面向低轨卫星的时空特征融合LSTM松弛测量方法

杨孟欣; 张青婷; 曾令昕; 顾忆宵; 曾丹; 夏斌

doi:10.11999/JEIT251146

面向低轨卫星的时空特征融合LSTM松弛测量方法

doi: 10.11999/JEIT251146 cstr: 32379.14.JEIT251146

1.
上海大学上海 200444
2.
中国星网网络应用有限公司重庆 400065
3.
上海交通大学上海 200240

基金项目: 重庆市自然科学基金项目 (CSTB2024NSCQ-LMX0024)

详细信息

作者简介:
杨孟欣：女，硕士生，研究方向为卫星通信、分布式学习

张青婷：女，工程师，研究方向为卫星通信、地面5G通信

曾令昕：男，高级工程师，研究方向为卫星通信、信号处理

顾忆宵：男，讲师，研究方向为卫星通信、边缘计算、人工智能等

曾丹：女，教授，研究方向为计算机视觉与模式识别

夏斌：男，教授，研究方向为卫星通信，通感一体网络

通讯作者:
顾忆宵　yxgu@shu.edu.cn

中图分类号: TN927.2
计量
- 文章访问数: 9
- HTML全文浏览量: 5
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2025-11-01
- 修回日期: 2025-11-22
- 录用日期: 2026-01-06
- 网络出版日期: 2026-01-10

A Spatio-Temporal Feature Fusion LSTM Relaxation Measurement Method for LEO Satellites

1.
Shanghai University, Shanghai 200444, China
2.
The China Satellite Network Exploration Co.Lt, Chongqing 400065, China
3.
Shanghai Jiao Tong University, Shanghai 200240, China

Funds: Natural Science Foundation of Chongqing, China (CSTB2024NSCQ-LMX0024)

