Ego vehicle trajectory prediction based on driver-vehicle coupling characteristics in diverse urban scenarios

This software package supports the reproduction of key numerical results reported in the study, including scripts that use driver-vehicle coupling characteristics as input to predict the ego-vehicle’s trajectory in diverse urban scenarios, as well as script for evaluating the performance of the predicted trajectories. The package includes processed data & normalization parameters, ego vehicle trajectory prediction deep learning training models (DriVETP, DriVETP-Ablation1, DriVETP-Ablation2, AM-LSTM, Transformer), visulaization results from multi-scenario testing, performance metrics (RMSE, ADE, FDE) radar plots, and model generalization validation.  Test scenarios cover typical urban driving conditions: left lane changing, right lane changing, roundabout navigation, left turning, right turning, and straight driving. Detailed instructions for reproducing the main results are provided in the README file.