Deep Learning-Based Channel Parameter Estimation for Free-Space Quantum Communication

BI Huicong; LIU Tao; ZHANG Rongxiang; ZHOU Yuncheng; YUAN Zibo

doi:10.3788/gzxb20265507.0727001

Chinese

您当前的位置：

首页 >

文章列表页 >

Deep Learning-Based Channel Parameter Estimation for Free-Space Quantum Communication

更新时间：2026-06-12

- Deep Learning-Based Channel Parameter Estimation for Free-Space Quantum Communication
- Acta Photonica Sinica Pages: 1-14(2026)
- 作者机构：
  
  1.河北大学物理科学与技术学院，保定 071002
  2.华北电力大学燕赵电力实验室，保定 071003
  3.华北电力大学电子与通信工程系，保定 071003
  4.华北电力大学河北省电力物联网技术重点实验室，保定 071003
- 作者简介：
- 基金信息：
  
  the Research and Practice Project for Higher Education Teaching Reform in Hebei Province(2025GJJG035);National Natural Science Foundation of China(62071180)
- DOI：10.3788/gzxb20265507.0727001
  CLC： TN929.12
- CSTR：32255.14.gzxb20265507.0727001
- Received：02 March 2026，
  
  Revised：2026-04-30，
  
  Accepted：30 April 2026，
- 稿件说明：
移动端阅览
毕慧聪，刘涛，张荣香，等. 基于深度学习的自由空间量子通信信道参数估计［J］.光子学报，2026，55（7）：0727001

BI Huicong， LIU Tao， ZHANG Rongxiang， et al. Deep Learning-Based Channel Parameter Estimation for Free-Space Quantum Communication［J］. Acta Photonica Sinica， 2026， 55（7）：0727001
毕慧聪，刘涛，张荣香，等. 基于深度学习的自由空间量子通信信道参数估计［J］.光子学报，2026，55（7）：0727001 DOI： 10.3788/gzxb20265507.0727001. CSTR： 32255.14.gzxb20265507.0727001.

BI Huicong， LIU Tao， ZHANG Rongxiang， et al. Deep Learning-Based Channel Parameter Estimation for Free-Space Quantum Communication［J］. Acta Photonica Sinica， 2026， 55（7）：0727001 DOI： 10.3788/gzxb20265507.0727001. CSTR： 32255.14.gzxb20265507.0727001.

摘要

本文围绕自由空间量子通信中大气信道参数精确估计问题，提出一种TCN-Transformer-CrossAttention（TTC）融合预测模型，构建面向量子通信应用的能见度与衰减预测框架。该方法充分发挥时间卷积神经网络对局部多尺度时序模式的提取能力与变换器对长程依赖关系建模的优势，通过交叉注意力实现不同特征表征间的对齐与自适应加权融合，从而实现端到端的信道衰减系数多步预测。实验以2020—2025年保定地区逐小时气象数据为基础，采用均方根误差、均方误差、平均绝对误差与决定系数作为评价指标，并与多种典型模型进行对比验证。结果表明，所提模型在不同预测步长下均表现出较高拟合度，一步预测决定系数达到95.68%，均方根误差为0.0915；在两步与三步预测任务中仍能保持稳定性能；与BPNN、LSTM、GRU、Transformer、TCN-Transformer以及Transformer-CrossAttentio等模型相比，所提方法在误差指标与拟合优度方面均取得最优结果，验证了其在自由空间量子通信信道衰减参数估计与链路状态预测中的应用潜力。

Abstract

Atmospheric channel parameter estimation is a key issue in free space quantum communication systems. Since transmission performance and link stability are very sensitive to changes in meteorological conditions. Among these factors， visibility and channel attenuation are particularly important， because they directly affect the transmission loss of photons， thus affecting the reliability of quantum links. Aiming at the problem of short-term accurate prediction of atmospheric attenuation parameters， this paper proposes a TCN-Transformer-Cross Attention （ TTC ） fusion model， and establishes a visibility an

