基于多模态网络事件的无线传感网自适应路由规划理论技术研究

Wireless sensor networks (WSNs) have been applied in a large variety of application domains, from environment monitoring, scientific observation, habitat tracing, to emergency detection, field surveillance, infrastructure protection, etc. As mentioned in many literatures, most of energy consumption is wasted while the radio is turned on. Therefore, how to effectively schedule the radio switch becomes the key factor to prolong the network lifetime. Low power listening (LPL) technology is widely used in network design for many WSNs. Traditional approaches adopt LPL as a static scheme to control the radios on sensor nodes. This project is to propose a dynamic energy-saving routing strategy for WSNs, by combining comprehensive network historical statuses and modeling network traffic with machine learning. Three main issues will be studied: 1) Statistic-Based network traffic modeling under low duty cycling, to predict the arrival time for each data packet; 2) analysis technology for unstrutured multi-dimentional network events, to excavate the key network parameters affecting routing performance; 3) network traffic based energy-saving routing protocol under LPL, assuring the compatibility with other existing network protocols. This study brings forth new approaches for traditional routing strategies, and offers both theoretical and technological support for protocol design in WSNs.

传感器网络将逻辑上的信息世界与真实的物理世界融合在一起，广泛地应用于环境监测、工农业控制、生物医疗、国防军事等诸多领域。根据大量研究表明，能量开销基本由无线收发器开启的时间决定。如何减少无线收发器开启时间成了延长无线传感器网络寿命的关键所在。传统基于低占空比的工作模式无法动态适应网络状态的改变，所有节点的路由配置都是静态的。本项目拟研究一种基于网络流量统计的动态节能路由策略。通过挖掘周期性数据采集无线传感器网络的特性，利用机器学习方法，完成网络数据流的实际建模过程。具体的研究内容包括：1）研究基于统计的低功耗异步网络数据流建模，以准确刻画节点发送和接收数据包的时间分布；2）研究非结构化多维网络行为分析技术，以挖掘影响网络性能的关键行为参数，为动态路由策略提供参照证据；3）研究面向低占空比的基于网络流量分布的路由节能策略，并保证与其他网络协议的兼容性。

随着通信技术、传感器技术和嵌入式计算技术的快速发展和日益成熟，具有通信、传感和计算能力的无线传感器节点开始出现，并且被逐渐的应用到生产 和生活之中。长时间无错误的运行是无线传感器稳定运行的基础。本项目提出的基于网络流量统计的无线传感网络节点动态路由节能策略。该方法挖掘数据周期性数特性，利用机器学习方法，完成无线传感器网络数据流的建模过程。该方法和传统方法比，对于网络数据流的预测比 传统的简单模型更加准确，能够为其他应用提供细粒度的服务。