This repository contains our work
Graph Neural Network for Distributed Beamforming and Power Control in Massive URLLC Networks, which is accepted by the TWC (early access).
For any reproduce, further research or development, please kindly cite our paper
@ARTICLE{G4U,
author={Gu, Yifan and She, Changyang and Bi, Suzhi and Quan, Zhi and Vucetic, Branka},
journal={IEEE Transactions on Wireless Communications},
title={Graph Neural Network for Distributed Beamforming and Power Control in Massive URLLC Networks},
year={2024},
volume={},
number={},
pages={},
note={early access},
}
Instructions:
- Simulation for GNN, WMMSE and EPA policies can be found in GNN and WMMSE and EPA.py.
- Simulation for the proposed G4U can be found in G4U.py.
- Simulation for the proposed PG4U can be found in PG4U.py.
- Note that we have developed a loss function for the training of URLLC networks. If one want to compare it with the utility function-based one, comment out line 186-189, and use line 192-197 in GNN and WMMSE and EPA.py for training. In addition, one may use other loss functions for training, such as error probability-based ones (not given in the codes but can be designed easily), but they may not achieve a good performance. Similar comments also apply for G4U.py and PG4U.py.
We thank the works "Graph Neural Networks for Scalable Radio Resource Management: Architecture Design and Theoretical Analysis" and "Spatial Deep Learning for Wireless Scheduling" for their source codes in creating this repository.