Hybrid Data-Model Driven Trajectory Prediction On Highways: Integrating Anticipatory Interaction Awareness and Personalized Driving Preferences

Strategic Insights -

New research has been conducted on the intersection of connected autonomous vehicles (CAVs) and human-driven vehicles (HDVs) on highways, emphasizing the importance of anticipating interactions between the two types of vehicles. The study, led by researchers from Zhejiang University, aimed to address the dual challenges of personalized driving preference modeling and anticipatory interaction awareness. To this end, a hybrid data-model driven framework was proposed, integrating physics-based behavioral calibration with data-driven interaction modeling.

Key Takeaways:

  • The proposed framework, Kepler Optimization-based Temporal Attention Fusion Transformer Network (KOTAFTN), combines physics-based behavioral calibration with data-driven interaction modeling to capture dynamic historical interactions, anticipatory interactions, and static driving preferences.
  • The framework's driving preference extraction module derives individualized behavioral traits using Kepler-based physical modeling, which are then encoded into context vectors via a static encoder and static enhancement layers.
  • A dynamic encoder and temporal attention fusion module jointly capture and fuse historical and anticipatory interactions by modeling temporal dependencies.
  • A multimodal trajectory prediction module generates diverse candidate trajectories reflecting potential future motion patterns of HDVs.
  • Experiments demonstrated that the proposed framework consistently outperformed benchmark methods in mixed traffic environments, particularly under complex and congested scenarios.
  • The study's findings underscore the framework's potential to improve interaction safety and trajectory accuracy during the transitional evolution of mixed traffic systems.

Statistics:

  • The study was financially supported by the National Natural Science Foundation of China (NSFC) and the Natural Science Foundation of Zhejiang Province.
  • The research was conducted by a team of researchers from Zhejiang University, including Sheng Jin, Congcong Bai, Xi Gao, Wentong Guo, Donglei Rong, Chengcheng Yang, and Mengdi Chen.
  • The study was peer-reviewed prior to publication.
  • The research was published in the journal Transportation Research Part C-emerging Technologies in 2025, Volume 180.

Sources:

  • National Natural Science Foundation of China (NSFC)
  • Natural Science Foundation of Zhejiang Province
  • Zhejiang University
  • Transportation Research Part C-emerging Technologies (Pergamon-elsevier Science Ltd, The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, England)
  • Information Technology Newsweekly (p. 359, November 4, 2025)