Artificial Pancreas under stable MPC: including the Physical Activity effect

Abstract: This work presents the application of Model Predictive Control (MPC) for the control of blood glucose concentrationin patients with Type 1 Diabetes Mellitus (T1DM). The main novelty of the algorithm is the inclusion of physical activity effects, i.e. one of the main critical aspects of the disease. The glycemic control is based on a physiological individualized long-term compartmental model. Leveraging on this model, the algorithm is able to control glycemia by performing prediction and estimation of patient’s next states, employing disturbance observers to compensate plant-model mismatches. The overall performance is tested in a cohort of in-silico patients from the FDA-approved UVA/Padova simulator.