Safe Stochastic Model-Based Policy Iteration with Chance Constraints

In this paper, we consider optimal control problems of stochastic discrete-time systems subject to additive disturbances. Safety of such systems is guaranteed in a probabilistic sense via chance constraints. We solve the corresponding chance constrained stochastic control problems by extending the unconstrained model-based Policy Iteration (PI), and thus chance constrained PI with safety guarantees is proposed. Additionally, the stability of generated control policies is analyzed in the mean square sense. Numerical simulations are provided to validate the proposed algorithm performance.