Event-triggered Output Feedback Control For Multi-agent Systems

The multi-agent systems is composed of multiple agents,in which the dynamics of each agent are expressed by differential equations or by a combination of differential equations and algebraic equations.In order to enhance the flexibility,autonomy,and intelligence of multi-agent systems,and to respond quickly to environmental changes through asynchronous communication while reducing the consumption of network resources,as well as to achieve consensus through collaborative cooperation,the event-triggered mechanism for multi-agent systems has emerged.Agents communicate with neighbors and autonomously make timely decisions upon perceiving event-triggered,thereby enhancing the system’s flexibility and its ability to rapidly adapt to environmental changes.While adhering to the fundamental principle of maintaining system performance,this approach reduces data transmission and controller update frequencies,thus decreasing communication overhead.Additionally,due to limitations in obtaining complete state information from physical systems,such as sensor technology constraints,the research on event-triggered output feedback consensus control of multi-agent systems holds significant theoretical value and practical significance.Considering potential issues like process noise,measurement noise,physical constraints,and denial-of-service attacks,as well as the inherent limitations of system models during the operation of multi-agent systems,this thesis designs new event-triggered mechanisms for several types of multi-agent systems.Moreover,by employing tools such as algebraic graph theory,Lyapunov stability theory,stochastic analysis theory,linear matrix inequalities,ordinary differential equations,event-triggered control,and model predictive control methods,the thesis conducts a detailed analysis and research for the stability of multi-agent systems.The main content of this thesis contains the following ones:Firstly,the problem of edge-based event-triggered mean-square consensus control for stochastic multi-agent systems is studied.An edge-based new dynamic event-triggered control protocol is designed for stochastic multi-agent systems.This protocol does not require clock synchronization among agents and ensures that the stochastic multi-agent systems can achieve exponential convergence in the sense of mean square consensus.It overcomes the drawback of the Laplacian matrix being asymmetric and addresses the problem of communication topology being directed graphs.Secondly,the problem of dynamic event-triggered output feedback control of stochastic nonlinear multi-agent systems is studied.An observer is constructed to estimate the state of each agent,then a observer-based new event-triggered control protocol is proposed.Finally,a new Lyapunov function is constructed,and Lyapunov theory is utilized to prove that stochastic multi-agent systems controlled by this event-triggered control protocol can achieve mean-square exponential consensus.Thirdly,the problem of event-triggered output feedback control for stochastic linear multiagent systems is studied,where the stochastic linear multi-agent systems are subject to process noise and measurement noise independent of the unknown states.Based on Kalman filtering theory,an observer that optimally estimates the state of each agent is constructed.Then,an edge-based event-triggered control protocol is proposed,which only uses the state information observed by the observer,does not require clock synchronization among agents,and maintains a positive lower bound on the interval between every two consecutive events to avoid continuous communication.Finally,a new Lyapunov function is constructed,and stochastic convergence analysis is used to prove that stochastic multi-agent systems controlled by this event-triggered control protocol can achieve almost sure consensus.Fourthly,the event-triggered output feedback model predictive control problem of multiagent systems with unknown states and multiple uncertainties is studied.A new distributed event-triggered output feedback model predictive control algorithm is proposed for multi-agent systems with unknown states and multiple uncertainties.This algorithm provides a computationally and communicationally efficient method for dealing with various uncertainties,including state noise,measurement error,and unknown initial states.With this algorithm,all agents in the multi-agent system eventually converge to a predetermined bounded set.In this algorithm,communication among agents is asynchronous,as agents only exchange information with their neighbors at event triggering times.Fifthly,the event-triggered output feedback control problem of networked multi-agent systems subject to denial of service attacks is studied.An interval observer is constructed for networked multi-agent systems to estimate the interval of each agent’s state,addressing the problem of unobservable system states.Then,a new event-triggered control protocol is proposed that saves communication resources and reduces controller update frequency,requiring only information observed by the interval observer.Sufficient conditions for achieving exponential consensus in networked multi-agent systems under denial of service attacks are provided.Sixthly,the problem of event-triggered output feedback tracking control is studied for uncertain singular multi-agent systems.An interval observer is constructed for uncertain singular multi-agent systems to estimate the interval of each agent’s state.Then,for disturbed singular multi-agent systems,a new static event-triggered control protocol is proposed,while for undisturbed singular multi-agent systems,a dynamic event-triggered control protocol is proposed to reduce communication resource consumption.Finally,the consensus problems of disturbed and undisturbed singular multi-agent systems are discussed separately,and sufficient conditions for achieving bounded consensus in disturbed singular multi-agent systems and exponential consensus in undisturbed singular multi-agent systems are provided.