DGPS/INS integration using neural network methodology

This paper presents an INS/DGPS land vehicle navigation system using a neural network methodology. The network setup is developed based on a mathematical model to avoid excessive training. The proposed method uses a KF-based backpropagation training rule, which achieves the optimal training criterion. The North and East travel distances are used as desired targets to train the two decoupled neural networks. The proposed method is suitable for INS and DGPS systems sampled at different rates. In addition, an online stochastic modeling method for the desired target is developed. This method facilitates the use of the extended Kalman filter trained backpropagation neural network approach whenever the desired target statistics are not available, or not reliable. The experimental results demonstrate the suitability of this method in developing an INS/DGPS land vehicle navigation method