神经网络辅助的GPS/INS组合导航滤波算法研究  

Study on GPS/INS integrated navigation filtering algorithm asststed by neural network

作　　者：陈鑫鑫 张复春[1] 郝雁中[1] 

机构地区：[1]中国人民解放军空军航空大学航空理论系,吉林长春130022

出　　处：《电子技术应用》2015年第5期84-87,共4页Application of Electronic Technique

摘　　要：在高空高速条件下,GPS信号失锁致使常规的卡尔曼滤波器发散,从而导致组合导航系统精度严重下降。以BP神经网络辅助技术手段对GPS/INS组合导航滤波算法实施精度补偿,即在GPS信号锁定时,对神经网络进行实时在线训练,而当在GPS信号失锁时,利用之前训练好的神经网络进行组合导航滤波,以解决精度严重下降问题。算法采用多神经网络并行结构,以减少神经网络在训练过程中的交叉耦合,提高训练速度。通过MATLAB仿真,验证了算法的可靠性与可行性,并证明其在GPS信号丢失时,精度较纯惯性导航系统有较大提高。In the conditions of high attitude and high speed, conventional Kalman filtering of GPS/INS integrated navigation systems is divergent because of the GPS outages. It leads to the accuracy of navigation system a serious decline. In order to solve the problem, an algorithm of BP neural network aided Kalman filtering information fusion is proposed. The BP is trained while GPS signals are available and output is predicted during the GPS outages. In the thesis, we take the parallel structure of multi neural network to reduce cross-coupling and improve the training speed. The test results indicate that the proposed methods can ef- ficiently compensate for GPS updates during short outages. The simulation results show the effectiveness of the method.

关 键 词：卡尔曼滤波 组合导航 BP神经网络 

分 类 号：V421.6[航空宇航科学与技术—飞行器设计] U666.1[交通运输工程—船舶及航道工程]