Research on the Trajectory Prediction Method Basing on Partially Observable Driving Scenario Data

Trajectory prediction is a critical component of automated driving systems within the broader context of intelligent transportation. This paper proposes a vehicle trajectory prediction model that operates on partially observable information. By utilizing intersection node data, the model facilitates vehicle re-identification and estimates driving states. It incorporates temporal state sequences of vehicle targets to visualize driving trajectories, thereby contributing to the development of a comprehensive vehicle perception system for digital road networks. The model demonstrates robust stability and reliability, effectively filtering noise during data acquisition and enhancing prediction accuracy. Using real-world road section data, the model predicts blind-spot trajectories within simulation software, with results showing minimal deviation from field-collected data.