A Receding Horizon Method for the Constrained Control of Insulin Infusion System

The paper deals with the problem of regulation of blood glucose for type I diabetic patients via continuous infusion of insulin. To this end, a receding horizon approach based on a nonquadratic optimization criterion is developed. By this approach the physical (linear) constraints on both the control and controlled variables as well as of their rate of variations can be easily incorporated in the overall optimization problem. AnMPC-like approach based on a receding horizon control strategy, along with a model reference technique is used to achieve input adaptation to disturbances and a smooth trajectory of the controlled variable. The control variable value at each time-period is determined by means of the solution of a linear optimization problem. The algorithm allows for the direct consideration of constraints and thus their exploitation to the maximum without having to use weighting of control and output variables to achieve feasible and acceptable solutions of the optimization problem