Abstract: We put forward a GPS/SINS integrated navigation algorithm based on neural network(NN) prediction.When GPS signal is available,this method uses the outputs of inertial sensors and the outputs of Kalman filter as the inputs and ideal outputs of NN respectively,and the NN is trained on-line.During GPS signal outage,the trained NN is used to predict the navigation parameter error to correct SINS.The ground static and dynamic car experimental results show the reliability and effectiveness of this method.