In this paper, we propose and design an adaptive dual controller for automatic glucose control of diabetic patients. The results could be used in the development of an artificial pancreas, which, while as yet unavailable, must consist of three major components: an insulin delivery device or pump, a continuous glucose sensor, and a control algorithm linking insulin delivery to measured glucose concentration. For improved performance the system would also include feedforward information about food intake, physical activity and other blood glucose perturbing inputs. A linear time-varying autoregressive model with exogenous inputs is constructed to characterize the kinetics of both glucose-insulin and glucosecarbohydrate interaction. Combined with a Kalman-filter based estimation scheme for online estimation of the time-varying model coefficients, we design an adaptive dual control that both excites the glucose dynamic system sufficiently to accelerate the parameter estimation and cautiously tracks the desired glucose level. Performance evaluation of the adaptive dual controller is accomplished via simulations on virtual patients constructed using clinical data from five different patients with type-1 insulindeficient diabetes using continuous subcutaneous insulin infusion for diabetes management during observation. Simulation results show both smaller glucose excursions and a reduction in the number of hypoglycemic events for all but one of the five subjects.