Search papers, labs, and topics across Lattice.
This paper introduces a novel Markov Chain Car-Following (MC-CF) model for traffic simulation that represents car-following behavior as a Markov process, sampling accelerations from empirical distributions within discretized state bins. Trained on the Waymo Open Motion Dataset (WOMD), MC-CF significantly outperforms traditional physics-based models in both one-step and open-loop trajectory prediction. The model demonstrates strong zero-shot generalization capabilities and achieves zero crashes in microscopic ring road simulations by integrating diverse datasets and employing a conservative inference strategy.
Forget hand-crafted physics: this data-driven car-following model learns driving behavior directly from real-world data and achieves zero crashes in simulated traffic scenarios.
Car-following behavior is fundamental to traffic flow theory, yet traditional models often fail to capture the stochasticity of naturalistic driving. This paper introduces a new car-following modeling category called the empirical probabilistic paradigm, which bypasses conventional parametric assumptions. Within this paradigm, we propose the Markov Chain Car-Following (MC-CF) model, which represents state transitions as a Markov process and predicts behavior by randomly sampling accelerations from empirical distributions within discretized state bins. Evaluation of the MC-CF model trained on the Waymo Open Motion Dataset (WOMD) demonstrates that its variants significantly outperform physics-based models including IDM, Gipps, FVDM, and SIDM in both one-step and open-loop trajectory prediction accuracy. Statistical analysis of transition probabilities confirms that the model-generated trajectories are indistinguishable from real-world behavior, successfully reproducing the probabilistic structure of naturalistic driving across all interaction types. Zero-shot generalization on the Naturalistic Phoenix (PHX) dataset further confirms the model's robustness. Finally, microscopic ring road simulations validate the framework's scalability. By incrementally integrating unconstrained free-flow trajectories and high-speed freeway data (TGSIM) alongside a conservative inference strategy, the model drastically reduces collisions, achieving zero crashes in multiple equilibrium and shockwave scenarios, while successfully reproducing naturalistic and stochastic shockwave propagation. Overall, the proposed MC-CF model provides a robust, scalable, and calibration-free foundation for high-fidelity stochastic traffic modeling, uniquely suited for the data-rich future of intelligent transportation.
