Elman网络辅助SINS/GNSS组合导航方法研究

Elman Network Aided SINS/GNSS Integrated Navigation

针对SINS/GNSS组合导航系统在GNSS信号失锁情况下定位误差迅速增大问题。提出一种基于Elman神经网络和自适应卡尔曼滤波的SINS/GNSS伪松组合导航方法。在GNSS信号可用时,利用SINS,GNSS数据训练Elman神经网络;当GNSS失锁时,利用训练好的网络模型预测GNSS观测值,保证组合导航的连续性。实验表明:GNSS信号失锁约3 min,该方法水平定位精度相较于纯惯导提高了64%,有效提高了纯惯导的定位精度。

In the SINS/GNSS integrated navigation system,the position error of pure inertial navigation will increase rapidly during GNSS outages.A method based on Elman neural network and adaptive Kalman filter to improve SINS/GNSS pseudo-loose integrated navigation is proposed.When GNSS signal is available,Elman neural network is trained by SINS and GNSS data.When the GNSS outage occurs,the trained network model is used to predict the GNSS observations to ensure the continuity of integrated navigation.Experiment showed that the horizontal position accuracy of the proposed method was 64%higher than that of the pure inertial navigation system when the GNSS signal was lost for about 3 min.The position accuracy of the pure inertial navigation system is effectively improved.