摘要
针对全球定位系统(GPS)信号中断引发的GPS/惯性导航系统(INS)组合导航定位性能下降问题,提出一种基于神经网络预测的双重优化车载定位算法。首先,构建卷积门循环神经网络(CNN-GRU)预测模型,通过CNN提取惯导数据特征,并利用GRU预测GPS中断期间的伪信号。其次,为提升神经网络预测性能,采用经验模态分解阈值滤波预处理惯性测量单元数据,通过白鲨算法优化网络超参数。然后,基于扩展卡尔曼滤波(EKF)内部参数与系统最优估计误差的关系,构建GRU误差补偿模型以进一步提升GPS中断期间的定位精度。结果表明,相较于CNN-GRU+EKF、CNN-LSTM+KF和GRU+AKF,所提算法在GPS中断30 s和150 s情况下,定位距离的均方根误差平均值分别降低了62.00%、68.31%、74.80%,验证了算法在GPS中断场景下的有效性。
To address the performance degradation of global positioning system(GPS)/inertial navigation system(INS)integrated navigation system caused by GPS signal outages,a dual optimization vehicle positioning algorithm based on neural network prediction is proposed.Firstly,a convolutional neural network-gated recurrent unit(CNN-GRU)prediction model is constructed,the INS data features are extracted by CNN,and the pseudo-signals during GPS outages are predicted by GRU.Secondly,to enhance the performance of neural network,the inertial measurement unit data is preprocessed by empirical mode decomposition threshold filtering,and the hyperparameters of the network are optimized by the white shark optimizer(WSO).Then,based on the relationship between the internal parameters of the extended Kalman filter(EKF)and the optimal estimation error of GPS/INS,error compensation model is established by GRU to further optimize the positioning accuracy during GPS outages.The experimental results show that compared with CNN-GRU+EKF,CNN-LSTM+KF and GRU+AKF,the proposed algorithm reduces the average root mean square error of positioning distance by 62.00%,68.31%,and 74.80%under 30 s and 150 s GPS outages,respectively,validating its effectiveness in scenarios with GPS signal denial.
作者
李志伟
宋博文
张树德
张秀宇
李昊林
LI Zhiwei;SONG Bowen;ZHANG Shude;ZHANG Xiuyu;LI Haolin(School of Automation Engineering,Northeast Electric Power University,Jilin 132012,China;Baishan Power Supply Company,State Grid Jilin Electric Power Co.,Ltd.,Baishan 134300,China)
出处
《中国惯性技术学报》
北大核心
2025年第5期462-471,共10页
Journal of Chinese Inertial Technology
基金
国家自然科学基金(62373092)。
