On Multi-Agent Localization Based On Robocup Four Legged Robot Group

This paper was based on RoboCup legged-robot group which uses Sony aibo robot as experimental object. With the platform of GT2004 published by GermanTeam in 2004, combining the image information taken by the robot's own camera with recent self-localization approach, the paper implemented real-time localization process under the real dynamic environment and met the need of RoboCup new rules. On the condition of wireless network communication and accurate pose estimation, the dynamic role assignment and coordination among multi-agents became possible.The paper first introduced the background of the project and the actuality about self-localization research. With that the scheme and content of the project were provided. Then some related coordinate systems were built according to the practical competition surroundings. The paper emphasized on the description of image processing and landmark parameters evaluating both in figures and with programs. Based on the processing results and Monte-Carlo approach, it gave a detailed explanation on the self-localization process under RoboCup legged-robot group soccer competition. Finally it built up the world model and the bid function for role reassignment, with the intention of making dynamic role assignment and coordination behaviors.Considering the practical application, the project took full advantage of the recent research and selected more feasible approaches. Moreover, it achieved some valuable success in character recognition, localization process and coordination behaviors, all of which having great significance in forthcoming related research especially on the humanoid cooperation among multi-agents.