Real-time Network Structure Identification using the Koopman Operator
Real-time Network Structure Identification using the Koopman Operator
复制标题
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Zhuanglin Mei;T. Oguchi
中科院分区：
文献类型：
--
作者：
Zhuanglin Mei;T. Oguchi
本文研究网络结构的识别问题。特别地，我们考虑由一组具有相同非受迫动力学的互连动力系统形成的网络，并将网络的结构建模为节点之间的耦合函数。对于这样的网络系统，我们提出了一种估计方法来确定所有节点的测量数据的耦合函数。该方法结合Koopman算子理论和稀疏识别方法来识别非线性耦合函数。此外，我们开发了一个在线算法来实现所提出的方法在实时。一个8节点振荡器网络的数值例子来证明所提出的方法的有效性。
: This paper considers the identification problem of network structures. In particular, we consider networks formed by a group of interconnected dynamical system with identical unforced dynamics and model the structure of the network as the coupling function among the nodes. For such network systems, we propose an estimation method to determine the coupling functions from measurement data of all the nodes. The proposed method combines the Koopman operator theory and the sparse identification method to identify nonlinear coupling functions. In addition, we develop an online algorithm to implement the proposed method in real-time. A numerical example of an 8-node oscillator network is presented to demonstrate the usefulness of the proposed method.