Self-organizing fuzzy system for vision-based position estimation

This paper presents a self-organizing fuzzy system for position estimation of a mobile robot based on raw vision data. Neither landmarks nor artificial symbols will be used. An onmidirectional local vision system offers images that reflect the environment. Due to its self-organization properties, curvilinear component analysis is used for features extraction from data images. In an initial learning step, the system is trained on the compressed input data, splitting the current environment in some areas. For each of these areas, a B-spline fuzzy controller will determine the position of the robot, based on a new acquired image. By incrementally learning a complex system, the robot is capable to locate itself and navigate in a natural environment.