Context-Aware Motion Diversi!cation for Crowd Simulation

L crowd simulation is important in computer games, movies, virtual training, and education applications. One popular approach, agent-based crowd simulation, considers the properties of each individual (agent) separately at every time step, enabling a highly realistic simulation of path navigation, cognitive reaction, collision avoidance, and animation control. Agent-based crowd simulation systems work according to a three-layer hierarchy. The highest layer provides the navigation waypoints through path!nding and decision-making. The intermediate layer achieves collision avoidance and collision response by computing high-level motion information for every update (that is, every time step) using perceptual rules or social forces. The lowest layer handles the detailed animation of each agent according to the parameters supplied by the higher layers. For each agent, these three layers answer three questions: