Simulation of crowd behaviors based on event reaction

We propose a new organization for virtual crowd responses to emergent events. In our approach, there are three main components: agent model, smart environment model and behavior model. The agent model is responsible for providing important information to the smart environment model and executing the corresponding actions. The smart environment model includes a set of smart objects and events and embeds the behavioral responses to inform the behavior model of behavior options. The behavior model adopts the rule-based module to generate appropriate behaviors according to the attributes and current state of agents. This scheme does not require complex reasoning and can run in real time. Furthermore, it can not only show plausible crowd behaviors and reflect the dynamic change of the emergent event, but is also much easier for extendibility and scalability. We demonstrate this approach in a scenario of agents reacting to a car accident.