Blood glucose regulation for diabetic mellitus using a hybrid intelligent technique

In this paper, we use a mathematical model for the interaction between the blood glucose and insulin concentration and investigates the on-line control schemes necessary to accomplish an external blood glucose regulation. The dynamic model is described in terms of blood glucose level, and net insulin level in blood as two state variables and external rate of blood glucose concentration as control variable. Using optimal regulation results, an objective is chosen to minimize the deviation of blood glucose from a preset level. The closed-loop optimal control scheme is developed for the biosystem in which a blood glucose sensor feeds information to a pump to release computed amount of insulin into the circulation system. The performance of the proposed optimal control scheme is compared with experimental results. Further to improve the closed-loop optimal performance, a soft control strategy based on adaptive neuro-fuzzy inference system (ANFIS) is devised leading to synergy of hard (optimal) and soft (artificial intelligent) control. ANFIS is a simple learning technique which is implemented in the framework of adaptive neural networks that provides the best optimization tool to find parameters that best fits the data. The application of this synergetic (hard and soft) control strategy to the diabetic regulation system shows good agreement between the experimental data and the theoretical/simulated results.