Contents 1. INTRODUCTION
Since the early results from [1], iterative learning control (lLC) has been extensively studied for about three decades. ILC has revealed a powerful capability in guiding dynamics systems without relying upon a full dynamic model [3]. Specifically, it is a practically beneficial algorithm in obtaining control signals for a precise motion planning with a partial knowledge of plants [4]. At the cost of these advantages, however, ILC has some fundamental requirements in its problem setup [2]. One of them is the initial reset condition of the states of the system; that is, at every iteration, ILC systems should begin at the same states. If the final states of the previous iteration are different from the fixed initial states, then the final states should be reset to the initial states every iteration. Even though, in many engineering applications that have repetitive dynamic behaviors, this initial reset condition may be not a serious issue [5], it still imposes a fundamental limitation in real-time applications; in particular, when the dynamics operates without stopping along the continuous time domain. Thus, though ILC has been extensively and widely researched thus far, in terms of real-time applications, it needs a further development, which is heavily related to the initial reset condition.
