Trajectory estimation algorithm for unmanned aerial vehicle based on LSTM-EKF
Aiming at the problem that the radar signal detection is too weak and greatly influenced by noise to conduct accurate tracking of the unmanned aerial vehicle(UAV),the paper proposed the combinatorial algorithm of long short-term memory(LSTM)neural networks and extended Kalman filter(EKF):EKF was integrated with radar data to predict the target,and the location and the speed of the target UAV were obtained;then,the shortcomings of single use of EKF algorithm were analyzed,LSTM networks in combination with filter location,speed and filter error were applied as inputs to carry out the training and learning,and the prediction error was gained as outputs to make up for EKF;finally,the frequently-used EKF algorithm was used to perform a comparative verification through the generated constant velocity and constant acceleration tracks.Simulational result showed that compared with the traditional EKF algorithm,the proposed LSTM-EKF algorithm would have higher tracking accuracy,which could control the tracking error within 10 m,and with better anti-noise ability.
NETL
NSTL
万方数据
