Statistical Analysis of GNSS Multipath Errors in Urban Canyons

Much of the prior work studying multipath interference and Non-Line of Sight (NLOS) reception literature studies precise pseudorange error estimation and/or improving a navigation filter’s performance in urban canyon environments. This paper instead studies statistical behaviors of pseudorange errors caused by multipath interference and NLOS reception. The error statistics are observed through histograms of known-NLOS measurements collected by a vehicle-mounted commercial Global Navigation Satellite System (GNSS) receiver traveling through a city center. The histograms formed from known-Line of Sight (LOS) measurements generally maintain a zero-mean bias and Gaussian shape, while the histograms from known-NLOS measurements exhibit both greater bias and heavier tails as the canyon depth increases. The histograms show that the NLOS measurements also suffer greater errors in canyons with similar depth on both sides as compared to canyons where the depth of one side is significantly greater. As modern state estimators often incorporate statistical error models to fuse measurements, these specific results offer insights into how to select statistical error models that better match the Radio Frequency (RF) environment which urban GNSS receivers experience.