Multi-population mean field games systems with Neumann boundary conditions (Q2345420): Difference between revisions

The author studies multi-population stationary mean field game equations on bounded domains. This system of nonlinear elliptic equations arises from modeling large populations of small players aiming to minimize a long term average cost subject to randomness. The models were proposed separately by [\textit{J.-M. Lasry} et al., Jpn. J. Math. 2, No. 2, 229--260 (2007; Zbl 1156.91321)] and [\textit{M. Huang} et al., Commun. Inf. Syst. 6, No. 3, 221--251 (2006; Zbl 1136.91349)] who derive them from taking limits of Nash equilibria of games. The system involves Hamilton-Jacobi-Bellman equations \[ - \nu_i \Delta u_i(x) + H^i(x,Du_i(x)) + \lambda_i = V^i[m](x) \] giving value functions and Fokker-Plank-Kolmogorov equations \[ -\nu_i \Delta m_i - \text{div}\big(D_pH^i(x,Du_i(x)) m_i\big) = 0 \] giving the distributions of the populations. Here the cost stems from the velocities of the agents and the functions \(V^i\) which depend on the distribution of the populations. The randomness is reflected in \(\nu_i\) from a term \(\sqrt{2\nu_i} B_t\) in the dynamics. While based on numerics from [\textit{Y. Achdou} et al., SIAM J. Numer. Anal. 48, No. 3, 1136--1162 (2010; Zbl 1286.91022)], this article differs in its treatment of systems and Neumann boundary conditions. By assuming that \(\{V^i\}\) have a regularizing effect on \(\{m_i\}\), the author proves existence through a fixed point argument that associates \(\{u_i\}\) to \(\{m_i\}\). He approximates by regularizing \(\{V^i\}\) to show existence with weakened assumptions which requires a careful study of equations' structure to obtain appropriate bounds. While the author proposes a sufficient condition for unique, nonuniqueness is expected for systems as shown through examples.