Consensus Problem Of Networked Multi-agent Systems Based On Event Triggering

With the development of large-scale communication networks and computation technologies,the coordinated control of multi-agent systems has attracted much attention in recent years.The consensus problem is one of the most fundamental and valuable tasks in the field of cooperative control of multi-agent systems.Consensus control requires all agents to reach common values on certain interests,based on a consensus protocol,namely an interaction rule of information transmission between neighboring agents.Existing results of consensus problems are mostly built on the common assumption of ideal communication.This means agents can communicate with their neighbors continuously,perfectly and in an error-free way.However,in reality,agents have to communicate with each other through wireless communication networks in multi-agent systems.Under the network bandwidth constraints,the state information should be encoded into finite-bit symbols before transmission.Meanwhile,the transmitted signals will be affected by unfavorable network factors such as network delay,packet loss,and network attacks.Therefore,more and more researchers pay attention to the consensus problems of networked multi-agent systems.In order to save network resources and further reduce the communication frequency between neighboring agents,the eventtriggered consensus control has received widespread attention as an important sub-issue of consensus control.Therefore,this thesis mainly focuses on the event-triggered consensus protocols of networked multi-agent systems with communication constraints.The main contents and contributions of this thesis are summarized as follows:1)To handle bandwidth constraints and save network resources,a model-based event-triggered strategy and a dynamic quantization mechanism are designed for the quantized consensus problem of discrete-time unstable multi-agent systems.Under the proposed strategy,the state estimation error of all agents is always limited by the eventtriggering function,which is also used to calculate a common upper bound on all control inputs.The sufficient bit rate condition to ensure asymptotic consensus is given.This condition is mainly determined by the dynamics and communication topology of agents.Since the event-triggered strategy uses the additional information extracted from triggering time of data packets,the derived consensus bit rate can be lower than the one of traditional periodic sampling strategy.2)For discrete-time second-order multi-agent systems,the combined effects of process noise,network delay and limited bandwidth on quantized consensus performance are quantitatively analyzed.Based on a dynamic quantization mechanism,it proposes an event-triggered sampling strategy that includes terms for handling noise,along with corresponding updating rules of the estimated state to adapt the network delay.By designing a time-dependent convergent event-triggering function,a sufficient bit rate condition to guarantee concerned consensus is derived.It also theoretically proves that the required consensus bit rate under bounded noise is the same as that under a noise-free system,i.e.,no additional bit rate is required to handle process noise.3)To guarantee consensus under bounded network delay and further reduce communication frequency between neighboring agents,a distributed adaptive consensus protocol based on a periodic dynamic event-triggered strategy is proposed for continuous-time general linear multi-agent systems.The auxiliary dynamic variables and weight parameters are introduced to all agents to construct the triggering conditions and select the sampling period.Meanwhile,time-varying coupling weights are also added to the control inputs.Through jointly designing the updating rules of dynamic variables,coupling weights and weighted parameters,the concerned consensus can be theoretically ensured by the Lyapunov method.Moreover,there exists a positive lower bound on triggering time intervals under periodic event-triggering strategy.4)To handle asynchronous network attacks,a distributed dynamic encodingdecoding quantization mechanism and an edge-based event-triggered strategy are proposed for the consensus problem of general linear multi-agent systems under limited bandwidth constraints.This means multiple Denial-of-Service attackers block different connected communication edges.To deal with the asynchronous quantization caused by network attacks,it independently determines the quantization range and occupied bits for each transmission according to the current attack state of the communication channel.Meanwhile,distributed event triggers associated with all communication edges are constructed to determine the information transmission instants.The event-triggered strategy can ensure that there is a shared upper bound on the state estimation error.A sufficient consensus bit rate condition is further derived,which mainly depends on the agent dynamics,the communication topology,and the parameters of network attacks.