Distributed Optimization Algorithm Design With Communication Packet Loss

In recent years,the research on multi-agent distributed cooperative control has attracted more and more attention.In the past decade,distributed optimization has gradually become a focus in the field of distributed cooperative control.The goal of distributed optimization of multi-agent system is to obtain the optimal value of the system objective function through the individual computation of the agents and the local communication between the agents and their neighbors.In this paper,based on the graph theory,matrix theory,probability theory,Lyapunov stability criterion and other related theoretical and mathematical tools,we investigate the distributed optimization problem of continuous-time multi-agent systems(MASs)with random communication packet loss.The main work of this thesis includes the following three aspects:Firstly,considering the identical packet loss among communication channels,a distributed optimization algorithm based on periodic sampling is designed to realize the states of all agents converge to the minimum point of system objective function in probability,that is,the optimal point of the distributed optimization problem.Compared with previous researches,the proposed algorithm relaxes the requirement that the gradient of the objective function is bounded.Secondly,considering the nonidentical packet loss among communication channels,another distributed optimization algorithm based on periodic sampling is designed to realize the states of all agents converge to the optimal value point in probability.It is proved that the proposed algorithm does not depend on the assumption of consistent packet loss,which extends the practicability of the proposed algorithm.Thirdly,in the case of nonidentical packet loss among communication channels,a distributed optimization algorithm is designed based on the mechanism of eventtriggering sampling,which realizes the states of all agents converge to the minimum point of system objective function in probability.By introducing the event-triggering sampling mechanism,there is no need for continuous time communication and periodic sampling,thus reducing communication consumption and communication resources.