A new method for compensating the errors of integrated navigation systems using artificial neural networks

Abstract

The integrated navigation system consists of Inertial Navigation System (INS) and receiver of Global Navigation Satellite System (GNSS). The combined system provides continuous and accurate navigation solution when compared to standalone INS or GNSS. However, the accuracy of navigation solution of the integrated system degrades during GNSS outages. Aiming to improve the position and velocity precision of the INS/GNSS system during GNSS outages, a novel method (namely UKF + NARX) that combines Unscented Kalman Filter (UKF) and nonlinear autoregressive neural network with external inputs (NARX) is proposed. The NARX networks are used to predict position and velocity errors during GNSS outages. The selection of inputs of NARX networks is performed using mutual information (MI) criterion and lag-space estimation (LSE). The performance of the proposed method is experimentally verified using real dataset acquired in a land-vehicle navigation test. Results show that the proposed method outperformed UKF and other methods that use different inputs for neural networks.