Biologically-Inspired Networking and Computation in Large-Scale Autonomous Sensor Networks

提供完整态势感知和普适计算环境的大规模自主无线传感器网络将成为未来网络基础设施的重要组成部分。节点协作是传感器网络成功的关键，因为每个节点本身受到传感范围、功率和处理能力的限制。另一方面，节点竞争是在有限的时间、空间和频率资源下使多个节点联网的结果。对于小规模网络，中心化方案可以访问所有节点信息并进一步控制整个网络，从而实现节点协作和竞争的最佳平衡。然而，在可能涉及数百万个节点和覆盖大地理区域的大型传感器网络中，不可能提供一个集中的方案来管理全网的功能协作和资源竞争。本研究计划专注于大规模传感器网络，并研究以纯分布式方式控制节点协作和竞争的基本机制，并联合考虑传感器网络设计中的三个关键要素：拓扑控制、信息传输和信息处理。这种设计方法的灵感来自于最近的一些生物学研究结果，这些研究是关于我们身体中数十亿细胞如何以合作和竞争的方式控制它们的生长和相互作用的。可交付成果是一般定理、性能界限和分析系统模型，它们捕获了大规模网络的各个方面的交互动态。这些模型不仅定义了相邻节点之间协作和竞争的基本原则，而且还定义了控制每个节点学习环境和调整自身行为的适应规则。它们对于嵌入到每个传感器节点的分布式网络协议的未来设计至关重要，这些节点模仿每个生物细胞中遗传的遗传密码。

Large-scale autonomous wireless sensor networks that provide complete situation awareness and ubiquitous computing environments will become an essential part of the future network infrastructure. Node collaboration is the key for the success of sensor networks due to the fact that each node itself is limited by sensing range, power, and processing ability. On the other hand, node competition is a consequence of networking multiple nodes with limited time, space, and frequency resources. For small-scale networks, node collaboration and competition can be optimally balanced by a centralized scheme that has the access to all the node information and further has the control over the whole network. However, in a large-scale sensor network, which may involve millions of nodes and cover a large geographic area, it is impossible to afford a centralized scheme that manages the network-wide functional collaboration and resource competition.This research program focuses on large-scale sensor networks and investigates the fundamental mechanism that controls node collaboration and competition in a purely distributed manner, with joint considerations of the three key elements in sensor networkdesign: topology control, information transmission, and information processing. The design methodology is motivated by some recent biological research results on how billions of cells in our body control their growth and interaction with each other in a bothcollaborative and competitive way. The deliverables are general theorems, performance bounds, and analytical system models, which capture the interactive dynamics of various aspects of large-scale networking. These models define not only the fundamental principles regarding collaboration and competition among neighboring nodes, but also the adaptation rules that control each node to learn the environment and adjust its behavior. They are crucial to the future design of distributed networking protocols that are embedded into each sensor node, which mimics the genetic code inherited in each biological cell.