Study On Iterative Learning Control For Several Classes Of Systems

Iterative learning control is a simple, but effective control technique, which is applied to the system that operates repetitively over a fixed time interval and improves its transient response performance. The idea of ILC is to gradually revise imperfect control input using the error between system output and the desired trajectory and realize perfect tracking in a finite time interval.In this dissertation, the convergence and robust issues of iterative learning control in response to different initial conditions for several plants are mainly studied。For uncertain linear and nonlinear continuous and discrete systems, the effects of the plant characteristics, disturbances and noises, initial conditions, time delays, uncertain modeling dynamic and learning algorithms on the convergence and performance of ILC are also investigated. A series of ILC laws and sufficient conditions guaranteeing the convergence of ILC are proposed, and the effectiveness of the proposed learning laws is ensured by the theoretical analysis and illustrated by the simulation examples.The main contributions of this dissertation are summarized as follows:(1) We study two cases of initial condition problems: 1). the initial states are the same but different from the desired initial states, 2). random initial states. For a class of linear systems, an open-closed-loop PID-type law is proposed. We analyze its robustness and convergence for two cases of initial conditions and present the convergence condition. The simulation results demonstrate Open-closed-loop PID-type algorithm is better than the Open-loop PID-type algorithm in terms of the convergent speed.(2) We consider a class of nonlinear multiple time-delay dynamic systems with external disturbance and output noise, and present PD-type ILC learning algorithm. The convergence and robustness issues are addressed in the means ofλ? norm theory and Bellman-Gronwall lemma. Meanwhile, the effect of the external disturbances and measurement noises on the convergence of tracking error and the effect of multiple time delays on ILC convergence are considered when there is no initial error。We propose the sufficient conditions to guarantee the system outputs, states and control inputs to converge to desired trajectories with bounded tracking errors. In the case when the external disturbances and measurement noises decay to zero, the bounds of the tracking errors also decay to zero。It is shown that the multiple time delay in state variables do not affect the ILC convergence property significantly. The simulation example indicates the effectiveness of the proposed ILC laws。(3) we design a high-order P-type iterative learning algorithm for improving the tracking performance and initial state learning scheme in order to relaxing the restriction on initial state for a class of linear discrete-time systems with multiple time delay whose initial condition is unknown at each iteration. At the same time, we derive the sufficient and necessary condition which guarantees asymptotic convergence of the actual output of system to the desired trajectory, and we also address the effect of multiple time delays on iterative learning process. The simulated results have shown that the proposed ILC scheme has a better tracking performance.(4) Several simulations have been made according to the proposed algorithms in this dissertation. These simulation results illustrate the feasibility and effectiveness of the proposed iterative learning controller.