Graph Neural Network for Location- and Orientation-Assisted mmWave Beam Alignment
Yuzhu Lei, Qiqi Xiao, Yinghui He, Guanding Yu
TL;DR
The paper tackles the high training overhead and performance degradation of location- and orientation-assisted mmWave beam alignment in massive MIMO. It introduces a graph neural network that models beams as nodes connected by angular correlations, enabling efficient learning with far fewer labeled samples. Empirical results show the GNN reaches equivalent accuracy with only 20% of the data and gains about 10% Top-1 accuracy on the same data compared with DNN baselines, while maintaining robustness to location and orientation errors and reducing model complexity. The method extends naturally to UPAs and remains scalable with increasing antenna counts, offering a practical pathway to fast, accurate beam alignment in dynamic mmWave environments.
Abstract
In massive multi-input multi-output (MIMO) systems, the main bottlenecks of location- and orientation-assisted beam alignment using deep neural networks (DNNs) are large training overhead and significant performance degradation. This paper proposes a graph neural network (GNN)-based beam selection approach that reduces the training overhead and improves the alignment accuracy, by capitalizing on the strong expressive ability and few trainable parameters of GNN. The channels of beams are correlated according to the beam direction. Therefore, we establish a graph according to the angular correlation between beams and use GNN to capture the channel correlation between adjacent beams, which helps accelerate the learning process and enhance the beam alignment performance. Compared to existing DNN-based algorithms, the proposed method requires only 20\% of the dataset size to achieve equivalent accuracy and improves the Top-1 accuracy by 10\% when using the same dataset.