Insulin on Board safety constraint effect in a CHoKI-based MPC for Artificial Pancreas

This work presents a learning-based Model Predictive Control (MPC) algorithm for the artificial pancreas able to autonomously manage basal insulin injections in type 1 diabetic patients. The main goal is to keep the blood glucose levels within the euglycemic range (70-180 mg/dL), trying to avoid hypoglycemia. To prevent this event, additional constraints are added that consider the Insulin On Board (IOB). The data collection and the testing of the proposal are performed on the virtual patients of the FDA-accepted UVA/Padova simulator. The final results seem promising since the proposed controller reduces the time in hypoglycemia with respect to the standard constant basal insulin therapy.