«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

期数:

2025年07期

页码:

48-54

栏目:

移动与物联网络

出版日期:

2025-07-10

文章信息/Info

Title:

Federated Learning Based ISAC Beamforming Method for Satellite Systems

文章编号:

1673-629X(2025)07-0048-07

作者:

赖海光1; 2; 朱邦兵1; 沈金海1; 万坤1; 王泽渝2

1. 南京控维通信科技有限公司,江苏 南京 211135;

2. 南京邮电大学 通信与信息工程学院,江苏 南京 210003

Author(s):

LAI Hai-guang1; 2; ZHU Bang-bing1; SHEN Jin-hai1; WAN Kun1; WANG Ze-yu2

1. Nanjing Kongwei Communication Technology,Nanjing 211135,China;

2. School of Communication and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China

关键词:

卫星通信; 通感一体化; 波束成形; 优化问题; 联邦学习

Keywords:

satellite communication; integrated of sensing and communication; beamforming; optimization problem; federated learning

分类号:

TP181

DOI:

10.20165/j.cnki.ISSN1673-629X.2025.0036

摘要:

卫星通信在 6G 发展中具有举足轻重的地位,是实现全球无缝连接和提供多样化服务的重要途径。 其中,通感一体化(Integrated of Sensing and Communication,ISAC)作为 6G 的核心技术之一,能够实现通信与感知的融合,提升频谱利用率,能够在资源有限的卫星系统内最大化通信与感知性能。 该文提出了一个以卫星系统的最大化通信信噪比为目标的优化问题,通过求解该优化问题的最优解,构建由信道矩阵生产波束成形矩阵的求解方案。 特别地,通过所获得的联邦学习数据集,创新性地提出了一种面向卫星系统通感一体化波束成形矩阵的优化求解方案。 考虑联邦学习训练过程,模型训练的参数包括信道矩阵与波束成形矩阵和功率分配系数等因素。 模型的训练则被部署至用户侧执行,而卫星侧作为服务器端执行优化模型的聚合。 仿真结果表明,与常用的通感一体化波束成形方案相比,采用联邦学习的波束成形方案可以带来更好的通感一体化性能。 因此,提出一种基于联邦学习的卫星系统通感一体化波束成形方法,简化基于优化问题的下行通感一体化波束成形矩阵的计算,有效提高了卫星系统的通感一体化性能,具备较好的泛用性。

Abstract:

Satellite communication plays a crucial role in the development of 6G and is an important way to achieve seamless global con-nectivity and provide diversified services. Among them,Integrated Sensing and Communication (ISAC),as one of the core technologies of 6G,can achieve the integration of communication and perception,improve spectrum utilization,and maximize communication and perception performance in resource limited satellite systems. We propose an optimization problem for maximizing the communication signal-to-noise ratio of a satellite system. By solving the optimal solution of this optimization problem,a solution scheme for producing beamforming matrices from channel matrices is constructed. Specifically, based on the obtained federated learning dataset, we innovatively propose an optimization solution for the integrated beamforming matrix of satellite system sensing. Considering the training process of federated learning,the parameters for model training include factors such as channel matrix,beamforming matrix,and power al-location coefficient. The training of the model is deployed to the user side for execution,while the satellite side serves as the server-side for optimizing the aggregation of the model. The simulation results show that compared with commonly used integrated sensing beamforming schemes,the beamforming scheme using federated learning can bring better integrated sensing performance. Therefore,we propose a federated learning based integrated beamforming method for satellite system sensing,which simplifies the calculation of the downlink sensing integrated beamforming matrix based on optimization problems, and effectively improves the sensing integration performance of satellite systems with good universality.

常用功能

更新日期/Last Update: 2025-07-10