8 May 2026 in IEEE Transactions on Vehicular Technology
Satellite communication networks, characterized by the advantages of seamless global coverage and ubiquitous access, have emerged as a main focus of current research. However, satellite network topology changes dynamically and burst traffic is spatio-temporal coupling, causing the satellite network traffic sensing problem with high overhead and low accuracy. Therefore, this paper proposes a spatio-temporal graph neural network-based satellite network traffic sensing approach (STSA) to collect data efficiently and predict traffic precisely. The study designs in-band network telemetry (INT)-based Hamilton cycle to reduce the number of repeated visits to satellite nodes, reducing the redundancy of data collection, and construct spatio-temporal graph of satellite network topology and traffic. STSA takes full advantage of the spatio-temporal dependence of topology and traffic to realize satellite network traffic sensing. The satellite network state view is generated by mapping the spatio-temporal characteristics of historical traffic to the next period. Simulation shows that STSA effectively reduces sensing overhead and improves accuracy compared to the traditional traffic sensing algorithm based on the Euler cycle and long short-term memory (LSTM). The proposed STSA provides traffic data for satellite network routing, enhancing the efficiency of satellite network management and control.