A data-driven model of pedestrian following and emergent crowd behavior

Pedestrian following is a common behavior, and may provide a key link between individual locomotion and crowd dynamics. Here, we present a model for following that is motivated by cognitive science and grounded in empirical data. In Experiment 1, we collected data from leader-follower pairs, and showed that a simple speed-matching model is sufficient to account for behavior. In Experiment 2, we manipulated the visual information of a virtual leader, and found that followers respond primarily to changes in visual angle. Finally, in Experiment 3, we use the speed-matching model to simulate speed coordination in small crowds of 4 pedestrians. The model performs as well in these small crowds as it did in the leader-follower pairs. This cognitivelyinspired, empirically-grounded model can serve as a component in a larger model of crowd dynamics.