基于增量探测的大规模传感网导向性诊断理论与技术研究

项目名称： 基于增量探测的大规模传感网导向性诊断理论与技术研究

项目编号： No.61303196

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 龚伟

作者单位： 清华大学

项目金额： 28万元

中文摘要： 无线传感器的故障诊断技术是保证无线传感器网络高效稳定运行的关键技术之一。现有无线传感器网络诊断方法中的大多数都存在信息收集与推理过程分离这一现象，这样会导致整个信息收集过程是静态和预定义的，无法适应无线传感器网络的高动态性特征。预定义的信息收集过程可能会造成两个不好的结果，过多的信息收集给网络带来了额外的通讯开销或是过少的信息收集使得原因推断产生过多的错误判断。基于这个观察，本项目组初步研究发现如果可以将信息搜集过程与诊断推理过程有效结合起来，将大大提高网络故障的诊断准确率。因此，本项目组提出了一种导向的无线传感器网络的诊断技术。在这个方法中，信息收集的过程是在概率推理模型的指导下进行的，随着信息收集不断完善，推理模型也不断得到细化，从而提高推理的准确率。同时，该方法通过动态获取潜在问题区域的拓扑关系，将诊断的范围限制在局部，从而有效地降低了网络诊断的开销。

中文关键词： 无线传感器；传感器网络；导向诊断；诊断建模；标签估计

英文摘要： Network diagnosis is crucial in managing a wireless sensor network (WSN) since many network-related faults, such as node and link failures, can easily happen. Diagnosis tools usually consist of two key components, information collection and root-cause deduction, while in most cases information collection process is independent with root-cause deduction. This results in either redundant information which might pose high communication burden on WSNs, or incomplete information for root-cause inference that leads false judgments. To address the issue, we propose DID, a directional diagnosis approach,in which the diagnosis information acquirement is guided by the fault inference process. Through several rounds of incremental information probing and fault reasoning, root causes of the network abnormalities with high credibility are deduced. We employ a node tracing scheme to reconstruct the topical topology of faulty regions and build the inference model accordingly. Combining an incremental probing scheme and a dynamic probabilistic inference model, DID effectively localizes the root causes of various network abnormalities. Additionally, the diagnosis is confined to topical area covering the problematic network elements in high potential. With a low overhead, a high accurate real-time network diagnosis service is thu

英文关键词： wireless sensor；sensor network；directional diagnosis；diagnosis modeling；tag estimatin