Adaptive continuous-time model predictive controller for implantable insulin delivery system in Type I diabetic patient

This paper presents a physiological model of glucose–insulin (GI) interaction and design of a Continuous‐time Model Predictive Controller (CMPC) to regulate the blood glucose (BG) level in Type I diabetes mellitus (TIDM) patients. For the designing of the CMPC, a nonlinear physiological model of TIDM patient is linearized as a ninth‐order state‐space model with an implanted insulin delivery device. A novel control approach based on Continuous‐time Model Predictive technique is proposed for the BG regulation with rejection of periodic or random meal and exercise disturbances in the process. To justify its efficacy a comparative analysis with Linear Quadratic Gaussian (LQG) control, and recently published control techniques like Proportional‐Integral‐Derivative (PID), Linear Quadratic Regulator with Loop Transfer Recovery (LQR/LTR) and H‐infinity has been established. The efficiency of the controller with respect to accuracy and robustness has been verified via simulation. The proposed controller performances are assessed in terms of ability to track a normoglycaemic set point of 81 mg/dl (4.5 mmol/l) in the presence of Gaussian and stochastic noise.