Anotation:
A new approach to design of multistage dual controller for stochastic systems with unknown parameters is proposed. This approach ensures the solvability of the multistage optimisation problem by utilising the Partial certainty equivalence technique (PCE) accompanied with modification of the criterion. The PCE technique is based on approximation of the joint probability density of the state augmented with unknown parameters where the certainty equivalence property is enforced on the state part of the augmented state. The modification of criterion lies in enhancing the criterion with term evaluating weighted prediction error of augmented state vector. The properties of the proposed criterion and of the derived control law are discussed and a numerical example comparing this new controller with other well known approaches is presented.