Terminal Iterative Learning Control for Non uniformly Distributed Pass Points

In industrial application often the challenge consists in obtaining an adequate model to be able to design a good tracking controller. This problem can be solved for unknown system with a repetitive task or movement by using an iterative learning controller. One essential condition to the system is to ensure that the system under control is at least locally Lipschitz to be able to guarantee convergence. Some of the additional requirements (e.g. the same starting position for each iteration,...) of this method has been lowered in the recent years to extend the application field of the method. Additionally the use of cost function has become popular even for iterative learning controllers, which offers new possibilities to reduce the needed number of samples in each iteration. The basis of this new class of iterative controllers is the norm optimal ILC. The aim of this paper is to present the potential of such methods, to demonstrate the actual drawbacks in the state of the art control method and to show how to cope with it. The presented method is based on the norm optimal ILC and instead of using all sampling points as the state of the art ILC, the number of needed sampling points has been reduced to the necessary minimum to fulfill the requirements.