Averaging principle for multiscale controlled jump diffusions and associated nonlocal HJB equations

Abstract: This paper studies the averaging principle of a class of two-time scale stochastic control systems with $\alpha$-stable noise. The associated singular perturbations problem for nonlocal Hamilton-Jacobi-Bellman (HJB) equations is also considered. We construct the effective stochastic control problem and the associated effective HJB equation for the original multiscale stochastic control problem by averaging over the ergodic measure of the fast component. We prove the convergence of the value function using two methods. With the probabilistic method, we show that the value function of the original multiscale stochastic control system converges to the value function of the effective stochastic control system by proving the weak averaging principle of the controlled jump diffusions. Using the PDE method, we study the value function as the viscosity solution of the associated nonlocal HJB equation with singular perturbations. We then prove the convergence using the perturbed test function method.