MPC-based artificial pancreas accounting for circadian variability of insulin sensitivity

ABUIN, PABLO; FERRAMOSCA, ANTONIO; GONZÁLEZ, ALEJANDRO HERNÁN

Marsella

Conferencia; 10th International Conference on Systems and Control (ICSC); 2022

In the context of closed-loop glycemic control, MPChas demonstrated the ability to improve glucose regulation inpatients with type 1 diabetes mellitus (T1DM) in silico. To reduceits complexity, many of the proposed control strategies have beendeveloped based on linear time invariant (LTI) models, withoutconsidering the well-know intraday parametric variability. In thiswork, a pulsatile Zone Model Predictive Control (pZMPC) isstudied under a realistic pattern of intraday insulin sensitivity(SI ), according to the recent updates of the FDA-approvedUVA/Padova simulator. Nominal updates of the postprandialinsulin sensitivity are explicitly included in the control-orientedmodel to improve the glucose predictions. The resulting controlleris tested in silico with the FDA-approved UVA/Padova simulator,while its performance is analyzed by comparing the usualstatistical metrics with the corresponding time invariant version.As expected, the results show significant improvements, whichjustifies the (reasonable) increment in the controller complexity.