Glucose Level Control for Type 1 Diabetes Patients Using Unknown Input Observers

Diabetes is a serious disease during which the body's production and use of insulin is impaired, which causes deviations of the glucose concentration level in the bloodstream from normal values. The most common treatment of diabetes type 1 (patients with defects in insulin production) is the measurement of the blood glucose level using suitable measurement devices and to regulate this level with an infusion of insulin. Advanced solutions are trying to apply continuous automatic feedback control for this process using glucose level sensors and insulin infusion pumps, forming a so called Artificial Pancreas. Currently, first solutions are in the stage of clinical experiments but all available solutions are far from being optimal. Therefore, the project DIABOB will contribute to the solution of the measurement problem as one of the major unsolved problems in the development of an artificial pancreas solution.

Fig. 1 Artificial pancreas (source:

The human body is a system very difficult to represent due to the large number of parameters, variables, constraints and disturbances or with inaccessible/unmeasurable inputs and often a state estimation and parameter identification will be necessary. For this reason the state observers to estimate these hardly measurable variables, play a very key role in the closed-loop control task. An observer is used to obtain an estimation of the full vector of state variables, using the available measurements of the system. Herein, we will apply different mathematical approaches to derive the unknown input observer such as the application of the nonlinear observer normal form concept.
In our project will apply an observer-based approach, based on extend mathematical models of the nonlinear insulin-glucose interaction such as Bergman’s minimal model with a suitable model of the subcutaneous measurement. We will apply an unknown input observer where the full vector of state variables can be estimated even if the glucose absorption from meals as the main disturbance is not measurable, with the subcutaneous in-tissue glucose as the only input.

Currently, the most promising control solutions in the artificial pancreas approach are controllers which are based on a dynamic state variable model of the glucose-insulin interaction in the human body and assume: (I) that continuous blood glucose measurements are available and exact, (II) that the full vector of state variables is available for a feedback and (III) that all disturbances such as glucose intake from meals are known. However, all these assumptions are not fulfilled in practice.

Fig. 2 Closed loop approach for Artificial Pancreas.

Another obstacle to satisfactory closed-loop control is the presence of significant disturbances (i.e., meals and physical activity). Moreover, the control system must satisfy constraints on both plasma glucose levels and insulin delivery rates.

In the DIABOB project, an obtained unknown input observer will be combined with a model predictive control approach for a first test in an advanced virtual diabetes patient simulator available at University of Luxembourg. Furthermore, patient data will be obtained with the help of the hospital in Ettelbruck as the national clinical partner.


Researchers: Dr. Adriana Aguilera-Gonzalez, Prof. Dr. Holger VoosProf. Dr. Mohamed DarouachDr. Jochen G. Schneider, Dr. Nicolas Knauf.