AGV Global Localization Using Indistinguishable Artificial Landmarks

In this paper we consider the global localization problem for an industrial AGV moving in a known environment. The problem consists of determining the pose of the vehicle without any prior information about its location. The vehicle is supposed to be equipped with a laser scanner that allows to measure the range and bearing of the vehicle with respect to a set of anonymous landmarks. A map with the positions of all landmarks in the environment is available to the localization system. We propose a novel algorithm for AGV self-localization based on landmarks identification that can take into account also false detections, very common in industrial environments. The pose is computed with a single scan (2D), without any sensor fusion. The performance of the proposed strategy is shown both by simulations and experiments on real industrial plants.