RBF神经网络在组合导航系统中的应用

Application of RBF Neural Network in Integrated Navigation System

针对Kalman滤波器自适应能力弱和组合导航系统存在外部干扰以及模型误差的问题。面向组合导航领域,该文提出了径向基函数神经网络(RBF)辅助联邦Kalman滤波的信息融合方案;结合神经网络和Kalman滤波2种方法共同提高系统的自适应能力,可以有效地消除外部的干扰和模型的不确定对系统的影响,使SINS/GPS/BDS组合导航系统有强大的鲁棒性。通过对仿真计算结果的分析,对比联邦Kalman和误差方向传播神经网络(BP)的信息融合方案,证明该方案能够有效地提高系统的导航精度和鲁棒性。

As for insufficient adaptive capability for Kalman filter and external interference and model error in integrated navigation system,an information fusion scheme is proposed by using radial basis function neural Network（RBF）-assisted federal Kalman filtering in the field of integrated navigation.Combining neural network and Kalman filtering methods together to improve the system’s self-adaptation capability,external interference can be effectively eliminated. Model uncertainty influence on system. Simulation results show that this proposed scheme can effectively improve system navigation accuracy and robustness in comparison with the information fusion scheme of federal Kalman and error direction propagation neural network（BP）.