一种基于SVR的定位误差修正算法

摘要

According to problem of recollecting received signal strength (RSS) fingerprint data when underground radio environment changed, this paper proposed a positioning error correction algorithm based on support vector regression (SVR).This algorithm used SVR in the offline phase, and established the nonlinear relationship between RSS value of to-be-trained points, positioning results and positioning error.In the online phase, the algorithm used the model to calculate the positioning error of RSS samples and fixed location results.The results show that, before and after the increase in the number of offline training points, the correction algorithm reduces the positioning error by 22% and 38% respectively when compared with the fingerprint matching algorithm.%针对井下无线信号传播环境发生改变需要重新采集RSS(received signal strength)指纹数据的问题,提出一种基于支持向量回归(SVR)的定位误差修正算法.该算法在离线阶段利用支持向量回归机,建立待训练点的RSS值和定位结果与定位误差之间的非线性关系,在线阶段利用该模型计算RSS样本的定位误差,并修正定位结果.实验结果表明,在离线训练点数量增加前后,该修正算法比指纹匹配算法的定位误差分别减少了22%与38%.