(118c) Performance Assessment and Modification of an Adaptive MPC for Closed-Loop Insulin Delivery Systems
MPC performance assessment is challenging because degradation in the closed-loop performance can arise due to model deficiencies, poor control design parameters, or inappropriate constraints. Complex nonlinear dynamical systems such as the metabolic processes in the human body are particularly challenging to control due to the time-varying characteristics of these systems and their nonlinear behavior, presence of stochastic and unknown disturbances, and uncertain time-varying delays.
An example of a system with noteworthy complexities that necessitates a safe and reliable controller integrated with a powerful control performance assessment and modification system is the regulation of glucose concentrations in people with type 1 diabetes (T1D). T1D is an autoimmune disease that destroys the insulin-producing beta-cells of the pancreas, and thus people with T1D must administer exogenous insulin to overcome the insulin deficiency and maintain the blood glucose concentration within a safe target range. The automation of insulin infusion through technologies that closed the loop between glucose sensing and insulin infusion pumps, termed the artificial pancreas (AP) system, is shown to improve glucose control and reduce the likelihood of developing diabetes related complications [9-10]. Despite the advancements in glycemic control algorithms, the complexity of the control problem in drug delivery has obstructed the reliable assessment of the closed-loop performance.
In this work, a controller performance assessment algorithm is developed to analyze the closed-loop behavior and modify the parameters of the control system used in automated insulin delivery. To this end, various performance indices are defined to quantitatively evaluate the controller efficacy in real-time. The controller assessment and modification module also incorporate on-line learning from historical operational data to anticipate impending disturbances and proactively counteract their effects.
To characterize the time-varying glucose-insulin dynamics, adaptive stable models identified through recursive subspace system identification techniques are integrated with physiological compartmental models . Assessing the performance of the recursively identified models is necessary to guarantee the model is able to provide accurate output predictions for use in model-based predictive control algorithms. So, we propose a performance assessment technique for the recursively identified models to check necessary key performance indexes (KPI) including the risk of under-predicting undesirably high glucose levels, risk of over-predicting dangerously low glucose levels, and mean-absolute-error between predicted and measured outputs.
Then, an adaptive model predictive control (MPC) algorithm is designed based on the recursively identified state space models with dynamic adjustments of constraints and objective function weights. The adaptive controller parameters, dynamic safety constraints and a feature extraction method that automatically detects the presence of known disturbances using qualitative descriptions of measured time-series data and modifies the constraints of the MPC empower the control system to effectively compute the optimal control action over diverse diurnal variations. Various KPIs are defined to evaluate the performance of the closed-loop system to modify the key controller parameters of the adaptive MPC if the poor performance is detected in real-time. An adaptive learning technique is also proposed to modify the controller set-point based on historical information. Historical data are used to deduce the probable times of unknown disturbances. This valuable information on daily life behaviors and habits can be used in the closed-loop insulin delivery systems to appropriately modify the controller set-point in advance for the anticipated periods of the disturbance effects.
 Garcia, Carlos E., David M. Prett, and Manfred Morari. "Model predictive control: theory and practiceâa survey." Automatica 25.3 (1989): 335-348.
 Rawlings, James Blake, and David Q. Mayne. Model predictive control: Theory and design. Madison, Wisconsin: Nob Hill Pub., 2009.
 Zavala, Victor M., and Lorenz T. Biegler. "The advanced-step NMPC controller: Optimality, stability and robustness." Automatica 45.1 (2009): 86-93.
 Harris, Thomas J. "Assessment of control loop performance." The Canadian Journal of Chemical Engineering 67.5 (1989): 856-861.
 Yu, Jie, and S. Joe Qin. "Statistical MIMO controller performance monitoring. Part I: Data-driven covariance benchmark." Journal of Process Control 18.3-4 (2008): 277-296.
 Huang, Biao, S. L. Shah, and E. K. Kwok. "Good, bad or optimal? Performance assessment of multivariable processes." Automatica 33.6 (1997): 1175-1183.
 Schäfer, Jochen, and Ali Cinar. "Multivariable MPC system performance assessment, monitoring, and diagnosis." Journal of process control 14.2 (2004): 113-129.
 Cinar, Ali, and Kamuran Turksoy. Advances in Artificial Pancreas Systems: Adaptive and Multivariable Predictive Control. Springer, 2018.
 Hajizadeh, Iman, Rashid, Mudassir, and Cinar, Ali. " Plasma Insulin Cognizant Predictive Control for Artificial Pancreas." Journal of Process Control (2019).
 Cinar, Ali, Kamuran Turksoy, and Iman Hajizadeh. "Multivariable artificial pancreas method and system." U.S. Patent Application No. 15/171,355.
 Hajizadeh, Iman, et al. "Incorporating Unannounced Meals and Exercise in Adaptive Learning of Personalized Models for Multivariable Artificial Pancreas Systems." Journal of diabetes science and technology 12.5 (2018): 953-966.