第22卷第5期 2014年5月
文章编号
1004-924X(2014)05-1304-08
光学精密工程
Optics and Precision Engineering
Vol.22No.5
May.2014
神经网络辅助卡尔曼滤波在组合导航中的应用
崔留争1，2，高思远，贾宏光1*，储海荣1，姜瑞凯1
（1.中国科学院长春光学精密机械与物理研究所，吉林长春130033；
2.中国科学院大学，北京100039）
摘要：为使基于微机电系统的捷联惯性导航/全球定位（MEMS-SINS/GPS)组合导航系统在GPS接收机无法正常工作时，仍能提供满足精度要求的导航信息，提出了径向基函数押经网络(RBFNN)辅助自适应卡尔曼滤波（AKF)的信息融合方法。首先，基于该方法设计了由神经网络训练与预测两种模式构成的组合导航系统。在GPS可用时，对RBFNN进行在线训练；在GPS失锁时，由RBFNN预测AKF更新过程的量测输人。然后，建立了RBFNN与AKF的数学模型，并设计了RBFNN的训练策略与AKF的自适应算法。最后，通过跑车实验验证了该信息融合方法的有效性。实验结果表明，在GPS断开时间为40s和100s时，系统的位置精度分别优于15m和90m。该信息融合方法能在GPS失锁时对导
航误差发散进行有效阻尼，是适用于小型无人机、制导炸弹与车辆的一种低成本、高鲁棒性、中等精度的导航方案。关键词：径向基函数神经网络；自适应卡尔曼滤波；信息融合；微电子机械系统；组合平航
中图分类号：V421.6;U666.11
文献标识码：A
doi;10. 3788/OPE. 20142205. 1304
Applicationof neuralnetworkaidedKalmanfilteringtoSINS/GPS
CUI Liu-zheng',2, GAO Si-yuan', JIA Hong-guang'', CHU Hai-rong',JIANG Rui-kai
(1.Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Sciences
Changchun130033，China;
2.University of ChineseAcademy of Sciences,Beijing1300039,China)
Correspondingauthor,E-mail:jiahg@ciomp.ac.cn
Abstract:To allow Micro-electro-mechanical System(MEMS)-based SINS/GPS integrated navigation systems to meet the accuracy requirements during GPS outages, a Radial Basis Function Neural Net work (RBFNN) aided Adaptive Kalman Filtering (AKF) information fusion method was proposed. Firstly,the system structure consisting of dual modes of RBFNN training and prediction was de-signed, The RBFNN was trained while GPS signals were available and the inputs for AKF measure-ment updates were predicted during the GPS outages. Then, the mathematic models for RBFNN and AKF were built and the training strategy for RBFNN and the adaptive algorithm for AKF were de-signed. Finally, the performance of the proposed information fusion method was validated using real field test data. Test and experiment results show that the position precisions are better than 15 m and 90 m during GPS outages at 40 s and 100 s,respectively. The proposed information fusion method can effectively damp the divergence of the navigation error during GPS outages and can provide a low-cost,
收稿日期：2013-02-04；修订日期：2013-04-01.
基金项目：中国科学院知识创新工程国防科技创新重要方向项目（No.YYYJ-1122)；中国科学院三期知识创新工程
资助项目