p. 61-66
In order to process a task by several agents, each agent must behave systematically. And there is an agreement issue for suitable decision making in multi-agent system. In real world, it is difficult to communicate to other agent because of noise or limited sensor. Focal points is a technique of sharing information without specified communication between multiple autonomous agents. For example, real soccer players can often do contactless passing and know where the teammate goes on the next situation as focal points. In this paper, we describe that acquisition and search of focal point using reinforcement learning method in order to process a cooperative task in multi-agent environment like a soccer.