Trajectory Prediction Model of Electric Vehicle Autonomous Driving Based on Hybrid Attention Transformer Network

Current electric vehicle trajectory prediction fails to fully consider the interaction between the target vehicle and other vehicles, resulting in poor prediction results. In order to solve this problem, this paper proposes a hybrid attention transformer network (HATN), which is designed for more accurate trajectory prediction. Firstly, based on the transformer network, this paper introduces a self-attention mechanism and a cross attention mechanism, and proposes a feature embedding and position encoding module as well as an interactive feature extraction module, so as to achieve accurate modelling of vehicle state information. With this approach, the interactive information between traffic participants can be fully extracted by effectively utilizing the map information. Secondly, a trajectory prediction decoder is proposed to expand the solution space of the model and enhance its ability to understand the real driving rules based on the driving intention recognization results of the surrounding vehicles, so that the prediction results can be more reasonable with stronger robustness. Thirdly, according to the experiments and analysis conducted based on the large-scale open datasets BDD100K and Waymo, the results show that the proposed model has a significant improvement in prediction accuracy compared with the comparison models, which verifies the effectiveness of the proposed model.