基于学习的线性多智能体系统弹性最优协同容错控制

针对一类线性多智能体系统,研究其在网络间歇性拒绝服务攻击下的最优同步控制问题.首先,在时变非对称通讯网络拓扑结构下,提出一种弹性最优协同容错控制策略,并优化多智能体的合作二次性能指标,然后证明全局跟踪误差在出现执行器故障和网络攻击时仍然渐进收敛.进一步,当考虑多智能体子系统模型参数未知,同时系统发生执行器故障的情况下,提出利用局部系统状态和输入信息的自学习迭代算法求解代数Riccati方程,计算子系统的反馈控制器增益,实现弹性协同容错控制目标.最后,通过Chua电路网络仿真算例验证所提出的控制方法的有效性.

Learning-based resilient optimal cooperative fault tolerant control for linear multi-agent systems

In this paper, the resilient optimal coordination control of linear multi-agent systems (MASs) under network intermittent denial of service (DoS) attacks is studied. Firstly, a resilient optimal fault-tolerant control strategy for MASs under a time-varying asymmetric communication network is proposed, and the cooperative quadratic performance index of the global agent network is optimized. Then, it is proved that the global tracking errors can asymptotically convergent to zero even in the case of actuator failures and network attacks. Further, when considering the unknown parameters in a MAS model with the actuator failures, a self-learning iterative algorithm based on local system state and input information is proposed to solve the algebraic Riccati equation(ARE). Each agent feedback controller gain can be obtained to achieve the flexible cooperative fault-tolerant control objective. Finally, the effectiveness of the proposed control method is verified by the Chua''s circuit network in a simulation example.