改进的动态差分优化网络的室内分级定位方法

Indoor hierarchical positioning based on dynamic differential optimization network

针对指纹定位方法中单一指纹信息特征有限,以及在提升定位精度时,因增大指纹库造成指纹匹配复杂度增加,导致定位效率不高等问题,提出一种改进的动态差分优化网络的室内分级定位方法:离线阶段,在无线保真(WiFi)指纹中引入三轴地磁强度特征,增加定位区域的局部特征信息,并将指纹样本通过k均值聚类算法(k-means)进行聚类切分,形成区域指纹库,各区域指纹库分别利用改进的动态差分优化埃尔曼神经网络(IDDE-ENN)模型,训练得到每个区域的精细定位模型;在线定位阶段,根据用户的定位数据与区域指纹库的聚类中心匹配,在减少指纹搜索复杂度的同时,实现了区域级定位,然后通过该区域的定位模型对用户进行精细定位。实验结果表明,该方法在线定位阶段的平均定位误差为1.67 m,验证了该室内定位模型的有效性。

Aiming at the problems of limited single fingerprint information features in fingerprint positioning methods,and high fingerprint matching complexity and low positioning efficiency due to the increase of fingerprint database when improving positioning accuracy,the paper proposed an indoor hierarchical positioning method based on dynamic differential optimization network:at the offline stage,three-axis geomagnetic features were introduced into Wireless Fidelity(WiFi)fingerprints to increase the local feature information of the positioning area,and the fingerprint samples were clustered and segmented to form a regional fingerprint database by the k-means algorithm,then the fingerprint database of each region was trained to obtain a fine positioning model for each region by using the Improved Dynamic Differential Evolution-Elman Neural Network(IDDE-ENN)algorithm;at the online positioning stage,the location data of the user were matched with the cluster center of the regional fingerprint database,and the region-level positioning was achieved with reducing the complexity of fingerprint searching,then the user was finely positioned through the regional positioning model.Experimental result showed that at the online positioning stage,the average positioning error of the proposed method would be 1.67 m,indicating the validity of the indoor positioning model.