EAGER: Identification and Design of Optimal Communication Topologies in Collaborative Networks

分布式交互代理的大型网络在现代社会和技术应用中无处不在。例子很广泛，包括社会政治网络、卫星和传感器星座、车辆编队和新兴的智能电网。随着网络规模和复杂性的增加，开发下一代定量基础设施以了解其集体行为并提高其性能是势在必行的。本提案中发展的理论和技术有望成为设计安全、有弹性和高效的未来网络的基础。建议的工作应用领域包括社会网络中的信息传播、多智能体系统中的分布式决策以及分布式发电网络的稳定性和效率提高。通过让研究生参与研究，该项目将有助于满足网络安全和网络科学领域的劳动力需求。在过去的十年里，随着理论和计算工具的发展，稀疏信号恢复领域也被称为压缩感知，在不同的科学界进行了大量的研究活动。在信号恢复中，揭示隐藏在欠采样信号和大数据集中的稀疏模式是一个很有意义的问题。然而，网络识别和设计问题虽然同样重要，但却没有受到如此多的关注。这些问题的重点是揭示关键的稀疏结构，这些结构负责大型网络的高效和弹性功能。本提案的研究目标是为协作网络的分析和设计开发一个分析和计算框架。采用动态网络模型，使其适合于拓扑变化和扩充。测量不确定性的放大是用来量化节点的集体行为跨越时间和空间。在凸优化的背景下，采用资源感知约束和扰动方法来确定关键的通信和交互拓扑，以平衡资源的使用和高集体性能。作为一个具体应用，该框架被用于寻找同步代理网络中的最优通信拓扑。

Large networks of distributed interacting agents are omnipresent in modern society and technological applications. Examples are wide-ranging and include sociopolitical networks, satellite and sensor constellations, vehicular formations, and the emerging smart grid. As networks increase in their size and complexity, it is imperative to develop the next generation of quantitative infrastructure to understand their collective behavior and to enhance their performance. The theory and techniques developed in this proposal are expected to serve as a foundation for designing secure, resilient, and efficient networks of the future. Application areas of the proposed work include information dissemination in social networks, distributed decision-making in multi-agent systems, and stability and efficiency improvement in distributed power generation networks. By engaging graduate students in research, the project will contribute to the workforce needs in cybersecurity and network science.The field of sparse signal recovery, also referred to as compressive sensing, has witnessed overwhelming research activity across different scientific communities over the past decade, accompanied by the development of theoretical and computational tools. In signal recovery it is of interest to expose sparse patterns hidden in under-sampled signals and large datasets. Problems of network identification and design however, while just as important, have not received as much attention. The focus of such problems is to uncover pivotal sparse structures that are responsible for the efficient and resilient functioning of large networks. The research objective of this proposal is to develop an analytic and computational framework for the analysis and design of collaborative networks. Models of dynamical networks are employed that lend themselves to topological variations and augmentation. Measures of uncertainty amplification are used to quantify nodes collective behavior across time and space. Resource-aware constraints and perturbation methods are employed in the context of convex optimization to identify pivotal communication and interaction topologies that balance the use of resources with high collective performance. As a concrete application, the proposed framework is used to find optimal communication topologies in networks of synchronizing agents.