McKean-Vlasov optimal control: the dynamic programming principle (Q2129699)

The paper is devoted to the problem of optimal control of McKean-Vlasov stochastic differential equations. The authors develope a general theory for the non-Markovian case of McKean-Vlasov stochastic control problem with common noise. They propose and investigate weak and strong formulations of the McKean-Vlasov optimal control problem and show their equivalence. The dynamic programming principle is established for weak and strong formulations as well as for strong formulation where the control is adapted to the common noise. Based on their results for the non-Markovian case, the authors obtain the dynamic programming principle for the Markovian control problem. Associated Hamilton-Jacobi-Bellman equations are derived for the common noise strong formulation and for the general strong formulation.