Feedforward compensation of exercise in diabetes

A nonlinear feedforward action based on the detection of exercise intensity is designed to prevent drastic blood glucose drops that may lead a diabetic patient to hypoglycemia. A physiological model is derived from the Bergman and the Sorensen models and the metabolic effects of physical activity. A statistical model for blood glucose rate is obtained through a discrete Wiener modeling technique. The statistical model is inverted for the calculation of an exogenous glucagon input that can cancel the blood glucose drop caused by exercise. Simulations show that the proposed feedforward control system satisfies the objective. Algorithm improvements as well as implementation with current commercial products are discussed.