Power Allocation Optimization Design for D2D Cellular Networks Based on Graph Neural Networks

This study focuses on the co-frequency interference issue between cellular users (CUs) and device-to-device users (DUs)., with the objective of maximizing the system sum-rate, in which a power allocation optimization strategy based on graph neural networks (GNN) is proposed. Leveraging the relative distances between transmitting and receiving nodes, a distance-based grouping and alternating channel allocation approach is devised. The aim is to identify suitable cellular user-shared channel resources for the device-to-device (D2D) source users within the system. In addition, this paper utilizes graph neural networks to address non-linear optimization problems, effectively overcoming the limitations of traditional optimization algorithms with low computational efficiency. Through simulation results, the efficacy of the proposed method in enhancing the cellular network's sum-rate is validated. In comparison to the weighted minimum mean square error (WMMSE) algorithm, the proposed approach demonstrates superior performance, such as higher generalization capability and faster response times.