Adaptive Cloud Simulation Using Position Based Fluids
Computer Animation and Virtual Worlds, Vol. 26, No. 3-4, pp. 367-375 (2015)
Proc. The 28th International Conference on Computer Animation and Social Agents (CASA2015) (2015-5)

Charles Welton Ferreira Barbosa 1 Yoshinori Dobashi 1,2 Tsuyoshi Yamamoto 1
1Hokkaido University,  2JST CREST 

     

Abstract
In this paper, we propose a method for the simulation of clouds using particles exclusively. The method is based on Position Based Fluids, which simulates fluids using position constraints. To reduce the simulation time, wefve used adaptive splitting and merging to concentrate computation on regions where it is most needed. When clouds are formed, particles are split so as to add more detail to the generated cloud surface and when they disappear, particles are merged back. We implement our adaptive method on the GPU to accelerate the computation. While the splitting portion is easily parallelizable, the merge portion is not. We develop a simple algorithm to address this problem and achieve reasonable simulation times.

Download
paper(draft):[pdf] video:avi presen:pptx

Link