d attenuation coefficient prediction framework for free space quantum communication applications. By combining the ability of Temporal Convolutional Networks （ TCN ） to extract local multi-scale temporal patterns and the advantages of Transformer architecture in modeling long-distance dependencies， and further introducing cross-attention mechanism， this method realizes end-to-end multi-step prediction of channel attenuation coefficient. Based on the hourly meteorological observation data of Baoding City from 2020 to 2025， the effectiveness of the proposed model is systematically evaluated through multiple prediction tasks and comparative experiments with several representative benchmark models.The attenuation coefficient of the free space quantum communication channel is closely related to the time evolution of atmospheric conditions， especially the meteorological variables related to visibility. Therefore， accurate prediction depends not only on capturing local fluctuations and short-term change patterns， but also on modeling the long-term correlation and feature correlation interactions in the observation sequence. In this work， we first use the hourly meteorological data of Baoding area from 2020 to 2025 to construct a multivariate time series dataset， and model the prediction of channel attenuation coefficient as a supervised sequence learning problem. On this basis， one-step， two-step and three-step prediction tasks are designed to evaluate the short-term prediction ability of the model in different horizons. From the perspective of modeling， the proposed TTC framework can be understood as a hierarchical fusion architecture with complementary timing representation mechanism. The TCN branch is responsible for extracting local and multi-scale temporal features from the input sequence. Through causal convolution and dilated convolution， the receptive field can be effectively expanded while preserving the temporal order of observations， so that the model can capture short-term fluctuations， abrupt local changes

and potential periodic structures in atmospheric data. At the same time， Transformer branch is used to learn global context dependence and long-range temporal interaction through self-attention mechanism. Compared with the simple convolution or circular structure， this branch is more suitable for describing the long-range correlation between meteorological observations distributed at different time intervals. In order to further utilize the complementarity between the two branches， a cross-attention module is introduced for feature alignment and adaptive fusion. The Cross-Attention mechanism is not a simple series of local features and global features， but dynamically adjusts the contribution of different representations and enhances the interaction between different time feature spaces， thereby improving the model 's ability to focus on the most informative components for attenuation prediction. In this way， the proposed TTC framework establishes an end-to-end mapping from meteorological time series input to future channel attenuation coefficient， which provides a reliable prediction architecture for free space quantum communication channel estimation.Subsequently， root mean square error （ RMSE ）， mean square error （ MSE ）， mean absolute error （ MAE ） and coefficient of determination （ R

） are used to evaluate the prediction performance of the proposed model， which together provide a comprehensive assessment of the error size and fitting quality. Several typical models such as BPNN， LSTM， GRU， Transformer， TCN-Transformer and Transformer-Cross Attention are compared to verify the effectiveness of the proposed fusion strategy. The experimental results show that the TTC model consistently achieves the best overall performance in different prediction ranges. In the one-step prediction task， the R2 of the proposed method is 95.68 % and the RMSE is 0.0915， indicating that the predicted value is highly consistent with the actual channel attenuation coefficient. The results show that the fusion of local tim

e feature extraction， global dependency modeling and adaptive feature association interaction can significantly improve the prediction accuracy of atmospheric channel parameters. In the two-step and three-step prediction tasks， although the prediction uncertainty increases naturally with the extension of the prediction time domain， the proposed model still maintains stable performance， relatively low error level and strong fitting ability. These results confirm that the TTC architecture not only performs well in real-time forecasting， but also maintains robust short-term trend forecasting capabilities under more challenging multi-step conditions. Further comparison with other neural network models reveals the specific advantages of the proposed design. Compared with BPNN， the TTC model has a stronger ability to characterize the nonlinear time evolution of meteorological data. Compared with loop architectures such as LSTM and GRU， it more effectively integrates local mutation learning and long-range dependence modeling， thereby reducing prediction errors and improving goodness of fit. Compared with the standard Transformer model， the introduced TCN branch enhances the extraction of local multi-scale patterns that may be underemphasized by using self-attention alone. In addition， compared with the mixed variants such as TCN-Transformer and Transformer-Cross Attention， the proposed method has achieved the best results in error-related indicators and fitting indicators， which further verifies the necessity and effectiveness of combining TCN， Transformer and Cross-Attention into a unified fusion framework for atmospheric attenuation coefficient prediction.In general， the TTC model proposed in this paper provides an effective solution for the estimation and short-term prediction of atmospheric channel attenuation coefficient in free space quantum communication. Its ability to jointly model local time structure， global sequence dependence and adaptive feature interaction can achieve accurate and stable prediction in dif

ferent prediction horizons. The results show that the proposed framework has considerable potential in channel attenuation estimation， link state prediction and transmission state evaluation， which can provide useful technical support for improving the reliability and environmental adaptability of free space quantum communication systems.

