Research On Fast Consensus OfMulti-agent Systems

Multi-agent system is composed of the collection of multiple agents, whose goal is to reconstruct a large and complex system into small, communicate and coordinate with each other and manageable systems. The multi-agent system coordinated control has been widely used in formation control, clustering, flocking, congestion control of communication network, distributed sensor network and so on. As one of the fundamental problems of the multi-agent system coordinated control, consensus problem has caught more and more attention from the researchers of various fields. Consensus means that one or some states of all agents in multi-agent system can converge to an agreement over time. In practical applications, whether the convergence speed of a system is fast or not has become an important performance to judge that a consensus protocol is good or not. Although it has been widely studied and got a lot of research achievements for rapid consensus problem, but there are still many questions worth discussing. For example, how to make full use of the known information of agents to improve the integrity of the system information, in order to allow the system to achieve consensus rapidly; In addition, under the influence of time delay or switch topology and so on, how to design a reasonable consensus protocol, which can make the system quickly converge to an agreement.In this thesis, the main research contents are summarized as follows:1) We studied fast consensus tracking problem of multi-agent systems with diverse communication delays. The fast consensus tracking problem for the first-order discrete multi-agent systems with both communication and input delays are studied by introducing the incremental PID algorithm, the sufficient condition of rapid convergence is obtained through the stability analysis. In addition, the convergence speed of the system is further accelerated by optimizing the consensus protocol with PID parameters optimization based on genetic algorithm;2) We studied the fast consensus problem of multi-agent systems based on local information feedback. For the fast consensus problem of first-order continuous multi-agent systems under directed network, a consensus protocol based on local information feedback is proposed. Based on matrix theory and the time-domain analysis, the relations of the closed-loop eigenvalues and the gains of these systems with single gain or double gains are analyzed respectively;3) We studied the applications of consensus in fuzzy control and filtering algorithm. First of all, a fuzzy PID consensus tracking algorithm for multi-agent system is proposed, considering the stability of the system under the influence of interference and time delay; Secondly, we analysis a consensus Kalman filtering algorithm by working on estimated value, through an example simulation analysis, the proposed algorithm can filter well to white Gaussian noise.