计算机与现代化Issue(1)：117-126,10.DOI:10.3969/j.issn.1006-2475.2026.01.016
基于注意力的LSTM模型和改进抗差算法的GNSS/INS组合导航系统误差抑制方法
Error Suppression Method Based on Attention-based LSTM Model and Improved Robust Algorithm for GNSS/INS Integrated Navigation System
摘要
Abstract
Global navigation satellite system(GNSS)and inertial navigation system(INS)have been widely studied and applied in the field of automatic driving.However,in the environment of GNSS signal rejection such as tunnels,the navigation perfor-mance of GNSS/INS integrated navigation system will be greatly reduced.In order to reduce the navigation error of GNSS/INS in-tegrated navigation system during GNSS outages,an error suppression method based on attention-based long short-term memory network(LSTM)model and improved robust algorithm is proposed.In this method,attention mechanism is introduced into LSTM network to build the prediction model,and GNSS pseudo-position increments are predicted to assist INS navigation.The model improves the prediction accuracy by dynamically adjusting the feature attention.In addition,since the prediction error of GNSS pseudo-position accumulates with time,the time fading weighting factor is introduced to construct the robust factor,and the measurement noise covariance matrix is expanded to reduce the impact of the cumulative error of pseudo-position information on the navigation performance.Land vehicle experiments show that compared with the advanced method,the root mean square er-ror of the position of this method is reduced by 57.5%,36.4%,and 47.3%in the north,east,and horizontal directions,respec-tively.Therefore,during GNSS signal outages,the proposed method can suppress the error divergence of the INS and effectively improve the navigation performance of the integrated system.关键词
GNSS中断/组合导航/注意力机制/LSTM/抗差卡尔曼滤波器/误差抑制Key words
GNSS outages/integrated navigation/attention mechanism/LSTM/robust Kalman filter/error suppression分类
信息技术与安全科学引用本文复制引用
庞贤瑞,江金光,谭宏彬,严培辉,孟小亮..基于注意力的LSTM模型和改进抗差算法的GNSS/INS组合导航系统误差抑制方法[J].计算机与现代化,2026,(1):117-126,10.基金项目
国家重点研发计划项目(2021YFB2501102) （2021YFB2501102）
2023年湖北省重大科技攻关项目(2023BAA025) （2023BAA025）
