Penalized stochastic optimal control problems