【作者】 李民；

【导师】 刘智伟；

【作者基本信息】 华中科技大学 ， 控制理论与控制工程， 2021， 硕士

【摘要】 多智能体网络分布式协同控制是复杂网络研究的一项重要分支,是一门先进的交叉学科,它包含仿生学理论、控制理论、图论、社会学理论以及统计学论的基本知识,相关研究已经日臻成熟。目前,该课题的研究已经从追求理论的创新性演化为追求与实际控制场景相结合的控制算法设计,这正是本文主要关注的内容。首先,本文对多智能体网络脉冲一致性问题进行研究,给出一致性控制方案。基于这一结果,进一步地对脉冲包含控制问题进行研究。最后,将多智能体一致性控制的基本思想应用于分布式感知环境下的线性系统协同镇定问题的研究。基于脉冲控制算法的二阶多智能体网络一致性研究。虽然早期的工作已经对具有一阶积分器、二阶积分器以及简谐振子节点动力学的多智能体网络一致性问题进行了研究,但是其节点动力学比较简单,不足以描述实际系统的复杂性。所以,对节点动力学更为一般的二阶多智能体网络协同行为进行研究是十分必要的。假设节点由一类可控的一般二阶节点动力学描述,它们相互通信组成多智能体网络,本文借助矩阵理论、离散系统稳定性理论以及代数图论,对此类多智能体网络进行脉冲一致性研究,得到了一致性的充要条件。在此基础上,进一步推导出多智能体网络实现具有确定收敛速度脉冲一致性的相关结论。基于脉冲控制算法的二阶多智能体网络鲁棒包含控制研究。考虑到多智能体网络实际运行中可能出现的节点进入或退出网络的现象,且多智能体网络的实际动力学和理论动力学往往具有一定的偏差,这些因素都会给网络运行带来不确定性,甚至是控制失效。针对这些问题,首先研究一般二阶多智能体网络包含控制的充要条件。以此为基础,对网络拓扑不确定性下的控制参数进行鲁棒性设计。并引入广义匹配系统的概念,对具有模型不确定性的脉冲包含控制参数进行鲁棒性设计。分布式感知环境下的线性系统协同镇定研究。线性系统被一组传感器监测时,每个传感器的输出往往无法包含线性系统的全局状态信息。那么,如何借助这样的输出来重构线性系统状态并使线性系统镇定,已经成为了一项很有挑战性的问题。针对这一问题,基于全局状态能观性假设,借助一致性控制思想,采用分布式观测器对线性系统状态进行协同观测。在此基础上,以现场控制结构的便利性起见,灵活选定部分观测器的状态来设计线性系统状态反馈镇定控制算法,可以实现节省通信资源的目的。通过构建增广系统,将协同观测问题和线性系统镇定问题综合为增广系统的稳定性问题。借助李雅普诺夫方法,对增广系统进行稳定性设计,进而得到相关控制参数的设计方法。最后,对全文主要工作进行了梳理和总结,并对多智能体网络协同控制问题以及分布式观测的相关工作进行了展望。更多还原

【Abstract】 Distributed coordinated control is an important branch of the research of complex networks,it is also an advanced interdisciplinary which includes principles of bionics theory,computer theory,control theory,graph theory,sociological theory,and statistics theory.Recent years,coordinated control has been actively researched.At present,relevant research has evolved from the pursuit of innovation in theory to applications in practical circumstances,which is also the main focus of this paper.Firstly,the impulsive consensus problem for multi-agent networks with general second-order node dynamics is studied.Based on this result,agents is divided into leaders and followers according to their sensory ability,and the impulsive containment control is studied.Finally,with the basic idea of consensus,this thesis studies the cooperative stabilization problem for LTI systems based on distributed observers.Consensus control of general second-order multi-agent networks based on impulsive control algorithms is studied.Although there are some impulsive consensus results of multiagent networks with node dynamics such as first-order integrators,second-order integrators,and harmonic oscillators,they are too simple to describe real systems.So this thesis studies the impulsive consensus control of general second-order multi-agent networks.By employing matrix theory,discrete system stability theory,and graph theory,some necessary and sufficient conditions for impulsive consensus are obtained.On this basis,some conditions are obtained which make the network achieve impulsive consensus with guaranteed convergence rates.Robust containment control problem for general second-order multi-agent networks is investigated in this thesis with impulsive control strategies.Considering there are always some uncertainties in real networks,for example,some nodes enter or exit the network,besides,the real dynamic and the theoretical dynamic of nodes always have some deviation.These factors may bring uncertainties to network operation,even control failure.So we would like to make the network run softly under a range of topologies and node dynamics.To solve this problem,based on necessary and sufficient conditions of impulsive containment control.Some robust design of control parameters with topology uncertainties is carried out.Besides,by employing generalized matching node dynamics,some robust design methods of control parameters with model uncertainties are obtained.Cooperative stabilization of linear systems is studied in distributed sensory environment.If a linear system is monitored by a group of sensors,the output of each sensor may not contain the global state of the linear system.Using such a set of outputs to reconstruct the state of the linear system and make it stabilized has become a challenging problem.To cope with this problem,based on the assumption of global observation and consensus strategies,distributed observers are designed to estimate the state of the linear system cooperatively.According to control requirements,a stabilization algorithm of linear system is designed by employing states of some number of observers rather than the states of all observers,which can help to save communication resources.By constructing augmented systems,the cooperative estimation problem and the stabilization problem are synthesized.By employing Lyapunov methods,some control parameters are designed to stabilize the augmented system.Finally,this thesis summarizes and evaluates the main work.Further research work of coordinated control of multi-agent networks and some related work of distributed observer networks is demonstrated.更多还原