面向多跳WSNs的基于LSSVR的节点定位算法

Least-Squares Support Vector Regression-Based Localization Algorithm in Multi-Hop Wireless Sensor Networks

多跳无线传感网络WSNs(Wireless Sensor Networks)中的多类应用均需要准确的位置信息。为此,提出面向多跳WSNs的基于最小二乘支持向量回归机定位算法LSSVR-LA(Least-Squares Support Vector Regression location algorithm)。LSSVR-LA算法先引用转发区域概念,并通过转发区域建立测距模型,然后再利用Secant算法估计传感节点与锚节点间距离,最后将这些距离作为LSSVR输入,建立了基于LSSVR定位算法模型。最终,估计未知节点的位置。实验数据表明,提出的LSSVR-LA算法的定位精度得到有效地提高。

In multi-hop wireless networks,location based applications require an accurate localization algorithm.Therefore,Least-Squares Support Vector Regression-based Localization Algorithm（ LSSVR-LA） is proposed in this paper. LSSVR-LA introduces a forwarding area,and construct ranging model by forwarding area. Then,the distance between sensors and anchor nodes are estimated by Secant algorithm. The distance is used as the input vector of LSSVR machine,and localization model-based LSSVR is constructed. Numerous simulation results show that LSSVR-GF-RSSI algorithm reduces at least 12% in average localization error compared with traditional localization algorithm.