基于信息耦合度的自组织分群控制方法研究

发布时间：2018-02-25 19:17

本文关键词： 群集系统分群行为自组织分群控制信息耦合度选择性交互隐式信息传播成对交互速度估计时延群集机器人　出处：《西北工业大学》2016年博士论文　论文类型：学位论文

【摘要】：群集系统的分群行为,表征为一个聚合的母群分裂成若干子群的现象。作为群集系统的固有运动模式,分群行为不仅弥补了单一组群运动难以对群集分裂现象进行有效表征的缺陷,而且赋予了群集更多的运动灵活性和行为多样性,是群集对环境适应能力的典型表现。因而,开展群集系统分群控制的研究,对于揭示群集"分裂-融合"行为的内在机制,深化对自组织群集运动机理的认知,以及促进分群行为的工程应用,均具有重要的理论价值和实践意义。本文以国家自然科学基金项目(No.51179156)为依托,针对一类不依赖于指派、协商等智能规划机制的典型自组织分群运动行为,在深入揭示其产生机理的基础上,对基于信息耦合度(Information coupling degree,ICD)的自组织分群控制方法及相关问题展开研究。论文的主要工作和研究成果如下:(1)群集系统自组织分群运动机理与建模研究以生物群集的分群行为作为参考,从刺激信息在群内定向传播的角度出发,研究了分群运动中群集内部信息的传播机制,指出分群行为是多元外部刺激信息在群内传播所引起的一种个体运动分化现象,并发现个体间关联强度的差异是导致分群行为的根本原因。基于该认识,定义了对个体间关联强度进行表征的指标—信息耦合度,并提出了一种基于信息耦合度的分群控制模型框架,为后续分群控制算法的设计奠定了基础。(2)基于信息耦合度的群集系统分群控制方法研究借鉴生物群集个体间的交互作用关系,综合利用个体位置和速度信息,建立了信息耦合度模型。然后,从刺激信息在群内传播最大化的角度出发,提出了一种基于"max-ICD"的分群控制策略。进而,将自组织"成对交互"规则融入传统的"分离/组队/聚合"群集协同规则中,设计了基于信息耦合度的分群控制算法,实现了外部刺激下群集系统的自组织分群行为。理论分析和仿真实验均验证了所提出的分群控制算法的有效性。(3)无速度测量下群集系统分群控制方法研究针对无速度测量下群集系统的自组织分群问题,提出了一种仅利用位置信息的分群控制方法。首先,通过构建分布式观测器,实现了利用个体间的相对位置信息对相对速度信息的实时估计。进而,以基于信息耦合度的分群控制理论框架为基础,结合速度信息未知的情形,利用个体间的相对位置信息和估计的相对速度信息,设计了无需速度测量的分群控制算法,实现了群集系统的自组织分群运动。理论分析和仿真实验均验证了所提出的无速度测量下自组织分群控制算法的有效性。(4)具有时延的群集系统分群控制方法研究针对分群运动中个体信息获取存在时延的问题,首先对前面提出的基于信息耦合度的分群控制算法的时延局限性展开讨论,发现时延的存在会显著降低分群算法的性能。进而,分定常时延和时变时延两种情况,对时延下群集系统的分群控制方法展开研究。对于定常时延情形,采用Lyapunov-Razumikhin定理得出了群集实现稳定分群运动的时延上界及相关收敛条件。对于时变时延情形,利用Lyaponov-Krasovskii泛函得出了线性矩阵不等式形式的时延上界及相关收敛条件。理论分析和仿真实验均验证了所提出的时延下群集系统分群控制算法的有效性。(5)群集机器人系统分群控制实验验证研究以E-puck群集机器人系统为实验平台,对基于信息耦合度的自组织分群控制方法展开实验验证工作。针对E-puck机器人的特点,首先建立其运动学模型,并结合所量测的邻居距离-方位信息,以基于信息耦合度的分群控制思想为基础,设计了面向群集机器人分群应用的控制算法,最后在Webots机器人仿真平台上开展的分群实验验证了所提算法的可行性和有效性。
[Abstract]:The clustering behavior of the cluster system, represented as a polymerization mother groups are divided into several subgroups. The phenomenon as an inherent motion mode of cluster system, cluster behavior not only overcomes the defects of single group movement difficult to carry out effective characterization of cluster splitting phenomenon, but also gives more flexibility and cluster motion behavior is diversity. The typical performance of cluster to adapt to the environment. Therefore, to carry out research on cluster clustering control system, to reveal the internal mechanism of cluster "split - Fusion" behavior, deepening of self organizing cluster motion mechanism of cognition, and the engineering application to promote clustering behavior, has important theoretical value and practical significance. Based on the National Natural Science Fund Project (No.51179156) as the basis, for a class does not depend on the typical assignment, consultation and other intelligent planning mechanism of self-organizing clustering behavior, in revealing its Based on the generation mechanism, based on information coupling (Information coupling degree, ICD) from the research methods and related issues of clustering in control. The main work and research results are as follows: (1) clustering behavior of self-organizing clustering mechanism and Modeling Research on biological cluster cluster system as the reference, from the stimulus information based on directional propagation angles within the group, the internal information communication mechanism of cluster grouping movement, pointed out that the clustering behavior is multiple external stimuli spread a differentiation phenomenon caused by the movement of the individual within the group, and found the difference correlation strength between individuals is the root cause of clustering behavior. Based on the understanding of the definition the characterization of the strength of the correlation between individual index information coupling degree, and