Embedded Real-Time Model Predictive Control for Glucose Regulation

This paper reports our investigation on a model predictive control (MPC) with constraints for continuous diabetes management, and its implementation on the microcontroller of our artificial pancreas. The operational constraints for the MPC are rate of change, amplitude and output constraints, while the associated optimization problem is solved using a primal-dual interior-point algorithm based on predicator-corrector method. Our real-time implantable close-loop system is able to achieve desired diabetes management by maintaining the blood sugar level at less than 140 mg/dl, and consistently within the range of 70–120 mg/dl. Experimental results demonstrate that the low power 16-bit microcontroller in our prototype artificial pancreas can provide sufficient computational power with our computational efficient embedded system solution.