基于事件驱动的多智能体系统参数辨识

从一致性原理出发，针对Box-Jenkins多模型表示的分布式多智能体系统参数辨识问题，提出一种事件驱动的异步通信协议来辨识系统参数。首先，采用智能体的参数估计信息设计事件触发机制，并引入激活函数决定智能体间的通信触发。其次， 采用辅助模型思想解决系统不可观测的噪声和中间变量。然后，在通信协议中引入多新息理论提高参数估计精度。在网络拓扑无向联通情况下，各个智能体以完全分布式的方式迭代更新模型参数。在此基础上，讨论了算法收敛性问题。最后，通过仿真实例验证所提方法的有效性。

Parameter identification for multi-agent systems under Event-triggered communications

Based on the consistency principle, an event-triggered asynchronous communication protocol is proposed to solve the parameter identification problem of distributed multi-agent systems represented by Box- Jenkins multi-models. Firstly, the event triggering mechanism is designed by using the parameter estimation information of agents, and the activation function is introduced to determine the communication triggering between agents. Secondly, the auxiliary model is used to solve the issue of the unobservable noise and internal variables. Then, the theory of multi-innovation is introduced into the communication protocol to improve the accuracy of parameter estimation. In the case of undirected connectivity of network topology, each agent updates the model parameters iteratively in a fully distributed way On the basis, the convergence of the algorithm is discussed. Finally, the effectiveness of the proposed method is verified by the results of the simulation example.