Abstract: According to the rough sets theories, rules can be extracted from the control data of a two-cell immune controller. With a new criterion proposed in this paper, the best rules were selected to compose the rough-regularized two-cell immune controller. From a normal two-cell immune controller to a rough-regularized two-cell immune controller, the characteristic of self-learning and anti-delay can be inherited, which is proved through a simulation. The better anti-disturb ability and faster response were also demonstrated.