Car following regime taxonomy based on Markov switching

Numerous regime classification models have been introduced to capture car following driving behavior and to simulate constrained longitudinal driving characteristics. However, these models disregard the switching process dynamics between different driving regimes. In this paper, we propose a model that incorporates stochastic Markov regime switching model to estimate individual drivers characteristics and extract different driving regimes features. The proposed model takes into consideration the time factor and analyzes sequences of observations on driving time series. Trajectory data such as: velocity, acceleration, and space gap between leader and follower drivers were used as well as switching features to learn the model. Evaluation of the proposed model using real car following data sets shows that the model is able to classify normal car following driving behavior, rare events, and short time events. More importantly, the model is able to determine the switching dynamics among different regimes by applying maximum likelihood estimates and Hamilton filter. Additionally, the proposed model can infer regime specific characteristics, such as: expected duration, the probability of moving from one regime to another, switching parameters and driving patterns. Application of the proposed model includes- but not limited to: crash predication, and driver assistance and assessment systems.