This report proposes a control system design for soccer robots based on the immune system. The immune system is a kind of engineering model imitating the human immune system, whose advantages are as follows: self-optimization by learning, and easiness in dynamic reconfiguration of themselves. Thus, the immune system can be effectively applied to robot control in dynamic environments. After applying the system to an obstacle avoidance problem, we extend it to 3 vs. 2 soccer plays. Simulation results show that the immune system is useful for robot control.