Comparision of Real-Time Crowd Simulation Methods Based on Parallel Architecture

Abstract

Crowd simulation plays an important role in virtual reality and it has been widely applied in various fields. The research always focuses on the simulation efficiency and rendering. We construct a real-time crowd simulation system with different methods based on parallel architecture. In this paper, we compare and analyze their different simulation efficiencies through the experiments. The method in combination of CPU and GPU with parallel computing on multi-core CPUs can get the highest efficiency in our simulation system. The GPU method is suitable for the crowd simulation system without frequent data transmission between CPU and GPU and a mass of branch statements and texture look-ups.