Identification of critical conditions for pedestrian-vehicle game modes at non-signalized crosswalks based on MCMC

The lack of traffic control and unclear right of way result in incomplete temporal and spatial separation of motor vehicles and pedestrians at non-signalized crosswalks. Consequently, crossing decisions rely on individuals' judgments of surrounding road conditions, environment, and perceived danger, leading to frequent interactions between vehicles, pedestrians. By collecting UAV video data from these crosswalks, we extracted data on pedestrian-motor vehicle interactions. We proposed an integrated index, PVGI (Pedestrian-Vehicle Game Index), which incorporates speed, acceleration, and distance factors to characterize the pedestrian-motor vehicle interaction process. Using the Markov Chain Monte Carlo (MCMC) simulation method, we identified the critical conditions that distinguish different game modes. The results indicate that PVGI can effectively classify pedestrian-motor vehicle interaction modes at non-signalized crosswalks. The MCMC simulations identified the interaction domain as [-4.0s, 2.0s], within which the game modes split into "pedestrian yielding-vehicle dominant" and "vehicle yielding-pedestrian dominant," corresponding to the intervals [-4.0s, 0s] and [0s, 2s], respectively. This study enhances the understanding of the decision-making mechanisms of motor vehicles and pedestrians at crossings and is significant for developing autonomous driving decision-making strategies.