Abstract
Iterative learning control (ILC) obtains a desired input that exactly generates a desired output through repetitions of the similar tasks. The adjoint-type ILC, which is based on the gradient method, can obtain the bounded desired input for a finite dimensional and non-minimum phase system. This aspect suggests that the adjoint-type ILC can achieve good output tracking for a flexible arm, which is a infinite dimensional and non-minimum phase system. In this paper, we propose an update law and give a convergence condition for the adjoint-type ILC applied to a single-link flexible arm. In order to verify the effectiveness of the adjoint-type ILC for the single-link flexible arm, an experiment is carried out. Several researchers have proposed a iterative learning controller for the flexible arm by using assumed modes method. In this paper, a simple model of the single-link flexible arm is used because ILC dose not require the accurate model. Experiments result show that, even if the single-link flexible arm is modeled by the simple method and has some uncertainties, good output tracking is achieved. Moreover, pre-actuation, which is remarkable aspect of the adjoint-type ILC, is observed.
