A fuzzy approach to build sonar maps for mobile robots

This paper addresses the problem of building environment models from sensor observations. In particular, real world data, acquired by the ultrasonic sensors of a mobile robot, is used to obtain a 2-D polygonal approximation of the environment outline. Since the system works with inaccurate and noisy data, a way to represent uncertainty and to propagate it through all the steps involved in map building is also considered, so that not only a plausible outline, but also the confidence on this outline, is obtained. Fuzzy sets are used to represent the uncertainty in the location of the object boundaries. The degrees of membership indicate the similarity of a given boundary to other boundaries located in the proximity. The approach is illustrated by experiments in an office environment.