Toward human-in-the-loop PID control based on CACLA reinforcement learning

Zhong, Junpei and Li, Yanan (2019) Toward human-in-the-loop PID control based on CACLA reinforcement learning. The 12th International Conference on Intelligent Robotics and Applications (ICIRA 2019), Shenyang, China, 8 - 11 August, 2019. Published in: Proceedings of the 12th International Conference on Intelligent Robotics and Applications. 11742 605-613. Springer ISBN 9783030275341

[img] PDF - Accepted Version
Restricted to SRO admin only until 2 August 2020.

Download (493kB)

Abstract

A self-tuning PID control strategy using a reinforcement
learning method, called CACLA (Continuous Actor-critic Learning Automata) is proposed in this paper with the example application of humanin-the-loop physical assistive control. An advantage of using reinforcement learning is that it can be done in an online manner. Moreover, since human is a time-variant system. The demonstration also shows that the reinforcement learning framework would be beneficial to give semi-supervision signal to reinforce the positive learning performance in any time-step.

Item Type: Conference Proceedings
Keywords: human-in-the-loop, reinforcement learning, adaptive control
Schools and Departments: School of Engineering and Informatics > Engineering and Design
Depositing User: Lucy Arnold
Date Deposited: 07 Aug 2019 09:46
Last Modified: 07 Aug 2019 09:46
URI: http://sro.sussex.ac.uk/id/eprint/85341

View download statistics for this item

📧 Request an update