Predicting Pedestrian Crossing Behavior and Violations Using a Pathfinding Approach

Increasing walking mode share is a widespread goal, but achieving this goal requires tools to proactively evaluate potential facility designs and control strategies. Existing tools do not provide the ability to simulate a pedestrian’s localized pathfinding decisions and road crossing behaviors and model pedestrian violation behaviors. This paper proposes a pedestrian simulation model in which pedestrians follow a minimum-cost route in which costs consider the surface type, violations, and interactions with vehicles. The model is calibrated and validated using field observations at a stop-controlled intersection and a midblock crossing. Results show that the model can predict the proportion of pedestrians performing a violation behavior on their way from a given origin to destination, including the trade-offs between waiting for a gap in traffic to cross versus diverting to the nearest designated crosswalk.