Deng Chao, Jin Xiao-Zheng, Che Wei-Wei, Wang Hai
IEEE Trans Neural Netw Learn Syst. 2022 Oct;33(10):5504-5513. doi: 10.1109/TNNLS.2021.3070869. Epub 2022 Oct 5.
In this article, we consider the distributed fault-tolerant resilient consensus problem for heterogeneous multiagent systems (MASs) under both physical failures and network denial-of-service (DoS) attacks. Different from the existing consensus results, the dynamic model of the leader is unknown for all followers in this article. To learn this unknown dynamic model under the influence of DoS attacks, a distributed resilient learning algorithm is proposed by using the idea of data-driven. Based on the learned dynamic model of the leader, a distributed resilient estimator is designed for each agent to estimate the states of the leader. Then, a new adaptive fault-tolerant resilient controller is designed to resist the effect of physical failures and network DoS attacks. Moreover, it is shown that the consensus can be achieved with the proposed learning-based fault-tolerant resilient control method. Finally, a simulation example is provided to show the effectiveness of the proposed method.
在本文中，我们考虑了异构多智能体系统（MASs）在物理故障和网络拒绝服务（DoS）攻击下的分布式容错弹性共识问题。与现有的共识结果不同，本文中所有跟随者都不知道领导者的动态模型。为了在DoS攻击的影响下学习这个未知的动态模型，利用数据驱动的思想提出了一种分布式弹性学习算法。基于所学习的领导者动态模型，为每个智能体设计了一个分布式弹性估计器来估计领导者的状态。然后，设计了一种新的自适应容错弹性控制器来抵抗物理故障和网络DoS攻击的影响。此外，结果表明，所提出的基于学习的容错弹性控制方法能够实现共识。最后，提供了一个仿真例子来说明所提方法的有效性。
