Intelligent Prediction Method for Updraft of UAV That Is Based on LSTM Network

Abstract

Updrafts widely exist in the nature and with the help of updrafts, an unmanned aerial vehicle (UAV) may improve its flight time performance. Some methods have been proposed for updraft prediction, such as the extended Kalman filter (EKF) or unscented Kalman filter (UKF). In this article, a prediction method based on the long short-term memory (LSTM) network is proposed. First, a flight simulation system is developed, which is used to generate the training data for the LSTM network. Then, an LSTM network is developed and trained for updraft prediction. Finally, some experiments are made to compare the LSTM network with traditional methods that are based on EKF and UKF. The results show that the LSTM network has a substantial advantage in terms of the prediction accuracy and convergence rate.