CIF: Small: Collaborative Research: Coordination and Cooperation in Networked Multi-Agent Systems

随着蜂窝和WiFi网络现在无处不在地出现在我们的日常生活中，联网的无线系统已经对社会产生了重大影响。一类重要的网络是分散式网络，其特点是缺乏固定的基础设施、多跳性质和节点移动性；这些功能提供了极大的灵活性，但也引入了各种设计问题。在分散的网络系统中，一个基本的问题是不同节点之间的活动协调，其中足够的通信资源的可用性对于代理之间的合作和协调至关重要。这个项目采取了一种独特的方法来发展对通信和合作之间的相互作用以实现协调的基本理解，并解决了联网多代理系统中概率动作的通信和协调。这可以被视为基于网络的随机开环控制的一个实例，其中，在没有反馈的情况下，向代理发送单个通信消息来控制其行为。该项目调查了两项相互关联的研究任务。首先，PI希望了解小型多终端网络中协调的基本信息论限制以及这些限制对通信网络拓扑的依赖性。第二个研究目标是根据纠错码设计显式协调码，并评估其在严格的延迟和复杂性约束下的性能。例如，用于探索未知拓扑的多代理系统、分布式监视应用、自动交通控制应用以及大型计算机网络和电网中的负载平衡。与该项目有关的其他活动包括将研究成果纳入本科生和研究生课程，以及对科学和技术领域的少数群体和妇女进行指导。

Networked wireless systems have significantly affected society, as cellular and WiFi networks are now ubiquitous in our daily life. An important class of networks is that of decentralized networks, which are characterized by a lack of fixed infrastructure, a multi-hop nature, and node mobility; such features provide significant flexibility, but also introduce a variety of design issues. A fundamental problem in decentralized networked systems is the coordination of activities of different nodes, where the availability of sufficient communication resources is critical in order for agents to cooperate and coordinate. This project takes a distinctive approach to develop a fundamental understanding of the interplay between communication and cooperation to achieve coordination and addresses the communication and coordination of probabilistic actions in networked multi-agent systems. This can be seen as an instance of network-based stochastic open loop control where a single communication message is sent to the agents to control their behavior in the absence of feedback. The project investigates two interrelated research tasks. First, the PIs wish to understand both the fundamental information-theoretic limits of coordination in small multi-terminal networks and the dependency of these limits on the communication network topology. The second research goal addresses the design of explicit coordination codes from error correction codes and to assess their performance under stringent delay and complexity constraints.The proposed research has the potential to enable advances in applications where actions must be communicated to ensure a specific behavior. Examples are multi-agent systems for the exploration of an unknown topology, distributed surveillance applications, automatic traffic control applications, and load balancing in large computer networks and in power grids. Additional activities related to the project include the integration of research outcomes in courses at the undergraduate and graduate level and the mentorship of minorities and women in science and technology.