为提高非视距场景下超宽带(ultra-wideband,UWB)定位精度,本文提出一种基于误差因子的改进加权最小二乘(weighted least square,WLS)算法.该算法利用测距值和实时信道冲激响应特征训练1维卷积神经网络,实现误差因子的准确预测;基于预测得到的误差因子设计改进WLS算法的加权矩阵,赋予不同基站合理的权重,以改善非视距场景下UWB定位性能.通过实测采集静态和动态定位数据对改进WLS算法进行性能验证.实验结果表明:视距场景下,改进WLS算法与最小二乘(least square,LS)算法、WLS算法定位性能相近;非视距场景下,改进WLS算法明显优于LS算法、WLS算法,能够有效抑制非视距误差.
Abstract
In order to improve the positioning accuracy of ultra-wideband(UWB)in non-line of sight(NLOS)scenarios,an improved weighted least square(WLS)algorithm based on error factor was proposed in this paper.A one dimensional convolutional neural network(1DCNN)is trained by using ranging values and real-time channel impulse response(CIR)features to achieve accurate prediction of error factor.Based on the predicted error factor,the weighting matrix of improved WLS algorithm is designed,and different base stations are given reasonable weights to improve the UWB positioning performance in NLOS scenarios.Static and dynamic measured data are collected from the real environment to verify the performance of the improved WLS algorithm.The experimental results show that the improved WLS algorithm has similar positioning performance to the least square(LS)algorithm and WLS algorithm in the line of sight(LOS)scenarios.In the NLOS scenarios,the improved WLS algorithm is obviously better than the LS algorithm and WLS algorithm,and can effectively restrain the NLOS error.
关键词
超宽带/到达时间/非视距/1维卷积神经网络/改进加权最小二乘算法
Key words
ultra-wideband/time of arrival/non-line of sight(NLOS)/one dimensional convolution neural network(1DCNN)/improved weighted least square(WLS)algorithm
维普
万方数据