Position improvement of GPS/INS using NAR neural network and robust Kalman filter during GPS outage

Precise guidance and navigation is one of the necessities of every moving vehicle in the transportation industry. Different methods of navigation has been used to determine exact location of the vehicle in each moment. Inertial navigation is a newton-based method that provides position of the vehicle regardless of any external communication equipment. Inertial navigation is always subject to different disturbing errors that consistently reduce the performance of the system, therefore, for long-term navigation purposes, there should be at least one navigation assisting system to maintain positioning accuracy. Consequently, a GPS/INS data fusion using a Robust Extended Kalman Filter (REKF) is investigated in this paper. When vehicles enter an area with a signal jammer, GPS position would be unavailable, and, filter observations will not be updated. Thus, a trained nonlinear neural network is used to predict position in this scenario. In order to test the algorithm in real-world circumstances, a custom designed board with military standards is employed. The results show about 70% of position improvement towards each axis. The proposed algorithm has improved the position accuracy in GPS/INS integrated system in defined scenario.