Development of a Model to Detect GNSS Reflected Signals Using Machine Learning and Application to Stand-alone Positioning

Abstract

GNSS positioning technology is used for various kinds of vehicles such as drones because of the convenient position detection regardless of location. However, there is a problem that the accuracy of GNSS positioning in an urban environment deteriorates significantly. One of the causes is that signals reflected by buildings are used for positioning. This problem must be solved to improve safety of automated vehicles using GNSS in urban areas. Therefore, in this research, we developed a model to detect reflected signals using machine learning. As a results of static and rover experiments, improvement of positioning accuracy was demonstrated by excluding the reflected signals from the positioning calculation.