Research On Iterative Learning Control Methods Based On Iteration-domain

Iterative Learning Control (ILC) is a new intelligent control method developed in recent twenty years, whose applicative object is the plant with repeated characteristic. Its basic idea is to improve the current control signal by using the tracking output error of the system and the previous control experience, and the objective is to make the output perfectly track the desired trajectory. In practice, the non-repetitiveness always exists in the system, it contains the iteration-variant disturbances, iteration-variant uncertainties and the iteration-variant desired trajectories. The baseline performance of ILC is limited mainly by the non-repetitiveness factors, because the traditional iterative learning algorithm cannot harness them well. So, the non-repetitiveness factors are harnessed by using new update laws.The basic knowledge of iterative learning control is introduced, and its development process and research status is deeply analyzed. The non-repetitiveness problems are considered in ILC. The feedback controller is designed by using a lifting technique and an iteration-domain complex frequency operator for discrete repetitive processes subjected to iteration-varying reference signals, where, iteration-domain complex frequency operator was given by K. L. Moore. Then, an internal model principle is given that ensures perfect tracking along-the-pass and asymptotically convergent tracking along the iteration axis. For the iteration-variant disturbance with a known repeating pattern is completely rejected by using internal model principle.When the iteration-variant disturbance is with an unknown repeating pattern, it is accommodated by borrowing the idea of adaptive feedforward compensation. For the iteration varying model uncertainty, an algorithm of analysis and design based on 2-D system theory is given. The appropriate gain matrixs are designed and chose by using interval concept and idea of 2-D theory.MATLAB simulation experiments show that the internal model principle and the adaptive feedforward compensation are more effective than the traditional first-order iterative learning algorithm. And using the 2-D theory can make the iteration varying model uncertainty keep better stability and good tracking performance.