A Trust-Region Method for Data-Driven Iterative Learning Control of Nonlinear Systems

Abstract

This letter employs a derivative-free trust-region method to solve the norm-optimal iterative learning control problem for nonlinear systems with unknown dynamics. The iteration process is composed by two kinds of trials: main and additional trials. The tracking error is reduced in each main trial, and the additional trials explore the nonlinear dynamics around the main trial input. Then the trust-region subproblem is constructed based on the additional trial data, and solved to generate the next main trial input. The convergence of the tracking error is proved under mild assumptions. Our method is illustrated in simulations https://github.com/JiaaaWang/TRILC .