具有切换拓扑的多智能体系统协同控制的差分隐私保护

网络信息技术的不断发展与普及使得各类数据的发布采集变得方便与便捷, 但数据的直接发布势必会造成个网络信息的泄露和敏感信息的失密, 因此敏感信息的保护成为了各行各业关注的问题. 本文研究了基于固定拓扑和切换拓扑的多智能体系统协同控制的差分隐私保护问题, 将差分隐私算法与传统平均一致性算法结合, 提出了具有隐私保护的协同控制算法, 分析了隐私保护算法对分布式协同控制闭环系统稳定性的影响. 基于所提算法, 应用矩阵论和概率统计对隐私保护协同控制算法的收敛性和隐私性进行理论分析, 该算法可以保护智能个体的数据隐私, 同时可以使得系统运动实现均方一致. 在系统拓扑结构动态变化的情况下, 本文对该算法的收敛性和隐私性进行理论分析, 讨论了切换拓扑对隐私保护的影响. 最后的仿真示例验证了理论结果的正确性.

Differential privacy protection for cooperative control of multi-agent systems with switching topologies

With the continuous development and popularization of network information technology, the release andcollection of all kinds of data become more and more convenient. However, the direct release of data will inevitably leadto the leakage of network information and the loss of sensitive information, so the protection of sensitive informationhas become a concern of all walks of life. In this paper, the differential privacy problem of cooperative control for firstordermulti-agent systems based on fixed topology and switched topology is studied. A cooperative control algorithmwith privacy is proposed by combining the differential privacy algorithm and the average consistency algorithm, and theinfluence of privacy protection algorithm on the stability of distributed cooperative control closed-loop system is analyzed.Based on the proposed algorithm, the convergence and the privacy of privacy protection cooperative control algorithmare theoretically analyzed by using matrix theory and probability statistics. The algorithm can protect the data privacyof intelligent individuals and make the system motion achieve mean square consistency. In the case of dynamic systemtopologies, the convergence and privacy of the algorithm is analyzed, and the impact of switching topologies on privacyprotection is discussed. Finally, numerical simulations are shown to verify the correctness of the results.