Andrew G. Zaferakis
Progress
Update March 30, 2000
Progress
Update April 20, 2000
Final
Report May 12, 2000
Project
Presentation May 12, 2000
GOAL
Using graphics hardware to fast compute the generalized voronoi diagram,
it is possible to use this information to determine between objects. Think
of particles falling through a scene that is cluttered with objects. The
particles will bounce off the objects and eventually come to a resting
position. To determine the collisions, I will use the depth buffer, created
by the voronoi diagram of the objects in the scene, to determine a particle's
distance from an object. Each particle will then know its distance
from the closest voronoi site (object) in the scene, when the distance
is very small it has collided with an object.
MOTIVATION
Using graphics hardware to fast compute the generalized voronoi diagrams
is a new topic in computer science. Since graphics hardware is extremely
fast, it can be used as an easy way to perform fast collision detection.
The ability to speed up a graphical rigid body simulation is useful for
many applications.
PERSONAL INTEREST
Interactive and dynamic 2D and 3D computer graphics are of particular interest
to me. I would like to explore the various ways to increase the speed
of simulations and other dynamic applications. Rigid body dynamics
is a method that is used in many applications, which will give me a good
perspective on how to tackle diverse problems. Using hardware to
fast compute voronoi diagrams is relatively new and not much research has
been published, giving me the opportunity to publish my findings.
PREVIOUS WORK
Fast computation of generalized
voronoi diagrams originated here at the UNC,
by Kenny Hoff, Dinesh
Manocha, and others in the Geometry
Group presented at SIGGRAPH '99. UNC has also done research in
interactive
motion planning, which uses the voronoi diagram and graphics hardware.
Kenny's previous particle demonstration uses the voronoi distance as a
repulsion field for the particles. As the particle comes closer to the
object, the repulsion is greater, that way never having to actually detect
the collision, which might be missed due to a jump in the timestep. I will
attempt to use newtoni I am also using Witkin and Baraff's SIGGRAPH course
notes on physically
based modelling.
DEMONSTRATION
I would like to show an interactive demonstration of 2D rigid body dynamics
using OpenGL. In addition I am planning on presenting my research
findings using powerpoint slides and possibly a technical paper.
346 Sitterson Hall, CB #3175, Chapel Hill, NC 27599-3175,
andrewz@cs.unc.edu