| An integrated, automated, adaptive control system, utilizing model predictive control, is developed for closed-loop control of blood glucose in type 1 diabetes, and is experimentally tested using insulin-glucagon infusion pumps to regulate blood glucose online in a diabetic swine model in vivo. As a preliminary study, the pharmacodynamics and stability at room temperature of glucagon in solution were tested in the swine model, thereby qualifying using glucagon in the control system. Subsequently, an empirical subject model, revealing dominant orders and delay, was obtained offline by pseudo-linear system identification performed on input-output data generated from in vivo experiments on the diabetic swine model. Indirect adaptive control was used in conjunction with the identified model structure to regulate blood glucose in vivo in closed-loop, with model parameters recursively updated using the extended least-squares estimation algorithm. The optimal control algorithm was formulated with integral effect, which rejects tracking offsets that can otherwise result from nonzero-mean disturbances, such as a carbohydrate intake. Furthermore, by augmenting the algorithm with a model that governs the accumulation of insulin in the subcutaneous depot, the controller was inculcated with knowledge of pending insulin action, which is critical to avoid excessive insulin dosing that would otherwise lead to episodes of impending hypoglycemia. Several closed-loop in vivo experiments were conducted in anesthetized pigs, with meal uptakes simulated by intravenous glucose loads, induced either by an impulse injection, by successive isolated injections, or in the form of a distributed uniform load from an intravenous drip. Control results consistently showed successful glucose regulation to euglycemic range within 60 minutes in response to an initial hyperglycemic state, and within 80--140 minutes in response to intravenously administered glucose loads. Results also demonstrated the potency of subcutaneous doses of glucagon in staving off episodic hypoglycemia, and revealed the efficacy of self-adaptation in coping with inter- and intra-subject response variations. The presented control system provides proof of concept and strikes an intricate balance between tight blood-glucose control and optimal drug consumption, while simultaneously maintaining emphasis on simplicity and reliability. |
