Characterization Method and Performance Verification of Space-Ground Integrated Networks Based on Graph Neural Networks

Unmanned Aerial Vehicles (UAVs) are increasingly used to augment ground networks, creating an integrated air-ground heterogeneous network architecture. In such networks, each link has unique transmission characteristics and performance metrics, making it particularly challenging to represent the entire network comprehensively. To address this issue, we have developed a unified representation method for heterogeneous communication networks based on Graph Neural Networks (GNNs). GNNs can simultaneously handle multi-dimensional features of nodes and edges, considering factors such as link type, Signal-to-Interference-plus-Noise Ratio (SINR), and transmission rate to accurately assess and optimize network performance. Experimental results demonstrate that this method reliably reflects link quality in real-world communication environments. Furthermore, the static routing algorithm based on the representation results outperforms traditional methods in key performance indicators such as throughput, packet drop rate, and latency.