Abstract

Despite continuous research effort, patients with type 1 diabetes mellitus (T1D) experience difficulties in daily adjustments of their blood glucose concentrations. New technological developments in the form of implanted intravenous infusion pumps and continuous blood glucose sensors might alleviate obstacles for the automatic adjustment of blood glucose concentration. These obstacles consist, for example, of large time-delays and insulin storage effects for the subcutaneous/interstitial route. Towards the goal of an artificial pancreas, we present a novel feedback controller approach that combines classical loop-shaping techniques with gain-scheduling and modern H ∞ -robust control approaches. A disturbance rejection design is proposed in discrete frequency domain based on the detailed model of the diabetic Göttingen minipig. The model is trimmed and linearised over a large operating range of blood glucose concentrations and insulin sensitivity values. Controller parameters are determined for each of these operating points. A discrete H ∞ loop-shaping compensator is designed to increase robustness of the artificial pancreas against general coprime factor uncertainty. The gain scheduled controller uses subcutaneous insulin injection as a control input and determines the controller input error from intravenous blood glucose concentration measurements, where parameter scheduling is achieved by an estimator of the insulin sensitivity parameter. Thus, only one controller stabilises a family of animal models. The controller is validated in silico with a total number of five Göttingen Minipig models, which were previously obtained by experimental identification procedures. Its performance is compared with an experimentally tested switching PI-controller.

Highlights

  • Diabetes mellitus collectively denotes a group of metabolic diseases with a worldwide increasing prevalence over the last decades

  • A subgroup of diabetes mellitus is type 1 diabetes mellitus (T1D), which is characterised by the destruction of pancreatic β-cells and a subsequent lack of the hormone insulin, which is important in the blood glucose regulation

  • A novel discrete blood glucose control design procedure was developed that is focused on disturbance rejection of meal uptake

Read more

Summary

Introduction

Diabetes mellitus collectively denotes a group of metabolic diseases with a worldwide increasing prevalence over the last decades. Diabetes might induce long term secondary diseases that may affect heart, vascular system, peripheral nervous system, and can lead to blindness, heart disease or limb amputation To this day, manual control has to be used for patients with T1D to reduce increased blood glucose levels to the normophysiological range. The glucose metabolic system is subject to external disturbance and internal dynamics, which affect the blood glucose level after insulin bolus application. Hernjak and Doyle [8] concluded that a simple PD controller structure is not able to stabilise the blood glucose level around a tight operating point as the nonlinear, time-varying metabolic system is subject to external disturbances. Gain scheduling of the discrete controllers is proposed on the measured blood glucose concentration and the estimated insulin sensitivity, guaranteeing an adaption to changing process conditions.

Göttingen Minipig Model
Nonlinear State Space Model
Model Structure
Controller Design
Model Linearisation
Discrete Controller Design Prerequisites
Disturbance Rejection Design
Initial Controller Gain for Disturbance Rejection
Integral Action
Lead–Lag Compensator
Experimental Animal Study
Model Identification
In Silico Feedback Control Study
Controller Implementation
Disturbance Rejection Performance
Findings
Conclusions and Discussion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call