WN: Collaboration of Networked Nodes through Belief Propagation: Where Computing Meets Communications

无线自组织/传感器网络在我们的生活中扮演着越来越重要的角色，分布式节点之间的协作对其成功至关重要。虽然大多数现有的节点协作方法仅限于局部和链路行为，但本研究采用了一种新的、全网范围的观点：采用一种通用的方法，即信念传播（BP），为无线网络中的协作信息处理和传播提供了一个系统、准确和灵活的框架。信念传播是一种运行在图形模型上的计算算法，而无线网络中存在反映连接拓扑的通信图。本研究探讨两者的协同作用：计算图与通信图的交集。首先，它提供了设计指南和分析工具，以促进BP在无线网络中的应用，从而实现分布式、鲁棒、可扩展和节能的通信和网络协议。同时，明确研究了实际通信约束对设计和分析的影响。对广义BP和混合体系结构的扩展也进行了讨论。在整个研究过程中，积极探索特定于应用程序和以数据为中心的方法以及跨层方法，以促进设计和分析并提高性能。本研究涉及网络、通信和计算的接口，主要依赖于信息论、通信理论、贝叶斯推理、图形理论和模型、通信/计算/信息复杂性等工具。其结果可能会推动这些领域的理论和实践，并有助于下一代无线网络的发展。该研究的多学科性质也有助于跨学科教育和学生的全面培训。

Wireless ad hoc/sensor networks are playing an increasingly important role in our lives, for which collaboration among distributed nodes is crucial for their success. While most existing approaches for node collaboration are restricted to local and link behavior, this research takes a new, network-wide view: a general methodology, belief propagation (BP), is employed to provide a systematic, accurate, and yet flexible framework for collaborative information processing and dissemination in wireless networks. Belief propagation is a computing algorithm operating on graphical models, while in wireless networks there is a communication graph reflecting connectivity topology. This research investigates the synergy of the two: where the computing graph meets the communication graph. First, it provides design guidelines and analytical tools to facilitate application of BP in wireless networks, leading to distributed, robust, scalable, and energy-efficient communication and network protocols. Meanwhile, the impact of real communication constraints on the design and analysis is explicitly studied. Extension to generalized BP and hybrid architectures is also addressed. Throughout this research, application-specific and data-centric approaches, and cross-layer approaches, are actively explored to facilitate design and analysis and improve performance. This research lies in the interface of networking, communications, and computing, and relies heavily on tools from information theory, communication theory, Bayesian inference, graphical theory and models, and communication/computation/information complexity. Its outcome may advance the theory and practice of these areas, and contribute to the evolution of next-generation wireless networks. The multi-disciplinary nature of this research also lends itself to cross-disciplinary education and well-rounded training of students.