Reconfiguration and Cooperative Control for Multi-Agent Networks

多智能体网络的重构与协作控制

• 批准号: DGDND-2017-00100
• 负责人: Aghdam, Amir
• 金额: $ 2.91万
• 依托单位: Concordia University
• 依托单位国家: 加拿大
• 项目类别: DND/NSERC Discovery Grant Supplement
• 财政年份: 2017
• 资助国家: 加拿大
• 起止时间: 2017-01-01 至 2018-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=617774
• 关键词: Reconfiguration; Cooperative; Control; Multi; Agent

There is an increasing interest in the deployment of multi-agent networks for emerging applications such as surveillance and target tracking. This proposal focuses on two research problems in the area of multi-agent networks. The first problem concerns network reconfiguration, which aims at improving the performance of a network by properly positioning its nodes and/or changing the weights of its links (e.g., by adjusting the communication/sensing power of the nodes). This is a problem which requires a strong practical background, and while the results will be developed for a general class of asymmetric networks, the applicant's prior experience in the design of experimentally validated distributed control schemes for underwater sensor networks puts him in a unique position to address some of the practical shortcomings of existing results for this type of systems. One of the main characteristics of asymmetric networks is that the graph representing them is directed, and in the case of underwater sensor networks, particularly, it is random too. However, there are not many results in the literature for the analysis of this type of graphs, due to their complexity compared to (deterministic) undirected graphs. This limits the extent of which certain observations in such networks can be justified analytically. For example, it is known that the relative positions of the acoustic nodes in an underwater sensor network can have a significant impact on data aggregation performance. In fact, similar observations have been reported in other types of asymmetric networks with no convincing theoretical justification. The applicant and his team have recently developed some theoretical results, validated by simulations, that relate the important properties of general weighted directed graphs (such as connectivity) to the configuration of the network, enabling the research community for the first time to justify these observations theoretically, and more importantly, use the results to further improve the performance of the network. These results will be used in the proposed research to find the optimal configuration for asymmetric networks. The results can also be used in other applications such as traffic network control systems, to justify some counter-intuitive observations reported in this type of systems (e.g., negative impact of the addition of some roads on the overall traffic flow in the network). The other problem investigated in this proposal is concerned with cooperative decision making for heading control of multiple vehicles, where it is desired to coordinate a group of vehicles in order to detect, localize, track and intercept a group of objects that arrive in a protected area (mission space) at random time instants. The proposed dynamic decision making and control design is based on a reward allocation strategy which directs the vehicles toward the objects in an optimal cooperative fashion.

对于新兴应用程序（例如监视和目标跟踪）的部署，对部署的多代理网络的部署越来越兴趣。该提案重点介绍了多代理网络领域的两个研究问题。第一个问题涉及网络重新配置，旨在通过适当定位其节点和/或更改其链接的权重（例如，通过调整节点的通信/传感能力）来提高网络的性能。这个问题需要强大的实践背景，尽管将为一般的不对称网络制定结果，但申请人在设计实验验证的水下传感器网络设计方面的先前经验，使他处于独特的位置，以解决这种类型的系统现有结果的某些实际缺陷。不对称网络的主要特征之一是指向代表它们的图形，并且在水下传感器网络的情况下，特别是它也是随机的。但是，由于与（确定性）无方向的图相比，文献中的分析结果并不多，用于分析此类图。这限制了该网络中某些观察结果的程度可以通过分析性地证明是合理的。例如，众所周知，声学节点在水下传感器网络中的相对位置可能会对数据聚合性能产生重大影响。实际上，在没有令人信服的理论理由的其他类型的非对称网络中已经报道了类似的观察结果。申请人及其团队最近通过模拟验证了一些理论结果，这些结果将一般加权有向图（例如连接性）的重要特性与网络的配置联系起来，使研究社区首次能够以理论上的理论证明这些观察结果是合理的，更重要的是，更重要的是，使用结果进一步提高了网络的性能。这些结果将在拟议的研究中使用，以找到非对称网络的最佳配置。结果也可以用于其他应用程序（例如流量网络控制系统），以证明在这种类型的系统中报告的一些违反直觉观察是合理的（例如，某些道路对网络的整体流量流的负面影响）。该提案中调查的另一个问题与对多个车辆进行控制的合作决策有关，在该决策中，需要协调一组车辆，以便在随机的时机中检测，本地化，跟踪和拦截一组在受保护区域（任务空间）到达的物体。提出的动态决策和控制设计是基于奖励分配策略，该策略以最佳的合作方式将车辆引导到对象。