| 引用本文: | 李芮,杨洪勇,潘龙硕.具有隐私保护的多智能体系统自适应分布式优化控制[J].控制理论与应用,2026,43(5):1011~1022.[点击复制] |
| LI Rui,YANG Hong-yong,PAN Long-shuo.Adaptive distributed optimization control for privacy-preserving multi-agent systems[J].Control Theory & Applications,2026,43(5):1011~1022.[点击复制] |
|
| 具有隐私保护的多智能体系统自适应分布式优化控制 |
| Adaptive distributed optimization control for privacy-preserving multi-agent systems |
| 摘要点击 364 全文点击 21 投稿时间：2024-12-28 修订日期：2025-11-16 |
| 查看全文 查看/发表评论 下载PDF阅读器 HTML |
| DOI编号 10.7641/CTA.2025.40646 |
| 2026,43(5):1011-1022 |
| 中文关键词 多智能体系统 分布式优化 隐私保护 自适应控制 梯度跟踪 |
| 英文关键词 multi-agent systems distributed optimizatio privacy preservation adaptive control gradient tracking |
| 基金项目 国家自然科学基金项目(61673200), 山东省自然科学基金项目(ZR2022MF231), 鲁东大学研究生创新项目(IPGS2025-073)资助. |
|
| 中文摘要 |
| 本文研究了具有隐私保护的多智能体系统分布式优化问题. 在实际应用中, 多智能体系统经常面临隐私泄
露的风险, 本文构造了一个状态加密函数, 用于保护相互通信的智能体之间的信息安全, 防止信息泄露; 设计了自适
应控制参数, 利用智能体之间的状态差异动态调整通信权重, 加快系统的收敛速度. 假设系统的全局目标函数是所
有智能体的局部目标函数之和, 构造了目标函数的梯度跟踪方法, 对平均梯度和进行局部估计. 提出了一个分布式
优化控制算法, 在没有全局信息的情况下, 智能体利用局部信息即可实现全局最优. 通过对系统的性能进行分析,
研究表明该方法能够保护多智能体系统的数据隐私, 同时使得系统的运动轨迹快速收敛至目标函数的最优解. 最
后, 通过仿真实验, 验证了隐私保护方法和分布式控制协议的有效性. |
| 英文摘要 |
| This paper investigates the distributed optimization problem for multi-agent systems (MASs) with privacy
preservation. In practice, MASs often face the risk of privacy leakage. To address this, this paper proposes a state encryption
function to protect the information exchanged between communicating agents and prevent potential data leakage. Adaptive
control parameters are designed to dynamically adjust communication weights based on the state differences between
agents, thereby accelerating the system’s convergence rate. Assuming that the global objective function of the system is
the sum of the local objective functions of all agents, the paper develops a gradient tracking method to estimate the average
gradient sum. A distributed optimization control algorithm is proposed, enabling agents to achieve global optimization
using local information without relying on global data. Through performance analysis, the proposed method is shown to
effectively protect the data privacy of the MASs while ensuring that the motion trajectories of the agents rapidly converge to
the optimal solution of the objective function. Finally, the effectiveness of the privacy-preserving method and the distributed
control protocol is validated through simulation experiments. |
|
|
|
|
|