摘要： 利用迭代最近等值点(ICCP)算法对重力图上的航迹进行匹配可以减小惯性导航系统误差，但计算量大。针对上述问题，通过修改激励函数并增加假饱和预防函数，提出一种改进的反向传播神经网络学习算法。仿真结果表明，该算法可以加快搜索最近等值点的速度，更好地满足重力辅助导航对匹配精度及匹配速度的要求。

关键词: 反向传播神经网络, 迭代最近等值点算法, 航迹匹配, 惯性导航, 梯度下降

Abstract: It can reduce the error of the inertial navigation system when using Iterative Closest Contour Point(ICCP) algorithm for track matching on the gravity map, however it brings large computation costs. By modifying the activation function and false-saturating prevention function, this paper presents an improved Back Propagation(BP) neural network learning method to search for the closest contour points. Simulation results show that the algorithm improves the searching speed of the closest point and meets the matching speed and accuracy demand of gravity-aided navigation.

Key words: Back Propagation(BP) neural network, Iterative Closest Contour Point(ICCP) algorithm, track matching, inertial navigation, gradient descent

中图分类号: 

• TP18