RoboCup 2013: Best Humanoid Award Winner JoiTech

摘要

This article presents the technical strategy employed by JoiTech in the RoboCup 2013 humanoid league adult size championship. Two features focused on by the team were the design of a versatile robot simulator and smart strategies for the robot player JoiTech Messi. The input for the versatile robot simulator in our system was data from the robot's camera and the output was a motor command given to the robot. Our system used real video data and a robot, as well as virtual data and video data recorded from the real robot's camera. Thus, we could select one of three inputs, i.e., real, virtual, and recorded, and one of two outputs, i.e., real and virtual. This combination of data allowed us to debug the codes used by Messi in an efficient manner. This reduced the number of real robot tests, which minimized damage to the robot. In the design of the smart strategies for the robot player JoiTech Messi, we developed a system that recognized opponent robots using background subtraction, which improved the accuracy of the striker and the goalkeeper. The detection of an opponent player was useful for finding the space to shoot and to block the goal when an opponent player attempted to score. These two features worked effectively during the competitions and our JoiTech team won the championship, as well as the best humanoid award ("Louis Vuitton Cup"), which demonstrated the success of our system.