-
空间信道连续变量量子密钥分发协议(Continuous-Variable Quantum Key Distribution,CV-QKD)工作在光学衍射极限、通信距离极限、光电探测极限条件下，协议参数（如调制方差VA）的优化选择会影响协议的可行性。而低轨卫星和地面站始终处于高速相对运动中，可视窗口时间有限且在轨计算资源受限，传统优化算法难以满足空间信道快速动态变化的实时优化需求。本文提出了空间信道高斯调制CV-QKD的Unet网络参数预测优化方法，搭建空间CV-QKD链路仿真平台，改变轨道高度、天顶角等组合参数生成126,575组训练数据集，利用Unet网络的对称结构和特征融合能力实现近实时地预测调制方差VA。仿真结果表明Unet网络在6,328组跨轨道高度(510-710km)和过量噪声水平(0.01-0.03)的测试数据中可以达到99.25%-99.41%的预测准确率，同时相较于局部搜索算法14,754秒的基准耗时，Unet将推理时间缩短至1.08秒（加速比达1.48×106），为后续空间信道CV-QKD实验的参数实时优化提供了理论支撑。
-
关键词:
- 连续变量量子密钥分发 /
- 参数优化 /
- 机器学习
The field of quantum communication has seen significant advancements with Continuous-Variable Quantum Key Distribution (CV-QKD), which operates under stringent conditions such as optical diffraction limits, maximum communication distances, and photoelectric detection limits. The optimization of protocol parameters, particularly the modulation variance (VA), is crucial for the feasibility of CV-QKD. However, in space-to-ground CV-QKD scenarios, the high-speed relative motion between low-earth-orbit (LEO) satellites and ground stations, coupled with limited on-board computing resources, poses challenges for conventional optimization algorithms to meet the real-time demands of rapidly changing space channels.
To address these challenges, this paper introduces a novel method for optimizing Gaussian-modulation CV-QKD in space channels using a Unet-based approach. We have developed a comprehensive simulation platform for CV-QKD links, generating a substantial training dataset of 126,575 samples by varying parameters such as orbital height and zenith angle. The Unet network, renowned for its symmetric architecture and powerful feature fusion capabilities, is utilized to achieve near-real-time prediction of modulation variance. Our simulation results demonstrate the effectiveness of the proposed method, with the Unet network achieving a remarkable prediction accuracy of 99.25%–99.41% on 6,328 datasets with orbital heights ranging from 510 to 710 km and excess noise levels between 0.01 and 0.03.
Compared to the local search algorithm, which takes 14,754 seconds, the Unet-based approach significantly reduces the inference time to just 1.08 seconds, representing a speed-up ratio of 1.48×106. These findings provide a solid theoretical foundation for real-time parameter optimization in future space-channel CV-QKD experiments, offering a significant advancement in the field of quantum communication. The proposed method not only enhances the efficiency of parameter optimization but also ensures the security and reliability of CV-QKD in dynamic space environments. -
[1] Luo Y Z, Ma L J, Sun M S 2024Acta Phys. Sin. 73 240301(in Chinese)[罗一振, 马洛嘉, 孙铭烁2024物理学报73240301]
[2] Liao S K, Cai W Q, Liu W Y, Zhang K, Li J, Ren J G, Yin J, Shen Q, Cao Y, Li Z P, Wu F C, Li W Q, Liu S L, Ren J G, Peng C Z, Chen X W, Pan J W 2017Nature 549 43
[3] Yin J, Li Y H, Liao S K, Yang M, Cao Y, Zhang L, Ren J G, Cai W Q, Liu W Y, Li S L, Shu R, Huang Y M, Deng L, Li L, Zhang Q, Liu N L, Chen Y A, Lu C Y, Wang X B, Peng C Z, Pan J W 2020Nature 582 501
[4] Ren J G, Xu P, Yong H L, Zhang L, Liao S K, Yin J, Liu W Y, Cai W Q, Yang M, Li L, Yang K X, Han X, Yao Y Q, Li J, Wu H Y, Wan S L, Liu L, Liu D Q, Kuang Y W, He Z P, Shang P, Guo C, Zheng K H, Tian K, Zhu Z C, Liu N L, Lu C Y, Shu R, Chen Y A, Peng C Z, Wang X B, Pan J W 2017Nature 549 70
[5] Dubey U, Bhole P, Dutta A, Goyal S K, Behera B K, Panigrahi P K 2023arXiv 2309.13417
[6] Harney C, Fletcher A I, Pirandola S 2022Phys. Rev. Appl. 18 014012
[7] Long N K, Malaney R, Grant K J 2023Information 14 553
[8] Zhou Z C, Guo Y 2024Electronics 13 1410
[9] Wang W, Lo H K 2019Phys. Rev. A 100 062334
[10] Jin D, Guo Y, Wang Y, Li Y B, Wang T Y 2021Phys. Rev. A 104 012616
[11] Long N K, Malaney R, Grant K J 2024Proc. SPIE 13106 1310602
[12] Liu Z P, Zhou M G, Liu W B, Wang P, Liu J Y, Guo Y 2022Opt. Express 30 15024
[13] Mao Y, Huang W, Zhong H, Liao Q, Zhang S L, Guo Y 2020New J. Phys. 22 083073
[14] Liao Q, Xiao G, Zhong H, Guo Y, Huang D 2020New J. Phys. 22 083086
[15] Liu W, Huang P, Peng J, Wang S Y, Huang D 2018Phys. Rev. A 97 022316
[16] Yin X H, Wang Y C, Li D Y 2021J. Software 32 519(in Chinese)[殷晓航, 王永才, 李德英2021软件学报32 519]
[17] Chen Y J, Cheng J, Sun X, Liu S H, Zhang Y M 2025Chin. J. Lasers 52 330(in Chinese)[陈宇杰, 程锦, 孙新, 刘胜豪, 张一鸣2025中国激光52330]
[18] Hornik K, Stinchcombe M, White H 1989Neural Networks 2 359
[19] Liao S K, Lin J, Ren J G, Liu C, Liang H, Yin J, Cao Y, Wu F C, Li S L, Li H, Shu R, Xue G, Li B, Shen Q, Jiang L, Yang L, Wang Z, You L X, Wang Z, Pan J W 2017Chin. Phys. Lett. 34 090302
[20] Pirandola S 2021Phys. Rev. Research 3 023130
[21] Weedbrook C, Lance A M, Bowen W P, Symul T, Ralph T C, Lam P K 2004Phys. Rev. Lett. 93 170504
[22] Katoch S, Chauhan S S, Kumar V 2021Multimedia Tools Appl.
计量
- 文章访问数: 63
- PDF下载量: 0
- 被引次数: 0
下载:
