事件驱动作为一种新兴的采样设计方法，在节约系统有限的控制资源、缓解数据传输带宽压力方面有着比时间驱动采样控制更显著的优势。多智能体系统由于个体数目较大，且具有需要个体间局部交互完成全局任务的特点，利用事件驱动控制来实现高效地配置多智能体系统的控制资源逐渐成为研究热点。然而，事件驱动系统属于混杂系统的范畴，理论分析工具尚不完善。在多智能体系统中，由于通讯拓扑的存在，使得对事件驱动条件下的分布式控制策略的设计难度加大；同时，对该控制策略下多智能体系统的稳定性分析带来新的挑战。本文在已有的研究基础上，寻求多智能体系统的事件驱动控制设计的新方法，并引入新的理论分析工具，主要的研究工作如下： 研究了有向通讯拓扑下的一阶、二阶多智能体系统边事件驱动控制问题。为了降低系统的控制器更新频率，我们设计了一个以指数衰减的时间函数作为事件触发阈值的事件驱动条件。该事件驱动条件有如下特性：(i)边事件测量误差依赖于有向的信息流给出，可解决有向通讯拓扑下的多智能体一致性控制问题。(ii)每个个体仅仅更新边事件触发对应边的状态信息，并且每个个体的控制执行是相互独立的。(iii)所提出的边事件驱动控制方法不会出现Zeno行为。基于图谱理论、矩阵分析中的方法，建立了一阶、二阶多智能体系统在含有有向生成树拓扑下的一致性条件。 讨论了耦合调和振子系统的事件驱动控制下的同步问题。基于节点状态，提出了中心式事件驱动控制策略，并在此基础上提出了仅依赖于局部状态信息的无中心的事件驱动控制算法。在基于节点设计的事件驱动控制策略下，振子的控制器仅当事件驱动测量误差超过状态依赖的阈值时才更新，有效地降低了系统控制更新代价。利用非光滑分析的工具，建立了该事件驱动控制策略下耦合振子系统达到渐近同步的充分条件。为了排除基于节点设计的事件驱动控制系统的Zeno行为，我们设计了无中心控制的固定阈值事件驱动策略。然后，我们以相互耦合的边为研究对象，构建了连接振子的边动力学模型。设计了边事件驱动控制策略实现对耦合调和振子系统的同步控制。通过结合周期事件检测和边事件驱动方法，确保边事件驱动策略下系统无Zeno行为的出现，并给出保证耦合振子同步的充分条件。同时，我们设计了两个边事件检测规则实现以较低的个体控制器更新频率和个体间的通讯次数完成同步控制。 研究了非线性弱耦合的Kuramoto振子同步问题，推导出Kuramoto振子网络通过间歇和异步的耦合就能达到同步的充分条件。首先，基于周期采样控制方法，我们通过对相差动力学的分析推导出Kuramoto振子同步的充分条件。为了放宽在周期采样下所有振子需要同时更新相差的限制，我们设计了基于边事件驱动的控制策略来实现异步更新相应相差。在系统同步分析中，采用不同于一般非线性系统在平衡点处线性化的处理方法和平均场理论，而重新构造关于边状态的Lyapunov函数并借助非光滑分析理论获得同步条件和系统同步的吸引域。 讨论了具有输入欠量无源性的异质非线性多智能体的输出同步控制问题。为了缓解智能体间的通讯压力和降低控制器的更新频率，我们设计了事件驱动控制下的同步协议，使得个体仅需要在事件时刻传输状态信息和更新控制器状态。基于无源性控制方法，我们证明了在所提出的控制协议下，只要耦合强度足够小，多智能体系统就能渐近地实现输出同步。 关键词：多智能体系统；同步；分布式控制；事件驱动控制；采样控制

 Event-triggered control is a new emerged sampled-data control method that outperforms the time-triggered control in terms of saving limited control resources and reliving communication burdens. In the multi-agent system, agents are required to interact with neighbors so as to achieve the global objective of the overall system. However, the agents are usually large-scaled and distributed in a wide region, and thus the efficient usage of control resources has attracted great attention in multi-agent systems. In recent years, the event-triggered control for multi-agent systems has become a hot research topic since it is an alternative strategy to solve these resource restriction problems. Many challenges in the design of distributed event-triggering condition and theoretical analysis arise from the underlying communication topologies of multi-agent systems. Moreover, the eventtriggered systems belong to a class of hybrid systems and the analytical tools for such systems are far from being completed. Based on the existing work in the event-triggered control, this paper devotes to proposing new event-trigged control designs and developing new analysis methods for multi-agent systems. The main contributions are listed as follows: The edge event-triggered consensus control for multi-agent systems with singleor double-integrator dynamics under directed communication topologies is investigated. With the edge event-triggered sampling strategies, we propose a class of novel consensus protocols to reduce the frequency of controller updates. The main features of the proposed edge event-triggered control include three aspects: (i) the measurement error is defined according to directed information flow and thus it allows that the communication topology is directed, (ii) each individual controller update its edge states only when the associated edge event occurs and the control actions are independent of each agent, and (iii) the proposed edge event-triggered control does not exhibit Zeno behavior. By graph theory and matrix analysis, the consensus conditions are established for the multi-agent systems with single- or double-integrator dynamics underlying the directed communication topologies with spanning trees. The synchronization problem of coupled harmonic oscillators is addressed by eventtriggered control. Based on node states, a centralized event-triggered control strategy is first developed and is further extended to a decentralized counterpart, in which the control protocol and event-triggering conditions only require local information. With the nodebased event-triggered control strategies, controllers update at the discrete instants when the related measurement errors exceed some proper state-dependent thresholds, which can reduce the computation costs of controllers. By the tools from nonsmooth analysis, it is shown that the proposed event-triggered control strategies synchronize all oscillator states asymptotically. Furthermore, a decentralized event-triggered strategy with a fixed threshold is proposed for the sake of excluding the Zeno behavior. Based on the formulation of edge dynamics with respect to two coupled oscillators, we propose a new edge-based event-triggered control strategy to address the synchronization problems of coupled harmonic oscillators. Combined with the periodic event-detecting and edge event-triggered approach, sufficient conditions that guarantee the synchronization of coupled harmonic oscillators are presented. Two edge event-detecting rules are given to achieve the synchronization of coupled harmonic oscillators with low resource consumption. The synchronization problem of Kuramoto oscillators under sampled-data updating law is considered. This study is motivated by the needs of synchronization of Kuramoto oscillators in the presence of intermittent and asynchronous coupling updates. Based on the periodically sampled-data method, a sufficient condition ensuring synchronization under periodic updates is derived. In order to relax the requirement of having all data updated simultaneously, an event-triggered law is designed to implement asynchronous coupling updates. Our synchronization analysis does not rely on any linearization technique around equilibrium points and the mean field theory, instead, we propose a new Lyapunov function based on the edge states and employ the nonsmooth analysis technique to deduce the synchronization conditions. The distributed output synchronization problem of heterogeneous discrete-time nonlinear multi-agent systems with input shortage passivity is studied. In order to reduce the communication burden and the update frequency of controllers, an event-triggered distributed synchronization protocol is proposed where the transmission of output states and the updates of controllers are only needed at event instants. By using the passivity-based control, we show that the multi-agent system can be driven to output state synchronization provided that the coupling strengths are weak enough. Keywords: multi-agent systems, synchronization, distributed control, event-triggered control, sampled-data control