Self-positioning and localization of a mobile robot using vision-based behaviors

This paper presents a method to localize a mobile robot in a topological map. This work enters a navigation method developed for mobile robots, which basic idea is to represent the robot spatial knowledge by a topological map and to use the behavioral approach to control the robot moves between fixed positions of the environment known as self-positioning sites. A feature of this method is that each physically grounded site is represented by a node in the topological map, and the localization problem is to find the node-site correspondences. In the paper we develop and analyze a localization method based on vision that compares images taken by a robot-mounted camera with references images. Special considerations is given to noise rejection by making the method robust against environmental changes. The localization method is implemented on a Nomad 200 mobile robot. Presented localization results illustrate its performance and degree of robustness.