Cooperative Control of Multi-Agent Systems by Event-Triggered Strategies

基於事件驅動策略的多智能體系統協作式控制

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

  • Wenfeng HU

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date6 Sep 2016

Abstract

Over the past decade, various cooperative control problems of multi-agent systems have been intensively studied, due to their wide applications. Typical applications include unmanned aerial vehicle formations, distributed wireless sensor networks, and cooperative surveillance, to name just a few. Consensus and cooperative output regulation are two of the most fundamental cooperative control problems under study.

In practice, individual agents are usually equipped with simple embedded microprocessors, onboard communication modules, and actuation modules, which have limited energy resources to perform such functions as gathering information, communicating with neighboring agents, and driving the agent. For most existing control schemes in cooperative control, an agent needs to measure its state, send the state information to its neighboring agents, and update its control signal continuously or in a fixed sampling rate. Such control schemes might become infeasible or impractical in many applications due to their excessive consumption of on-board energy resources. It is thus desirable to design novel control schemes, such that the load of communication among agents and/or controller updates for each agent can be reduced significantly. In this way, limited energy resources of agents can be greatly saved and operational lifespan of multi-agent systems can be thus prolonged.

This thesis aims to address the cooperative control problems by event-triggered strategies, such that the communication among neighboring agents can be reduced. We will focus on the event-triggered control of two fundamental cooperative control problems: consensus and cooperative output regulation. For the first problem, the main results are summarized as follows:

1. The leaderless consensus problem of multi-agent systems with general linear dynamics is studied by event-triggered strategies. We propose a novel event-triggered control scheme with some desirable features, namely, distributed, asynchronous and independent. It is shown that consensus of the controlled multi-agent system can be reached asymptotically. The feasibility of the event-triggered strategy is further verified by the exclusion of Zeno behavior. Moreover, a self-triggered algorithm is developed, where the next triggering time instant for each agent is determined based on its local information at the previous triggering time instant. Continuous monitoring of measurement errors is thus avoided.

2. The leader-following consensus problem of multi-agent systems with homogeneous dynamics is addressed by event-triggered strategies. In this problem, the considered multi-agent system is of general linear dynamics, which include the mostly studied single-or double-integrator multi-agent systems as special cases. The derived results can be regarded as an extension of the event-triggered control scheme from the leaderless consensus problem to the leader-following consensus problem.

3. The output consensus problem of linear multi-agent systems is addressed by proposing a novel distributed event-triggered control scheme. The challenge of this work is that the dynamics of the agents are allowed to be nonidentical, and the usually studied state consensus problem can be viewed as a special case of this work. It is shown that, with the proposed control scheme, the output consensus problem can be solved if and only if two matrix equations are satisfied. Then a novel self-triggered control scheme is further proposed, with which continuous monitoring of measurement errors can be avoided. By introducing a fixed timer into both event- and self-triggered control schemes, Zeno behavior can be ruled out for each agent.

The second part of this thesis addresses the cooperative output regulation problem of linear multi-agent systems by event-triggered strategies. The considered problem is more general than the leader-following consensus problem in that it can handle multi-agent systems with nonidentical dynamics, external disturbance rejection, and reference tracking at the same time. Furthermore, the model uncertainty, directed communication topology, and switching communication topology are taken into consideration, which makes the problems even more challenging. Our objective is to address all such challenging problems by developing novel event- and self-triggered control schemes, such that the communication among neighboring agents is intermittent. The main results of this part are summarized as follows:

1. The cooperative output regulation problem of heterogeneous linear multi-agent systems with directed communication topology is addressed, requiring only intermittent communication. A unified framework of time- and event-triggering strategies is proposed. Then based on the unified triggering framework, a novel triggering mechanism utilizing the self-triggering strategy is developed. It is shown that with the proposed triggering mechanism, the cooperative output regulation problem can be solved by a distributed control law with only intermittent communication. It is further shown that with the unified framework of time-and event-triggering strategies for each agent, a positive minimum inter-event time can be explicitly given and Zeno behavior can be thus excluded.

2. The robust event-triggered cooperative output regulation problem of heterogeneous uncertain linear multi-agent systems is studied. In this problem, the system uncertainty is dealt with by utilizing the internal model principle. Then, novel distributed event-and self-triggered control schemes are developed. The feasibility is also discussed by excluding Zeno behavior for each agent.

3. The event-triggered cooperative output regulation problem of heterogeneous linear multi-agent systems with switching topology is investigated. Novel event-and self-triggered control schemes are proposed. It is noted that our result only assumes that the communication topology is jointly connected, then the cooperative output regulation problem can be solved by the proposed control schemes, which requires only intermittent communication. The feasibility of the control schemes is also discussed.