Adaptive Model Predictive Control with Particle Filter for Artificial Pancreas

Controller takes an important role for artificial pancreas (AP) to regulate insulin infusion rate according variable requirements of diabetic patients. In this research, an adaptive model predictive control (MPC) algorithm is proposed to overcome the parameter uncertainty induce by inter and intra variability. Firstly, a glucose-insulin dynamic model is established to describe the integrated metabolism of glucose and insulin, in which the time-varying parameters can be extended to observable state variables. Then, particle filtering technology is introduced to track and adjust the parameters. Meanwhile, the glucose and insulin concentration in plasma (PGC and PIC) are also estimated. Finally, imbedding the dynamic model with personalized parameters, an adaptive MPC algorithm is proposed based on the estimated PIC and PGC. For validation, the in-silico experiments are carried out on the 30 virtual patients of the UVa/Padova simulator. The proposed algorithm shows promising performances. It shows that the proposed method has the potential for artificial pancreas in clinical treatment.