Strategically Efficient Exploration in Competitive Multi-agent Reinforcement Learning

TL;DR

This paper investigates exploration strategies in multi-agent reinforcement learning, revealing limitations of optimistic exploration in zero-sum games and proposing new algorithms for more sample-efficient learning.

Contribution

It introduces the concept of strategically efficient exploration in Markov games and develops algorithms that outperform optimistic methods in sample efficiency.

Findings

Optimistic exploration can waste samples in zero-sum games.

Proposed algorithms are more sample-efficient than optimistic exploration.

Strategically efficient exploration reduces irrelevant sampling.

Abstract

High sample complexity remains a barrier to the application of reinforcement learning (RL), particularly in multi-agent systems. A large body of work has demonstrated that exploration mechanisms based on the principle of optimism under uncertainty can significantly improve the sample efficiency of RL in single agent tasks. This work seeks to understand the role of optimistic exploration in non-cooperative multi-agent settings. We will show that, in zero-sum games, optimistic exploration can cause the learner to waste time sampling parts of the state space that are irrelevant to strategic play, as they can only be reached through cooperation between both players. To address this issue, we introduce a formal notion of strategically efficient exploration in Markov games, and use this to develop two strategically efficient learning algorithms for finite Markov games. We demonstrate that these methods can be significantly more sample efficient than their optimistic counterparts.