Applied Research Of Adaptive Iterative Learning Control Based On Characteristic Model

With the expansion of mechanical large-scale production, plants in industrial become more complex, the requirement of control accuracy becomes higher. Import issues, such as how to model and control plants, stand out with the preceding trends. As one of the important members in industrial production, the manipulator, is a MIMO and highly coupled non-line system. There exists a lot of uncertainties in the practical application, such as random disturbance, load change, which lead to model the system difficultly. Characteristic model, which is built upon input, output and requirements of the target plant, can efficiently model manipulators. Iterative learning control and adaptive control both have the characteristic of learning and without depending on accurate models in the control, and thus provide a solution to the problem of manipulators control. Discussion and study of plants modeling and control are of real significance in the industrial production.Based on the theory of characteristic model, this thesis combines with iterative learning control and adaptive control, and designs two kinds of adaptive learning controllers to control manipulator and inverted pendulum systems. At the same time, aiming at the manipulator system with repeated interference, an adaptive controller is designed with double iterative learning algorithm. The main contributions of this work are listed as follows.1. This thesis introduces the theory of characteristic model and proposes a method of characteristic modeling for nonlinear systems. Based on this theory, this article deduces characteristic models of single-link manipulator and double links manipulator. Theoretical analyses and results of simulation show that when the sampling time interval is small enough, outputs of real systems and characteristic models are equivalent. Such results verify the characteristic model’s effectiveness.2. The Least squares iterative identification algorithm in this article is developed and improved, which employs variant forgetting factors. Simulation shows that the improved algorithm has a better accuracy of identification.3. An optimal adaptive learning controller based on performance indicators and a simple adaptive learning controller based on perfect tracking are designed in this work against characteristic models. Simulation for the manipulator and inverted pendulum systems show that both of them work well and can reach satisfactory results. Compared with the optimal adaptive learning controller, the sample adaptive learning controller designed in this article has better tracking capabilities and smoother inputs.4. An adaptive double iterative learning controller is designed in this article against the system which has repeated interference. The proof of the convergence is provided. From the simulation results, we can conclude that this controller is robust and can obtain the desired trajectory with the increase of iteration number.