A Transformer-based Trajectory Prediction Method for a Near-space Flight Vehicle

The accurate prediction of trajectories for near-space flight vehicles is crucial for effective defense strategies. A transformer-based method aimed at enhancing the precision of trajectory prediction is proposed in this paper. This method models the maneuvering behaviors of the target and incorporates the flight mechanics of the target. Initially, a behavior recognition network based on LSTM is designed to identify target maneuvers using measurement data. Subsequently, a behavior prediction network based on Transformer is formulated to predict future behaviors based on recognized behaviors. Finally, a trajectory prediction algorithm is developed to integrate multiple models for trajectory prediction. Simulation experiments demonstrate the notable prediction accuracy achieved by the proposed method.