Prediction-based strategies for robust near real-time GPS signal anomaly detection on a global scale

Abstract

During an anomalous signal deformation event, off-nominal distortions of signals with respect to each other are produced. This leads to abnormal receiver-dependent pseudorange biases that pose a threat for high integrity applications. The best characterization of signal deformations is obtained using high-gain antennas. However, due to their limited tracking capabilities and costly operation, they are less suitable for continuous multi-satellite monitoring. Although using the Wide Area Augmentation System (WAAS) monitor network makes it more likely to detect signal anomalies when they occur, this approach is limited to monitor only roughly half of the GPS constellation in a permanent way. This contribution presents a methodology for continuous and global GPS L1 and L2 signal quality monitoring, aimed at the early detection of anomalous signal distortions. It is based on the estimation of very high-rate (5 min samples) intra-frequency satellite differential code biases (DCB) using observations from the International GNSS Service (IGS) network. Time series of estimates are used as monitoring metrics. Detection of non-nominal estimates is done based on an unsupervised learning mixture model using results from classical, Bayesian and neural network-based autoregressive models. To quantify the confidence of detection, epoch-wise probabilities are computed using a Markov chain model. Two anomalous signal deformation events, occurred to GPS IIF SVN66 and SVN73 satellites in 2021 and 2022, respectively, were analyzed. To evaluate the effect of such events on individual signal components and observables, precise point positioning (PPP) residuals of selected IGS stations were computed. The obtained results show that the proposed methodology is suitable for the detection on a global scale in near real-time (few minutes to few hours) of such type of events. Through an early detection mechanism, the presented strategies aim at contributing to the prompt characterization of anomalous signal deformation events.