Multipath is one of the main causes of degraded position accuracy in the Global Navigation Satellite System (GNSS) because portions of the signals can be reflected by high buildings in dense urban areas. Multipath mitigation techniques based on hardware enhancement or signal processing help to improve GNSS accuracy for high-precision surveying. Geographic Information System (GIS) is also used in the signal propagation model to predict multipath effects. In addition to these existing approaches, we found that spatial statistical methods are useful in multipath mitigation because a unique spatial distribution of user positions can be produced by the multipath. In this paper, we present a spatial statistical method for mitigating multipath and improving the accuracy in GNSS positioning. Multipath tends to be associated with spatial outliers in simulated user positions (SUPs) and contributes little to the spatial clustering of SUPs. Using these spatial characteristics, we developed a method for identifying multipath satellites, which consists of the components of deviation distance, deviation load, and deviation ratio. Once the identified multipath satellites are excluded, a user position is determined using a mean spatial center of the SUPs from the remaining satellites. The effects of such multipath mitigation were validated by examining whether our method correctly identified multipath satellites and by comparing the position errors with and without the method. We demonstrated the applicability of our solution with a simulation experiment for Shinjuku, using a precise ephemeris for the Global Positioning System (GPS) and the orbital parameters for the proposed constellations of the GALILEO and the Quasi-Zenith Satellite System (QZSS).
