General Linear-Quadratic Mean Field Stochastic Differential Game with Common Noise: A Direct Method

TL;DR

This paper develops a comprehensive approach to linear-quadratic mean field games with common noise, deriving Nash equilibria using direct and fixed-point methods, and demonstrates their equivalence and practical application.

Contribution

It introduces a unified framework for solving mean field games with common noise using both direct and fixed-point methods, including explicit Riccati equation solutions.

Findings

Decentralized strategies are asymptotically optimal.

The direct and fixed-point methods yield identical strategies.

Application to a production planning example demonstrates practicality.

Abstract

This paper investigates a class of general linear-quadratic mean field games with common noise, where the diffusion terms of the system contain the state variables, control variables, and the average state terms. We solve the problem using both the direct method and the fixed-point method in the paper. First, by using the variational method to solve a finite $N$-players game problem, we obtain the necessary and sufficient conditions for the centralized open-loop Nash equilibrium strategy. Subsequently, by employing the decoupling technique, we derive the state feedback representation of the centralized open-loop Nash equilibrium strategy in terms of Riccati equations. Next, by studying the asymptotic solvability of the Riccati equations, we construct the decentralized open-loop asymptotic Nash equilibrium strategies. Finally, through some estimates, we prove the asymptotic optimality of the decentralized open-loop Nash equilibrium strategies. Moreover, we find that the decentralized open-loop asymptotic Nash equilibrium strategies obtained via the fixed-point method are identical to those derived using the direct method. Finally, we apply the main results of this paper to a concrete production planning example.