A Fast Satellite Selection Method Based on Graph Attention Network for Improving GNSS Positioning Accuracy

Xiuxun Liu, Jiaolong Wei, Zuping Tang, Chuang Zhou

Peer Reviewed

Abstract:	The rapid development of multi-constellation global navigation satellite systems (GNSS) has led to a significant increase in the number of visible satellites. However, traditional satellite selection algorithms based on iterative optimization or heuristic search suffer from significantly increased computational complexity with the number of satellites, and struggle to explicitly model the complex nonlinear geometric relationships between satellites. Therefore, a satellite selection method based on graph attention network (GAT) is proposed. The proposed method models the satellites of each epoch as a fully connected directed graph, where nodes represent satellites and contain their associated observation information, and edges represent the spatial distribution relationships between satellites. The GAT model adaptively learns the impact of different nodes in the graph on positioning accuracy through a multi-head attention (MHA) mechanism and assigns different weights to the nodes, thereby assessing the importance of each satellite. Finally, the optimal satellite subset is obtained. Once the model is trained, the presented method can select satellites with just one forward propagation, without any matrix inversion or iterative search. This effectively avoids the complex iterative calculations present in traditional satellite selection algorithms. Experiments were conducted on the proposed method using data actually collected via a receiver. Experimental data were collected in an urban canyon environment. The comparative experimental results show that the positioning accuracy of the proposed method is 21.9%, 17.6%, and 15.5% higher than that of the genetic algorithm, particle swarm optimization algorithm, and long short-term memory network model, respectively. Furthermore, the presented method exhibits high satellite selection efficiency through computational complexity analysis. This indicates that the proposed method strikes a good balance between satellite selection efficiency and improved positioning accuracy in GNSS satellite selection, and has the potential for wide application in various GNSS terminals.
Published in:	Proceedings of the 2026 International Technical Meeting of The Institute of Navigation January 26 - 29, 2026 Hyatt Regency Orange County Anaheim, California
Pages:	8 - 21
Cite this article:	Liu, Xiuxun, Wei, Jiaolong, Tang, Zuping, Zhou, Chuang, "A Fast Satellite Selection Method Based on Graph Attention Network for Improving GNSS Positioning Accuracy," Proceedings of the 2026 International Technical Meeting of The Institute of Navigation, Anaheim, California, January 2026, pp. 8-21. https://doi.org/10.33012/2026.20559
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In