基于神经网络的INS/GPS组合导航误差补偿研究

Research on Error Compensation for INS/GPS Integrated Navigation Systems Based on Neural Networks

针对飞行条件下INS/GPS组合导航系统在GPS失锁时解算精度下降甚至发散的问题,采用RBFNN(径向基神经网络)辅助组合提高导航解算精度。该方法在GPS信号有效时对神经网络进行训练,在GPS失锁时利用神经网络对导航系统的误差进行预测并修正,实现组合的不间断进行。飞行试验数据仿真表明,该方法能够在一定程度上抑制惯性解算的发散。采用神经网络的辅助组合导航方法可以为GPS失锁时的导航事后误差补偿提供一种相对有效的途径。

To compensate navigation error accumulations of INS (inertial navigation system) when GPS (global positioning system) receiver loses signal in INS/GPS integrated navigation system, a RBFNN (Radical Base Function Neural Network) was applied to aid integrated navigation algorithm. This method trained the RBFNN when GPS was valid, and ran the network to estimate and then correct navigation errors when GPS failed. This algorithm was tested by a set of experimental data from a flight test, and simulation results showed that it was somewhat liable and valid to prevent inertial navigation from diverging. Over all, this RBFNN-aided navigation algorithm is a relatively effective way to compensate navigation errors of INS/GPS integrated navigation systems offline when GPS fails.