LSTM Intelligent Trajectory Prediction for Hypersonic Vehicles Based on Attention Mechanism

The near-space hypersonic vehicle has a non-inertial trajectory form and large-scale and strong maneuvering penetration capabilities. The accurate prediction of the target's flight trajectory can provide strong technical support for the effective interception of the missile interception system. To address the problem of glide and skiptrajectory prediction of hypersonic aircrafts, this paper proposes a Seq2Seq trajectory prediction model based on attention mechanism, which employsthe LSTM network to design the encoder and decoder, and uses the information extracted by attention mechanism to performdecodingand prediction. The network takes the six-dimensional feature sequence of the target trajectory's position, velocity, trajectory inclination angle and attack angle as the input network, and the continuous trajectory sequence during a certain period in the futureis the network output. The trajectory data of the target aircraft obtained by the trajectory simulation model is used as the training setto train and optimize the network. The experimental results show that the proposed network can effectively predict the various flight trajectories of hypersonic aircrafts with small prediction errors, which can provide some insights into the missile interception system. © 2022 China Ordnance Society. All rights reserved.