An enhancement of 3D fluid control using adaptive kernel estimation

Abstract

In computer animation, animation tools are required for fluid-like motions which are controllable by users or animators. One of the popular and widely used methods for simulating fluid flow in computer graphics is Smoothed Particle Hydrodynamics (SPH). Although SPH is practical for applying to fluid movement control and also preserves efficiently properties of fluid flow, it is complicated to simulate various details in same flow. This study proposes an enhanced method based on Smoothed Particle Hydrodynamics (SPH) which is controllable and automatically kernel length adjustable in order to preserve small details; lost by fixed kernel. Reeb graph is used to construct a structure of an input object. To control fluid flow, we use a kind of particle control methods called Skeletal Particles. Skeletal Particles or control particles are placed on the center of smoothing kernel in computation step. To solve the problem caused by fixed kernel length, we implement the Adaptive Kernel Density Estimation (AKDE) technique into the computation step. Therefore, the smoothing length is dynamically adapted to all given points to be appropriate with density of their regions. As a result, the errors of unstable mean values of particles in low density regions are decreased