CPS-Based Vehicle Trajectory Prediction in Highway Interchanges under Mixed Traffic Conditions

With the rapid advancement of intelligent connected vehicle technology, ensuring safe and efficient travel in highway weaving areas has become a significant research topic. Traditional trajectory prediction methods inadequately consider real-time traffic situation changes and lack generalizability. Consequently, this study introduces a trajectory prediction method for mixed vehicles in highway weaving areas based on Cyber-Physical Systems (CPS). Moreover, an enhanced Transformer model is employed to model the dynamic spatial interactions between vehicles. The well-trained model is deployed in the cyber layer, and by training with data from the physical layer, it facilitates real-time prediction of manually driven vehicle trajectories. Experimental results demonstrate that compared to prevalent approaches, our method offers more accurate and efficient trajectory prediction in complex weaving area scenarios.