Improving Prediction of GNSS Satellite Visibility in Urban Canyon Based on Graph Transformer

Signals from global navigation satellite systems (GNSS) suffer from serious multipath errors in urban areas caused by building blockages and reflections. The use of deep neural networks offer great potential for predicting and eliminating complex multipath/non-line-of-sight (NLOS) errors. However, existing methods for predicting the original signals face two remaining challenges. The first is the inability to exploit effectively the irregular GNSS dataset because of inconsistent numbers of visible satellites in different epochs. The second is degradation in the generalization performance of the multipath/NLOS prediction model when using data collected from different locations and periods. To address these challenges, this paper proposes a novel graph transformer neural network for predicting satellite visibility that effectively learns environment representations from the irregular GNSS measurements to both alleviate multipath interference and improve the generalization performance of the multipath prediction model. To learn from the irregular GNSS measurements, a sky satellite graph is constructed as the input to a graph neural network by using the satellites captured in the same epoch, which can represent the spatial relationships between the satellites and enhance the model to enable learning of satellite-related features sufficiently well. To improve generalization ability of our multipath prediction model, a multihead attention mechanism is introduced to aggregate satellite node information by computing the correlation between satellites for extracting the environment representation around the receiver. Based on the constructed sky satellite graph and the multihead attention mechanism, we develop a novel graph transformer neural network (GTNN) for predicting satellite visibility, which can not only handle irregular GNSS measurements but also learn an environment representation via graph attention. Comparative experiments were carried out on real-world GNSS measurement data in urban areas, which showed that the proposed method could achieve an accuracy exceeding 96% for satellite visibility prediction and obtain better generalization performance than existing multipath prediction methods. Moreover, the attention weights among the satellites were visualized to demonstrate the environment representation learned by the GTNN from the sky satellite graph.