Towards Robust Collaborative DGNSS in the Presence of Outliers

Kurzfassung

This paper extends the code-based collaborative differential GNSS (C-DGNSS) framework to provide robustness under non-ideal conditions. The C-DGNSS functional model is integrated into the M-estimator based on Huber's loss function, aiming to improve positioning performance in the presence of heavy-tailed noise. Two key research aspects are addressed: (i) the impact of outliers on the performance of the C-DGNSS framework, particularly their effect on both users directly affected and those indirectly influenced through centralized processing, and (ii) the effectiveness of robust statistical methods in mitigating this impact. We conduct an experiment addressing multipath propagation in urban environments with limited satellite visibility and another focused on faulty measurements caused by jamming or Byzantine attacks. Results demonstrate the superior performance of the robust C-DGNSS framework, achieving a reduction in positioning root mean square error (RMSE) of up to 30 meters for urban users under moderate multipath conditions, and an improvement of over 25 meters in worst-case error when the central node receives severely faulty measurements. This is achieved while effectively preventing error propagation to unaffected users with favorable geometries, even in networks with a high proportion of faulty nodes. Ultimately, this work marks a pivotal step in redefining the limits of collaborative GNSS performance, proving that robust estimation can transform vulnerable networks into reliable systems.