Graph embedding-enhanced GNN for scalable D2D power control

As mobile communication and device numbers surge, wireless resources face increasing demand. D2D communication conserves base station resources but creates interference, making power control a crucial aspect of the system. Traditional algorithms struggle to strike a balance between performance and computational cost in non-convex scenarios. Data-driven methods, particularly GNNs, now excel at modelling complex relationships between devices. This study utilizes GNNs and graph embedding to enhance power control, yielding robust results across diverse and even unseen environments, with only a modest increase in computational time.