Centroidal Particles for Interactive Crowd Simulation

Abstract

Real-time crowd simulation is a challenging task that demands a careful consideration of the classic trade-off between accuracy and efficiency. Existing particle-based methods have seen success in simulating crowd scenarios for various applications in the architecture, military, urban planning, robotics, and entertainment (film and gaming) industries. In this paper we focus on local dynamics and present an area-based penalty force that captures the infringement of each entity's personal space. This method does not necessitate a costly nearest-neighbor search and allows for an inherently data-parallel implementation that is capable of simulating thousands of entities at interactive frame-rates. The algorithm successfully reproduces personal space compression around motion barriers for moving crowds and around points of interest for static crowds.