GNSS signal classification to LOS and NLOS signals is of great value for conventional ranging‐based and shadow matching algorithms. The most common attribute for performing this classification is the signal strength. Alas, such classification is often insufficient, in particular, in urban environments. In this paper, we present a novel approach for LOS/NLOS classification utilizing supervised machine learning algorithms. Provided with a sufficiently large labeled training set, the proposed approach is able to predict with high certainty (>85 percent) the satellites' visibility status in dense urban regions. This achievement was possible due to the vast raw measurements supplied for the algorithm and using sophisticated feature‐selection techniques. Copyright © 2017 Institute of Navigation.

No abstract

No abstract