Context-Dependent Adaptability in Crowd Behavior Simulation

Abstract

This paper applies a context-dependent behavior selection architecture to agent-based simulation of crowd behavior. In order to generate realistic humanlike crowd behaviors, we consider adaptability of an autonomous agent at two different levels: a behavior level where the agent makes adaptive selections among multiple competitive behaviors, and a behavioral context level where the agent modulates its behavior patterns according to different psychological and/or social states. A case study example of emergency response simulation is provided. The behavior and behavioral context models are described and simulation results are presented