Integrating economic model predictive control and event-triggered control: Application to bi-hormonal artificial pancreas system

Abstract

Type 1 diabetes mellitus is a lifelong disease with very high morbidity, and its medical expenses are also very alarming. It not only seriously threatens the health of patients, but also brings heavy financial burden to their families. It is necessary to seek effective and low-cost diabetes treatment therapies. Currently, the existing artificial pancreas (AP) systems emphasize only on the blood glucose (BG) control performance. This paper attempts to reduce both the economic cost and the computational burden while guaranteeing the control performance in a bi-hormonal AP system. Since economic model predictive control (EMPC) could improve the economic profit and event-triggered control could save computational resources, they are integrated and then the integrated method is applied to the bi-hormonal AP. The proposed method achieves the desired objective: 1) it guarantees the control performance, i.e., BG is maintained within 70-180mg/dL, and the average tracking error and BG risk index are small enough (21.23 and 1.59); 2) it significantly reduces the economic cost, where the total price of the hormone is reduced by 67.15% compared with the switching PID and 56.22% compared with the switching MPC; and 3) it substantially improves the computational efficiency, where the running time is reduced by 51.14% compared with the switching EMPC. It also performs well in robustness tests. Future studies will involve the clinical evaluation of the proposed scheme.