A data-driven design of optimal ILC for nonlinear systems

This paper presents a new data-driven optimal design framework of iterative learning control (ILC) for a class of general nonlinear systems. The presented data-driven optimal ILC consists of a control input iterative updating law and a gradient matrix iterative estimate law based on two quadratic criterions, respectively. A major contribution of the presented optimal ILC mechanism is that it only uses the real-time measured I/O data without any model information of the plant for the controller design and analysis. Rigorous mathematical analysis is developed to illustrate the effecience of the proposed approaches.