基于自适应距离ADWKNN室内定位算法

Adaptive Distance-based ADWKNN Indoor Positioning Algorithm

针对基于蓝牙接收信号强度(RSS)的加权K-近邻(WKNN)位置指纹室内定位算法中由于信号波动存在的单一距离估算标准导致定位精度偏低的问题,提出了一种基于自适应距离的ADWKNN定位算法。离线阶段利用K-means聚类算法对指纹数据库进行划分,减少数据查询量以保证定位的时效性。在线定位阶段首先对待定位点处采集到的RSS信号值进行卡尔曼滤波,减弱随机噪声的干扰;然后采用ADWKNN算法计算曼哈顿距离与欧氏距离的标准差,自适应的选择距离估算方法并实现K值的动态变化。实验结果表明,ADWKNN算法的平均定位精度为1.22 m,与使用余弦距离、曼哈顿距离、欧氏距离和Sorensen距离的单一距离的WKNN算法相比,ADWKNN算法的平均定位精度有明显提高。

In the weighted K-nearest neighbor(WKNN)position fingerprint indoor positioning algorithm based on Bluetooth received signal strength(RSS),signal fluctuations can lead to low positioning accuracy when using a single distance estimation stan-dard.This article proposes an ADWKNN localization algorithm based on adaptive distance to address this issue.In the offline stage,K-means clustering algorithm is used to partition the fingerprint database to reduce the amount of data queries and ensure the timeli-ness of localization.In the online positioning stage,the RSS signals collected at the location point are Kalman filtered to reduce the interference of random noise,and then the ADWKNN algorithm is used to calculate the standard deviation of Manhattan distance and Euclidean distance,to select the distance estimation method adaptively and to realise the dynamic change of K value.The experi-mental results show that the average positioning accuracy of the ADWKNN algorithm is 1.22 m,which is a significant improvement compared with the WKNN algorithm using a single distance of cosine distance,Manhattan distance,Euclidean distance and So-rensen distance.