Research On Event-Based Consensus Of Networked Multi-Agent Systems

With the development of advanced control technologies,the emergence and burgeon-ing of distributed large-scale complex networks is widely used in various control fields,such as wireless sensor networks,formation control,power systems and so on.Multi-agent systems,as one of complex networks,has gained increasingly attention from researchers owing to its wide applications in practice.In multi-agent cooperative control,the consensus is the fundamental issue and its aim is to develop an appropriate control protocol such that states or outputs of agents subject to local communication converge to a common value.In traditional consensus control,consensus can be achieved by assuming that continuous and secure information exchange among agents is required.However,such an assumption requires sufficient computing resources and the support of an ideal communication envi-ronment in a digital platform.Especially when the internal devices of agents are powered by batteries and the communication network consisting of agents is easy,to be attacked by adversaries.Obviously,it cannot be realized in practical applications due to the limited communication bandwidth and communication channels.To take use of the limited band-width in multi-agent networks,the event-triggered control,as an effective control approach,is introduced to investigate the consensus problem of multi-agent systemsIn this paper,we mainly discuss a series of problems of multi-agent systems based on event trigger control strategy,such as fixed/switching topologies,occurrence of fault-tolerant problem,secure communication among agents subject to unkown/known attack strategies.The detailed work includes the following four aspects:Firstly,event-triggered consensus analysis of multi-agent systems with fixed/switching topologies.Considering that each agent is difficult to obtain its own real state due to com-plex communication environments,the model-based control approach is employed to esti-mate the real state of each agent.Then,the event-triggered consensus protocol and event-triggered condition consisting of the estimated states are developed to study the consensus of multi-agent systems under fixed/switching topologies.It is shown that designed event-triggered controllers and triggering conditions can not only guarantee the consensus of multi-agent systems without any continuous communication but also avoid the occurrence of "Zeno behavior".Secondly,observer-based distributed adaptive event-triggered fault-tolerant consen-sus analysis of multi-agent systems.In multi-agent networks,there may exist some unpre-dictable factors including unavailable system state,the occurrence of controller failure and the chattering phenomenon arising from the process of switching devices-To overcome the above factors that affect system performances,this chapter proposes the observer-based distributed adaptive event-triggered fault-tolerant communication protocols and distribut-ed adaptive triggering conditions.Note that the existence of chattering phenomenon makes the designed controller discontinuous.Then,a state boundary layer method is employed to turn the discontinuous controller to continuous controller.Theoretical analysis shows that the consensus of multi-agent systems can be realized without continuous communication among agents for both the developed controllers and triggering conditions.Thirdly,event-triggered secure consensus analysis of multi-agent systems with known DoS attacks strategy.In multi-agent networks,the communication network between agents may be attacked by energy-limited adversaries,which leads to lower the system's per-formance and even destroy the stability of the controlled system.It is necessary to study the secure consensus of multi-agent systems from the view of cyber-security.The entire time-sequence can be divided into two different periods,communication period and attack period,where the communication period exists triggering events without adversaries;the attack period has adversaries such that all communication edges among agent are interrupt-ed and the control inputs of agents turn to zero.For the communication period,the secure event-triggered communication protocol and triggering condition are proposed based on the combinational measurement method to study the consensus problem.For the attack pe-riod,considering that the partial information of adversaries have been detected in advance.The estimated approach is adopted to analyze the evolution of the controlled system.For the entire time-sequence,Lyapunov function and the iterative approach are adopted to pro-vide the sufficient condition of achieving secure consensus of multi-agent systemsFinally,event-triggered secure consensus analysis of multi-agent systems with un-known DoS attacks strategy.When the DoS attack strategy is unknown,for the commu-nication period,model-based approach is employed to estimate the relative state among agent at intermittent instants.Then,the event-triggered communication protocol and the input-based triggering condition are designed based on the estimated relative states,where the triggering condition is based on the control input of each agent instead of the mea-surement error.For attack period,assume that the existence of the backup mechanisms,the controlled system subject to no control input can be restored to communication period after a period of time.Thus,the whole system can be regarded as a"switching systems"framework.By introducing the attack frequency and attack duration,the stability of switch-ing systems is analyzed to give the sufficient condition of achieving secure consensus by discussing and giving the upper bounds of the attack frequency and attack duration.