Large deviation limit for discrete-time, totally observed stochastic control problems with multiplicative cost (Q5961572)

The authors investigate some relation between risk sensitive stochastic control problems and dynamic games for discrete time, using large deviation methods. Considering controlled Markov kernels, they determine a large class of controlled stochastic processes, for which the high risk small noise limit turns out to be a dynamic game, and show how the dynamics and the cost function of the game are related to large deviation properties of controlled Markov kernels.