Research On Resilient Consensus For Heterogeneous Multi-Agent Systems

In recent years,with the wide application of multi-agent systems in many fields such as unmanned aerial vehicles formation,traffic control,wireless sensor networks,and the improvement of theoretical principles,it has become one of the research hotspots that attracts widespread concerns of researchers form various fields.The consensus control problem,as the fundamental problem of cooperative control of multi-agent systems,is also the basis of other distributed control and estimated problems.Through the information interaction between neighbor agents,it aims to guarantee that all agents arrive a common state in the absence of central coordination and global communications.Due to the severe network security situation,agents would be inevitably suffered from mailicious and even be control by attackers.Hence,the study of resilient consensus problem of multi-agent systems has vital theoretical and engineering value.This thesis investigates the resilient consensus problems of the heterogeneous multi-agent systems under various situations of the fixed topologies,the switching topologies,and setting trusted nodes mechanism.The main contents of this thesis are summarized as follows:(1)The resilient consensus problem for heterogeneous multi-agent systems under the fixed topologies is investigated.Compared with homogeneous multi-agent systems,the dynamic characters of heterogeneous multi-agent system composed of first-and second-order agents are more complicated.There are some agents suffered form adversary attacks sending error information to their neighbors insteading of updating state by the predefined control protocol.In order to ensure the other normal agents can converge to a consistent state,a distributed control algorithm using the local information between agents is proposed.This algorithm selectively removes neighborhood agents to eliminate the influence of error message of malicious agents.A novel graph-theoretic property named topology robustness is introduced to enable the normal agents in systems can achieve consensus without extra global information of network topology and the specific identity of malicious neighbor.Then,based on the different mailicous attack models,main results about multi-agent systems can achieve resilient consensus are provided.(2)The resilient consensus problem for heterogeneous multi-agent systems under the switching topologies is investigated,and the trusted node mechanism is added to reduce the network resource usage and improve the robustness of network topology.With the consideration of various factors as data packet loss,physical obstacle and network congestion in the channel,communication links in the network can not be guaranteed to be connected at every time-step.For this case,a security control algorithm which collects relevant neighborhood information in a fixed period is designed.This algorithm doest not require the network topology must have certain algebraic network connectivity at every time-step.Meanwhile,the normal consensus algorithm has high requirement of network connectivity which limits the performance of multi-agent systems in a large-scale distributed application.Setting some agents as absolutely trusted individuals can significantly improve the network topology properties without adding extra links.Then,based on the different attack models and the control parameters,the consensus results of multi-agent systems under switching topologies and trusted node mechanism are given.