Dynamic Target Zone Modulation for Physical Activity Management in Artificial Pancreas Using Pulsatile Zone Model Predictive Control

Physical activity (PA) presents a significant challenge for artificial pancreas (AP) systems in managing Type 1 Diabetes, primarily by elevating the risk of hypoglycemia both during and after exercise. This work introduces and evaluates a novel exercise-aware pulsatile zone model predictive control (pZMPC) strategy designed to enhance glycemic safety and control around PA; in this proposal, the controller proactively mitigates this risk by dynamically modulating glucose target zones, elevating the target in anticipation of, during, and following announced PA. The performance of this strategy was systematically evaluated through in silico simulations across a cohort of ten virtual adult patients, under various scenarios designed to mimic real-life events, and using two control-relevant metabolic models, one of which explicitly accounted for exercise effects as an uncontrollable input. Compared to a standard pZMPC controller without exercise-specific adaptations, the proposed pZMPC with dynamic target zone modulation demonstrated an improvement in glycemic safety, reducing the incidence and duration of exercise-induced hypoglycemia. Furthermore, the strategy successfully decreased overall glycemic variability and maintained robust control even in the presence of uncertainties in PA information.