Graph Attention Network-Based Precoding for Reconfigurable Intelligent Surfaces Aided Wireless Communication Systems

In this paper, we propose a novel precoding algorithm based on an unsupervised learning graph attention network (ULGAT) to improve the performance of the sum rate in reconfigurable intelligent surface (RIS) assisted communication systems. Specifically, an unsupervised learning scheme is used to optimise the precoding of RIS-assisted multi-user downlink systems, which significantly reduces the difficulty of obtaining samples. Then, a graph attention network (GAT) based on the multi-head attention is employed to refine the performance of the precoding, where the underlying topology formed is fully utilized by the channel matrix and phase shift matrix. The proposed ULGAT algorithm can obtain better performance by exploiting the learning capability of the GAT and making full use of the network topology with unsupervised learning to produce datasets at a low cost. The results of the simulation demonstrate that the proposed ULGAT algorithm significantly outperforms the existing algorithms in terms of the sum rate and generalization capability.