A Trajectory Prediction Method for High-Speed and High-Maneuverability Glide Vehicle Based on Mid-Terminal Guidance Handover Point Identification

Abstract

Addressing the challenges of unclear mission scenario definition and insufficient intent prior information utilization in the trajectory prediction process for high-speed and high-maneuverability glide vehicle, a trajectory prediction method for high-speed and high-maneuverability glide vehicle based on mid-terminal guidance handover point identification is proposed. Firstly, a mission scenario involving the gliding of high-speed and high-maneuverability vehicle towards multiple typical guidance handover points is constructed. A quasi-equilibrium glide guidance method is employed to generate trajectory datasets. Secondly, a guidance handover point recognition method based on long short-term memory network is proposed, utilizing tracking data to construct feature sequences for preliminary classification of glide trajectory. Finally, the self-attention mechanism is introduced to improve the feature extraction performance of sequence-to-sequence prediction networks, and encoder-decoder method is employd to predict the classified glide trajectory in the long term. Simulation results show that the trajectory prediction method based on mid-terminal guidance handover point identification exhibits high accuracy. For prediction times of 120 s, 180 s, and 240 s, the trajectory errors remain within 18.77 km, 36.91 km, and 57.75 km, respectively. Compared to directly utilizing a deep learning mo-del for prediction, the proposed prediction method demonstrates a reduction of 37.61% in average prediction error and 37.34% in maximum prediction error within a prediction time of 240 s.

Keywords

• |high-speed and high-maneuverability vehicle|long short-term memory network|mid-terminal gui-dance handover point identification|trajectory prediction