Reinforcement Learning-Based Disturbance Rejection Control

Abstract

We address the disturbance suppression control using reinforcement learning. The control object handled in this paper is a pseudo-simulator of a chemical plant, which is subject to the influence of disturbances. The traditional control method for chemical plants, namely PID control, generally does not exhibit high control performance under disturbances, and various control methods have been proposed to address the disturbance suppression. In this paper, we aim to suppress disturbances in chemical plants using reinforcement learning. Reinforcement learning is a method of learning control inputs based on feedback data of control inputs and outputs. By using reinforcement learning to pre-learn input-output data under disturbances, it is expected to learn a controller with high disturbance suppression capability.