关键词

Keywords

references

WANG Xiangbin ， PAN Jianwei . The Nobel Prize in Physics 2022： Quantum Entanglement ［J］. Bulletin of National Natural Science Foundation of China ， 2022 ， 36 （ 6 ）： 928 - 930 .

王向斌，潘建伟 . 2022年诺贝尔物理学奖：量子纠缠［J］. 中国科学基金， 2022 ， 36 （ 06 ）： 928 - 930 .

Gisin N ， Ribordy G ， Tittel W ， et al . Quantum cryptography ［J］. Reviews of Modern Physics ， 2002 ， 74 （ 1 ）： 145 - 195 .

Guo guang can . Research status and future of quantum information technology ［J］. Scientia Sinica （Informationis）， 2020 ， 50 （ 9 ）： 1395 - 1406 .

郭光灿 . 量子信息技术研究现状与未来［J］. 中国科学：信息科学， 2020 ， 50 （ 9 ）： 1395 - 1406 .

Lopez-Leyva J A ， Talamantes-Alvarez A ， Ponce-Camacho M A ， et al . Free-Space-Optical Quantum Key Distribution Systems：Challenges and Trends ［M］. Quantum Cryptography in Advanced Networks ， 2019 .

Pan Jianwei . Progress of the quantum experiment science satellite （QUESS） Micius Project ［J］. Chinese Journal of Space Science ， 2020 ， 40 （ 5 ）： 643 - 647 .

Cao Yuan ， Li Yuhuai ， Yang Kuixing ， et al . Long-distance free-space measurement-device-independent quantum key distribution ［J］. Physical Review Letters ， 2020 ， 125 （ 26 ）： 260503 .

BENNETT C H ， BRASSARD G . Quantum cryptography： Public key distribution and coin tossing ［J］. Theoretical Computer Science ， 2014 ， 560 （ 1 ）： 7 - 11 .

BENNETT C H ． Quantum cryptography using any two nonorthogonal states ［J］. Physical Review Letters ， 1992 ， 68 ： 3121 - 3124 .

ARTUR E K . Quantum cryptography based on Bell's theorem ［J］. Physical Review Letters ， 1991 ， 67 （ 6 ）： 661 .

MULLER A ， ZBINDEN H ， GISIN N . Quantum cryptography over 23km in installed under-lake telecom fibre ［J］. EPL （Europhysics Letters）， 1996 ， 33 ： 335 - 339 .

KURTSIEFER C ， ZARDA P ， Halder M ， et al . A step towards global key distribution ［J］. Nature ， 2002 ， 419 （ 6906 ）： 450 .

URSIN R ， TIEFENBACHER F ， SCHMITT MANDERBACH T ， et al . Entanglement-based quantum communication over 144 km ［J］. Nature Physics ， 2007 ， 3 （ 7 ）： 481 - 486 .

YIN Juan ， CAO Yuan ， LI Yuhuai ， et al . Satellite-based entanglement distribution over 1200 kilometers ［J］. Science ， 2017 ， 356 （ 6343 ）： 1140 - 1144 .

YIN Juan ， LI Yuhuai ， LIAO Shengkai ， et al . Entanglement-based secure quantum cryptography over 1120 kilometres ［J］. Nature ， 2020 ， 582 （ 7813 ）： 501 - 505

Liu Jianlong ， Luo Xiyu ， Yu Yong ， et al . Creation of memory-memory entanglement in a metropolitan quantum network ［J］. Nature ， 2024 ， 629 ： 579 - 58

Liu Jingyang ， Ma Xiao ， Ding Huajian ， et al . Experimental demonstration of five-intensity measurement-device-independent quantum key distribution over 442 km ［J］. Physical Review A ， 2023 ， 108 （ 2 ）： 022605 .

Liu Yang ， Zhang Weijun ， Jiang Cong ， et al . Experimental twin-field quantum key distribution over 1000 km fiber distance ［J］. Physical Review Letters ， 2023 ， 130 （ 21 ）： 210801 .

HAN Liqiang ， YOU Yahui . Performance of multiple-input multiple-output free space optical communication under atmospheric attenuation and turbulence effects ［J］. Chinese Journal of Lasers ， 2016 ， 43 （ 7 ）： 0706004 .

韩立强，游雅晖 . 大气衰减和大气湍流效应下多输入多输出自由空间光通信的性能［J］. 中国激光， 2016 ， 43 （ 7 ）： 0706004 . DOI: 10.3788/cjl201643.0706004 http://dx.doi.org/10.3788/cjl201643.0706004

LIU Tao ， ZHU Cong ， SUN Chunyang ， et al . Influence of Different Weather Conditions on Performance of Free-Space Quantum Communic ation System ［J］. Acta Optica Sinica ， 2020 ， 40 （ 2 ）： 0227001 .

