复合网络攻击下离散时间多智能体系统的云端预测控制  

Cloud Predictive Control of Discrete-time Multi-agent System under Compound Cyber Attacks

作　　者：贾新春[1] 张媛 池小波[2] 阮艺琳 JIA Xinchun;ZHANG Yuan;CHI Xiaobo;RUAN Yilin(School of Automation and Sofrware Engineering.Shanri University,Taiyuan 030013,China;School of Mathematical Sciences.Shanxi University,Taiyuan 030006 China)

机构地区：[1]山西大学自动化与软件学院,山西太原030013 [2]山西大学数学科学学院,山西太原030006

出　　处：《山西大学学报（自然科学版）》2020年第3期499-507,共9页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(61803244,61973201);山西省回国留学人员科研资助项目(2017-024)。

摘　　要：文章研究复合网络攻击下离散时间多智能体系统(DMASs)的云端预测控制设计问题,其中不同智能体的输出采用不同速率的传感器来采样,且云端控制器到执行器间的网络传输通道遭受重放攻击和欺骗攻击的共同影响。首先设计一个基于异步采样输出的观测器来估计系统状态,考虑到网络因素对多智能体系统的影响,引入云端模拟攻击机制来补偿观测器和预测控制器所用的控制输入不一致情形,一种新颖的云端预测控制协议被提出用来补偿网络诱导时延和复合网络攻击。其次,将带有预测控制机制的DMASs建模为依赖于阈值误差的时滞系统模型,并且通过Lyapunov稳定性定理和LMI方法给出了DMASs估计状态一致性的判据。最后,通过一个数值例子验证了所提方法的有效性。The paper studies the cloud predictive controller design problem of discrete-time multi-agent systems(DMASs)under compound network attacks,where the output vectors of different agents are sampled using sensors with different rates,and the network transmission channel between the cloud controller and the actuator is affected by both replay attacks and deception attacks.An asynchronous sampling outputs based observer is designed to estimate system states.During this process,we consider the influence of network factors on the multi-agent system and introduce a cloud-based simulated attack mechanism to compensate for the inconsistencies in the control inputs used by the observer and the predictive controller.Moreover,a novel cloud predictive control protocol is proposed to compensate for network-induced delays and compound network attacks.Second,the DMAS with predictive control protocol is modeled as a threshold error dependent time-delay system,and the criterion of the state consensus of the closed-loop DMAS is given by the Lyapunov stability theorem and LMI method.The effectiveness of the proposed method is verified by a numerical example。

关 键 词：多智能体系统 网络攻击 异步采样 云端预测控制 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]