Abstract: | A new design methodology for iterative learning control systems is developed. It is based on the convergence condition for systems operating on an infinite time interval which is of the H∞ type. The principal idea of the design technique is to design a learning controller such that the speed of convergence is maximized, with a compromise to robustness. The issue of finite versus infinite trial lengths is addressed, as well as limitations on the best achievable rate of convergence due to structural properties of the plant. |