ADAPTIVE CONTROL STRATEGY FOR GLUCOSE REGULATION USING RECURSIVE LINEAR MODELS

Abstract

Abstract Using frequently sampled blood glucose measurements (at 5 min intervals), low-order recursive linear time series models have been developed for the prediction of future blood glucose concentrations. Such predicted glucose values are then integrated with model based control algorithms, such as GPC and LQC, for adjusting the required insulin infusion rates with an automated insulin pump. Since the models are derived from patients’ own glucose data, the proposed algorithm can dynamically adapt to inter- and intra-subject variability.