Dual control for stochastic systems with multiple uncertainties

Uncertainties commonly exist in stochastic systems, and they may seriously affect control quality of the stochastic system. In this paper, a novel dual control is proposed, where the uncertainties caused by uncontrolled excitations, model parameter fluctuations and system noises are all taken into consideration. First, an online outlier detection algorithm is proposed and integrated in the control scheme to find the outliers in the uncontrolled excitations. Second, based on the model with uncontrolled excitations, two cost functions showing the dual properties are used to derive the control signal: one cost function represents the system output tracking performance, the other one represents the unknown parameter estimation performance. The minimization of the two cost functions forms a double-objective optimization problem and is solved by a designed perturbation domain around the nominal value. Experiments are carried out, and the results show that the proposed controller bears better performance, especially in cases where outliers exist.