MAPT: A Near Real-Time Adaptive Multipath Modeling and Clustering Algorithm for GNSS Urban Positioning

Position estimation in urban canyons using Global Navigation Satellite Systems (GNSS) often suffers from reduced accuracy due to signal interactions with buildings. One example is multipath, where signals reflect off surfaces and introduce delays in the received signal (often resulting in pseudorange estimation errors). In this work, we introduce MAPT (Multipath Adaptive Positioning Techniques) to improve positioning accuracy. MAPT accomplishes this goal by integrating: 1) the NavQ line-of-sight (LOS) performance monitor, 2) a simplified multipath model within NavQ with a non-line-of-sight (NLOS) masking technique, 3) a satellite combination and position solution processor, and 4) a clustering algorithm. These techniques are collectively leveraged to identify trustworthy satellite signals for accurate positioning. MAPT evaluated across 5,475 data points from a representative urban canyon in Columbus, Ohio reduced the 95th percentile of horizontal error from 47.8 meters (without processing) to 4.4 meters. Additionally, this was accomplished by processing each epoch in under 0.24 seconds on average.