Artificial pancreas under stable pulsatile Model Predictive Control: Including the physical activity effect

Type 1 diabetes management presents significant challenges due to the complex interplay between insulin delivery, carbohydrate intake, and physical activity. This paper introduces an enhanced artificial pancreas system based on stable pulsatile Zone Model Predictive Control (pZMPC) that explicitly addresses the metabolic effects of physical activity. The proposed approach integrates (i) a novel dynamical model that captures physical activity’s impact on glucose–insulin dynamics, (ii) the incorporation of a non-standard insulin-on-board constraint to prevent hypoglycemia, and (iii) the implementation of an asymmetric cost function that penalizes hypoglycemia more severely than hyperglycemia. The control strategy was evaluated using in silico trials with a virtual population of ten patients under various realistic scenarios, including meal and exercise announcement errors and circadian variations in insulin sensitivity.