大规模人群的并行仿真研究

Research on Parallel Simulation of Large Scale Crowd

本文针对大规模人群仿真对仿真真实性和实时性的仿真要求,提出了一种基于Agent的并行人群仿真方法,通过CPUGPU混合并行的方式提高了仿真系统的性能和可扩展性。通过层次化的人群建模,使得个体更适合并行环境下的数据交互,仿真中通过感知区域视角约束对人群的运动进行控制,实现了人群平滑运动。实验部分对室外环境的人群漫游行为进行了仿真,验证了系统具有较高的性能和可扩展性。

To meet the authenticity of the simulation of large - scale crowd simulation and real - time simulation require- ments, this paper proposes the Agent based parallel crowd simulation method. Through CPU - GPU hybrid parallel simula- tion approach, it improves the performance of the system and scalability. Hierarchical modehng of the individuals in the crowd makes the individual more suitable for data exchange in parallel environments. Simulation by perceptual constraints on regional perspectives to control the movement of people achieves a smooth movement of the crowd. Experiments of be- havior of the crowd roaming the outdoor environment verify that the system has high performance and scalability.