Robust polygon-based localization

The current work presents an approach that is able to estimate the pose of a robot even when all rangefinder measurements are originated by unknown obstacles. This approach exploits the shape of the boundaries of an environment, displaced according to the rangefinder measurements, to determine the set of possible poses the robot may have. By using interval analysis to compute its estimation, this method suits well to scenarios where no knowledge about the probabilistic density function of the measurements is provided. Comparing to classical interval approaches, this method does not expect that the robot has a perfect knowledge of the environment. This way, this approach can compute a feasible estimation even in presence of unknown obstacles without using time-consuming techniques such as Q-Relaxed Intersection or GOMNE. For validation, we present a comparison between the proposed method in this paper to a classical interval approach to the robot's localization problem.