摘要

摘要: 为降低低轨(LEO, Low Earth Orbit)卫星系统的高动态性导致的终端频繁链路测量，现有松弛测量方案主要采用基于静态阈值或常规的时空预测模型的方法。然而，在卫星的高动态性导致不同历史时刻的测量数据和不同测量指标的重要性随之动态演化，而传统方法难以捕获这种时空重要性的动态变化。为此，针对上述挑战，该文提出一种面向低轨卫星的时空特征融合长短期记忆递归神经网络(Long Short-Term Memory, LSTM)松弛测量方法。首先，建立了LEO卫星通信系统模型，以获取所需要的历史测量数据。之后，构建了集成双重注意力机制的LSTM预测模型，实现对关键历史时刻与动态重要特征的精准聚焦，从而完成对测量频点集合与测量周期的精准预测。随后，在选择的测量频点集合和松弛周期下，执行自适应的链路测量。仿真结果表明，相较于基线方法，该文所提出的松弛测量方案在确保链路可靠性的同时，可以显著降低终端的测量频次。并且，与基线相比，在更大的速度和松弛周期下，具有更强的自适应性。
- 低轨卫星通信 /
- 松弛测量 /
- 频点选择 /
- 周期松弛
Abstract: Objective The high dynamics of Low Earth Orbit (LEO) satellite communication systems lead to frequent link measurements. Existing schemes mainly adopt threshold-based or standard spatio-temporal prediction measurement relaxation strategies to alleviate this issue. However, these approaches fail to effectively capture the dynamic evolution in the importance of both historical data and multiple measurement metrics caused by satellite high mobility, making it difficult to adapt to the highly time-varying satellite-ground link environment. To address this challenge, a spatio-temporal feature fusion Long Short-term memory (LSTM) relaxation measurement method for LEO satellite communication is proposed. The algorithm constructs a LSTM recurrent neural network integrated with a dual-attention mechanism. Specially, the LSTM is employed to extract correlations among historical measurement data, while the temporal attention mechanism and variable attention mechanism are respectively utilized to focus on key historical moments and significant features. On this basis, the proposed method jointly predicts the measurement frequency set and the number of relaxation period. Under the selected measurement frequency set and relaxation cycle, intelligent link measurement is then performed, achieving adaptive and energy-efficient link monitoring in LEO satellite systems. Methods The proposed spatio-temporal feature fusion LSTM–based relaxation measurement method for LEO satellite communication employs a dual-attention LSTM (DA-LSTM) model to adaptively reduce measurement overhead while maintaining reliable link monitoring. By leveraging historical link quality indicators, including reference signal receiving power (RSRP), reference signal receiving quality (RSRQ), and Doppler shift, together with satellite ephemeris information, the model captures both temporal and spatial variations to enable the joint prediction of an subset of measurement frequency points and their corresponding relaxation periods. Based on the predicted results, the terminal either performs adaptive frequency selection and dynamic period adjustment or executes full-band measurements with a fixed period. This process enables adaptive and energy-efficient link monitoring while maintaining reliable communication performance in LEO satellite systems. Results and Discussions The proposed relaxation measurement method adopts a DA-LSTM model to predict frequency points and the number of relaxation period based on historical link quality information. Simulation results show that the proposed method achieves higher convergence efficiency and higher training accuracy and lower loss in both frequency point selection and relax period selection (Fig.4 and Fig.5) than baseline. The proposed measurement algorithm achieves an average measurement frequency under 30% with minimal performance loss (Table 3). This is because that the algorithm can adaptively select high-quality frequency points and adjust measurement period. Then we further investigate the trade-off between measurement frequency and communication performance (Fig.6 and Fig.7), showing that the proposed method achieves a better balance than baselines and the comparison between different speeds. Besides, the simulations are further validated under different terminal speeds (Fig.8) and maximum relaxed periods (Fig.9), demonstrating its ability to maintain high energy efficiency and communication performance across different operational scenarios. Conclusions This work addresses the key challenge of dynamic spatio-temporal importance evolution caused by satellite mobility, which limits the effectiveness by proposing a relaxed measurement algorithm based on a DA-LSTM model. The proposed method predicts both the set of measurement frequency points and the number of relaxed measurement period by extracting spatio-temporal correlations from historical link quality. The algorithm is evaluated through simulations under various scenarios, and the results show that: (1) The proposed algorithm achieves higher convergence efficiency and training accuracy than baseline; (2) The algorithm can adaptively select high-quality frequency points and adjust measurement periods, maintaining an optimal balance between measurement frequency and communication reliability; (3) The method remains effective across different terminal speeds and maximum relaxed periods, demonstrating its scalability and robustness in dynamic operational environments. However, the current study focuses on simulations and does not consider practical hardware limitations, atmospheric effects, or real-time processing constraints, which should be investigated in future work.
- Low earth orbit satellite communication /
- Relaxation measurement /
- Frequency point selection /
- Measurement period relaxation

HTML全文

图 1 星地直连通信系统

下载: 全尺寸图片幻灯片

图 2 链路松弛测量流程

下载: 全尺寸图片幻灯片

图 3 DA-LSTM预测模型架构

下载: 全尺寸图片幻灯片

图 4 预测模型训练准确率变化

下载: 全尺寸图片幻灯片

图 5 预测模型训练损失值变化

下载: 全尺寸图片幻灯片

图 6 测量频次与通信性能之间的权衡

下载: 全尺寸图片幻灯片

图 7 测量频次与通信性能之间的权衡

下载: 全尺寸图片幻灯片

图 8 不同终端速度下算法性能对比

下载: 全尺寸图片幻灯片

图 9 不同最大松弛周期下算法性能对比

下载: 全尺寸图片幻灯片

表 1 具体仿真参数配置

系统参数	参数设置
轨道面数、每个轨道卫星数	6、10
相位因子	1、2
倾角	55°、89°
卫星频段	n510、n511、n512
偏心率	0.001020
轨道高度	875 km、925 km
天线增益、带宽B	16 dBi, 380 MHz

下载: 导出CSV

表 2 神经网络模型参数

系统参数	参数设置
损失函数	均方误差(MSE)
优化器	Adam
激活函数	ReLU、Softmax
可学习参数量	308.5 K
学习率	0.00001

下载: 导出CSV

表 3 终端测量平均性能(%)

