Research On Consensus Of Multi-agent Systems Under Non-ideal Networks

Multi-agent systems consist of a large number of autonomous or semiautonomous agents.Agents accomplish a task or achieve a common status by means of cooperation or competition.Consensus of multi-agent systems is a fundamental problem in the study,which means that all agents are able to converge to a common state by local information exchange and collaboration.With wide applications of multi-agent systems in the formation control of unmanned aerial vehicle(UAV)groups,attitude control of satellites,automatic highway system scheduling,air traffic control and so on,the consensus of multi-agent systems has attracted considerable interest of research.The agents of a multi-agent system exchange their information through a network.However,the communication delays and packet losses are inevitable,due to the limitation on network bandwidth and transmission rate which are induced by network.The existence of communication time-delay or packet losses may degrade the performance of systems or even destroy the stability of systems.Moreover,with the development of technology,the increased number of signal transmission intensifies the contradiction between the communication traffic and the limited network resource.It is desirable to reduce traffic as much as possible while maintaining the performance of system.Therefore,the consensus of multi-agent systems under the unideal network setting such as communication timedelays,packet losses,limited resources,is investigated in this thesis.The main contributions of this thesis are listed as follows:1.The background,research significance and research status of the consensus of multi-agent systems are reviewed,in which the consensus of multi-agent system in network setting is emphatically elaborated.2.The scaled consensus of linear multi-agent systems with nonuniform communication time-delay is investigated.With the help of the connectivity of the communication topology,the multi-agent system is decoupled into several subsystems.Based on the solution of Riccati matrix equations,the design method of the distributed controller is proposed.From the analysis,it is easy to see that the system can tolerate arbitrary bounded time-delay when all eigenvalues of the systems are zeros.We provide the upper bound of time delay in the case that when there exists non-zero eigenvalue of the system.3.For the periodic sampled multi-agent systems,the influences of packet loss and time-delay on the consensus of multi-agent system are considered.By distinguishing the inducement of different types of time-delay,full information of time-delay function is utilized in order to obtain lower conservative results.With the aid of the theory of switched system and Halanay inequality,we obtain allowable rate of packet loss and upper bounded of time-delay for the systems with deterministic packet losses and stochastic packet losses.An example is provided to demonstrate the effectiveness of our results.4.The leader-following consensus of multi-agent systems with unknown state of the leader is considered.Due to the limitation on communication resource,a distributed observer based on the event-triggered mechanism is constructed.Utilizing the input time-delay method,the consensus of multi-agent systems with event-triggered mechanism can be transformed into the consensus of time-delay multi-agent systems.In order to obtain a sufficient condition for the consensus of multi-agent system,the property of linear systems is utilized.5.The influence of packet loss on the consensus of multi-agent system under eventtriggered mechanism is studied.It is supposed that once the packet loss happens,the control input will be set to zero.Under this assumption,the influence of packet losses on the consensus of multi-agent systems is discussed.By modelling the multi-agent systems with such packet loss as switched system,the relationship among packet losses,event-triggered mechanism and convergence rate is constructed by utilizing the Lyapunov functional and the theory of switched system.The proposed periodic event-triggered mechanism can greatly reduce the amount of communication as well as avoid the Zeno behavior effectively.6.A new event-triggered mechanism which is named hybrid event-triggered mechanism is proposed.The proposed hybrid event-triggered mechanism processes the character of both periodic sampling and continuous event-triggered mechanism.The consensus of nonlinear stochastic multi-agent systems with hybrid event-triggered mechanism is studied as well as the influence of time-delay.Due to the existence of stochastic factor,a measurement method based on mathematical expectation is proposed.7.The above mentioned event-triggered mechanisms depend on the continuous monitoring of the system.To propose an event-triggered mechanism which is independent of the real-time monitoring,a self-triggered mechanism is utilized.Different from the continuous measurement for event-triggered mechanism,the next sampling instant for self-triggered mechanism is calculated by a set procedure.The self-triggered mechanism is proposed for nonlinear stochastic complex systems with or without time-delay,as well as the allowable upper bound of time-delay.