The major problem in controlling multi-agent system is how to arbitrate each agent's action so that the overall group behavior tends to achieve the group objective as efficiently as possible. Without interaction with other agents, an agent's action selection with high individual performance index may not lead to high group achivement. The characteristics of stimulation and suppression of immune networks, by which the concentration of antibodies in veterbrate animals is regulated, are suitable at this arbitration problem. In this paper the decision making scheme for cooperating multi-agent systems with artificial immune networks is suggested. For computer simulaton a two-on-two soccer game is hired.
