Will I cross when I see you crossing? Towards illegal pedestrian road crossing simulations

Abstract

The temporal dimension (how long the pedestrian has been in a situation) and the social influence from the neighbor's behaviors are important factors in street crossing decisions. In this paper, we propose a preliminary agent-based model of the street crossing decisions at a pedestrian light without traffic based on three priors. First, a pedestrian agent is willing to wait a certain amount of time before crossing a street (we call it Accepted Waiting Time, AWT) and beyond that, it crosses even at a red light. The next two priors are based on Rosenbloom's work, who suggests that a pedestrian waiting at a red light may be influenced by those who are crossing or waiting. Combining these two priors mimics some aspect of social influence, and we assume this social influence modulates the AWT of the pedestrian agent. Thus, in our simulations, agents supposed to cross at the red light may wait for the green light, influenced by waiting neighbors. Conversely, the window of opportunity of the neighbors crossing at the red light influences the waiting agents to cross as well. These results mean that the crossing decision would be different if the agents were alone. Moreover, agents with similar characteristics (in terms of AWT) and perceiving the same situation (color of the pedestrian traffic light and number of neighbors) but arriving at the crossing location at different times, will take different decisions (cross/wait). Furthermore, the behaviors produced by our model are quite consistent with those reported in the literature, in terms of violation rate and waiting time. Further research is needed in order to include additional influence factors to this proof-of-concept model, and to take traffic into account, which is needed before using the proposed model in virtual reality applications dedicated to children learning to cross roads.