基于EMD-LSTM耦合预测模型的BDS多路径误差削弱方法研究  

BDS multipath errors reducing method based on EMD-LSTM coupled prediction model

作　　者：徐小汶 陶远 XU Xiaowen;TAO Yuan(School of Geomatics,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学测绘学院,安徽淮南232001

出　　处：《全球定位系统》2020年第2期98-104,共7页Gnss World of China

基　　金：国家自然科学基金(1704008);安徽理工大学研究生创新基金(2019CX2077)。

摘　　要：北斗卫星导航系统(BDS)在短基线测量中存在的多路径误差是影响定位精度的主要误差项.针对多路径误差的非线性以及坐标序列的非平稳特性,拟采用经验模态分解(EMD)与长短期记忆网络(LSTM)结合的方法,构建EMD-LSTM耦合预测模型,对多路径误差进行预测,削弱多路径误差的影响.实验结果表明,EMD-LSTM耦合预测模型能够有效地削弱了多路径误差影响,E、N、U方向精度分别提高了22%、36%、40%.The multipath errors in short baseline measurement of BeiDou Navigation Satellite System(BDS)were the main errors affecting positioning accuracy.Aiming at the nonlinearity of multipath errors and the non-stationarity of coordinate series,a coupled prediction model of EMD-LSTM was proposed by combining empirical mode decomposition(EMD)with long short-term memory(LSTM)to predict multipath errors and weaken the influence of multipath errors.The experimental results show that the EMD-LSTM coupled prediction model can effectively reduce the multipath errors,and the E,N,and U directions were respectively improved by 22%,36%,and 40%.

关 键 词：北斗导航卫星系统 多路径误差 经验模态分解 长短期记忆网络 

分 类 号：P228.4[天文地球—大地测量学与测量工程]