A Motion Planning Framework with Integrating Transformer-based Trajectory Prediction on Highway

In the realm of autonomous vehicles (AVs), ensuring safe and effective motion planning is a paramount and challenging task, particularly in complex and dynamic scenarios. To enhance driving safety, a low-risk, high-efficiency motion planning framework is established for AVs based on surrounding vehicles (SVs) trajectory prediction in highway scenarios. Firstly, a Transformer-based predictive model is constructed for trajectory prediction of SVs. Subsequently, a driving risk assessment method is designed based on the trajectory prediction results. Furthermore, a motion planning framework utilizing polynomial curve technique is proposed while considering driving risk, efficiency and comfort. Finally, the predictive model is trained and tested by collecting the highway dataset on the Crala simulation experiment platform, and the effectiveness of our constructed motion planning framework is verified through two lane-changing events. The simulation results affirm that our proposed method achieves both safe and efficient motion planning for AVs.