AI 中文总结
该综述聚焦自适应、非地面及应用集成QKD系统，梳理了ML、RL及QML在QKD多类特定场景的应用，分析了各主题的问题、方案与增益，并指出了相关开放挑战。
AI 中文摘要
量子密钥分发（QKD）基于量子力学定律提供信息论安全，但其实际部署正日益扩展至传统点对点光纤链路之外。多个快速兴起的QKD方向常被分开研究，包括自适应协议与参数支持；自由空间、卫星、无人机（UAV）及高空平台（HAP）信道；与物联网（IoT）和6G网络的集成；量子安全联邦学习；量子机器学习（QML）辅助决策支持；以及面向单侧设备无关QKD的可操纵性感知估计。本综述研究机器学习（ML）、强化学习（RL）和QKD如何解决这些特定场景，并将文献组织为五大主题支柱：（I）自适应协议与参数支持；（II）自由空间、卫星、UAV及HAP辅助QKD；（III）面向IoT、6G及量子安全联邦学习的QKD；（IV）QML辅助QKD功能；（V）可操纵性感知与单侧设备无关QKD安全估计。针对每个主题，我们遵循统一的问题、传统解决方案及ML/RL/QML解决方案结构，并使用准确率、平均绝对百分比误差、量子比特错误率（QBER）降低、密钥生成率提升等指标总结报告的增益。我们还提供主题内及跨主题比较表，并确定了开放挑战，包括数据集稀缺性、天气与移动条件下的可迁移性、可解释性、可信QML，以及基于ML的决策支持与安全认证之间的边界。本综述为自适应、非地面及应用集成QKD系统提供了聚焦的参考。
英文摘要
Quantum Key Distribution (QKD) provides information-theoretic security grounded in the laws of quantum mechanics, yet practical deployment increasingly extends beyond conventional point-to-point fiber links. Several rapidly emerging QKD directions are often studied separately, including adaptive protocol and parameter support; free-space, satellite, UAV, and high-altitude platform (HAP) channels; integration with IoT and 6G networks; quantum-secured federated learning; Quantum Machine Learning (QML) assisted decision support; and steerability-aware estimation for one-sided device-independent QKD. This survey examines how Machine Learning (ML), Reinforcement Learning (RL), and QML address these specialized scenarios and organizes the literature into five thematic pillars: (I) adaptive protocol and parameter support; (II) free-space, satellite, UAV, and HAP-assisted QKD; (III) QKD for IoT, 6G, and quantum-secured federated learning; (IV) QML-assisted QKD functions; and (V) steerability-aware and one-sided device-independent QKD security estimation. For each theme, we follow a consistent problem, conventional solution, and ML/RL/QML solution structure and summarize reported gains using metrics such as accuracy, mean absolute percentage error, QBER reduction, and secret key rate improvement. We further provide thematic and cross-theme comparison tables and identify open challenges, including dataset scarcity, transferability across weather and mobility conditions, interpretability, trustworthy QML, and the boundary between ML-based decision support and security certification. This survey serves as a focused reference for adaptive, non-terrestrial, and application-integrated QKD systems.
