Improved postprandial glucose control with a customized model predictive controller

Meal compensation in blood glucose control of people affected by type 1 Diabetes is an open challenge. The proposed Model Predictive Controller (MPC) is equipped with an asymmetric quadratic cost function, postprandial (pp) input integral and pp output soft constraints. The controller is synthesized with a linear glucose-insulin model customized on the basis of the patient clinical knowledge. An in-silico study on 100 adult virtual patients of the UVA/Padova simulator is performed to evaluate the achieved controller performance. This is compared with the performance obtained with a previously developed MPC. The evaluation is performed in perturbed scenario in which the controller is not aware of random variations of insulin sensitivity in each virtual patient. The proposed controller is shown to be able to significantly increase the average control performance and to reduce both hyper- and hypoglycemia phenomena. A nonlinear version of the proposed MPC, whose performance is evaluated in a nominal scenario, is also considered as ideal reference.