基于BP人工神经网络的GPS／SINS组合导航算法

基于扩展Kalman滤波的GPS／SINS组合导航算法，需要对原始的非线性连续系统模型进行线性化和离散化处理，要求系统噪声和测量噪声为零均值的高斯白噪声，且易于出现滤波器发散。BP人工神经网络毋需对所求解的问题建模，能够很好地逼近系统非线性特性，获得较高精度的导航定位信息；还具有计算过程稳定，不涉及矩阵求逆，不需要迭代逼近，以及容易实现并行处理等优点。本文设计适用于GPS／SINS组合导航系统的BP网络模型，并在标准的BP算法基础上，采用共轭梯度法改进网络训练速度及精度。最后，通过仿真算例说明BP网络方法用于GPS／SINS组合导航计算的可行性。

Integrated GPS/SINS Navigation Algorithm Based on the Artificial Neural Network BP

The integrated GPS/SINS navigation algorithm based on extended Kalman filtering requires that the original non-linear and continuous system models have to be made linearization and discretization,and that the system noises and measurement noises have to be the white ones with zero-mean.Thus the filter is liable to get divergence in the practice.The artificial neural network BP needn't establish the system model to solve problem,can approximate well to the non-linear system features and obtain high-accurate navigation and positioning solutions.Moreover,the BP network still has a lot of advantages.For instance,its competing procedure is very stable,it isn't concerned to the inverse matrix solution and the iterative approximation calculation,it is easy to make parallel processing for the BP network,and so on.Therefore,the BP network structure applied to the integrated GPS/SINS navigation system will be designed in this paper.On the basis of the standard BP algorithm,meanwhile,the conjugate gradient algorithm is used to improve the training velocity and accuracy of BP network.In conclusion, the feasibility of the BP network method applied to the integrated GPS/SINS navigation calculation can be illustrated from a simulating example.