非视距误差神经网络改正的超宽带定位模型研究

非视距环境是造成超宽带定位系统精度下降的主要原因。由于非视距环境的测距精度下降难以通过常规计算方法建立改正模型,提出了一种基于反向传播算法的神经网络改正的超宽带稳健定位模型。该方法通过反向传播神经网络的自适应学习方法建立了一种超宽带非视距误差改正的稳健定位模型,实现了在非视距环境下超宽带定位精度的提升。首先采集非视距环境下超宽带测距值,提取超宽带在非视距环境下的坐标序列,计算得到误差序列,然后通过反向传播神经网络建立误差改正模型预测得到标签的误差改正值,最后使用超宽带Kalman滤波定位模型进行超宽带定位,从而消除非视距环境对定位精度的影响。通过对比实验分析,本模型较多项式拟合模型超宽带测距精度提高46.8%,定位精度提高43.4%;较多面函数拟合模型超宽带测距精度提高28.2%,定位精度提高26.2%。实验结果表明,反向传播算法的神经网络对超宽带非视距定位模型的误差改正有很好的效果,对超宽带定位精度的改正效果显著。

Research on UWB Positioning Model Corrected by Non-Line-of-Sight Error Neural Network

The non-line-of-sight (NLOS) environment is the main reason for the accuracy degradation of UWB positioning system. It is difficult to establish a correction model by conventional calculation methods because of the decrease of ranging accuracy in NLOS environment. In this paper, a robust UWB positioning model based on back-propagation neural network correction is proposed. By using the adaptive learning method of back-propagation neural network, a robust UWB positioning model with NLOS error correction is established, which can improve the positioning accuracy of UWB in non-line-of-sight environment. First, the UWB ranging value in NLOS environment is collected, the coordinate sequence of UWB in NLOS environment is extracted, and the error sequence is calculated. Then, the error correction model is established by back-propagation neural network to predict the error correction value of the tag. Finally, UWB positioning is carried out by using UWB Kalman filter positioning model, so as to eliminate the influence of non-line-of-sight environment on positioning accuracy. Compared with the polynomial fitting model, the model proposed improves the precision of UWB ranging by 46.8%, positioning accuracy by 43.4%, and positioning accuracy by 28.2% and 26.2% respectively. Experiment results show that the back-propagation neural network has a good effect on the error correction of UWB NLOS positioning model, as well as a significant effect on the correction of UWB positioning accuracy.