Simultaneous Iterative Learning and Feedback Control Design

Iterative learning controllers aim to produce high precision tracking in operations where the same tracking maneuver is repeated over and over again. Model-based iterative learning control laws are designed from the system Markov parameters which could be inaccurate. Chapter 2 examines several important learning control laws and develops an understanding of how and when inaccuracy in knowledge of the Markov parameters results in instability of the learning process. While an iterative learning controller can compensate for unknown repeating errors and disturbances, it is not suited to handle non-repeating, stochastic errors and disturbances, which can be more effectively handled by a feedback controller. Chapter 3 explores feedback and iterative learning combination controllers, showing how a one-time step behind disturbance estimator and one-repetition behind disturbance estimator can be incorporated together in such a combination. Since learning control applications are finite-time by their very nature, frequency response based design techniques are not best suited for designing the feedback controller in this context. A finite-time feedback controller design approach is more appropriate given the overall aim of zero tracking error for the entire trajectory, even for shorter trajectories where the system response is still in its transient phase and has not yet reached steady state. Chapter 4 presents a combination of finite-time feedback and learning control as a natural solution for such a control objective, showing how a finite-time feedback controller and an iterative learning controller can be simultaneously synthesized during the learning process. Finally, Chapter 5 examines different configurations where a combination of a feedback controller and an iterative learning controller can be implemented. Numerical results are used to illustrate the feedback and iterative controller designs developed in this thesis.