Tutoring Reinforcement Learning via Feedback Control

TL;DR

This paper presents CTRL, a control-tutored reinforcement learning algorithm that improves learning efficiency by integrating control strategies, demonstrated on the inverted pendulum stabilization problem.

Contribution

The paper introduces a novel control-tutored reinforcement learning method that reduces learning time using limited system knowledge.

Findings

Enhanced learning speed with control tutoring

Effective stabilization of inverted pendulum

Reduced learning rate compared to traditional methods

Abstract

We introduce a control-tutored reinforcement learning (CTRL) algorithm. The idea is to enhance tabular learning algorithms by means of a control strategy with limited knowledge of the system model. By tutoring the learning process, the learning rate can be substantially reduced. We use the classical problem of stabilizing an inverted pendulum as a benchmark to numerically illustrate the advantages and disadvantages of the approach.