Abstract:Accurate flight trajectory prediction is essential for trajectory-based air traffic management (TBO). To address the inherent complexity of the trajectory prediction problem and the limitations in current prediction accuracy, this study proposes an enhanced Transformer-based prediction method incorporating a Spatial Attention Mechanism (SAM). The encoder integrates the SAM structure to enable joint modeling of spatiotemporal features within flight trajectory data. The flexible design of the encoder allows it to capture global dependencies while simultaneously enhancing local pattern awareness. This improved architecture significantly enhances the model’s ability to learn trajectory-related features, thereby improving the overall prediction accuracy. Experimental results demonstrate that the proposed method achieves superior performance in both longitude and latitude predictions, indicating that the integration of the spatial attention mechanism effectively enhances the precision of flight trajectory forecasting.
