Individual Time Stepping for SPH Fluids

Abstract

We present a novel adaptive stepping scheme for SPH fluids, in which particles have their own time-steps determined from local conditions, e.g. Courant condition. These individual time-steps are constrained for global convergence and stability. Fluid particles are then updated asynchronously. The approach naturally allocates computing resources to visually complex regions, e.g. regions with intense collisions, thereby reducing the overall computational time. The experiments show that our approach is more efficient than the standard method and the method with globally adaptive time steps, especially in highly dynamic scenes.