基于BP神经网络辅助的惯性/天文组合导航方法  

Inertial/celestial integrated navigation method based on BP neural network

作　　者：孙洪驰 穆荣军[1] 杜华军[2,3] 崔乃刚 SUN Hongchi;MU Rongjun;DU Huajun;CUI Naigang(School of Astronautics,Harbin Institute of Technology,Harbin 150001,China;Beijing Aerospace Automatic Control Institute,Beijing 100854,China;National Key Laboratory of Science and Technology on Aerospace Intelligence Control,Beijing 100854,China)

机构地区：[1]哈尔滨工业大学航天学院,哈尔滨150001 [2]北京航天自动控制研究所,北京100854 [3]宇航智能控制技术国家级重点实验室,北京100854

出　　处：《中国惯性技术学报》2019年第4期486-491,共6页Journal of Chinese Inertial Technology

基　　金：国家高技术研究发展计划（863计划）（2015AA7026083）;载人航天四批预研项目（060201）

摘　　要：针对高动态环境下惯性/天文组合导航精度下降的问题,提出一种基于神经网络辅助的惯性/天文组合导航方法。首先以组合导航滤波估计过程中的增益矩阵和动态环境下的惯性器件量测信息构建特征向量;然后,采用导航估计误差对BP神经网络进行训练;最后,利用BP神经网络的输出结果辅助修正组合导航系统。计算机仿真验证结果表明,相较于传统方法,基于BP神经网络辅助的惯性/天文组合导航系统的姿态估计精度可提高30%以上,在动态环境下姿态精度可以保持在5″(1σ)以内。所提出的方法对提高动态环境下惯性/天文组合导航系统的精度和适应能力具有一定的参考价值。Aiming at the problem that the precision of inertial/celestial navigation(ICN)is decreased in high dynamic environment,an ICN method based on neural network is proposed.First,the eigenvectors of BP neural network are constructed by the gain matrix in the ICN filtering estimation process and the angular velocity in dynamic environment.Then,the BP neural network is trained by navigation estimation errors.Finally,the filtering results of the ICN system is corrected by the output of BP neural network.Simulation results show that,by the proposed method,the attitude estimation accuracy of the ICN system can be improved by more than 30%compared with the traditional method,and the attitude accuracy can be maintained within 52(1s)in high dynamic environment,which has important reference value for improving the accuracy and adaptability of the ICN in dynamic environment.

关 键 词：天文导航 BP神经网络 惯性导航 组合导航 

分 类 号：V249.322[航空宇航科学与技术—飞行器设计] U666.11[交通运输工程—船舶及航道工程]