Soccer robot system is a typical multi-agent system and multi-robot cooperative system. Furthermore, it provides a new standard test-bed and theoretical research model for real-time artificial intelligence with multi-agent systems. The problem of robot path-planning is an important task in intellectual robot research. Soccer robot system is a dynamic, uncertain and real-time platform, and in such an adversarial and competitive environment, how to realize the real-time robot path-planning is a challenge problem. At present, many researchers apply various technologies to the soccer robot path-planning, such as artificial potential field, grads modeling method, evolutionary algorithms and artificial neural network. However, the application of these theories is not perfect in high dynamic and real-time environment, and some more effective algorithms should be investigated.This thesis is based on soccer robot system. It mainly focuses on the problem of robot path-planning and explores effective methods to solve this problem. Firstly, this paper gives the architecture of the soccer robot system and its key technologies. Especially, it discusses the design of Decision-making subsystem and analyzes the importance of path-planning in this system. We also investigate the system model of robot soccer and collision characteristics of this platform. On the basis of the above work, the author studies the four representative methods which are artificial potential field, grads modeling method, evolutionary algorithms and artificial neural network and discusses their feasibility in real-time system. Finally, a soccer robot path-planning system is designed based on evolutionary artificial potential field. The system has the merit of combining with platform and includes the basic path-planning subsystem, boundary path-planning subsystem and forbidden zone path-planning subsystem. The method is effective by the testimony on the Robot Soccer Simulator. |