Research On MAS Of Structure And Cooperation Mechanism And It's Application In The Robocup

Distributed Artificial Intelligence is a branch of Artificial Intelligence; and now it has been a hotspot in Artificial Intelligence. The most efficient solution of distributed Artificial Intelligence is to build a model of agent. Generally, it needs to design the structure of agent and use machine learning or other methods to get the intelligence of the agent. Finally use corresponding methods and plannings to reach the Cooperation of the agents .In this thesis, we design a Robocup simulation team based on agent technique and multi-agent system (MAS) theory. The MAS takes into account of real time and the noises of the world, and every agent has a reasonable structure and higher intelligence and can cooperate with other Agents for a given aim.First we shows some typical agent technique, Multi-Agent system and important Robocup soccerserver models. Secondly, we design a layered agent structure, which we call HfutAgent structure. While, some algorithms of machine learning are introduced to get the intelligence of the individual of HfutAgent which makes individual skills in the Robocup. Finally, we realize the multi-Agent cooperation mechanism using the knoledge of soccer experts. In our system, a typical cooperation method in Robocup called SBSP is used, and we explains how to use reinforcement learning method to reach the goal of local cooperation, and the offense and defense strategy system is build by Decision-Theoretic.