Benefits of CNN-Based Multipath Detection for Robust GNSS Positioning

Anthony Guillard, Paul Thevenon, Carl Milner, Christophe Macabiau

Abstract: Detecting GNSS multipath has always been extensively researched as it is one of the hardest error sources to predict and model. However, detecting multipath by itself does not lead to an improved positioning solution. To do so, the multipath affected observable(s) must either be excluded or their effect mitigated in the positioning solution. The approach taken in this paper was to mitigate their effect but to keep the degraded observable in the positioning solution. Indeed, in challenging GNSS environments, multipath is unavoidable. Hence, excluding multipath-riddled measurements could lead to availability issues of the GNSS position. This paper proposes a fusion of a robust estimator and a classical weighted least squares estimation (WLSE) algorithm. This fusion is used to mitigate the effect of multipath by using the optimal WLSE algorithm when no multipath is detected and by using robust estimators when multipath is detected. The multipath is detected using a trained convolutional neural network (CNN) to detect which code pseudorange is affected by multipath. For the presented data, the proposed algorithm was shown to improve the horizontal positioning accuracy by 20% and 11% for 20% and 60% of the time with respect to using robust MM estimators or WLSE solutions alone.
Published in: Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023)
September 11 - 15, 2023
Hyatt Regency Denver
Denver, Colorado
Pages: 283 - 297
Cite this article: Guillard, Anthony, Thevenon, Paul, Milner, Carl, Macabiau, Christophe, "Benefits of CNN-Based Multipath Detection for Robust GNSS Positioning," Proceedings of the 36th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2023), Denver, Colorado, September 2023, pp. 283-297. https://doi.org/10.33012/2023.19421
Full Paper: ION Members/Non-Members: 1 Download Credit
Sign In