A Robust Fully Sampled High-Resolution Vehicle Trajectory Reconstruction Method Fusing Aggregate Traffic State and Microscopic Vehicle Behavior

Reconstructing high-resolution fully sampled vehicle trajectories is very challenging due to the limited deployment of fixed and moving sensors. Existing studies are mostly based on microscopic car-following models that require high-resolution (HR) trajectory data with a certain penetration rate, making their methods vulnerable in low-penetration data environments. This study proposes a data-robust and scenario-flexible method for high-resolution trajectory reconstruction (TR), in which the macroscopic traffic state and microscopic vehicle behavior are both exploited. First, the macroscopic module recovers high-resolution spatiotemporal traffic velocity, providing constraints for subsequent trajectory reconstruction. Then, the microscopic module generates candidate vehicle trajectories, which are further fused to reconstruct high-resolution trajectories through an error-resistance method. The proposed method significantly improves the flexibility and robustness of high-resolution trajectory reconstruction on freeways and achieves a 57.54% improvement (in MAPE) even under an extremely low penetration rate of probe vehicles.