An inverse optimal problem in discrete-time stochastic control