Integrating Extreme Learning Machine with Kalman Filter to Bridge GPS Outages

Abstract

Nowadays, the low-cost SINS/GPS integrated navigation system has been widely used. Generally, the integrated system works well in providing reliable navigation information. However, the GPS signal may be lost easily and the navigation accuracy will deteriorate badly without compensation. In order to overcome the limitation, a hybrid prediction method that combines the neural network and extended Kalman filter (EKF) is proposed. The neural network is trained when GPS signal is available and then it is used to forecast the measurement of EKF during GPS outages. In recent years, extreme learning machine (ELM) has attracted much attention and interest among scientific community for its extremely fast learning speed and superior generalization performance. Thus in this paper, the ELM is adopted and compared with Radial Basis Function neural network (RBFNN). The simulation result shows that the accuracy of the integrated navigation system is significantly improved by applying the proposed method and the ELM performs better in real-time capacity and generalization ability.