Partitioned Dynamics

Abstract

Most physically-based simulation systems for computer graphics target only a single simulation domain such as particle systems, rigid bodies, cloth, and liquids. By using "layering" techniques, one-sided interactions between different domains are easily produced. For example, one-sided particle system/rigid-body simulation is achieved by first running a rigid-body simulator and then injecting the rigid body motions into a particle-system simulator; in the particle system, particles rebound off the (possibly moving) solid objects without affecting the solid objects' motion. This paper address the problem of combining disparate simulations so as to allow two-way interactions, such as a jet of particles that both deflects off a stack of solid objects, but also causes the stack to topple over realistically. Incorporating multiple simulation domains within a single simulation system is difficult because each simulation problem is best attacked using specialized techniques. Instead, we propose a method that treats each simulator as a "black box" with a simple generic interface. We present a technique called interleaved simulation that achieves geometrically accurate behavior with minimal overhead, compared to the cost of running the simulations independently. The method works best where the component systems' masses differ significantly. We show that interleaved simulation is equivalent to taking one matrix-solving gradient step per simulation time step, with a very good preconditioner. We include an easy-to-implement description of the method, and present a variety of simulation results.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Mar 01, 1997
Accession Number
ADA594838

Entities

People

  • Andrew Witkin
  • David Baraff

Organizations

  • Carnegie Mellon University

Tags

Communities of Interest

  • Ground and Sea Platforms

DTIC Thesaurus Topics

  • Collisions
  • Computer Graphics
  • Computer Programs
  • Computers
  • Detection
  • Differential Equations
  • Dynamics
  • Equations
  • Errors
  • Fabrics
  • Geometry
  • Iterations
  • Particle Collisions
  • Particles
  • Simulations
  • Simulators
  • Triangles

Fields of Study

  • Computer science

Readers

  • Computational Fluid Dynamics (CFD)
  • Computational Modeling and Simulation
  • Parallel and Distributed Computing.