State Estimation of ALV Integrated Navigation System Based on BP Neural Network

In this paper, a data fusion method based on BP neural network and DR is proposed for ALV´s navigation. Mathematical models have been established for data fusion. When GPS is valid, neural network is adopted, which is three layered network with 5 input/ 3 output neurons with a single hidden layer. When GPS is invalid, DR is introduced by using outputs of IMU and milometer and initial information from BP network. Training and simulation are made with this BP network using the BP network program compiled with MATLAB, and experimental results indicate that the accuracy of integrated system has been improved in comparison with separate GPS and DR (navigation accuracy of DR has not been shown in this paper).