Self-localization of omni-directional soccer robot based on multi-sensor
基于多传感器的全向足球机器人自定位

Because of the increasing intelligence level of soccer robot, the international committee of robot soccer removes from the playing field the landmark information of column and goal which had been used for self-localization by many algorithms. We propose a new self-localization algorithm which utilizes the information from the compass and the odometer, and adopts the image feature-match method. First, the robot takes a possible pose via compass and odometer; and then, the vision system makes use the possible pose as a conversion factor to transform the real-time shooting image. By comparing the white lines of transformed images with the white lines of the reference image, the algorithm determines the conversion factor which makes the two images most similar as the self-localization pose. The results show that this algorithm provides high precision and meets the strong real-time requirements.