Research on flight trajectory prediction method based on transformer

As an essential component to realize the concept of Trajectory-based Operation, accurate trajectory prediction plays a crucial role in enabling the Air Traffic Management system to anticipate potential hazards and ensure safe operations. However, the trajectory prediction task faces significant challenges due to the multi-dimensional characteristics of trajectory data and the susceptibility of flight performance to external factors, which result in high uncertainty. To address these challenges, this paper proposes a Transformer-based trajectory prediction model that leverages the attention mechanism to identify key factors, enhancing its ability to extract diverse information from the data. The model is thoroughly evaluated using real trajectory datasets, and the simulation results demonstrate its effectiveness in predicting four-dimensional trajectories. Notably, compared to improved attention-based models and single recurrent neural network algorithms, the proposed model demonstrates higher prediction accuracy and superior performance in parallel sequential data processing. This lays a robust foundation for subsequent conflict detection and decision-making processes.