针对SINS/GPS/MCP组合导航系统初始对准中GPS失锁时系统精度下降的问题,引入基于神经网络和小波技术的解决方案,将相关特征量经小波去噪后作为神经网络的训练样本.基于该方案建立了系统滤波模型,包括捷联惯性系统失准角、速度和位置误差方程以及速度航向匹配的状态方程和量测方程.为了验证该方案的有效性,分别对GPS失锁、加入神经网络修正和引入小波去噪技术3种情况进行了仿真,结果发现：神经网络修正可解决速度精度变差的问题;经小波去噪后对神经网络重新训练,各项指标都有所提高.可见,基于小波的神经网络方法可提高神经网络逼近模型的程度,进而提高对准精度.

In order to avoid the system accuracy degradation in the initial alignment due to the GPS outage in SINS/GPS/MCP integrated navigation system,a scheme based on neural network and wavelet is proposed,in which the relevant characteristic components denoised by wavelet are treated as the training samples of neural network.Then,a filtering model based the scheme is established,which consists of the equations of misalignment angle,velocity error and position error in SINS system,as well as the state and measurement equations of Kalman filter with velocity and heading matching.Moreover,in order to verify the proposed scheme,simulations are carried out in the conditions of GPS outage,neural network introduction and wavelet denoising.The results indicate that the introduction of neural network avoids the velocity accuracy degradation,and that the retraining of neural network using wavelet denoising effectively improves all the indexes of SINS/GPS/MCP system.It is thus concluded that,as the adoption of wavelet denoising can improve the approximation of neural network to the actual model,the alignment accuracy is further improved.