Realistic Modeling of Agents in Crowd Simulations

Simulation of crowd behaviors is a widely discussed topic for over a decade for computer games, architecture and entertainment. The most important thing is that agents should execute the human like behaviors in real life scenarios. This situation becomes very complex with dense crowds. Our proposed Realistic Behavior Agent Simulation (RBAS) model simulates the agents from the knowledge of empirical findings of crowd physical behaviors and crowd psychology. This paper presents a novel technique to design the agents to with following modules 1) Path planning behavior for collision avoidance 2) Situation awareness during herding behavior and turbulent flow in high density crowds. 3) Personal Reaction bubble (PRB) based response and perceptions. The evaluation with real life situations is performed to validate the RBAS model. The RBAS model allows the agents to use cognitive understandings to plan ahead their path using the visual perception information. This model encapsulates the emergent crowd behaviors such as self-organization behavior in herding situation. During turbulent high density flows, most of the existing models fail to predict the behavior of agents, the evaluations show the RBAS model mimics the same behavior of the crowd in different situations.