A Cooperative Self-localization Method Based on Group Robot Information Sharing

In the RoboCup MSL (Middle Size League), each robot should be autonomous and have fundamental abilities like obstacle avoidance, path planning, and cooperative behavior. Self-localization is one of the most basic and important functions for mobile robots, especially, multirobot system with cooperative behavior. We aim at improvement of self-localization accuracy of each robot of multirobot system by information sharing. To enhance the robot position accuracy, the soccer ball is used as the common landmark that just one exists in the playing field. In general, measured data, such as distances and angles to the white lines on the field, the ball, have measurement errors in the actual environment, so that we assume that the existence possibility of landmark position follows the Gaussian distribution. In our multirobot system, all robots share the group robot information such as estimated self-location and likelihood, distance and angle to the ball, role (defender, forward) via wireless LAN. Each robot evaluates the position likelihood of the landmark based on the landmark's existence probability. Then each robot corrects its position using fed-back the landmark position. In order to evaluate the efficiency of the proposed method in the real environment, the estimated self-location has been evaluated through experiments using the soccer robots "Musashi".