Comparative analysis of pseudorange multipath mitigation performance using K-means and Fuzzy c-means clustering techniques

Abstract Low satellite signal quality in urban areas is caused by multipath interference resulting from the abundance of various obstacles that limit satellite visibility performance, leading to poor satellite geometry and an increase in dilution of precision (DOP). The chance of encountering multipath significantly increases when receiving reflected signals from non-line-of-sight (NLOS) satellites, causing large positioning errors in pseudorange measurements. Enhanced identification of the multipath error source can improve positioning accuracy. This study first detected and then minimized multipath effects in pseudorange measurements using two unsupervised learning techniques – K-means and Fuzzy c-means (FCM) – which executed clustering across diverse multipath conditions and different combinations of GNSS, handled unlabeled quantitative data, and defined a certain number of clusters. Results indicated comparable performances of the K-means and FCM algorithms, with horizontal positioning accuracy improved by up to 35 % and vertical accuracy enhanced by up to 27 %.