Proof of Concept Control of a T1DM Model Using Robust Fixed-Point Transformations via Sliding Mode Differentiators

Abstract

Type 1 Diabetes Mellitus (T1DM) is a disease where insulin production is obstructed in the pancreas, and exogenous administration of the hormone must be utilized. Automatic control of the administration can be achieved using the Artificial Pancreas (AP) concept, whose performance is heavily reliant on the underlying control algorithm. A Robust Fixed-Point Transformations (RFPT)-based control strategy was designed to automate the insulin delivery process, which incorporates a Sliding Mode Differentiator (SMD) to provide higher order derivatives of the blood glucose level. Inter-patient variability, carbohydrate disturbances, and real-life sampling were included in the validation of the method. Results showed that the algorithm could regulate the blood glucose level, with a significant overshoot at the beginning of the control action due to the adaptive nature of the controller. Results indicate that the design requires additional modifications to be feasible in practice, including an extended validation with more virtual patients and realistic simulation settings in the future. Nevertheless, the current control algorithm has several attractive features, which are discussed with respect to PID and Model Predictive Control (MPC).