Driving Reinforcement Learning with Models

TL;DR

This paper introduces MPRL, a novel algorithm that combines reinforcement learning with model predictive control to leverage their strengths, demonstrated by outperforming individual methods in the Atari Pong game.

Contribution

The paper presents a new algorithm, MPRL, integrating RL and MPC in a way that enhances performance over each method alone.

Findings

MPRL outperforms standalone RL and MPC in Atari Pong

The combined approach leverages strengths of both RL and MPC

Demonstrates effectiveness of hybrid control strategies

Abstract

In this paper we propose a new approach to complement reinforcement learning (RL) with model-based control (in particular, Model Predictive Control - MPC). We introduce an algorithm, the MPC augmented RL (MPRL) that combines RL and MPC in a novel way so that they can augment each other's strengths. We demonstrate the effectiveness of the MPRL by letting it play against the Atari game Pong. For this task, the results highlight how MPRL is able to outperform both RL and MPC when these are used individually.