A multiagent system is a research field that tackles problems that are difficult to solve by one agent by running a large number of agents. My research focus is teamwork behavior in the simulated soccer domain. We are studying behavior learning algorithms for soccer tasks using the RoboCup Soccer Simulator. Achieving better methods of environmental recognition and decision making through multiagent adversarial games such as soccer is expected to be applied to real human society. As applications in the real world, we are also conducting research such as collecting and analyzing information that humans pay attention to when deciding actions and building environmental maps using multiple mobile robots.