Accelerate Reinforcement Learning with PID Controllers in the Pendulum Simulations

TL;DR

This paper introduces a PID controller-based coaching scheme to accelerate reinforcement learning in pendulum simulations, aiming to improve learning efficiency and convergence speed.

Contribution

The novel contribution is integrating PID controllers as a coaching mechanism to enhance reinforcement learning performance in pendulum tasks.

Findings

PID coaching significantly speeds up RL training.

Improved convergence rates observed in simulations.

Enhanced stability during learning process.

Abstract

We propose a Proportional Integral Derivative (PID) controller-based coaching scheme to expedite reinforcement learning (RL).