ON ITERATIVE LEARNING CONTROL

Abstract

This thesis concentrates on iterative learning control (ILC). Based upon former results which obtained by several scholars, novel modified or improved iterative learning control algorithms have been proposed in the thesis. Simulation results which correspond to the presented ILC laws demonstrate their functionality, feasibility and effectiveness. Discrete-time ILC algorithms for systems with direct coupling between the system input and the output are first analyzed and discussed in the thesis. When current cycle feedback iterative learning control algorithm was implemented, the implementation obstacle that concurrent output error signal is not available has been successfully solved through certain adjustment. Initial instability related to iterative learning control was then observed and investigated. This phenomenon occurs due to the sampling delay when designing ILC in continuous­ time domain but implementing it in discrete-time domain. Through analysis and illustration based upon the convergence results of Chapter 2, a modification approach is presented to eliminate the initial invariant discrepancy. Although convergence of discrete-time iterative learning control schemes with direct coupling was verified and their effectiveness was demonstrated, those ILC schemes, generally called P-type or D-type iterative learning control, are basic and preliminary. Hence their applicable scopes are rather limited. In Chapter 4, a robust nonlinear learning control scheme is developed. The underlying idea of the new nonlinear learning control is to apply Lyapunov 's direct method such that global asymptotic convergence can be achieved for more general classes of nonlinear dynamic systems. The iterative learning control and robust control are made to function in a complementary manner. Learning control redounds to learning and eliminating state-independent uncertainties as much as possible while robust control contributes to ensuring global stability of the control system.