[Paper Review] Linear-Quadratic-Gaussian Mixed Mean-field Games with Heterogeneous Input Constraints

Ying Hu, Jianhui Huang+1 more|arXiv (Cornell University)|Oct 9, 2017

Stochastic processes and financial applications24 references3 citations

TL;DR

This paper develops a linear-quadratic-Gaussian mean-field game framework with a major agent and heterogeneous minor agents, where individual controls are constrained in closed convex sets. It establishes the well-posedness of the consistency condition system via forward-backward stochastic differential equations with projection operators, and proves the existence of an ε-Nash equilibrium using contraction mapping and discounting methods under local and global conditions.

ABSTRACT

We consider a class of linear-quadratic-Gaussian mean-field games with a major agent and considerable heterogeneous minor agents in the presence of mean-field interactions. The individual admissible controls are constrained in closed convex subsets $Γ_{k}$ of $\mathbb{R}^{m}.$ The decentralized strategies for individual agents and consistency condition system are represented in an unified manner through a class of mean-field forward-backward stochastic differential equations involving projection operators on $Γ_{k}$. The well-posedness of consistency system is established in both the local and global cases by the contraction mapping and discounting method respectively. Related $\varepsilon-$Nash equilibrium property is also verified.

Motivation & Objective

• To model large-population stochastic systems with a major agent exerting significant influence on minor agents.
• To address control constraints in minor agents by modeling individual admissible controls within closed convex subsets Γk ⊂ ℝm.
• To develop a decentralized strategy design method based on local information and mass-effect quantities.
• To establish the well-posedness of the consistency condition system for the mean-field equilibrium in both local and global settings.
• To verify the ε-Nash equilibrium property of the derived strategies, ensuring approximate optimality in large-population settings.

Proposed method

• Formulates the problem using a forward-backward stochastic differential equation (FBSDE) system with projection operators onto individual control constraint sets Γk.
• Introduces a two-step mapping: first, solve the forward SDE with given mean-field terms; second, solve the backward SDE using the projected control strategy.
• Applies the contraction mapping principle under a weighted norm ‖·‖λ to prove local well-posedness of the consistency system.
• Employs a discounting method to establish global well-posedness by adjusting the exponential weight λ to ensure contraction.
• Uses projection operators to enforce heterogeneous input constraints, ensuring admissible controls remain within Γk for each minor agent.
• Derives the consistency condition by requiring that the empirical distribution of minor agents' states matches the mean-field limit generated by the equilibrium strategy.

Experimental results

Research questions

• RQ1 How can a mean-field game model be extended to include a major agent with significant influence on a large number of minor agents with heterogeneous control constraints?
• RQ2 What conditions ensure the existence and uniqueness of a solution to the consistency condition system in such a mixed mean-field game?
• RQ3 How can decentralized strategies be constructed using only local state information and mass-effect terms, while respecting individual control constraints?
• RQ4 What mathematical techniques can be used to prove the well-posedness of the forward-backward stochastic differential equation system under heterogeneous constraints?
• RQ5 Does the derived strategy profile constitute an ε-Nash equilibrium, and under what conditions is this approximation valid?

Key findings

• The consistency condition system is well-posed in the local case via the contraction mapping principle applied to a weighted norm space.
• Global well-posedness of the consistency system is established using a discounting method, ensuring existence under broader parameter conditions.
• The solution to the FBSDE system with projection operators characterizes the decentralized strategies for all minor agents and the major agent.
• The derived strategies form an ε-Nash equilibrium, with the approximation error ε vanishing as the number of minor agents tends to infinity.
• The convergence rate of the ε-Nash equilibrium is quantified through exponential decay terms involving parameters such as k11, k12, k4, k5, and the time horizon T.
• The proof relies on bounding the difference of solutions via weighted norms, with sufficient conditions on λ, K1, K2, K3, K4 ensuring contraction.