	传统松弛测量方法	LSTM松弛测量方法	本文松弛测量方法
测量频次降低率	60.38	62.92	72.57
最佳频点RSRP相对值	97.57	97.68	99.89

下载: 导出CSV

参考文献(23)

[1]	何元智, 肖永伟, 张世杰, 等. 全球泛在连接新模式: 手机直连卫星关键技术及挑战[J]. 电子与信息学报, 2024, 46(5): 1591–1603. doi: 10.11999/JEIT240032. HE Yuanzhi, XIAO Yongwei, ZHANG Shijie, et al. A novel pattern for global ubiquitous interconnection: Key technologies and challenges of direct-to-smartphone[J]. Journal of Electronics & Information Technology, 2024, 46(5): 1591–1603. doi: 10.11999/JEIT240032.
[2]	IMT-2030(6G)推进组. 6G总体愿景与潜在关键技术白皮书[R]. 2021. (查阅网上资料, 未找到本条文献报告编号和出版年以及作者信息, 请确认). IMT-2030(6G) Promotion Group. 6G overall vision and potential key technology white paper[R]. 2021.
[3]	苏昭阳, 刘留, 艾渤, 等. 面向低轨卫星的星地信道模型综述[J]. 电子与信息学报, 2024, 46(5): 1684–1702. doi: 10.11999/JEIT230941. SU Zhaoyang, LIU Liu, AI Bo, et al. Survey of satellite-ground channel models for low earth orbit satellites[J]. Journal of Electronics & Information Technology, 2024, 46(5): 1684–1702. doi: 10.11999/JEIT230941.
[4]	YUE Pingyue, AN Jianping, ZHANG Jiankang, et al. Low earth orbit satellite security and reliability: Issues, solutions, and the road ahead[J]. IEEE Communications Surveys & Tutorials, 2023, 25(3): 1604–1652. doi: 10.1109/COMST.2023.3296160.
[5]	3GPP. TR 38.331-2022 NR Radio Resource Control (RRC) protocol specification[S]. Sophia Antipolis, France: 3GPP. 2022. (查阅网上资料, 未找到本条文献信息, 请确认).
[6]	3GPP. TR 38.821-2020 Solutions for NR to support non-terrestrial networks (NTN)[S]. Sophia Antipolis, France: 3GPP. 2020. (查阅网上资料, 未找到本条文献信息, 请确认).
[7]	PARK S and KIM J. Trends in LEO satellite handover algorithms[C]. 2021 Twelfth International Conference on Ubiquitous and Future Networks (ICUFN), Jeju Island, Korea, 2021: 422–425. doi: 10.1109/ICUFN49451.2021.9528738.
[8]	JUAN E, LAURIDSEN M, WIGARD J, et al. Handover solutions for 5G low-earth orbit satellite networks[J]. IEEE Access, 2022, 10: 93309–93325. doi: 10.1109/ACCESS.2022.3203189.
[9]	Nokia Siemens Networks and Nokia Corporation. Small cell signal based control of inter-frequency measurements[R]. TSG-RAN WG2 Meeting #77bis, 2012. (查阅网上资料, 未找到本条文献信息, 请确认).
[10]	郑景仁. NTN中的测量和小区重选[P]. 中国, 118216179A, 2024. ZHENG Jingren. Measurement and cell reselection in NTN[P]. CN, 118216179A, 2024.
[11]	孔繁华, 邝奕如, 徐海博, 等. 测量放松方法及装置[P]. 中国, 115460660A, 2022. KONG Fanhua, KUANG Yiru, XU Haibo, et al. Measurement relaxation method and device[P]. CN, 115460660A, 2022.
[12]	LUO Hejia, WANG Yu, WANG Xiaolu, et al. The present application provides a communication method and apparatus, relates to the technical field of wireless[P]. CN, WO2023061166A1, 2023.
[13]	何燃燃, 张艳霞, 肖潇. 测量控制方法、终端及网络侧设备[P]. 中国, 117440395A, 2024. HE Ranran, ZHNAG Yanxia, and XIAO Xiao. Measurement control method, terminal and network side equipment[P]. CN, 117440395A, 2024.
