Automatic NLOS Classification from Virtual 3D City Models

The paper describes the use of publicly available 3D models of buildings for possible mobile positioning accuracy enhancements. The lack of the line of sight (LOS) between base stations (gNB) and user equipment (UE) is one of main reasons for poor performance of positioning algorithms. If the lack of LOS is automatically detected then given measurement can be simply discarded. Alternative option is mitigation of the lack of LOS to the final positioning estimate. Method uses virtual 3D city models in CityGML (Geography Mark-up Language) format to automatically detect if the straight line from gNB to UE intersects any surface of the surrounding buildings or not. Method was tested with the measurement results collected with the portable 5G test scanner. Classification results were verified visually.