Farahmand Bahareh, Dehghani Maryam, Vafamand Navid, Mirzaee Alireza, Boostani Reza, Pieper Jeffrey Kurt
School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran; Schulich School of Engineering, University of Calgary, Calgary, Alberta, Canada.
School of Electrical and Computer Engineering, Shiraz University, Shiraz, Iran.
ISA Trans. 2023 Feb;133:353-368. doi: 10.1016/j.isatra.2022.07.009. Epub 2022 Jul 13.
Recent advances in the artificial pancreas system provide an emerging treatment option for type 1 diabetes. The performance of the blood glucose regulation directly relies on the accuracy of the glucose-insulin modeling. Sorenson model involves the behavior of different organs and offers precise representation. However, the high complexity of such a model makes the controller design procedure a hard task. Therefore, the high-order nonlinear Sorensen model as a popular high-fidelity physiological model is opted in this paper to analyze the glucose-insulin interactions in great detail, and a new robust nonlinear approach to regulate the blood glucose concentration (BGC) in Type-I diabetic patients is proposed. Inspiring the backstepping technique, for designing an acceptable controller, the model is divided into three main subsystems such that in each subsystem, the virtual control input laws are obtained using both Lyapunov stability and input-to-state theorems. Since the measurement of the parameters in the glucose-insulin system is not accurate, parametric uncertainties are defined in the investigated model. Furthermore, owing to the fact that the only measurable state variable is blood glucose, the estimation of inaccessible state variables is an important issue that is properly considered by the unscented Kalman filter (UKF) estimator. The suggested approach is compared to H, robust H, and linear parameter-varying control approaches. The comparison results on 500 simulated patients imply a remarkable superiority of the proposed controller approach to the compared methods in terms of the BGC tracking and the algorithm robustness in the presence of food intake disturbance patterns.
人工胰腺系统的最新进展为1型糖尿病提供了一种新兴的治疗选择。血糖调节的性能直接依赖于葡萄糖-胰岛素建模的准确性。索伦森模型涉及不同器官的行为，并提供精确的表征。然而，这种模型的高度复杂性使得控制器设计过程成为一项艰巨的任务。因此，本文选择高阶非线性索伦森模型作为一种流行的高保真生理模型，以详细分析葡萄糖-胰岛素相互作用，并提出一种调节I型糖尿病患者血糖浓度（BGC）的新型鲁棒非线性方法。受反步法技术的启发，为了设计一个可接受的控制器，将模型分为三个主要子系统，使得在每个子系统中，使用李雅普诺夫稳定性和输入到状态定理获得虚拟控制输入律。由于葡萄糖-胰岛素系统中参数的测量不准确，在所研究的模型中定义了参数不确定性。此外，由于唯一可测量的状态变量是血糖，不可达状态变量的估计是一个重要问题，无迹卡尔曼滤波器（UKF）估计器对此进行了适当考虑。将所提出的方法与H、鲁棒H和线性参数变化控制方法进行了比较。对500名模拟患者的比较结果表明，在所提出的控制器方法在BGC跟踪和存在食物摄入干扰模式时的算法鲁棒性方面，相对于比较方法具有显著优势。
