基于NARX神经网络辅助组合导航方法研究  

Integrated Navigation Method Based on NARX Neural Network

作　　者：张帅 郑龙江[1] 侯培国[1] ZHANG Shuai;ZHENG Long-jiang;HOU Pei-guo(School of Electrical Engineering,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《测控技术》2022年第11期119-125,共7页Measurement & Control Technology

摘　　要：为解决短时全球导航卫星系统(GNSS)失效造成车载组合导航系统导航精度下降的问题,提出一种NARX神经网络辅助的组合导航方法。对神经网络辅助导航的原理进行了分类,并分析了神经网络可利用的输入输出信息,提出一种根据惯性测量单位(IMU)测量信息和惯性导航解算信息对GNSS位置速度增量进行预测的方法。通过实测数据实验验证了方法的有效性,GNSS失效60 s期间,导航最大位置误差5.1 m、最大速度误差0.15 m/s。In order to solve the problem of navigation accuracy degradation of integrated navigation system caused by GNSS(global navigation satellite system)failure,a NARX(nonlinear autoregressive with external input)neural network assisted integrated navigation method is proposed.The principle of neural network aided navigation is analyzed,and the available input and output information of neural network is analyzed.A method to predict the position and velocity increment of GNSS according to IMU(inertial measurement unit)information and inertial navigation information is proposed.The effectiveness of the method is verified by experiments.During 60 s of GNSS failure,the maximum position error of navigation is 5.1 m and the maximum speed error is 0.15 m/s.

关 键 词：组合导航 卫星导航失效 神经网络 非线性自回归模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]