In this thesis, the iterative learning control system(ILCS) including feedback control is considered to improve the control accuracy for the repetitive desired trajectory. For the ILCS, the convergence and robustness to modeling uncertainty are analyzed in the frequency domain.
At first, a design method of learning filters for a class of stable non-minimum phase system is proposed by the model-matching theory. The iterative learning controller consists of two learning filters acting on both the previous input signal and the previous error signal. To design learning filters, the convergence condition of the iterative learning controller is converted into the model-matching problem and a stable and proper learning filter is obtained by solving the Nevanlinna-Pick algorithm.
Meanwhile, in order to analyze the effects of modeling uncertainty, the closed-loop system is assumed as a family of interval systems. Under the assumption, the concept of robust convergence of ILCS is proposed and some robustness results are obtained by use of the Kharitonov's theorem. Based on the results, design methods of learning filters are proposed from the convergence condition of the ILCS. To show the effectiveness of the proposed design method, some simulation study and experiments for a robot manipulator are executed. The proposed method is regarded as an efficient method to handle the parametric modeling uncertainty in the design of learning controller.