论文部分内容阅读
针对传统高斯 RBF网络应用于惯性导航初始对准建模时 ,对处于两基函数中心点之间的值拟合效果不太理想的情况 ,提出了一种将增量余弦 RBF网络用于惯性导航初始对准建模的方法 .该方法采用增量余弦函数作为 RBF网络的基函数 ,对多变量非线性系统有很好的拟合能力 .相对于传统高斯 RBF网络 ,增量余弦 RBF网络的基函数具有更强的局域性 ,解算时同时参与运算的基函数数量更少 ,有效地降低了网络的解算时间 .仿真结果表明 ,增量余弦 RBF网络用于惯性导航的初始对准 ,既可获得较高的对准精度 ,又有效地降低了系统的解算时间 ,提高了系统的实时性 .
Aiming at the problem that the traditional Gaussian RBF neural network is used in the initial alignment modeling of inertial navigation, the fitting effect of the value between the center points of two basis functions is not ideal, a new method is proposed to use incremental cosine RBF neural network for inertial navigation Initial alignment modeling method using incremental cosine function as the basis function of RBF network for multivariate nonlinear systems have good fitting ability.Compared with the traditional Gaussian RBF network, incremental cosine RBF network basis The function is more localized and the number of basis functions involved in the solution is less, which effectively reduces the solution time of the network.The simulation results show that incremental cosine RBF neural network is used for initial alignment of inertial navigation, It not only achieves high alignment accuracy but also effectively reduces the solution time of the system and improves the real-time performance of the system.