基于多Agent系统的网络协同控制研究

发布时间：2018-07-24 18:06

【摘要】：随着科学技术在各个方面的迅速发展,多Agent系统所研究的分布式问题是当前研究的热点问题。在所研究问题呈现出复杂性时和分布性时,单体解决问题时通常会变现出受资源与能力的制约,因此通过单体间的组合协同实现能力的提升和资源的扩充将使系统拥有更好的有效性。在国防军事、环境监控以及交通管理等诸多方面都是无线传感器网络的相关应用领域,多传感器网络目标跟踪技术作为其中的一方面应用存在着与多Agent系统系统研究相似的问题。单个传感器受到感知、处理等节点资源的严重制约,很难完成目标的跟踪实现,通过与多Agent系统的网络协同处理来充分利用资源避免资源限制实现更好的跟踪能力将会是很好的解决办法。本文对多Agent系统网络的协同控制问题进行分析研究,开始介绍了多Agent系统相关的基础理论和相关的研究方向。对于协同控制部分首先进行了任务分解相关内容的研究,分析了一般通过任务关系图来描述任务之间的约束和依赖关系的有向无环图,综合现在的网络环境和通信需求,将其通过XML任务树描述进行任务的分解算法实现。然后进行了任务协同过程的具体研究,在综合分析几种协作过程的基础之上改进了一种基于动态任务分配的任务协同过程,并通过围捕过程验证了任务协同的有效性。最后通过对传感器网络特点和目标跟踪需求进行分析,结合前几章的Agent协同原理提出基于动态集的目标跟踪策略,通过基于目标状态信息的变化和网络节点状态信息建立成员和集首都随目标动态变化的动态跟踪集,目的使资源的调用更加合理、目标的跟踪更加有效,最后在Visual C++6.0环境下进行的仿真分析也验证了其有效性和资源调用合理性。
[Abstract]:With the rapid development of science and technology, the distributed problem of multiple Agent systems is a hot issue. When the problem is complex and distributed, the problem solving often appears to be constrained by resources and capacity. Therefore, the enhancement of capability and the expansion of resources will make the system more effective. In defense, military, environmental monitoring, traffic management and many other aspects of wireless sensor networks are related applications, As one of the applications of multi-sensor network target tracking technology has similar problems as the research of multi-sensor Agent systems. Single sensor is severely restricted by node resources such as perception and processing, so it is difficult to achieve target tracking. It will be a good solution to make full use of resources and avoid resource constraints to achieve better tracking ability by cooperating with multiple Agent systems. In this paper, the cooperative control problem of multiple Agent system network is analyzed, and the basic theory and research direction of multi-Agent system are introduced. In the part of collaborative control, the task decomposition related content is studied, and the directed acyclic graph which describes the constraints and dependencies between tasks by task relation graph is analyzed, which synthesizes the current network environment and communication requirements. The task decomposition algorithm is implemented by describing the XML task tree. Then the task coordination process is studied and a task coordination process based on dynamic task assignment is improved on the basis of comprehensive analysis of several collaboration processes. The effectiveness of task coordination is verified by the capture process. Finally, by analyzing the characteristics of sensor networks and the requirement of target tracking, combined with the principle of Agent collaboration in the previous chapters, a target tracking strategy based on dynamic set is proposed. Based on the change of the target state information and the network node state information, the dynamic tracking set of the member and set capitals with the dynamic change of the target is established in order to make the resource transfer more reasonable and the target tracking more effective. Finally, the simulation analysis under Visual C 6.0 environment also verifies its validity and rationality of resource transfer.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18;TP13

【参考文献】