网络化多智能体系统的时变编队跟踪控制

发布时间：2018-01-09 18:11

  本文关键词：网络化多智能体系统的时变编队跟踪控制　出处：《哈尔滨工业大学》2017年博士论文　论文类型：学位论文

  更多相关文章： 网络化多智能体系统 时变编队跟踪 时变一致性 广义一致性 网络化预测控制

【摘要】：运动协作和编队问题是多智能体系统的一个重要应用领域。在实际工程应用中,期望的多智能体队形中各个智能体之间的关系通常是随时间变化的,甚至是时变非线性的。然而,现有文献中只能解决一部分具有特定条件的多智能体系统时变编队问题,对于多智能体的一般性的时变编队问题尚没有成熟的解决方案。多智能体系统一般性的时变编队问题还是一个空白领域。因此,本文的主要研究目的是为了提出一套适用于有参考输入的网络化多智能体时变编队控制策略。在多智能体系统控制问题中,智能体之间的网络通信受限问题是一个不可避免的问题。智能体之间的通信延迟和丢包会影响系统的控制性能,甚至是破坏系统的稳定性。在现有文献中,从随机过程的角度分析网络延迟和通过数学模型对网络延迟进行补偿是两种常用的研究手段。其中,从随机过程进行研究的控制策略通常需要假设网络特性符合特定的随机分布,这使得控制策略的工程应用范围受到很大的限制。网络化预测控制是一种基于数学模型的对网络通信延迟和丢包进行补偿的工程方法。然而,在现有文献中,利用网络化预测控制策略解决智能体间的通信延迟问题通常采用一种“被动补偿”的方式,这使得整个系统中需要多次重复的实现用于延迟补偿的动态补偿器,进而使得系统整体的运算效率低下。为了解决多智能体的时变编队跟踪控制问题,本文首先将现有的多智能体一致性问题扩展到多智能体的时变一致性问题,分别针对一阶平面网络化多智能体系统和一般线性网络化多智能体系统提出了时变一致性协议设计方法和稳定性分析方法。齐次,为了解决更一般性的时变编队控制问题,本文提出了广义一致性控制问题,并针对一阶平面网络化多智能体系统给出了广义一致性协议设计方法和稳定性分析方法。在解决智能体间通信受限问题上,我们针对智能体间网络通信延迟和丢包问题设计了基于模型的主动预测补偿策略。相比于现有的“被动补偿策略”,“主动补偿策略”有效的解决的多智能体系统整体计算效率低下的问题。在编队方面,为了解决传统“一致性问题”在解决时变编队问题的瓶颈,我们提出了“时变一致性问题”,并给出了时变一致性协议设计方法和稳定性分析方法。时变一致性问题有效的描述和解决了期望队形中智能体间的输出差(或状态差)随时间变化时的编队问题。利用本文提出的控制策略,能够确定多智能体系统在编队过程中不随队形时变特性变化的稳态误差的上界,从而确保智能体不会在队形保持过程中发生碰撞事故。在时变一致性的基础上,为了描述和解决更一般性的时变编队问题,我们提出了广义一致性问题,并给出了一阶平面多智能体系统的广义一致性协议设计方法和稳定性分析方法。在理论成果的工程验证方面,我们在自主研发的新型网络化控制器NetcontrollerDM3730的基础上,并实现了新型的网络化控制系统仿真与实验平台。同时,为了提高实验过程中轮式机器人对定位信号延迟和数据丢失的鲁棒性,我们研究了反馈通道存在延迟和丢包的轮式机器人跟踪控制问题。在理论研究成果和工程验证平台的基础上,本文分别在局域网环境和万维网环境下实现了充分的实验和实时仿真。实验和实时仿真的结果验证了本文提出的控制策略能够有效解决网络化多智能体的时变编队跟踪控制问题。据我们所知,本文中实现的轮式机器人环绕编队以及轮式机器人在给定曲线路径上的固定车距编队均为在本文中首次实现。
[Abstract]:The movement coordination and formation is an important application of multi-agent system. In practical application, the relationships among the intelligent multi-agent formation in the body is usually expected to change over time, and even time-varying nonlinear. However, variable formation in the existing literature can only solve part of the the specific conditions of the multi-agent system, the general multi-agent time-varying formation problem is still not a mature solution. The general multi-agent system time-varying formation problem is still a blank field. Therefore, the main research purpose of this paper is to put forward a set of suitable for formation control strategy the reference input of networked multi-agent. The time-varying control problem in multi-agent systems, the problem of Limited network communication between the agents is an unavoidable problem. The communication delay and lost between the agents The package will affect the control performance of the system, and even undermine the stability of the system. In the existing literature, from the view of stochastic process analysis by mathematical model of network delay and network delay compensation are the two commonly used research methods. The research from the stochastic process control strategy usually needs to assume the network characteristics are consistent with the random distribution specific, the application scope of control strategy is limited. Networked predictive control is a mathematical model of Engineering compensation method for network communication based on delay and packet loss. However, in the existing literature, the network usually uses a "passive compensation" predictive control strategy to solve the communication agent the delay problem, which makes repeated implementation for dynamic compensator delay compensation needs of the entire system, which makes the whole system The operation efficiency is low. In order to solve the multi agent time-varying tracking control problem of formation, the multi-agent consensus problem existing extended to multi-agent time-varying consistency problem, design method and stability analysis of consistency protocol method respectively according to the first-order planar networked multi-agent systems and general linear network multi agent system is proposed. The time-varying homogeneous, in order to solve the more general time-varying formation control problem, this paper proposes a control problem of generalized consistency, and multi-agent system are given the broad consensus protocol design method and stability analysis method for the first-order planar network in solving the problem of limited communication. Among the agents, we aim at the intelligent body network communication delay and packet loss problems in the design of active compensation strategy based on prediction model. Compared with the existing "passive compensation strategy", "main The whole multi-agent system to solve the dynamic compensation strategy of the problem of low computation efficiency. In the formation, in order to solve the bottleneck of traditional "consistency" formation when solving the problems, we proposed a time varying consistency problem, design method and stability analysis method and consistency protocol are given. Variable. Time varying consistency problems effectively describe and solve the expected output of the difference between the formation of agent (or worse) formation changes with time. The control strategy proposed in this paper, the upper bound can determine the multi-agent system in the process of shape formation with no steady-state error time-varying characteristics of change intelligent, so as to ensure the body does not keep in formation collision process. Based on the variable consistency in the formation, in order to describe and solve problems become more general, we propose a generalized uniform The problem, and gives the analysis method of generalized consistent first-order planar multi-agent system design method and stability. In the theoretical results of the engineering verification, based on independent research and development of our new network controller NetcontrollerDM3730, and the realization of the network model and simulation control system experimental platform. At the same time, in order to to improve the robustness of wheeled robot in the process of positioning signal delay and data loss, we study the existence of wheeled robot feedback channel delay and packet loss tracking control problem. Based on theoretical research and engineering verification platform, this paper in the LAN environment and web environment to achieve adequate experimental and real-time simulation. Experiment and real-time simulation results verify the proposed control strategy can effectively solve the networked multi-agent time-varying formation tracking Control problem. According to our knowledge, the wheeled robot's surrounding formation and wheeled robot's fixed vehicle distance formation on given curve path are first implemented in this paper.

【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP18;TP13
，

本文编号：1402110