On terrain-aided navigation for unmanned aerial vehicle using B-spline neural network and extended Kalman filter

Abstract

Terrain-aided navigation technology estimates position information based on the terrain elevation data, and corrects the inertial navigation system (INS) error. A terrain matching algorithm based on B-spline neural network and extended Kalman filter (EKF) is proposed for unmanned aerial vehicle (UAV). In order to improve the accuracy of traditional terrain linearization method, B-spline neural network is applied to fit terrain data. Terrain-aided navigation (TAN) system often need to preload digital elevation map (DEM), so offline training of the neural network using the actual terrain data is practical. The neural network calculates the continuous terrain elevation and terrain gradient. Then these data are used in EKF. The simulation results show that the B-spline neural network can calculate the high-accuracy linearized terrain data on the DEM, and the performance of TAN system is better by using EKF combined B-spline neural network method.