Control of Diabetes Mellitus by Advanced Robust Control Solution

Abstract

The current article investigates the possibilities of designing a robust control solution for a control problem related to type 1 diabetes mellitus. The proposed control methodology exploits linear parameter varying, linear matrix inequality, tensor product model transformation and extended Kalman filtering, four advanced control methods in order to guarantee hard safety control constraints for diabetic patients. In this research we have applied an extension of the minimal model to simulate the glucose-insulin dynamics and the glucose and insulin absorption of a diabetic patient. We have validated our results on numerical simulations by using realistic patient data. During the evaluation we have used randomized glucose intakes both in time and amount (as 'unfavorable' disturbance signals). Furthermore, we did a long-term (30 days) assessment to confirm the stability of the proposed framework. The results have shown that the developed controller effectively intervenes into the process and provides appropriate control action by avoiding hypoglycemia thereby satisfying the predefined quantity and quality requirements.