Performance Comparison of Blood Glucose Controllers for Diabetic Patients

Abstract

Diabetes remains a significant health problem that demands serious attention and costly to maintain in the western world and developing countries adopting western lifestyles and diets. Since the development of insulin in the 1920s, there have been myriad problems in developing suitable technology for optimal administration of correct dosage to maintain normoglycaemic state in both type 1 and type 2 diabetic patients. A promising direction is the development of artificial pancreas (AP), a control engineering approach that mimics the pharmacokinetic counterbalancing action of the pancreas in producing optimal insulin and glucagon for blood glucose regulation. However, the optimal controller design to properly handle postprandial disturbances has been a significant challenge in AP design. Although there is a plethora of controller design techniques in the literature, there is no generally agreed benchmark criteria for comparing these controllers. Therefore, in this paper, an experimental testbed where the popular control algorithms can be compared and their response can be studied is proposed. This was done by simulating a virtual patient using a well-designed mathematical model. Using the testbed, the performance of three controllers was studied and rich insight was gleaned from these controllers' behaviour on how they handled the postprandial disturbances.