Robust vehicle-infrastructure localization using factor graph and probability data association

This paper presents a robust graph-based optimization framework for vehicle-infrastructure cooperative localization. Compared to the state-of-the-art approaches, the proposed solution keeps high performance in presence of unknown data association environments. In this paper, the association probability of each measurement is calculated, and then assigned to the corresponding edges on the graph, in which the nonlinear least square method is utilized to optimize the state. Thus the proposed approach presents a robust framework in the presence of high association uncertainty during vehicle-infrastructure cooperative localization, in which the corresponding weights from outliers are lower than the true vehicles. The experimental results demonstrate the good robustness in simulated data.