A Novel Elman Network Based INS/GPS Fusion Filter to Enhance Tracking Accuracy in UAVs

The development of a highly accurate position tracking technique for extremely nonlinear dynamic systems such as unmanned aerial vehicles (UAV) using integrated global positioning system (GPS) and inertial navigation system (INS) is a challenging problem. During GPS outages, the existing systems experience considerable errors. In this paper, we propose a novel fusion algorithm based on an extended Kalman filter (EKF) - Elman neural network (ENN) that is capable of enhancing the tracking accuracy during GPS outages. The proposed technique relates the INS outputs from the sensors to the GPS position increment. ENN corrects the system during GPS outages by predicting the pseudo GPS position. The time information is also considered to obtain precise estimates and also to reduce the computational complexity. Simulation studies are performed on a UAV trajectory using low-cost MEMS-INS sensors to test the robustness of the algorithm under extreme varieties. The proposed system shows a reduction in root mean square error (RMSE) of the estimated position compared to the existing back propagation neural network (BPNN) model.