Evaluation of ML-based classification algorithms for GNSS signals in ocean environment

In the maritime environment, multipath interference exhibits a significantly pronounced influence, resulting in GNSS system performance degradation. Enhancing system performance involves the identification and elimination of NLOS signals. This study focuses on the analysis of multipath data induced by sea waves, collected off the coast of Kakinada Sea (16.98° N, 82.29° E), to categorize signals as Line-of-Sight (LOS), Non-Line-of-Sight (NLOS) and Multipath (MP). A machine learning (ML) approach is employed to identify the presence of LOS, NLOS and MP signals in a coastal environment, both before and after the advancement of tidal waves. In the proposed approach, ML algorithms are trained using 3 key parameters namely elevation angle, signal strength and pseudorange residuals. This study involves the implementation of 14 prominent supervised classification algorithms and their accuracies and computational times are compared. The results due to GPS (L1) and IRNSS (L5 and S1) are considered. Decision Tree and its ensemble function AdaBoost, exhibited exceptional performance of accuracy (99.99 %) and computational time (0.45 s).