K. Doki, N. Isetani, A. Torii, and A. Ueda (Japan)
Self-position Estimation, Anytime Sensing, Image tem- plate Generations Autonomous mobile robot, Genetic algorithm
In this paper, we propose a new self-position estimation method for an autonomous mobile robot. This l'l'tethtxl is based on anytime sensing which harmonizes the time for the sensing and time for the action search in the action ac- quisition of an autonomous mobile robot by changing the processing time for sensing. In the self-msition estimation systems the position ol- the robot is estimated by matching the input image at the current situation with the stored im- age templates which indicate certain positions. As a cri- terion of the template matching, the normalized correla- tion coefficient is applied. In order to realize the previ- ous idea, the size of the image template can be varied in order to change the time for the self-position estimation. To realize a stable and efficient self-position estimation. an image template is generated with Genetic Algorithm. In this paper, the usefulness ot- the proposed lnethod is shown through some experimental results with a real-robot.