A stochastic jump-process driving dynamic model with application to traffic safety

Emerging high frequency, high resolution, large scale driving data reflect driving dynamics and provide valuable information on driving risk. Microscopic driving models are one of the fundamental tools for analyzing driving kinematic data and understanding temporal patterns within instantaneous driving decisions. Safety-critical events, characterized by abrupt and severe jerky behavior, lie outside the capabilities of such models. This paper proposes a stochastic driving model incorporating a novel jump process to capture aggressive driving behaviors related to crashes. The model parameters are estimated using a nonparametric approach. A new safety evaluation metric, Jump Size, is derived based on the stochastic model, accounting for the cumulative sum of squared jump sizes. The proposed safety metric demonstrates superior out-of-sample performance with solid theoretical foundations. We demonstrated the proposed stochastic model using the data from the Second Strategic Highway Research Program Naturalistic Driving Study. The proposed stochastic model outperforms its state-of-the-art counterparts and can benefit microscopic driving simulations and safety evaluations of automated driving systems.