Research On Consensus Of Networked Multi-agent Systems

A lot of interesting biological phenomena, such as flocking of birds, school offish, rendezvous of ants, synchronous flashing of fireflies and formation of wild gooseetc, are found in nature. These seemingly simple collective behaviors can accomplishthe complex tasks what the individuals are incapable of. Multi-agent systems haveattracted considerable attentions of researchers recently due to the motivation of thecoordination of animals. Network is the key point for the coordination in multi-agentsystems, however, the communication networks cannot be considered ideally anymore in reality owing to underlying communication imperfections and constraints.Thus, problems caused by the real network should be taken into account. Consensus ofnetworked multi-agent systems under constraint conditions is investigated, and themain contents of this thesis are as follows:This paper investigates the synchronous sampled-data consensus problem ofgeneral linear multi-agent systems. Under fixed topologies, the consensus analysis isbased on a continuous-time model which considers the property of discretization ofsampled-data. The advantage of this approach is not only the avoidance ofuncertainties in terms of exponents caused by the discretization, but also the reductionin the conservatism due to the neglect of the property of samplings. Under switchingtopologies, the sampled-data consensus problem is converted into the consensusproblem in continuous-time with time delay. After that, it is proved by averagingmethods that the consensus could be achieved under switching topologies as long asthe consensus problem in continuous-time with time delay can be solved under theaveraged topology and the switching of topologies is fast enough. It provides a newway to deal with the consensus problem of general linear multi-agent systems underswitching topologies, especially for the case of exponentially unstable systems,although the conservatism will be produced by the system transformation.This paper investigates the asynchronous sampled-data consensus problem ofgeneral linear multi-agent systems. Under fixed topologies, the system is convertedinto the synchronous sampled-data switching system with multiple input delays. Theconsensus analysis is based on the property of discretization of sampled-data.Compared with the existing approaches, the system is converted into the synchronoussampled-data system with sampling delays which keeps more system characteristics. In common with the synchronous sampling case, the averaged approach is also appliedto the asynchronous sampling case under switching topologies. The only differencefrom the synchronous sampling case is that the asynchronous sampled-data system isconverted into a continuous-time system with multiple input delays.This paper investigates the event-triggered sampled-data consensus problem ofgeneral linear multi-agent systems. In order to design distributed event-triggeringstrategies, the general linear dynamic is converted into the single integrator model.Then the distributed event-triggering schemes are obtained under periodic eventdetections with fixed and switching topologies. For the exponentially unstable systems,the model-based event-triggered controller is proposed, whose design is more flexiblethan the first one. It is convenient to design the controller according to theevent-triggering conditions or develop co-design solutions.This paper investigates consensus problem of general linear multi-agent systemswith communication delays. The controller with communication delay compensationsis proposed based the idea of truncated compensations and model transformations. It isshown that the proposed consensus protocol has the robustness of arbitrarily boundedcommunication delays by regulating the coupling gain. Unfortunately, this approach isnot applicable to the exponentially unstable systems. For the general output feedback,the system is decoupled and converted into a parameter uncertain system with timedelays. By virtue of the small gain theorem in interconnection systems, the controlleris solved.For more generalized coordination of multi-agent systems, this paper discussesthe group consensus problem of multi-agent systems from the views of couplingweights and network topologies, respectively. The effect of group consensus isdecomposed into two parts, which are intra-subgroup and inter-subgroup.Consequently, a sufficient condition of group consensus is obtained, which is feasibleto more general balanced conditions of coupling weights. Moreover, this paperanalyses the connectivity vectors of network topology with non-negative couplingweights and proposes a distributed estimator to estimate these vectors. By virtue ofthese connectivity vectors, a topology selection algorithm is designed to achieve thegroup consensus.