Edge-Featured Graph Attention Network for Power Control in Cell-Free and D2D Heterogeneous Networks

Cell-free and device-to-device (D2D) heterogeneous networks can effectively enhance spectral efficiency. However, spectrum reuse among different transmission links leads to severe interference to limit the growth of network capacity. In this paper, we propose a power control algorithm based on graph attention network for interference coordination in cell-free and D2D heterogeneous networks. Specifically, a heterogeneous graph is utilized to model the power control problem, where desirable transmissions and interference are represented as edges of different types. The edge-featured attention mechanism is applied over the constructed graph to capture the relationships among various edges. Simulation results show that our proposed algorithm can improve network capacity and exhibit generalization performance with respect to network scales.