Event-Triggered Consensus Schemes of Multi-Agent Systems

事件驅動策略下的多智能體系統一致性研究

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date5 Jun 2017

Abstract

Over the past decade, the study of multi-agent systems has attracted increasing research interests due to their broad applications ranging from wireless networks, unmanned aerial vehicles, automated military systems, robotic teams, power systems, to social networks. Consensus is one of the most typical collective behaviors of interacting agents. Its objective is to make the states of multiple agents converge to a common value via information transmission through network structure. To achieve consensus behavior of multi-agent systems, the information transmission is a key factor for coordinated movements of agents, and an effective consensus protocol is constructed on the basis of local information exchange.

The time-triggered communication strategies have been widely investigated for multi-agent consensus behaviors, which assume that each agent exchanges information with its neighbors as time lapses. Recently, some alternative kinds of communication schemes, i.e. event-triggered communication schemes, are proposed with an advantage of saving communication cost. Under these kinds of schemes, information is transmitted only when a well-defined condition is satisfied. Thus the communication frequency is reduced significantly, and then the updates of the consensus protocols are also lowered. In this thesis, it is highly desirable to develop a framework of event-triggered schemes for the consensus behavior of multi-agent systems. One of the difficult issues in event-triggered schemes is to avoid Zeno behavior, and Zeno behavior is an undesired phenomenon in which an event condition is infinitely triggered in a finite time period.

The following topics will be presented in this thesis in detail: (a) three kinds of event-triggered schemes for leader-following consensus; (b) an impulsive framework for event-triggered schemes with respect to different types of information transmission; (c) a layered event-triggered control analysis under a multi-layered directed network topology; (d) a novel edge event-triggered scheme via a communication buffer with/without the quantization effect; (e) an observer-type event-triggered control for discrete time-varying multi-agent systems with redundant channels.

The main contributions of this thesis are listed as follows:
• The leader-following consensus problem is studied for multi-agent systems with general linear dynamics by Event-Triggered Schemes (ETSs). Three types of schemes, namely, distributed ETS, centralized ETS, and clustered ETS are proposed with a novel update strategy, and the corresponding event-triggered consensus protocols are constructed. All of these protocols guarantee that all followers can track the leader eventually. The exclusion of Zeno behavior is assured due to the existence of positive inner-event time intervals. Furthermore, the case of switching topology is investigated under centralized ETS.

• A novel impulsive framework is developed to analyze event-triggered schemes of multi-agent systems from a new perspective. Under the impulsive framework, various types of information transmissions are discussed, and hence, different event-triggered control protocols are proposed based on state feedback and output feedback. In addition, the effect of the external disturbance on the desired collective behavior is also studied. Furthermore, one application is used to demonstrate the effectiveness of the proposed theoretical results.

• A multi-layer directed network topology is investigated in multi-agent systems. A layered event-triggered scheme (LETS) is constructed to cater for the characteristics of multi-layer network topology. Under the LETS, agents in different layers achieve asymptotical consensus eventually and the Zeno behavior is successfully eliminated. Then, an algorithm is provided to avoid continuous event detection such that the communication traffic of the whole network is further lowered.

• A distributed edge event-triggered (DEET) scheme of multi-agent systems is proposed via a communication buffer. The introduction of the communication buffer reduces unnecessary update of controllers induced by fast information transmission. This edge scheme avoids unnecessary cost induced by the synchronous information transmission in node event-triggered mechanism. Then, the effect of DDET scheme on consensus behavior is discussed. In addition, the quantized information transmission is further investigated, and a quantized edge event-triggered control protocol is constructed. In this situation, asymptotical consensus is eventually achieved and improves the results on bounded consensus in most of the existing literature.

• A discrete time-varying multi-agent system is investigated by means of event-triggered schemes. Both the external disturbances and missing measurements are taken into consideration for the multi-agent system, and the influence of the missing measurements on the consensus behaviors is thoroughly examined. Then redundant channels are adopted during the communication process with intention of improving the reliability of network communication. Moreover, a novel observer-type event-triggered control protocol is designed for the time-varying multi-agent system to reduce both the frequency of transmission and the number of the control protocol updates. Furthermore, a constrained recursive Riccati Difference Equation approach is developed to guarantee the H∞ consensus performance over a finite horizon, and to design the controller parameters.