这份文本介绍了一种用于超大规模天线阵列（ELAA）系统的联邦学习（FL）辅助预测波束赋形框架。该框架旨在通过结合速率拆分多址接入（RSMA）和新颖的补丁混合方法，解决ELAA系统在用户移动性和信道状态信息（CSIT）不完美下管理干扰的挑战。研究人员设计了一个预测波束赋形协议，并提出FL辅助训练过程以提高效率和模型鲁棒性，最终使ELAA技术在实际应用中更具可行性。仿真结果表明，与传统方法相比，该方法在几何平均用户速率和处理时间效率方面表现出卓越的性能。

Federated Learning-Assisted Predictive Beamforming for Extremely Large-Scale Antenna Array Systems With Rate-Splitting Multiple Access

Abstract:

Achieving perfect Channel State Information at the Transmitter (CSIT) is often infeasible in Extremely Large-scale Antenna Array (ELAA) systems due to user mobility and feedback/processing delay. This results in severe multi-user interference. Therefore, how to effectively and efficiently manage interference with partial/historical CSIT is one of the most important challenges for implementing ELAA. In this paper, we propose a Federated Learning (FL)-assisted predictive beamforming framework for ELAA systems to address this challenge. Specifically, we introduce Rate-Splitting Multiple Access (RSMA) to relax the sensitivity to imperfect CSIT while still benefiting from the spatial resolution. Moreover, a predictive beamforming protocol is designed to optimize the precoder design under the imperfections in the channel estimate quality originating from user mobility and latency. To calculate the beamformers, we first propose a lightweight patch-mixing approach to split the historical CSIT data samples into smaller manageable segments. Then, we propose an FL-based training method that enables parallel processing of these CSI segments, thereby accelerating the training process. Simulation results show the effectiveness and efficacy of the proposed FL-assisted predictive beamforming framework, which paves the way for real-world implementation of ELAA.

Published in: IEEE Journal of Selected Topics in Signal Processing ( Volume: 19, Issue: 2, March 2025)

Page(s): 461 - 476

Date of Publication: 24 January 2025

ISSN Information:

Print ISSN: 1932-4553

Electronic ISSN: 1941-0484

DOI: 10.1109/JSTSP.2025.3532040

Publisher: IEEE

Funding Agency:

10.13039/501100001381-National Research Foundation Singapore

Infocomm Media Development Authority under its Future Communications Research & Development Programme

10.13039/501100001809-National Natural Science Foundation of China (Grant Number: 62201347)

Shanghai Sailing Program (Grant Number: 22YF1428400)

10.13039/100014013-UK Research and Innovation (Grant Number: EP/X040569/1, EP/Y037197/1, EP/X04047X/1 and EP/Y037243/1)