Decentralized ϵ$$ \epsilon $$‐Nash strategy for linear quadratic mean field games using a successive approximation approach