Robust localization for autonomous vehicles in dense urban areas

GPS signals in dense urban areas usually get blocked by buildings or tunnels, making the localization task challenging. Therefore, robust positioning strategies are needed for safe autonomous driving operation. This paper proposes a localization framework for dense urban areas that includes a robust alternative localization algorithm based on a computationally low demanding map matching algorithm. The approach includes an error detection module and a decision layer to determine the most reliable source of positioning: GPS, map matching algorithm or local positioning algorithm. The approach is validated on a simulation environment, presenting an average mean positioning error of 3.11m for the longitudinal axis and 0.629m for the lateral axis, in a wide range of different variable failure scenarios.