Robust Message-Passing Based Cooperative Positioning for VANETs Using GNSS and UWB Measurements

Compared to single-vehicle positioning, GNSS-based cooperative positioning improves accuracy and robustness, making it a key technology for future Vehicular Ad Hoc Networks. This paper proposes a robust Bayesian cooperative positioning approach that integrates double-differenced GNSS pseudorange and ultra-wideband ranging measurements. The method explicitly accounts for inter-vehicle state coupling and the correlated noise characteristics introduced by cooperative differencing, and constructs the global joint posterior distribution of all vehicle states, which is then factorized for distributed inference. We develop a message-passing algorithm based on belief propagation, enhanced with Cauchy M-estimation. This algorithm enables each vehicle to adaptively reweight abnormal measurements, modify the nominal measurement likelihood functions accordingly, and compute the messages exchanged with its neighbors. As a result, each vehicle is able to independently estimate its global posterior in a fully distributed and fault-tolerant manner. Experimental results in challenging urban environments demonstrate that the proposed method significantly outperforms existing algorithms in both estimation accuracy and robustness, particularly in the presence of measurement anomalies.