Mean field games: numerical methods for the planning problem (Q2884587)

A finite difference semi-implicit scheme for Lasy-Lions mean field games and a parametrized `penalized' version thereof is introduced. By relating the resultant equations to a discrete control problem, the existence and uniqueness issues are analyzed using convex duality theory. For the penalized case, some key estimates that capture the parameter dependence are derived. A Newton iteration based strategy for solving the semi-implicit scheme in the penalized case is proposed and numerical experiments are given.