<ul><li>A new deep learning framework is introduced to model intra-driver heterogeneity in car-following dynamics.</li><li>The framework incorporates discrete driving regimes and combines Gated Recurrent Units with Long Short-Term Memory networks.</li><li>Driving regimes are identified using a segmentation algorithm and Dynamic Time Warping for robust characterization of behavioral states.</li><li>Comparative analyses show that the framework reduces prediction errors for acceleration, speed, and spacing metrics.</li></ul>

A Driving Regime-Embedded Deep Learning Framework for Modeling Intra-Driver Heterogeneity in Multi-Scale Car-Following Dynamics

Discover more