Consensus Control Of Linear Multi-agent Systems Based On Data Sampling Approach

In recent years,due to the potential application for distributed cooperative control of the multi-agent system(MAS)in sensor networks,UAV formation,intelligent robots and other fields,many scholars have paid attention to this research area and studied a variety of control problems,such as formation control,containment control,group consensus control.These control are extended on the basis of consensus control,so the theory of consensus control is the research basis of MAS cooperative control,which is a very important research content.In order to reduce the communication load of the system,this paper mainly studies the consensus of MAS based on data sampling method.The linear system is the main research object.The specific works are as follows:1.Firstly,based on the data sampling method,the event-triggered consensus of the linear MAS is considered,as well the random packet losses at the sampling instants are taken into account.The Zeno behavior of the systems and continuous monitoring of the states can be avoided simultaneously.The stochastic variable subjected to Bernoulli dis-tribution is introduced to describe the packet losses probability.Based on the combined measurement of the agents,a novel distributed controller is proposed,and a distribut-ed event-triggered condition is also designed by using Riccati inequalities and linear matrix inequalities to achieve mean square consensus.In the trigger condition,fewer parameters are involved.In theoretical analysis,the stabilization problem of the linear system is considered instead of the consensus problem of the agents.At the same time,the interaction among the control parameter matrix,sampling interval and packet losses probability is analyzed.The maximum sampling interval is also presented explicitly.2.Secondly,for the directed tracking of continuous linear MAS,the data sampling is adopted as an approach to solve this problem with iterative learning control(ILC).For the MAS,the limited duration is pre-defined to perform each iteration task,and the sampling interval is periodic.The control law is designed based on the open-loop MAS.That is,the state information from the last task is fully utilized in the current task,and the current state values are not used.As a result,we avoid using these error messages when communication jams and time-delay occur in the current iteration task.At the same time,these reduce energy consumption of the MAS in the real-time communica-tion process.Based on Zero-order Hold(ZOH),the discretization of continuous linear MAS is used to construct a error system.The sufficient condition to achieve tracking and the novel monotonous convergence condition are given based on Schur comple-ment theorem and linear matrix inequalities.It is can be found that the input errors and the state errors are monotonous to zero explicitly as the number of the tasks tends to infinity.