EAGER: Distributed Network Optimization and Consensus with Event-Driven Communications

本研究的目的是开发一种技术，允许多个决策者通过单独的计算和决策者之间有限的通信来共同解决优化和学习问题，从而促进分布式优化和学习。在这里，每个决策者都被视为一个智能代理，被建模为代理网络中的一个节点。代理之间的通信只有在它们的节点通过网络图的链接连接时才能发生。该框架允许对由多个代理组成的大型系统进行建模，其中每个代理从环境及其相邻代理收集数据。代理处理接收到的数据，并随着时间的推移相互通信。本研究项目旨在通过减少节点必须与相邻节点通信确认的要求，并通过使用事件驱动通信来减少通信量，从而扩展现有知识，其中相邻节点之间的通信由预定义事件的发生触发。考虑的应用包括无人机网络和电网。该项目将对大学学生和当地高中学生产生影响。该研究项目将为分布式通信代理网络提出新的问题，这些通信代理的任务是在使用有限通信的情况下实现全局定义的目标。该项目还旨在定义新的应用驱动的分布式网络问题，在这些问题上取得理论进展，并详细探讨该理论的几种应用。这项工作的主要目标是将现有算法扩展到事件触发框架。这些事件可以定义为满足特定条件的本地系统变量，也可以定义为从外部环境接收特定类型信息的到达事件。该研究项目将模拟这两种类型的事件以及它们对分布式决策动态的影响。对于外生到达信息，研究人员将考虑外生数据偶尔到达较小的特定节点集。本文将从到达过程、网络动态和图结构等方面分析所设计的分布式算法的稳定性。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The objective of this research is to develop techniques to allow multiple decision makers to jointly solve optimization and learning problems through individual computations and with limited communications among the decision makers contributing to an area known as distributed optimization and learning. Here, each decision maker is viewed as an intelligent agent, modeled as a node in a network of agents. Communication between the agents can occur only if their nodes are connected by a link of the network graph. This framework allows modeling of large systems consisting of multiple agents where each agent collects data from the environment as well as from its neighboring agents. The agents process the received data and communicate with one another over time. This research project aims to extend existing knowledge by reducing the requirement that nodes must communicate acknowledgments from neighboring nodes and reducing the amount of communication by using event-driven communications, where communication between neighboring nodes is triggered by the occurrence of pre-defined events. Applications considered include networks of unmanned aerial vehicles and electric power grids. The project will provide an impact on students at the university and through outreach to local high schools.The research project will pose novel problems for distributed networks of communicating agents tasked with achieving goals defined globally while using limited communication. The project also aims to define new applications-motivated distributed network problems, to make theoretical headway on these problems, and to pursue several applications of the theory in detail. A main goal of this work is to extend existing algorithms to the event-triggered framework. These events can be defined in terms of local system variables meeting certain conditions, or as arrival events where certain types of information are received from the external environment. The research project will model both types of events as well as their influence on the dynamics of distributed decision making. For exogenous arriving information, the researchers will consider occasional arrivals of exogenous data at a smaller set of specialized nodes. Stability of the designed distributed algorithms will be analyzed relative to the arrival processes, the network dynamics, and the graph structure.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.