Research On Finite-Time Iterative Learning Control Method

Iterative learning control is particularly suited to complete control tasks with repeated movement characteristics, and it can achieve tracking the desired trajectory completely. But for each iteration, the limitation on the initial state is strict that it demands that initial position of the system must be located at the starting point of the desired trajectory. Because of the limitation of actual repetition accuracy of systems, the initial repositioning error is inevitable. In the presence of arbitrary initial state errors, the improvement of the performance of completely tracking under the iterative learning control is a topic worth studying.Considering the case of arbitrary initial states, a strategy of iterative learning control with the initial corrective action is explored deeply in this thesis. Aiming at different systems and combined with different methods, it accomplished the practical completed tracking. The main work of this thesis are following:1. As for the case that the control gain is relevant to the system state, based on Lyapunov-Like method, it deals with the finite time iterative learning control problem of parametric uncertain nonlinear systems and non-parametric uncertain nonlinear systems, and the finite time iterative learning control method is extended to more general case that under arbitrary initial state that effective control, the completely tracking on a limited range [Δ, T], of all kinds of systems can be obtained.2. For general nonlinear systems, and combined with the method of limiting dynamic performance, a finite time iterative learning control method which ensures the dynamic performance is designed, and the PCT performance is achieved, whilst the dynamic performance of tracking error can be ensured in each iteration.3. For a class of non-parametric uncertain nonlinear systems, using the neural network algorithm’s ability of approximating to nonlinear function, and cooperating with the adaptive ILC’s advantage in dealing with time-varying parameters, a kind of finite time iterative learning control method, which is based on a RBF neural network, is studied. In the treatment of the nonparametric uncertainties, it does useful trials in promoting the finite time iterative learning control.4. For time-varying robot systems, combined with finite time iterative learning control strategy used in this thesis, the feasibility of realizing of PCT tracking performance under arbitrary initial state is studied. A kind of finite time iterative learning controller which can handle with unknown constant parameters and unknown time-varying parameters at the same time is designed, and the effectiveness of the proposed finite time iterative learning control method is validated from the perspective of the experiment.