A Preliminary Study on the Relationship Between Iterative Learning Control and Reinforcement Learning ?

摘要

Iterative learning control is a control system design method that is able to achieve high tracking performance by repeatedly executing a task and learning the best input from previous attempts of performing the task. Reinforcement learning is a machine learning method that determines the best action such that some utility function (reward) is maximised by repeatedly interacting with the environment (system) and learning the best action policy based on the reward received from such interactions. These two methods belong to different subject disciplines but share a number of similarities. The relationship between these two design approaches, however, has not been investigated in detail. This paper presents a preliminary study on the relationship between iterative learning control and reinforcement learning, hopefully shedding some light on how these two areas can benefit each other in future research.