Automatic Detection of Ionospheric Scintillation-like GNSS Oscillator Anomaly Using a Machine Learning Algorithm

In this paper, we propose a machine learning-based approach to automatically detect satellite oscillator anomaly. A major challenge is to differentiate an oscillator anomaly from ionospheric scintillation. Although both scintillation and oscillator anomalies cause phase disturbances, their underlying physics are different and, therefore, show different carrier frequency dependency. By using triple-frequency signals, distinct features are extracted from the disturbed signals and applied to the radial basis function (RBF) support vector machine (SVM) classifier to identify an oscillator anomaly. The results show that the proposed RBF SVM displays superior performance and outperforms several other classification methods. The proposed approach is applied on a database to conduct automatic satellite oscillator anomaly detection. Preliminary detection results show that there are on average 1.7 and 0.2 anomaly events observed on PRN1 and PRN25 each day, respectively.