基于神经网络的GPS深度融合弹载惯导模型

GPS deep fusion missile-borne inertial navigation model based on neural network

针对传统弹载惯性导航系统误差修正采用线性化模型,无法对非线性误差进行修正的缺点,提出一种基于神经网络的GPS深度融合弹载惯导模型。构建IMU六路脉冲数到发射惯性坐标系下视速度增量的非线性映射关系,利用GPS信息对网络误差模型进行训练,为GPS信号失效后的长航时惯性导航系统提供更高精度的误差修正模型。仿真实验结果表明,相较于纯惯性导航系统,所提模型能够在GPS信号失效后抑制惯导系统误差发散,速度精度提高76.56%,位置精度提高91.61%。

Aiming at the shortcomings that the traditional error correction model of missile-borne inertial navigation system adopts the linear model,which cannot correct the nonlinear errors,a GPS deep fusion missile-borne inertial navigation model based on neural network is proposed.The nonlinear mapping relationship between IMU six channel pulse data and the apparent velocity increment in the launch inertial coordinate system is constructed.The network error model is trained by GPS information,so as to provide a more accurate error correction model for the long-endurance inertial navigation system after GPS signal failure.The simulation results show that compared with pure inertial navigation system,the proposed model can suppress the error divergence of inertial navigation system after GPS signal failure,and the speed accuracy and position accuracy can be improved by 76.56%and 91.61%.