Data-Driven Disturbance Estimation and Control With Application to Blood Glucose Regulation

Abstract

A data-driven control approach for nonlinear systems is proposed, called data-driven estimation and control (D2EC), which combines a disturbance estimator and a nonlinear control algorithm. The estimator provides a signal representing the unknown disturbances affecting the plant to control. This signal is used by the control algorithm to improve its performance. A real-data study is presented, concerned with the regulation of blood glucose concentration in type 1 diabetic patients. Preliminary tests of the D2EC approach are also carried out using a diabetic patient simulator, obtained from a revised version of the well-known University of Virginia/Padova model. Both the real-data and the simulator-based studies indicate that the proposed approach has the potential to become an effective tool in the context of diabetes treatment and, more in general, in the biomedical field, where accurate first-principle models can seldom be found and relevant disturbances are present.