This paper studies asymptotic solvability of a linear quadratic (LQ) mean field social optimization problem with controlled diffusions and indefinite state and control weights. We employ a rescaling approach to derive a low dimensional Riccati ordinary differential equation (ODE) system, which characterizes a necessary and sufficient condition for asymptotic solvability. The decentralized strategies obtained from the mean field limit ensure a bounded optimality gap in minimizing the social cost of the magnitude O(N ), which translates to an O(1/N ) optimality loss per agent. We further quantify the efficiency gain of the social optimum with respect to the mean field game solution.
CONTENTS1. Introduction 1.1. Contributions and organization 1.2. Notation 2. State feedback for LQ social optimization 2.1. The formal derivation of the Riccati equation 3. Asymptotic solvability 3.1. Main result 3.2. Solvability of the limiting ODE system 3.3. Interpretation of the limiting Riccati ODEs 4. Closed loop dynamics and mean field limit 5. Decentralized control 5.1. Social cost under decentralized control 5.2. Upper bound of optimality gap 5.3. Performance comparison with the mean field game solution 6. Numerical examples 6.1. Asymptotic solvability 6.2. Performance 6.3. Comparison between social optimum and the mean field game 7. Conclusion Appendix A. Proof of Lemma 3.1 Appendix B. Proof of Lemmas 3.2 and 3.4 Appendix C. Mean field game ODEs References