Evaluating and Mitigating GNSS NLOS Effects for Autonomous Navigation in Urban Environments

In this study, we experimentally evaluate the impact of non-line-of-sight (NLOS) signals on Global Navigation Satellite Systems (GNSS) positioning accuracy in complex urban environments and examine various satellite selection strategies to mitigate these effects. We utilise Light Detection and Ranging (LiDAR)-generated point clouds to build an environmental model that classifies satellite signals as either line-of-sight (LOS) or NLOS. In an effort to reduce pseudorange measurement errors, we also apply a reflection model to estimate the additional propagation path introduced by NLOS signals. Our comprehensive dynamic outdoor experiments demonstrate that, when a sufficient number of visible satellites is available, relying solely on LOS signals can significantly enhance positioning accuracy. However, our static experiments also indicate that NLOS signals not only introduce extra path delays but also cause substantial fluctuations in consecutive pseudorange observations. Furthermore, since existing model rely solely on environmental geometry without adequately considering key factors such as multiple reflections and material reflection coefficients, adopting a simplistic reflection model with fixed-value NLOS corrections may not reliably enhance performance in environments with severe multipath interference.