PrefaceThe essence of adaptive control is to observe an unknown process, derive a hypothesis of how to control the process in order to obtain best performance and continuously update this hypothesis based on new information. In this thesis, I investigate ways to improve performance even as the information is corrupted by disturbances affecting the process.I would like to thank Professor D. Schröder for having introduced me to this exciting field of research and for having provided his continuous support and interest during the years of my doctoral work. His open-mindedness and vision was the incentive for me to pursue new ways of intelligent control for mechatronic systems.To my teacher and mentor, Professor K. S. Narendra, I express my profound gratitude for having accepted me as his graduate student even though I never officially enrolled at Yale.I thank him for having opened my eyes to the complex nature of research, for being such a brilliant thinker and such a dedicated teacher.The research reported in this doctoral thesis was made possible by Siemens AG, which generously provided me with an Ernst von Siemens-stipend over a period of three years. In
AbstractThe emphasis of this thesis is on developing a theoretical framework for the practical design of adaptive systems. The state of the art in industrial adaptive control is to combine a parameter estimator with an (existing) linear control-loop. The success of the approach depends upon the knowledge of the physical parameters after completion of the estimation process. In the presence of disturbances, though, a correct estimation of the parameters is impossible. It is first shown how a system with unknown parameters can be controlled adaptively, without relying on parameter convergence. Following this, three classes of disturbances and the corresponding methods to reject them are developed: external disturbances, unmodelled dynamics and time-variations. In the latter case, a new algorithm based on multiple adaptive models is presented which was developed at Yale University during regular visits of the author at the Center for Systems Science. An experimental study of an adaptively controlled two-mass system concludes the thesis.