A Single Online Agent Can Efficiently Learn Mean Field Games

TL;DR

This paper introduces a novel online, model-free learning approach for a single agent to efficiently find mean field Nash equilibria in large-population systems, avoiding complex iterative methods.

Contribution

It proposes the first online single-agent, model-free algorithms for learning mean field equilibria without prior knowledge of the environment.

Findings

The algorithms effectively approximate fixed-point iteration.

Sample complexity guarantees are established.

Numerical experiments confirm the methods' efficiency.

Abstract

Mean field games (MFGs) are a promising framework for modeling the behavior of large-population systems. However, solving MFGs can be challenging due to the coupling of forward population evolution and backward agent dynamics. Typically, obtaining mean field Nash equilibria (MFNE) involves an iterative approach where the forward and backward processes are solved alternately, known as fixed-point iteration (FPI). This method requires fully observed population propagation and agent dynamics over the entire spatial domain, which could be impractical in some real-world scenarios. To overcome this limitation, this paper introduces a novel online single-agent model-free learning scheme, which enables a single agent to learn MFNE using online samples, without prior knowledge of the state-action space, reward function, or transition dynamics. Specifically, the agent updates its policy through the value function (Q), while simultaneously evaluating the mean field state (M), using the same batch of observations. We develop two variants of this learning scheme: off-policy and on-policy QM iteration. We prove that they efficiently approximate FPI, and a sample complexity guarantee is provided. The efficacy of our methods is confirmed by numerical experiments.