to each kind of machine learning is compared with conventional algorithms to prove that the positioning performance is greatly improved.