Attention based feature extraction and fusion for autonomous vehicle trajectory prediction

Vehicle trajectory prediction is a key technology in advancing autonomous driving and intelligent transportation systems, which is of great significance to improve traffic safety and efficiency. This paper addresses the challenges posed by complex traffic scenarios and vehicle interactions by proposing ABFEF, a precise and dependable vehicle trajectory prediction model. The feature extraction network captures multi-scale interactions between vehicles and between vehicles and the surrounding environment, while the feature fusion network further integrates global semantic information to enhance prediction accuracy and robustness. We evaluate ABFEF on Argoverse 1, a benchmark in motion forecasting. Our experiments demonstrate that ABFEF outperforms other models in terms of prediction accuracy and reliability. Experimental results validate the effectiveness of our approach in improving autonomous vehicle trajectory prediction under complex traffic conditions.