摘要
WSN定位中的每个节点提供的数据会对其他节点的位置估计产生重要影响,然而在坐标真值未知的情况下难以对其进行有效度量。为此,基于无监督机器学习方法,在真实值未知的前提下设置评估基准,对节点的数据质量评估问题进行合理建模;同时,结合节点的历史行为,设计基于阈值判断的异常检测方案,剔除不合格的定位数据。仿真证明,所提方案能够有效评估节点的定位数据质量,结合异常节点检测方案,能够极大程度地提高网络的定位精度。
The data provided by each node in wireless sensor network(WSN)localization has a significant impact on the position estimation of other nodes,but it is dificult to effectively evaluate the data quality when the ground truth of the coordinates is unknown.Therefore,based on the unsupervised machine learning methods,this paper establishes the evaluation benchmark under the premise of unknown true value,and the data quality evaluation problem of nodes is reasonably modeled.Meanwhile,according to the historical behavior of nodes,an anomaly detection scheme based on threshold judgment is designed to eliminate unqualified location data.Simulation results show that the proposed scheme can effectively evaluate the location data quality of nodes,and the location accuracy of the network can be significantly improved with the combination of the abnormal nodedetection scheme.
作者
邱颖
代玉琢
王琛
朱亚萍
QIU Ying;DAI Yuzhuo;WANG Chen;ZHU Yaping(School of Software Engineering,Tongji University,Shanghai 201804,China)
出处
《移动通信》
2023年第10期44-50,共7页
Mobile Communications
基金
国家自然科学基金青年科学基金“基于测距的室内协作定位激励机制研究”(62101385)。
关键词
无线传感器网络定位
无监督学习
数据质量评估
异常检测
wireless sensor network localization
unsupervised learning
data quality assessment
anomaly detection