刘涛，朱聪，孙春阳，等 . 不同天气条件对自由空间量子通信系统性能的影响［J］. 光学学报， 2020 ， 40 （ 2 ）： 0227001 . DOI: 10.3788/aos202040.0227001 http://dx.doi.org/10.3788/aos202040.0227001

NIE Min ， CHEN Wei ， ZHANG Meiling ， et al . Influence of Haze on Free-Space Quantum State Transmission ［J］. Journal of Xi'an University of Posts and Telecommunications ， 2015 ， 20 （ 1 ）： 19 - 23 .

聂敏，陈伟，张美玲，等 . 雾霾对自由空间量子态传输的影响［J］. 西安邮电大学学报， 2015 ， 20 （ 1 ）： 19 - 23 .. DOI: 10.13682/j.issn.2095-6533.2015.01.004 http://dx.doi.org/10.13682/j.issn.2095-6533.2015.01.004

Xu Jiaxin ， Ma Xiao ， Liu Jingyang ， et al . Automatically identifying imperfections and attacks in practical quantum key distribution systems via machine learning ［J］. Science China Information Sciences ， 2024 ， 67 （ 10 ）： 202501 .

Liu Jingyang ， Jiang Qinging ， Ding Huajian ， et al . Highly efficient twin-field quantum key distribution with neural networks ［J］. Science China. Information Sciences ， 2023 ， 66 （ 8 ）： 189402 .

Ren Ziang ， Chen Yipeng ， Liu Jingyang ， et al . Implementation of machine learning in quantum key distributions ［J］. IEEE Communications Letters ， 2020 ， 25 （ 3 ）： 940 - 944 .

Liu Jingyang ， Ding Huajian ， Zhang Chunmei ， et al . Practical phase-modulation stabilization in quantum key distribution via machine learning ［J］. Physical Review Applied ， 2019 ， 12 （ 1 ）： 014059 .

WANG Kai ， ZHAO Hong ， LIU Aixia ， et al . Atmospheric visibility prediction based on risk neural network ［J］. China Environmental Science ， 2009 ，（ 10 ）： 1029 - 1033 .

王恺，赵宏，刘爱霞，等 . 基于风险神经网络的大气能见度预测［J］. 中国环境科学， 2009 ，（ 10 ）： 1029 - 1033 .

Boudala F S ， Isaac G A ， Crawford R W ， et al . Parameterization of Runway Visual Range as a Function of Visibility：Implications for Numerical Weather Prediction Models ［J］. Journal ofAtmospheric & Oceanic Technology ， 2011 ， 29 （ 2 ）： 177 - 191 .

CHEN Qu ， XIA Bin ， WANG Song . Visibility prediction based on NARX neural network ［C］. The 8th Yangtze River Delta Science and Technology Forum——Meteorological Science and Technology Development Forum .

陈曲，夏斌，王淞 . 基于NARX神经网络的能见度预测［C］. 第八届长三角科技论坛——气象科技发展论坛 .

Zhou Binbin ， Du Jun ， Gultepe I ， et al . Forecast of Low Visibility and Fog from NCEP：Current Status and Efforts ［J］. Pure & Applied Geophysics ， 2012 ， 169 （ 5-6 ）： 895 - 909 .

GUAN Qin ， LI Qingping ， HUANG Lingguang ， et al . Visibility forecast of fog day in Taiyuan ［J］. Journal of Meteorology and Environment ， 2013 ， 29 （ 4 ）： 47 - 53 .

管琴，李青平，黄铃光，等 . 太原雾天能见度预报［J］. 气象与环境学报， 2013 ， 29 （ 4 ）： 47 - 53 .

MA Chuyan ， ZU Jian ， FU Qingpan ， et al . Prediction of atmospheric visibility based on genetic neural network model ［J］. Chinese Journal of Environmental Engineering ， 2015 ， 9 （ 4 ）： 1905 - 1910 .

马楚焱，祖建，付清盼，等 . 基于遗传神经网络模型的空气能见度预测［J］. 环境工程学报， 2015 ， 9 （ 004 ）： 1905 - 1910 .

Zhu Lei ， Zhu Guodong ， Han Lei ， et al . The application of deep learning in airport visibility forecast ［J］. Atmospheric and Climate Sciences ， 2017 ， 7 （ 03 ）： 314 .

