Researchers Yao Peng, Tingting Liu, and Chenyang Yang from the Beijing University of Posts and Telecommunications have developed a novel approach to improve spectral efficiency in cell-free multi-antenna systems, a technology that is increasingly relevant to the energy sector as it seeks to enhance the efficiency and reliability of wireless communication networks for smart grids and other applications.
In user-centric cell-free multi-antenna systems, pilot contamination is a significant challenge that degrades spectral efficiency. Existing solutions that jointly optimize pilot assignment and power allocation assume a fixed pilot length, which fails to balance pilot overhead against contamination. To address this, the researchers have proposed a method that jointly optimizes pilot length, pilot assignment, and power allocation using deep learning. This approach aims to maximize net spectral efficiency, a critical factor in the energy sector where efficient data transmission is essential for managing and monitoring energy systems.
The researchers designed size-generalizable graph neural networks (GNNs) to cope with the challenge of pilot length being a variable, which affects the size of the pilot assignment matrix. They proved that pilot assignment policy is a one-to-many mapping, and improperly designed GNNs cannot learn the optimal policy. To tackle this issue, they introduced feature enhancement. Additionally, they designed a contamination-aware attention mechanism for the GNNs to improve learning performance.
Given that pilot assignment and power allocation depend on large- and small-scale channels, the researchers developed a dual-timescale GNN framework to explore this potential. To reduce inference time, they also designed a single-timescale GNN. Simulation results showed that the designed GNNs outperformed existing methods in terms of net spectral efficiency, training complexity, and inference time, and could be well generalized across problem scales and channels.
The research was published in the IEEE Transactions on Wireless Communications, a prestigious journal in the field of wireless communications and networking. The findings have significant implications for the energy sector, where efficient and reliable wireless communication is crucial for smart grid management, remote monitoring, and control of energy systems. By improving spectral efficiency, the proposed method can help reduce energy consumption and costs associated with data transmission, making it a valuable tool for the energy industry.
This article is based on research available at arXiv.