Self-Localization in Dynamic Environments based on Laser and Vision Data
Conference paper (Refereed)
Proc. of the IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS)
Engineering and Technology
Artificial Intelligence & Integrated Computer Systems
For a robot situated in a dynamic real world environment the knowledge of its position and orientation is very advantageous and sometimes essential for carrying out a given task. Particularly, one would appreciate a robust, accurate and efficient selflocalization method which allows a global localization of the robot. In certain polygonal environments a laser based localization method is capable of combining all these properties by correlating observed lines with an a priori line model of the environment  . However, often line features can rather be detected by a vision system than by a laser range finder. For this reason we propose an extension of the laser based approach for the simultaneous use with lines detected by an omni-directional camera. The approach is evaluated in the RoboCup domain and experimental evidence is given for its robustness, accuracy and efficiency, as well as for its capability of global localization.