Brief Review of Reinforcement Learning Control for Cable-Driven Parallel Robots

Abstract

Reinforcement Learning (RL) is a branch of machine learning applied to many applications, such as mechatronics and robotics. RL allows more challenges to be resolved in robotics due to the high capacity for problem definition and active environmental interaction. Another notable property of RL is the capacity to include a variety of mathematical models, which the best-fitted model can be employed for solving any challenge. This characteristic of RL is highly beneficial for cable-driven parallel robots (CDPR) with applications in industry, construction and rehabilitation, in which the robot’s task space contains uncertainties. Recently many researchers have employed RL for various control tasks in CDPRs. Since RL has been applied to CDPRs’ different control challenges and satisfying results have been achieved, the future of RL in CDPR applications seems promising. In this paper, the applications of RL in CDPR control will be studied, and the most reported RL methods for these robots in the literature will be discussed. The promising future research subject will be described as well.