Iterative learning control of MIMO systems with less model knowledge

To design a stable iterative learning control, it often requires some prior knowledge about the unknown systems. In some applications, such as uncalibrated visual servoing, the knowledge is too hard to be gained. This paper proposed an iterative learning control for a class of MIMO systems. The controller consists of a Nussbaum-type gain selector for roughly probing proper control gain matrix and a refined compensator learned through repetitive tracking. It is able to guarantee convergence of the learning control even without any knowledge about the system. Stability of the proposed controller is proved and simulations are carried out to verify the proposed method.