基于果蝇——广义回归神经网络优化的WSN节点定位算法  

Node Localization Algorithm of WSN Based on Fruit Flies Optimization and Generalized Regression Neural Network

作　　者：陈璟[1] 虞继敏[2] 

机构地区：[1]广西科技师范学院数学与计算机科学学院,广西来宾546199 [2]重庆邮电大学工业物联网与网络化控制教育部重点实验室,重庆400065

出　　处：《南京师大学报（自然科学版）》2017年第2期31-38,共8页Journal of Nanjing Normal University(Natural Science Edition)

基　　金：2014年度广西高校科学技术研究项目(LX2014489);2016年广西高校重点实验室建设项目;重庆市自然科学基金(cstc2013jcyjC0013)

摘　　要：针对无线传感器网络(WSN)基于测距的定位算法中,利用节点坐标计算方法获得的节点坐标位置存在较大误差的问题,提出一种无需进行坐标计算的果蝇—广义回归神经网络(FOA-GRNN)优化的WSN节点定位算法.该算法利用广义回归神经网络(GRNN)较快的学习速度和较强的逼近能力建立WSN节点定位模型,通过果蝇优化算法(FOA)调整广义回归神经网络的平滑参数,降低调整平滑参数时人为因素的影响,由神经网络直接输出未知节点坐标.仿真实验表明,通过果蝇算法优化的FOA-GRNN模型的节点定位精度比未经优化的GRNN模型的节点定位精度高.同时,比较了FOA-GRNN模型与BP神经网络模型、虚拟节点BP网络模型(VNBP)在WSN节点定位中效果,表明FOA-GRNN模型在WSN节点定位精确性方面具有明显优势.In order to improve the accuracy of WSN node location and avoid the deficiency of the distance measurement based localization algorithm using node coordinates to calculate the location of unknown nodes, a new WSN node localization algorithm of Fruit Flies Optimization-Generalized Regression Neural Network（FOA-GRNN） was presented. The proposed FOA-GRNN algorithm builds WSN positioning model by the fast learning speed and strong approximation ability of Generalized Regression Neural Network （GRNN） and adjusts GRNN＇s smoothing parameter by using Fruit Flies Optimiza-tion Algorithm（FOA）to reduce the impact of human factors on selecting GRNN smoothing parameter to minimum. Finally, the coordinates of unknown nodes in WSN can be directly obtained from the output of FOA-GRNN model. Simulating results show that the localization accuracy of FOA-GRNN optimized by FOA is better than that of GRNN model. In addi-tion, the FOA-GRNN algorithm was compared with BP algorithm and VNBP algorithm in WSN nodes localization. Simula-ting results further show the FOA-GRNN algorithm has obvious advantages in the accuracy of WSN node localization.

关 键 词：无线传感器网络 节点定位 广义回归神经网络 果蝇优化算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]