[14]	胡奕, 李海涛. NTN中非连接态终端的RRM测量方式确定方法、装置及介质[P]. 中国, 202180101902.2, 2024. HU Yi and LI Haitao. Method and device for determining RRM (Radio Resource Management) measurement mode of non-connected terminal in NTN (Network Temporary Network), and medium[P]. CN, 202180101902.2, 2024.
[15]	WANG Juan, SUN Lijuan, ZHOU Jian, et al. An adaptive dynamic channel allocation algorithm based on a temporal–spatial correlation analysis for LEO satellite networks[J]. Applied Sciences, 2022, 12(21): 10939. doi: 10.3390/app122110939.
[16]	李学华, 廖海龙, 张贤, 等. 面向低轨卫星通信网络的联邦深度强化学习智能路由方法[J]. 电子与信息学报, 2025, 47(8): 2652–2664. doi: 10.11999/JEIT250072. LI Xuehua, LIAO Hailong, ZHANG Xian, et al. Federated deep reinforcement learning-based intelligent routing design for LEO satellite networks[J]. Journal of Electronics & Information Technology, 2025, 47(8): 2652–2664. doi: 10.11999/JEIT250072.
[17]	彭亮, 闫杰, 魏鹏, 等. 低轨卫星网络中基于时空相关性的不完全时间序列流量预测[J]. 信息与电子工程前沿(英文版), 2025, 26(5): 788–804. doi: 10.1631/FITEE.2300873. PENG Liang, YAN Jie, WEI Peng, et al. Spatio-temporal correlation-based incomplete time-series traffic prediction for LEO satellite networks[J]. Frontiers of Information Technology & Electronic Engineering, 2025, 26(5): 788–804. doi: 10.1631/FITEE.2300873.
[18]	ORTIZ F, VASQUEZ-PERALVO J A, QUEROL J, et al. Harnessing supervised learning for adaptive beamforming in multibeam satellite systems[C]. 2024 IEEE International Conference on Machine Learning for Communication and Networking (ICMLCN), Stockholm, Sweden, 2024: 386–392. doi: 10.1109/ICMLCN59089.2024.10624792.
[19]	SU Yongtao, LIU Yaoqi, ZHOU Yiqing, et al. Broadband LEO satellite communications: Architectures and key technologies[J]. IEEE Wireless Communications, 2019, 26(2): 55–61. doi: 10.1109/MWC.2019.1800299.
[20]	王瀚萱, 陈子博, 孙耀华. 低轨卫星通信中的波束跳变和频率复用方法研究[J]. 电子技术应用, 2023, 49(5): 24–29. doi: 10.16157/j.issn.0258-7998.233990. WANG Hanxuan, CHEN Zibo, and SUN Yaohua. Study on beam hopping and frequency reuse in LEO satellite communication[J]. Application of Electronic Technique, 2023, 49(5): 24–29. doi: 10.16157/j.issn.0258-7998.233990.
[21]	3GPP. TR 38.811-2018 Study on New Radio (NR) to support non-terrestrial networks[S]. 2018. (查阅网上资料, 未找到本条文献出版信息, 请确认).
[22]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]. Proceedings of the 31st International Conference on Neural Information Processing Systems, Long Beach, USA, 2017: 6000–6010. doi: 10.5555/3295222.3295349. (查阅网上资料,未找到本条文献doi信息,请确认).
[23]	张弓, 马福建, 聂欣, 等. PULSAR低轨卫星导航星座性能分析研究[J]. 航天器工程, 2024, 33(1): 1–10. doi: 10.3969/j.issn.1673-8748.2024.01.001. ZHANG Gong, MA Fujian, NIE Xin, et al. Study on performance of PULSAR low earth orbit constellation satellite for navigation[J]. Spacecraft Engineering, 2024, 33(1): 1–10. doi: 10.3969/j.issn.1673-8748.2024.01.001.