Research On The Predictive Control For Networked Multi-agent System Based On Time/Event-triggered Scheme

Multi-agent system(MAS)is a research hotspot in the field of artificial intelligence and control.However,the current research on MAS focuses more on the consistency of the system,and the packet loss among agents is rarely studied.Therefore,this thesis takes the networked multi-agent system(NMAS)as the research object,takes linear matrix inequality(LMI)as the main research tool,combines with the Lyapunov stability theory,adopts the research idea from time-triggered scheme to event-triggered scheme,from state feedback to output feedback,and systematically studies the synthesis approach of distributed model predictive control(DMPC).The work of this thesis mainly includes the following three aspects:1.The synthesis approach of distributed state feedback model predictive control for MAS with packet loss and actuator saturation is studied.Firstly,a unified model of packet loss and actuator saturation is established.It is assumed that packet loss obeys a Bernoulli distribution with probability being known,and the actuator saturation is dealt with by the Nth-step set invariance approach.Then,the compatibility constraint is designed to ensure the stability of the global system.Finally,by solving the optimization problem online at each sampling moment,a distributed model predictive controller based on state feedback is obtained,which ensures both the recursive feasibility of the optimization problem and the closed-loop stability of NMAS.2.The synthesis approach of based on event-triggered DMPC for MAS is studied.Firstly,in order to save the limited computing and communication resources,the event-triggered condition is designed according to the local stability of the system.Then,the compatibility constraint is provided to ensure that the global system is asymptotically stable,and the DMPC algorithm based on event-triggered scheme is obtained.Finally,the recursive feasibility of the optimization problem and the stability of the closed-loop system are proved.3.The synthesis approach of distributed output feedback model predictive control for NMAS via event-triggered scheme is studied.The off-line state observer is designed to obtain the estimated state value.Then,based on the estimated state,a distributed output feedback controller is designed and the compatibility constraint is given.Finally,the recursive feasibility of the optimization problem and the stability of the closed-loop system are proved.