Synchronization Control of Networked Two-Timescale Dynamic Agents: A Singular Perturbation Approach
Kun Liang,Wangli He,Feng Qian,Guanrong Chen
DOI: https://doi.org/10.1109/tcns.2023.3292054
IF: 4.347
2023-01-01
IEEE Transactions on Control of Network Systems
Abstract:This paper is intended to solve the synchronization control problem for a group of agents with two-timescale characteristic, described by singularly perturbed systems (SPSs) with a small singular perturbation parameter $\varepsilon$ . Three fundamental and yet challenging questions are addressed: 1) How to design a distributed controller to guarantee synchronization of coupled two-timescale agents for $\varepsilon \in (0,\varepsilon ^{*})$ , in which the stability bound $\varepsilon ^{*}$ has to be determined; 2) How to enlarge the stability bound $\varepsilon ^{*}$ for a given feedback gain matrix; 3) How to cope with the situation when the singular perturbation parameter exceeds the prescribed stability bound $\varepsilon ^{*}$ . First, a decoupled method is applied to transfer the synchronization problem of networked two-timescale dynamic agents to the stability problem of SPSs associated with the eigenvalues of the network Laplacian matrix. Second, based on a specially constructed Lyapunov function, sufficient conditions are derived for simultaneously stabilizing the decoupled systems and computing the stability bound. Furthermore, a particle swarm optimization (PSO)-assisted method is utilized to find a larger stability bound. Third, with the derived feedback gain matrix, some criteria are established to further enlarge the stability bound $\varepsilon ^{*}$ . Then, an integrated algorithm is developed to design a distributed controller with an even larger stability bound $\varepsilon ^{*}$ . Fourth, a novel network design method, incorporated with the concept of synchronization region, is proposed as an alternative and powerful solution to deal with the case when the feedback gain matrix is incapable of stabilizing the system. Finally, two examples are given to verify the theoretical results.