摘要
针对通信网络可能遭受多邻居联合窃听的多智能体系统,研究其基于全过程隐私保护的平均一致性问题,具体包括保护智能体的初始状态以及智能体在实现平均一致性整个过程中的实时状态.不同于现有的隐私保护平均一致性算法仅能保护智能体的初始状态且无法抵御联合窃听,提出基于虚拟子网和非消失扰动的全过程隐私保护平均一致性算法.在所提算法下,即使智能体的所有信道都被窃听,仍然可以实现多智能体系统的平均一致性且智能体的状态可以得到全过程保护.最后,通过几个数值仿真实验验证了算法的有效性.
This paper investigates the average consensus problem with whole-process privacy protection for multiagent systems facing potential collaborative eavesdropping from multiple neighbors.The research focuses on protecting both the initial states of agents and their real-time states throughout the entire process of achieving average consensus.Different from existing privacy-preserving average consensus algorithms that only safeguard initial states and cannot resist collaborative eavesdropping,a novel whole-process privacy-preserving average consensus algorithm based on virtual subnetworks and non-vanishing perturbations is proposed.Under the proposed algorithm,even if all communication channels of agents are eavesdropped,the average consensus of the multi-agent system can still be achieved while ensuring whole-process protection of agent states.Finally,several numerical simulation experiments verify the effectiveness of the algorithm.
作者
纪良浩
唐少洪
郭兴
解燕
JI Liang-Hao;TANG Shao-Hong;GUO Xing;XIE Yan(Chongqing Key Laboratory of Image Cognition,Chongqing University of Posts and Telecommunications,Chongqing 400065)
出处
《自动化学报》
北大核心
2025年第6期1359-1370,共12页
Acta Automatica Sinica
基金
国家自然科学基金(62276036)
重庆市自然科学基金创新发展联合基金重点项目(CSTB2024NSCQ-LZX0118)
重庆市教委科技重大项目(KJZD-M202100602)
重庆市教委科学技术研究项目(KJQN202400627)资助。
关键词
多智能体系统
平均一致性
隐私保护
全过程隐私
联合窃听
Multi-agent systems
average consensus
privacy protection
whole-process privacy
collaborative eavesdropping
