Analytical Modeling of Dynamic Decision-Making Behavior of Drivers During the Phase Transition Period Based on a Hidden Markov Model

A flashing green of 3s followed by a yellow of 3s is commonly applied to end a green phase, even at the rural high-speed intersections in many Chinese cities. As a result, the perception-and-reaction process during the phase transition period becomes rather complex and dynamic due to the flashing green and the insufficient yellow time. This paper thus proposed an analytical model based on the Hidden Markov Model theory to interpret this dynamic decision-making process, based on high-resolution vehicle trajectory data. In the authors’ model, the hidden states are driver’s time-dependent decisions whether to stop or pass, which cannot be directly observed until the subject vehicle passed the intersection or stopped before the stop-line; the observable states are instantaneous vehicle speeds and acceleration/deceleration rates, implying driver’s decision tendency. Video data were collected at four typical high-speed intersection approaches with a speed limit of 80km/h in Shanghai and 698 high-resolution vehicle trajectories including 179 trucks and 519 passenger cars were obtained and used for model estimation and validation in the study. It was found that the developed model could predict the stop-pass decisions with a high accuracy. In addition, approximately 50% of drivers experienced a two-step decision process. The conventional dilemma zone theory failed to explain the phenomenon due to its flawed hypothesis of one-step decision process. It was also found that those drivers are most likely to modify their decisions made at the onset of flashing green, 0~1.2s after encountering a yellow indication, based on the perceived surrounding traffic environment. The findings have significant implications on the proper design and safety countermeasures for high-speed signalized intersections with flashing green and insufficient yellow time.