| 引用本文: | 宁宝玲,李文博,毕强.支持隐私保护的自主无人系统分布式智能运维框架[J].控制理论与应用,2024,41(8):1351~1360.[点击复制] |
| NING Bao-ling,LI Wen-bo,BI Qiang.A distributed intelligent operating framework supporting privacy protection for autonomous unmanned systems[J].Control Theory and Technology,2024,41(8):1351~1360.[点击复制] |
|
| 支持隐私保护的自主无人系统分布式智能运维框架 |
| A distributed intelligent operating framework supporting privacy protection for autonomous unmanned systems |
| 摘要点击 2950 全文点击 147 投稿时间：2022-12-08 修订日期：2024-03-05 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2023.21068 |
| 2024,41(8):1351-1360 |
| 中文关键词 智能运维 自主无人系统 联邦学习 隐私保护 区块链 |
| 英文关键词 Intelligent Operating Autonomous Unmanned System Federated Learning Privacy Protection Blockchain |
| 基金项目 科技部重点研发项目(2021YFB1715000), 国家自然科学基金项目(U1811461, 62022013, 12150007, 62103450, 61832003, 62272137), 黑龙江省高 校大学专项科研资金项目(2022–KYYWF–1122) |
|
| 中文摘要 |
| 自主无人系统的智能运维是提升系统自主性以及安全可靠运行能力的有效手段, 面临分布式运维和隐私
保护两个方面的挑战. 现有工作未充分考虑自主无人系统的计算资源受限特点及对量化隐私保护的需求, 无法应对
上述挑战. 因此, 本文提出了一种支持隐私保护的自主无人系统分布式智能运维框架, 基于联邦学习、区块链和隐
私保护技术, 采用离线学习与在线监测、本地模型与全局模型相结合的方式, 利用基于联邦学习的运维模型训练机
制, 基于区块链的运维模型共享机制, 以及基于差分隐私的局部模型共享机制, 有效解决系统面临的资源受限和隐
私保护问题, 为系统提供安全可靠的运维环境以及支持多端融合、自主更新的运维能力. 实验结果表明, 所设计的
框架可以支持高效智能运维方法的构建, 针对框架所设计的联邦学习和隐私保护方法是有效的. |
| 英文摘要 |
| Constructing intelligent operation method is crucial to the autonomy, safety and reliability of autonomous
unmanned systems, and is challenged by the issues of distributed operation and privacy protection. Without considering the
limitation of computation resources and the requirement of quantitative privacy protection, the existing works can not solve
the above issues for autonomous unmanned systems. Therefore, in this paper, a distributed intelligent operating framework
supporting privacy protection for autonomous unmanned systems is proposed. The essential idea is combining offline
learning and online monitoring to build both local and global operating models, building operation models by federated
learning, sharing models based on the blockchain network, and protecting local models by differential privacy techniques,
such that the issues caused by the limited computing resource and privacy protection are resolved. The experimental results
show that the proposed framework can support the construction of efficient intelligent operation methods for autonomous
unmanned systems and the method designed within the framework is effective |
|
|
|
|
|