Graphics Reference
In-Depth Information
or brick objects cannot break without explicit simulation rules for creating new
objects from their pieces.
How much complexity do we need to abstract the behaviors of scenes that we
might want to simulate? A tumbling crate retains a rigid shape relative to its own
reference frame, but that frame moves through space. A walking person exhibits
underlying articulated skeleton of rigid bones connected at joints that are then
covered by deforming muscle and skin. The water in a stream lacks any rigid
substructure. It deforms around obstacles and conforms to the shape of the stream
bed under forces including gravity, pressure, and drag.
In each of these scenarios, the objects involved have varying amounts of state
needed to describe their poses and motion. The algorithms for computing changes
of that state vary accordingly.
Some object representations commonly employed in computer graphics (with
examples) are
1. Particles (smoke, bullets, people in a crowd)
2. Rigid body (metal crate, space ship)
3. Soft rigid body (beach ball)
4. Articulated rigid body (robot)
5. Mass-spring system (cloth, rope)
6. Skinned skeleton (human)
7. Fluid (mud, water, air)
These are listed in approximate order of complexity and algorithmic state. For
example, the dynamic state of a particle consists of its position and velocity. A
rigid body adds a 3D orientation to the particle representation.
Why are so many different representations employed? As an alternative, a
single unified representation would be much more theoretically appealing, be eas-
ier from a software engineering perspective, and automatically handle the tricky
interactions between objects of different representations.
One seemingly attractive alternative is to choose the simplest representation as
the universal one and sample very finely. Specifically, all objects in the real world
are composed of atoms, for which a particle system is an appropriate representa-
tion. Although it is possible to simulate everything at the particle level [vB95], in
practice this is usually considered awkward for an artist and overwhelming for a
physical simulation algorithm.
The authoring method and representation do not always match what is perceived
by the viewer. For example, many films and video games move the root frame of
seemingly complex characters as if they were simple rigid bodies under physical
simulation. In both cases, the individual characters are also animated by key pose
animation of skinned skeletons relative to their root frames. Thus, at different
scales of motion the objects have different specifications and representations. This
avoids the complexity of computing true interactions between characters while
retaining most of the realism.
This is analogous to level-of-detail modeling tricks for rendering. For exam-
ple, a building may be represented by boxlike geometry, a bump map that
