(Disney Meets Darwin)
1: Introduction
Animation, in the tradition of Disney, is "The Illusion of Life".
New computational techniques such as those practiced in the
field of artificial life have begun to contribute techniques
to the art of animation for the simulation of life. "Life"
can mean something quite different to a cartoon animator than
it can to a theoretical biologist. I have been exploring these
differences, and also the similarities. The ideas described
here are derived from recent explorations in applying genetic
algorithms to a variety of visual design domains, including
character animation [Ventrella 94, 1].
The use of the proper names "Disney", and "Darwin" in the
title are not a claim that this work represents the achievements
of either individual, but simply to imply that the work applies
the concept of natural selection to the art of expressive animation.
How do animators create expressive, humorous motion? It rarely
happens in one sweeping act of design, but rather in stages of refinement,
involving trial and error, often bringing in techniques which have worked
well in the past and combining them with new techniques, or taking aspects
of one character and applying them to another. The creative act is quite
often an evolutionary process, involving bottom-up emergence, discovery,
and serendipity - mixed in with the more understood top-down design schemes.
Can the evolutionary aspect of design activities be supported by computational
tools?
In this paper I present a system which was built with this
question in mind. It is manifest as an experimental evolution-based
tool for exploring variations in articulated stick figures, serving
as moving-skeletons for three dimensional forms. These figures move
in real-time on a computer screen and inhabit a virtual world of simple
physics, with gravity, momentum, and friction. Their anatomies and
styles of wiggling around come in many varieties, as determined by
their genetic makeup, and the ways in which their genes interact with
each other and the environment. The genetic component of these figures
can evolve in response to an animator's "breeding" activities
(interactive evolution), and also by way of automatic evolution - driven
by multiple fitness criteria.
(go to beginning of document)