| This dissertation studies robust, monotonically-convergent iterative learning control (ILC) and periodic adaptive learning control (PALO). ILC has been recognized as an efficient method that offers promising performance improvement of controlled systems. The robustness of ILC systems in the iteration domain, however, has not been well addressed. In particular, ILC stability analysis and ILC synthesis related to monotonic convergence (MC) for uncertain systems have not been investigated. This dissertation proposes a unified analysis and design framework for robust monotonically-convergent ILC, using parametric interval concepts, an algebraic Hinfinity approach, and Kalman filtering, all in the iteration domain, using the so-called super-vector framework. Parametric interval concepts enable us to design a less conservative monotonically-convergent ILC system, while considering all possible model uncertainties. By casting Hinfinity techniques and Kalman filtering algorithm into the super-vector framework, the fundamental baseline error of ILC is analytically established. In PALC, the learning control idea is used to solve challenging electromechanical, hard nonlinearity compensation problems. Particularly, non-Lipschitz, time-periodic and state-periodic external disturbances are rejected in the repetition domain. Finally, to make this dissertation self-contained and technically complete, four appendices are included. Appendix A presents an interesting taxonomy of ILC literature. Appendix B shows how to determine the exact bound of the maximum singular value of an interval matrix in general form. Appendix C offers a new scheme for checking the robust stability of interval polynomial matrix systems. Appendix D describes how to better estimate the bound of the power of an interval matrix. Appendices B to D document novel solutions to some fundamental robust interval computational problems that directly support the interval ILC investigation of this dissertation. |