Lu Zhenyu ， Lu Bingjian ， Zhang Hengde ， et al . A method of visibility forecast based on hierarchical sparserepresentation ［J］. Journal of Visual Communication & Image Representation ， 2019 ， 58 ： 160 - 165 .

ZHANG Jie ， XIONG Quan . Visibility forecast of expressway based on BP artificial neural network ［J］. Science and Technology & Innovation ， 2019 ， 18 ： 76 - 77 .

张杰，熊权 . 基于BP人工神经网络的高速公路能见度预报［J］. 科技与创新， 2019 ， 000 （ 018 ）： 76 - 77 .

Lu Zhenyu ， Zheng Cheng ， Yang Tingya . Application of offshore visibility forecast based on temporal convolutional network and transfer learning ［J］. Computational Intelligence and Neuroscience ， 2020 ， 2020 （ 1 ）： 8882279

YU Dongchang ， ZHAO Wenfang ， NIE Kai ， et al . Visibility forecast model based on LightGBM algorithm ［J］. Journal of Computer Applications ， 2021 ， 41 （ 4 ）： 1035 - 1041 .

余东昌，赵文芳，聂凯，等 . 基于LightGBM算法的能见度预测模型［J］. 计算机应用， 2021 ， 41 （ 4 ）： 1035 - 1041 .

Peláez-RodríguezC. ， Pérez-AracilJ. ， de Lopez-DizA. ， et al . Deep learning ensembles for accurate fog-related low-visibility events forecasting ［J］. Neuroco puting ， 2023 ， 549 ： 126435 .

ZHENG Chen ， WANG Xuan ， WANG Lijie ， et al . Atmospheric extinction characteristics and visibility prediction model of typical haze processes at airport ［J］. Environmental Engineering ， 2024 ， 42 （ 3 ）： 215 - 224 .

郑宸，王璇，王立婕，等 . 机场典型霾过程大气消光特性及能见度预测模型［J］. 环境工程， 2024 ， 42 （ 03 ）： 215 - 224 .

MA Yuzhao ， ZHANG Yuhang ， WANG Lingfei . Long-term airport visibility prediction based on TieGAN-Informer ［J］. Journal of Safety and Environment ， 2024 ， 19 （ 3 ）： 1 - 14 .

马愈昭，张宇航，王凌飞 . TieGAN-Inforer长时机场能见度预测［J］. 安全与环境学报， 2024 ， 19 （ 03 ）： 1 - 14 .

HAN Liqiang ， WANG Qi ， XINTAI Kegui . Performance of free space optical communication in Gamma-gamma atmospheric turbulence ［J］. Infrared and Laser Engineering ， 2011 ， 40 （ 7 ）： 1318 - 1322 .

韩立强，王祁，信太克归 . Gamma-gamma 大气湍流下自由空间光通信的性能［J］. 红外与激光工程， 2011 ， 40 （ 7 ）： 1318 - 1322 .

Kim I I ， McArthur B ， Korevaar E J . Comparison of laser beam propagation at 785 nm and 1550 nm in fog and haze for optical wireless communications ［C］. Proceedings of SPIE ， 2001 ， 4214 ： 26 - 37 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Research on Structured Light Wrapped Phase Unwrapping Algorithm Based on Deep Learning

Spatial-spectral Collaborative Unrolling Network for Pansharpening

Classification Method of Breast Tissue OCT Images Based on a Double Filtering Residual Network

Remote Sensing Image Fusion Based on Two-branch U-shaped Transformer

Object Detection Method of Optical Remote Sensing Image with Multi-attention Mechanism

Related Author

YANG Simin

ZHANG Yandong

WANG Yang

SHANG Jianhua

ZHENG Jianwei

XIA Hongyi

XU Honghui

DING Lihao

Related Institution

School of Information and Intelligent Science， Donghua University

Joint Laboratory of High Power Laser and Physics， Shanghai Institute of Optics and Fine Mechanics， Chinese Academy of Sciences

School of Computer Science and Technology， Zhejiang University of Technology

School of Electronic Engineering and Optoelectronic Technology ， Nanjing University of Science and Technology

College of Data Science， Taiyuan University of Technology

AI问答

Postal code：100079
Tel：（010）53879206 Email：tmw@bjxintong.com.cn
Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09082226号-64 京公网安备11010602201714号
It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.

⁰