MAMBPO: Sample-efficient multi-robot reinforcement learning using learned world models

TL;DR

This paper introduces MAMBPO, a multi-agent model-based reinforcement learning algorithm that significantly improves sample efficiency in multi-robot systems by using learned world models within a centralized training and decentralized execution framework.

Contribution

The paper presents a novel multi-agent model-based RL algorithm, MAMBPO, that enhances sample efficiency over existing model-free methods like MASAC using learned world models.

Findings

MAMBPO achieves similar performance to MASAC with fewer samples.

MAMBPO demonstrates effective learning in simulated multi-robot tasks.

The approach advances the feasibility of real-world multi-robot learning.

Abstract

Multi-robot systems can benefit from reinforcement learning (RL) algorithms that learn behaviours in a small number of trials, a property known as sample efficiency. This research thus investigates the use of learned world models to improve sample efficiency. We present a novel multi-agent model-based RL algorithm: Multi-Agent Model-Based Policy Optimization (MAMBPO), utilizing the Centralized Learning for Decentralized Execution (CLDE) framework. CLDE algorithms allow a group of agents to act in a fully decentralized manner after training. This is a desirable property for many systems comprising of multiple robots. MAMBPO uses a learned world model to improve sample efficiency compared to model-free Multi-Agent Soft Actor-Critic (MASAC). We demonstrate this on two simulated multi-robot tasks, where MAMBPO achieves a similar performance to MASAC, but requires far fewer samples to do so. Through this, we take an important step towards making real-life learning for multi-robot systems possible.