Scalable scheduling and transmission for sensor systems

传感器网络的一个很好的原型是一种由大量独立设备组成的技术，通过通信，这些设备可以就它们监控的区域内发生的事件即时达成共识，与它们的大小和通信范围相比，这个区域很大。该项目有三个主要的研究重点，将有助于建立这样的基础设施：1)传感器传输的调度；2)同步；3)物理层传输体系结构。我们提出的思想是分布式的、协作的和数据驱动的，挑战了当今用于构建通信网络的一些传统智慧，这些传统智慧忽略了问题的基本方面，例如：1)传感器不是最终用户；2)传感器可以协作而不是竞争媒介来传递关键信息；3)数据在空间和时间上的聚合熵/复杂性低。事实上，部署传感器网络是为了记录意外的分散事件、突然但局部的变化、不一致或很少发生的不连续。目前应用于这一问题的网络技术在一个根本方面是失败的：可扩展性。节点之间的交互在许多分层的功能中高度结构化。可扩展性将通过简化大量设备之间的交互，并将它们的操作和消息交换限制为通过广播、噪声、有限范围的接口来实现。本研究最具原创性的贡献是找到了一类在这些限制范围内运行的调度策略，它由传感器的数据驱动，实现了从分布式来源的高效传输，而不需要在它们之间进行任何事先的数据交换。

A good archetype for a sensor network is that of a technology that consists of large set of independent devices that, through communications, can reach consensus on the fly on events that occur in an area they monitor, a region, which is large, compared to their size and communication range. This project has three main research thrusts that will contribute to build such infrastructure: 1) the scheduling of the sensor transmissions; 2) the synchronization 3) the physical layer transmission architecture. The ideas we propose are distributed, cooperative and data driven and challenge some of the conventional wisdom used in building communication networks today which neglect fundamental aspects of the problem, such as: 1) the sensors are not end users; 2) the sensors can cooperate rather than contend for the medium to deliver critical information; 3) the aggregate entropy/complexity of the data in space and time is low. In fact, sensor networks are deployed to record unexpected dispersed events, sudden but localized changes, inconsistencies or discontinuities that occur rarely. Current network technology applied to this problem fails in one fundamental aspect: the scalability. The interactions among nodes are heavily structured in a number of layered functionalities.Scalability will result from un-complicating the interactions between thelarge population of devices and constraining their operations and message exchange to be through a broadcast, noisy, limited range interface. The most original contribution of this research is in finding a class of scheduling strategies that operate within these restrictions and is driven by the sensors' data, achieving efficient transmission from the distributed sources without requiring any prior exchange of data among them.