LAUSR

Abstract Continuous glucose monitoring (CGM) sensors are a critical component of artificial pancreas (AP) systems that enable individuals with type 1 diabetes to achieve tighter blood glucose control. CGM sensor signals are often afflicted by a variety of anomalies, such as biases, drifts, random noises, and pressure-induced sensor attenuations. To improve the accuracy of CGM measurements, an on-line fault detection method is proposed based on sparse recursive kernel filtering algorithms to identify faults in glucose concentration values. The fault detection algorithm is designed to effectively handle the nonlinearity of the measurements and to differentiate the normal variability in the glycemic dynamics from sensor anomalies. The effectiveness of the proposed recursive kernel filtering algorithm for sensor error detection is demonstrated using simulation studies.