Self-position Estimation of Autonomous Mobile Robot with Size-variable Image Template

We propose a new image template generation method for the self-position estimation of an autonomous mobile robot. In the proposed method, an image template is generated with genetic algorithm. Then, in the process of the self-position estimation, the size of the image template can be varied in order to change the time for the self-position estimation according to the situation around the robot. Therefore, a suitable image template is searched by GA search as the size of the image is varied. The position of the robot is estimated by matching the input image at the current situation with the stored image templates which indicate certain positions. As a criterion of the template matching, the normalized correlation coefficient is applied. This method is sensitive to the position shift of the image. Therefore, in order to realize the robust self-position estimation for the position shift, the amount of the position shift between the image template and the input image is compensated before the template matching. The usefulness of the proposed method is shown through some experimental results