Norhafiza Hamzah*, Norsuzila Yusof & Z.A.Omar
School of Science and Technology,
Universiti Malaysia Sabah, Jalan UMS, 88400 Kota Kinabalu, Sabah, Malaysia

ABSTRACT. This research is focused on flocking behaviour algorithm to simulate the crowd on parallel GPU using CUDA technology. The analysis of frame rates is conducted to compare the crowd simulation on parallel GPU platform and on a single processor. The result shows that the crowd simulation on a parallel GPU platform is 15 frames per second for 16, 384 characters. This result is equivalent with the number of frames per second for crowd simulation on a single processor with 576 characters. Thus, the results demonstrate that crowd simulation is more efficient on the parallel GPU platform especially for the large scale data. KEYWORDS. Crowd simulation; Flocking behaviour; GPU computing; CUDA



  • Jiang, H., Xu, W., Mao, T., Li, C., Xia, S. & Wang, Z. 2010. Continuum crowd simulation in complex environments. Computers & Graphics, 34(5): 537-544.
  • Reynolds, C. 2006. Big fast crowds on PS3. Proceedings of the 2006 ACM SIGGRAPH symposium on Videogames. Boston, Massachusetts: ACM, (pp.113-121).
  • Saboia, P. & Goldenstein, S. 2012. Vis Comput. Crowd simulation:applying mobile grids to  the social force model, pp.1039-1048.
  • Serrano, M. I. 2011. Flock Implementation for the Blender Game Engine. The Florida State University.
  • Passos, E., Joselli, M., Zamith, M., Rocha, J., Clua, E., Montenegro, A., Conci, A. & Feijó, B. 2008. Supermassive crowd simulation on GPU based on emergent behavior. Proceedings of the Seventh Brazilian Symposium on Computer Games and Digital Entertainment (SBGames’08), Sciedade Brasileira de Computação, SBC, pp.70-75.

Download Full Paper Here (Right-Click and Save As.. )