Design And Implement Of RoboCup 3D Simulation Team Based On Layered Learning

Presently, along with the rapid development of computer network technology, the computer changes to the opening, network platform, cooperation working mode rapidly. Agent based theory and technical, especially multi-agent systems, bring us a brand-new mode to design and implement software systems which run in a distributed and opening environment.RoboCup (The RobotWorld Cup Initiative) is an attempt to promote intelligent robotics research by providing a common task for the evaluation of various theories, algorithms and agent architecture. RoboCup, a typical multi-agent system, has currently chosen soccer as its standard task.The RoboCup simulation team divides into 2D simulation and 3D simulation. The 2D is only two-dimensional simulation, also its physics is not realistic, agent can predict too easily, moreover, the simulator can only simulate soccer. But the 3D is three dimensional simulation and uses the realistic physics which make it difficult to predict for agent, and it is easy to simulate other simulation by change its configure, with these benefits, its research results is easy to be used in the practical application, therefore, 3D become the hot spot of research in MAS. To establish a RoboCup 3D simulation team, I think there are two main work to be done: a good program framework and good choice of study algorithm for the question to be solved. Main work I have done list as below:(1) We propose a layered framework for our 3D team. The system has been divided into three layers: communation layer, technical layer and control layer. The lower layer provides services for the upper layer, and is transparent for the upper, thus, the framework is easy to expand. The communation layer provides agent with the ability to exchage information with server, agent receives information from server, and sends action to the server. The technical layer provides player's personal skill and team's cooperation skill. We begin with player's personal basic skill, which is studied in primitive state space. Then, based on these basic skills, we try to learn the cooperation skill, like pass ball to teammate. If the player has the ability to control...