proposes an information coupling clustering control model based on the framework for further clustering control Laid the foundation of the algorithm design. (2) cluster clustering control method based on the biological cluster interaction relationship between individual information based on coupling degree, the comprehensive utilization of position and speed information of individuals, to establish the information model of coupling degree. Then, starting from the stimulation of information dissemination in the maximum angle, put forward a based on the "max-ICD" grouping control strategy. Then, the self-organization "interactions" rules into the traditional "separation / team / polymerization" cluster coordination rules, designed the information coupling degree of clustering algorithm based on self organizing clustering behavior is realized under external stimulation of the cluster system. Theoretical analysis and simulation experiments verify the validity of the proposed control algorithm of clustering. (3) without velocity measurement under cluster self-organization groups system clustering control method for speed measurement of the cluster system is proposed. A method using only the location information of the cluster control method. Firstly, through the construction of distributed observer, the relative position between the individual information real-time estimation of the relative speed of information. Then, the information coupling clustering framework based on control theory as the basis, combined with the speed of the information is unknown, the relative position of relative velocity information estimation and information between individuals, designed without clustering control algorithm of velocity measurement, realizes the self organization clustering movement cluster system. Theoretical analysis and simulation experiments show that the speed measurement proposed since the effectiveness of clustering control algorithm. (4) with clustering control method for cluster system the individual information clustering movement access latency problems delay, firstly some delay limitation information coupling control algorithm based on cluster expansion The discussion, found delays will significantly reduce the performance of clustering algorithm. Then, divided into constant delay and time-varying delay two cases, to study the time delay of the cluster system clustering control method. For constant time, using the Lyapunov-Razumikhin theorem that the delay bound cluster realize stable cluster movement and related convergence conditions for time-varying delay, delay bound by the Lyaponov-Krasovskii functional form of linear matrix inequalities and convergence conditions are obtained. The theoretical analysis and experimental results show the effectiveness of the cluster clustering algorithm of the control system under the proposed delay. (5) the cluster grouping control robot system and experimental studies of the E-puck cluster system as the experimental robot based on the information platform, the coupling degree of self organization clustering control method experiment. According to the characteristics of E-puck robot, Firstly, the kinematic model, and combined with the measured neighbor distance azimuth information, the information coupling degree of clustering based on control theory, design oriented cluster control algorithm of robot application clustering, clustering experiment finally carried out in simulation platform of the Webots robot verified the feasibility and effectiveness of the algorithm.

【学位授予单位】：西北工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP13

【相似文献】