CHoKI-based MPC for blood glucose regulation in Artificial Pancreas with probabilistic constraints

This work presents a Model Predictive Control (MPC) algorithm for the Artificial Pancreas. In this work, we assume that an a-priori model is unknown and the Componentwise Holder Kinky Inference (CHoKI) data-based learning method is used to make glucose predictions. A stochastic formulation of the MPC with chance constraints is considered to have a less conservative controller. The data collection and the testing of the proposed controller are performed by exploiting the virtual patients of the FDA-accepted UVA/Padova simulator. The simulation results are quite satisfying since the time in hypoglycemia is reduced.