利用主成分分析解决定位中的复共线问题  

SOLVING MULTICOLLINEARITY IN LOCALISATION ESTIMATION USING PRINCIPAL COMPONENT ANALYSIS

作　　者：严筱永[1,2] 钱焕延[2] 杨万扣[3] 

机构地区：[1]金陵科技学院信息技术学院,江苏南京211169 [2]南京理工大学计算机科学与技术学院,江苏南京210094 [3]东南大学自动化学院,江苏南京210008

出　　处：《计算机应用与软件》2013年第1期55-58,78,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(61005008);江苏省教育厅省属高校自然科学研究基金项目(11KJD510002;12KJD510006);南京科技发展指导性计划项目(2011ZD011)

摘　　要：DV-Hop定位方法用跳距代替测距,而后利用信标节点进行定位。它具有硬件要求低、计算和通信开销不高且容错性能较好的特性,因而被广泛运用到各个定位系统中。但其定位常常受到信标节点摆放位置和噪声的影响,使得定位结果较差。通过对DV-Hop定位过程的分析,利用机器学习中的主成分分析方法处理信标节点随机摆放所产生的复共线性问题,同时可以消除部分噪声,使得未知节点估计值具有较小的均方差,进而提高定位的精度。仿真实验结果验证该改进后的算法同样具有原先算法优良特性,且定位精确度有所提高。DV-hop positioning algorithm utilises the hop distance instead of the ranging and then uses the beacon nodes for positioning. It has the properties of low hardware requirements, low overhead in computing and communication, and better fauh-tolerant performance, and thus is widely applied to various positioning systems. However, its positioning is often subject to the impact of beacon node placement and noise, this leads to poor positioning results. In this paper, we analyse the DV-Hop localisation process, and then utilise the principal component analysis in machine learning to overcome the multicollinearity problem caused by random placement of the beacon nodes, as well as to eliminate some noise, this makes the estimation value of the unknown nodes has less mean square deviation and further improves the precision of the positioning. Simulation results show that the improved algorit.hm has the excellent characteristics the same as the original algorithm has, and the precision of localisation is improved as well.

关 键 词：无线传感器网络 定位 DV-HOP定位算法 主成分分析 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]