Measurement-driven event-triggered LQR control strategy for artificial pancreas systems with sustainable battery

Type-1 diabetes is a chronic condition that requires external insulin administration to regulate blood glucose (BG) levels. Hyperglycemia caused by Type-1 diabetes can cause serious health problems such as heart attack, stroke, kidney disease, and organ failure. An automated system for insulin delivery is a viable option due to its precision and fewer needle insertions. In modern artificial pancreatic (AP) systems, which rely on battery-powered components, extending battery life is crucial for continuous and reliable operation. Since frequent updates to the insulin infusion rate consume significant energy, reducing the number of control actions becomes essential. This paper proposes a novel event-triggered (ET) linear quadratic regulator (LQR) strategy for BG regulation using a simplified Bergman model. Unlike ET-model predictive control, which necessitates online optimization and incurs substantial computational costs, the proposed scheme employs a robust LQR framework. The control objective is to maintain BG within the safe physiological range of 70–180 mg/dL while minimizing the frequency of insulin delivery updates. Additionally, compared with existing ET-LQR and/or conventional schemes that are mostly dependent on full-state error-based triggering conditions, the proposed scheme is driven by three physiological events: initial hyperglycemia, significant deviations in BG, and hypoglycemia prevention. A state feedback LQR controller is designed accordingly, which is further investigated for ET-LQR design via an extended Lyapunov stability approach. The proposed optimal approach ensures a suboptimal LQR performance, which considers the realistic environment of meal perturbations and event-driven mechanism, relative to the optimal continuous-time LQR solution. Simulation results demonstrate that the proposed ET-LQR control approach achieves effective glucose regulation with substantially fewer controller activations compared to continuous-time LQR as well as conventional ET control methods, thus improving energy efficiency in embedded medical devices.