Modeling Within-Driver Heterogeneity in Car-Following through Coupled Response–Scenario Dynamics

Chengyuan Zhang, Cathy Wu, Lijun Sun
Artificial Intelligence for Transportation (accepted, 2026). Pages/article 100096. DOI in uncorrected proof: 10.1016/j.ait.2026.100096.

Earlier preprint · BibTeX · All publications

How are latent driving regimes different from traffic scenarios?

PS-HMM-IDM represents driving response regimes and traffic scenarios as paired states connected by a joint Markov process. Each response regime has its own IDM response relationship.

Research summary

A single set of IDM parameters can average over distinct driving responses. PS-HMM-IDM represents response regimes and traffic scenarios as paired states, allowing the same response relationship to recur across scenarios and a scenario to accompany multiple responses.

The regimes are inferred model states, not direct measurements of a driver’s psychology. The study examines their temporal structure in highD car-following episodes.

What is the distinction between a behavior regime and a traffic scenario?

A response regime describes how acceleration relates to driving inputs; a traffic scenario describes the distribution of observed kinematics. Their paired states allow the same response regime to recur across scenarios and a scenario to accompany multiple responses.

Earlier preprint abstract

Accurate and interpretable car-following models are essential for traffic simulation and autonomous vehicle development. However, classical models like the Intelligent Driver Model (IDM) are fundamentally limited by their parsimonious and single-regime structure. They fail to capture the multi-modal nature of human driving, where a single driving state (e.g., speed, relative speed, and gap) can elicit many different driver actions. This forces the model to average across distinct behaviors, reducing its fidelity and making its parameters difficult to interpret. To overcome this, we introduce a regime-switching framework that allows driving behavior to be governed by different IDM parameter sets, each corresponding to an interpretable behavioral mode. This design enables the model to dynamically switch between interpretable behavioral modes, rather than averaging across diverse driving contexts. We instantiate the framework using a Factorial Hidden Markov Model with IDM dynamics (FHMM-IDM), which explicitly separates intrinsic driving regimes (e.g., aggressive acceleration, steady-state following) from external traffic scenarios (e.g., free-flow, congestion, stop-and-go) through two independent latent Markov processes. Bayesian inference via Markov chain Monte Carlo (MCMC) is used to jointly estimate the regime-specific parameters, transition dynamics, and latent state trajectories. Experiments on the HighD dataset demonstrate that FHMM-IDM uncovers interpretable structure in human driving, effectively disentangling internal driver actions from contextual traffic conditions and revealing dynamic regime-switching patterns. This framework provides a tractable and principled solution to modeling context-dependent driving behavior under uncertainty, offering improvements in the fidelity of traffic simulations, the efficacy of safety analyses, and the development of more human-centric ADAS.

Full abstract from the earlier arXiv version, checked 2026-09-13. This preprint uses the title "Markov Regime-Switching Intelligent Driver Model for Interpretable Car-Following Behavior" and an earlier model formulation; the accepted manuscript has been substantially revised.

Code · Technical note on the earlier preprint · Earlier research poster · Stochastic driving behavior