Consensus Control of Networked Multi-Agent Systems and Its Applications

本课题提出了基于共识的多智能体系统稳定化研究。特别地，研究了存在非线性执行器故障的多智能体系统的基于记忆的采样数据一致性框架。为了减少状态交换和节约能源资源，相邻代理之间的通信仅基于可变采样间隔的状态样本。作为作动器的两种一般约束，在问题的表述中考虑了有界非线性部分有效性损失和偏置失效。导出了在一定情况下保证一致的充分条件，即线性矩阵不等式条件。与现有的基于lyapunov - krasovskii的方法不同，本文提出的设计框架基于循环函数方法，该方法在设计所需的共识控制增益时减少了保守性。这种不太保守的方法允许更大的采样间隔以及更严重的执行器故障，增加了所提出方法的实用性。为了获得仿真结果，本项目有效地使用了MATLAB的almip解析器和SDPT3求解器。基于隧道二极管电路和非完整移动机器人MAS的仿真结果量化了该方法的有效性。此外，通过求解导出的线性矩阵的不等式，得到采样数据控制器的增益矩阵。此外，本文还研究了基于H∞性能的网络控制系统的鲁棒指数稳定性和Takagi-Sugeno模糊控制综合。

Consensus-based stabilization studies of multi-agent systems are proposed in this project. In particular, a memory-based sampled-data consensus framework for multi-agent systems in the presence of nonlinear actuator faults is studied. To reduce state exchanges and conserve energy resources, communication between neighboring agents is based solely on state samples at variable sampling intervals. As two general constraints of the actuator, both bounded nonlinear partial loss of effectiveness and bias failure are considered in the problem formulation. Sufficient conditions to guarantee consensus under certain circumstances are derived as linear matrix inequality conditions. Unlike existing Lyapunov-Krasovskii-based methods, the design framework proposed in this brief is based on a loop-functional approach that reduces conservatism in the design of the required consensus control gains. This less conservative approach allows for larger sampling intervals as well as more severe actuator failures, increasing the utility of the proposed approach. To obtain simulation results, the MATLAB Yalmip parser and SDPT3 solver are effectively used in this project. Simulation results based on tunnel diode circuit and a nonholonomic mobile robot MAS quantifies the effectiveness of the proposed approach. Additionally, the gain matrix of the sampled data controller is obtained by solving the inequality of the derived linear matrix. Moreover, robust exponential stability and Takagi-Sugeno fuzzy control synthesis for networked control systems via H∞ performance has been investigated.