Research On Cooperative Control Of Multi-agent Systems Under Communication Constraints

In recent years,with the rapid development of computer science technology,artificial intelligence technology and networked control technology,multi-agent systems(MASs)have been widely used in smart grids,smart transportation,autonomous driving,sensor networks,etc.Because multi-agent systems have the characteristics of autonomy,fault tolerance,flexibility and collaboration,how to coordinate with each other and complete tasks together in a complex environment has become an important prerequisite for achieving multi-agent collaborative control.Note that the information exchange between agents is through the network,and the existence of the network will bring brand new problems.Considering that the existence of network delay makes the interactive information unable to arrive in time,the existence of network attacks makes it impossible to guarantee the reliability of multi-agent systems and the impact of uncertain environments such as noise and disturbances in the system.The collaborative control of the intelligent body system has brought great challenges.The main research contents of this dissertation are as follows:First,the basic theories of cooperative control of multi-agent systems are introduced,including graph theory,Lyapunov stability,and event-triggered control theory.For a continuous homogeneous linear multi-agent system,an appropriate event-triggered mechanism is designed to reduce the frequency of control updates and ensure system consensus.The rationality of the result is verified by a simulation example.Second,the event-triggered bounded consensus control problem of a stochastic multi-agent systems with process,measurement noise and communication delay is studied.According to the Kalman filter theory,the best state estimate of each agent is obtained and sent to its neighboring agents through an undirected network topology.Correspondingly,an event-triggered controller related to communication delay is proposed to effectively reduce the update frequency of the controller.Then,the feedback gain and trigger parameters are designed and solved cooperatively through linear matrix inequality.The sufficient conditions for establishing a bounded consensus in mean square for fixed communication delay and time-varying communication delay are respectively given.Finally,the predictive cloud control problem of multi-agent systems with network delay and unknown disturbance under bandwidth allocation is studied.Firstly,an extended state observer based on prediction is designed to obtain the estimated values of agent states and disturbances.In order to make full use of communication resources,the bandwidth allocation problem of multi-agent systems is solved by using the Stackelberg game method,which includes the non-cooperative game between cloud controllers and the evolutionary game between agents.Then,the proposed predictive cloud control scheme can actively compensate for network delay and disturbance,and the Nash equilibrium solution of Stackelberg game and the stability condition of multi-agent systems are given respectively.