Return to Session D3

Session D3: GNSS Augmentation and Robustness for Autonomous Navigation

A Real-Time Anomaly Monitoring Method based on GLRT for Satellite Clock Offsets
Lin Zhao, Nan Li, College of Intelligent Systems Science and Engineering, Harbin Engineering University; Hui Li, College of Intelligent Systems Science and Engineering, Harbin Engineering University & HEU Qingdao Ship Science and Technology Co., Ltd; Xue Liu, Wenzhen Peng, College of Intelligent Systems Science and Engineering, Harbin Engineering University

Peer Reviewed

The real-time satellite clock offsets which are estimated based on observation data from global GNSS stations play an essential role in real-time precise point positioning (PPP). Since some anomalies from the observation data and the satellite clock itself are inevitable, there will be some abnormal errors in satellite clock offsets, such as gross errors, phase jumps, etc. A single nanosecond deviation of clock offset is equivalent to a user position error of 30 cm. Anomaly monitoring for real-time estimated satellite clock offsets is of significance for real-time PPP. The real-time PPP usually requires high-frequency satellite clock offsets to improve its accuracy. Therefore, it is desirable to provide an anomaly monitoring algorithm with tolerable computational complexity and detection time delay. The generalized likelihood ratio test (GLRT) is an effective anomaly monitoring method for satellite clocks based on the satellite clock frequency offsets. It is a powerful test statistics method for monitoring data mutation in a distribution, which also can be used in anomaly monitoring for satellite clock offsets. As the determination of test threshold obtained by the complicated Monte Carlo experiment in traditional GLRT increases the computational burden and is not suitable for real-time satellite clock offset anomaly monitoring, an improved GLRT which simplify the test threshold determination is proposed. For anomaly monitoring, the satellite clock offset prediction model is constructed based on Kalman filter, and the innovation sequence of Kalman filter which presents an independent Gaussian distributed is given as detection statistics. The improved GLRT takes into account different statistical hypotheses to describe the variation in the innovation sequence, which can convert the likelihood ratio test statistics to the well-known t-distribution. In this way, the test threshold values can be obtained by t-test. The null hypothesis states that the mean of innovation sequence presents a Gaussian distributed with an unchanged mean, while the alternative hypothesis states that the mean has changed. The maximum-likelihood estimates (MLE) of the mean and standard deviation of an assumed variation are calculated based on the innovation sequence, through which the test statistic t of a change versus no change can be evaluated. The fixed-length sliding window is used to detect the anomaly status of the latest innovation calculated by the real-time updated satellite clock offset. To ensure continuous anomaly monitoring, the improved GLRT is iteratively applied to the innovation sequence in the sliding window. The proposed satellite clock offset anomaly monitoring method is verified using the real-time satellite clock offsets provided by the analysis centers (ACs) of WHU. The result shows that the proposed method can quickly and effectively detect, identify and repair the anomalies in satellite clock offsets including gross errors and phase jumps, without increase the computational burden. The gross error can be replacement by the satellite clock offsets predicted by the Kalman filter without re-processing the anomalistic data, and the replaced error can be negligible. The real-time satellite clock offset anomaly monitoring method proposed in this paper combines the advantages of Kalman filter and GLRT, which is helpful to improve the reliability of real-time PPP.



Return to Session D3