Research On Robot Soccer Match System Based On Vision

Multi-Agent System (MAS) is a hot problem in artificial intelligence reaseach, and it is an importance branch in Distributed Artificial Intelligence (DAI). The research content of Multi-agent cooperation theoretics is that the multi-agent groups how to organize, cooperate and complete the complex task in a complex dynamic environment.Robot soccer match system is a typical multi-agent cooperation system; therefore it is a standard platform for researching multi-agent cooperation problem. FIRA (Federation of International Robot soccer Association) MiroSot (Micro-Robot Soccer Tournament) group is one of the most pop robot soccer matchs, whose system includes vision subsystem, decision-making subsystem, wireless communication subsystem and soccer robot subsystem. MiroSot is involved in many reseach domains, such as computer vision, picture engineering, MAS theoretics, robotics motion control, wireless communication, micro electromechanical system and electrical driver. The major content of tHSI paper is to reseach MiroSot vision subsystem and decision-making subsystem. Vision subsystem works as robot 's eye -recognise and trace chromatic image object. The information content of chromatic image is is more abundant than gray-level image. Therefore, there will be more challenge to realization fast recognising and tracing in chromatic image than in gray-level image. The decision-making subsystem is the core of MAS cooperation theoretics research and application; its target is to resolve cooperation and motion problem of robot. The research task is financed by national 863 planning project "Soccer Robot Development" (512-9935-08), "Micro Group Robot Soccer Pivotal Technology Research" (2001AA422270) and Guangdong province s&t reseaching project "Multi-agent System Experiment Platform, Simulaition System and Application" (2KM00502G)The research contents include:(1) MiroSot system architecture and principle. Research in the structure, function and hardware design of all MiroSot subsystems.(2) Chromatic image recognition. It discusses the feature of color, color model, image segment principle and method, and then analyzes the demands and characters of robot soccer vision system. An adaptive RGB threshold value color image segment segmentation method based on RGB and HSI model is presented, and it realize a realtime, veracity, adaptability and intelligent chromatic image recognition.(3) Muti-target visual tracing. To improve visual tracing precision and environment adaptability, it presents a method of soccer robot visual tracing based on action-vision coordination. The method validates the coherence of target action and command by vision. It can keep uninterrupted tracing in visual blind area, and correct the target tracing error.(4) Image rectification. In order to improve the positioning precision, it analyzes the image distortion in vision subsystem, and a simple distorted image rectification method is provided.(5) Multi-agent cooperation. To overcome the disadvantage of general centralized decision-making subsystem, according to the concept of software agnet, a distributed soccer robot decision-making subsystem based on multi-agent cooperation system and its structure, function and arithmetic are offered. At the same time, the software agent and cooperation mechanism of that cooperation system are researched. The cooperation system can organize cooperating relationship dynamically and flexibly to adapt to cadence of robot soccer. Finally, it studies on the software technologies of realizing the robot soccer multi-agent cooperation system on single host, including synchronous module and Multi-thread technology.(6) Soccer robot dynamic anti-collision and motion control. After analyzing the disadvantage of general anti-collision method, it offers a soccer robot anti-collision control method based on evolutionary artificial potential field. Combining bi-wheel robot motion model with robot tra...