Advanced model predictive control (MPC) for type I diabetic patient blood glucose control

Abstract
Advanced control strategies for delivering insulin to type I diabetic patients are developed. Utilizing the model predictive control (MPC) framework, an asymmetric objective function is tailored to address the inherent performance requirements of the physiological problem. This technique is then compared to an MPC controller incorporating prioritized objectives. In an in-patient setting, a second manipulated variable (glucose infusion) could be made available. This multiple-input single-output (MISO) problem is reformulated as a multiple-input multiple-output (MIMO) output regulator problem and is solved using the above techniques. Performance in all cases was demonstrably superior to previous results based on standard MPC formulations using symmetric objective functions.