Leader-follower model for agent based simulation of social collective behavior during egress

There is growing consensus that individuals egressing during emergencies engage in social collective behavior that is ordered and cooperative. Despite this fact however, a great many of the existing models either do not account for this type of behavior or model it in an inadequate manner. In this paper, a leader-follower agent-based model is proposed to interpret local social interactions and collective behavior and then use this information to mimic three particular scenarios: lining up in counter-flow, queuing, and collective mobility. To achieve this, a pedestrian agent can establish informal and transient leader-follower relationships with others while adjusting its behavioral patterns as warranted by the situation. The proposed model is calibrated to existing field data and then validated using another set of field data, where it is shown that the new model is capable of reasonably simulating social collective behavior during egress.