Genetic Programming Evolution of Controllers for 3-D
Character Animation
Larry Gritz James K. Hahn
Pixar Animation Studios
1001 W. Cutting Blvd.
Richmond, CA 94804
lg@pixar.com
The George Washington University
801 22nd St. NW
Washington, DC 20054
hahn@seas.gwu.edu
ABSTRACT
The dominant paradigm for 3-D
character animation requires an
animator to specify the values for all
degrees of freedom of an articulated
figure at key frames. Specifying motion
that is physically believable and
biologically plausible is a tedious
practice requiring great skill.
We use evolutionary techniques
(specifically Genetic Programming) as a
means of controller synthesis for
character animation. Controllers which
drive a dynamic simulation of the char-
acter are evolved using the goals of the
animation as an objective function,
resulting in physically plausible motion.
We discuss the development of objective
functions used to guide the controller
evolution, making reusable skill
controllers, and comparisons of the
convergence rates for different
parameters of the evolutionary runs.
1.
Introduction
In this investigation, we are concerned with a particular
subset of computer animation:
character animation
. In an
animation, many objects may be moving around.
Characters are those objects whose movements contribute
to telling the story and which express thinking, volition,
and emotion.
An example of animation which does not fall into the
category of character animation would be
effects
animation
. This would include movement of objects other
than characters, or which is not designed to convey
emotion or intent. Examples of such animation include
falling rain, leaves blowing in the wind, flying spaceships,
rotating business logos, or a flock of bats flying around.
Methods have been developed to successfully automate
several specific cases of effects animation, largely due to
the fact that many effects animation tasks can be solved by
straightforward physical simulation.
As the art and craft of character animation was refined,
particularly at Disney Studios, animators identified specific
properties of superior animation. Thomas and Johnston
[Thomas and Johntson, 1984], and later Lasseter
[Lasseter, 1987], identified “principles of animation”
which describe important aspects of realistic character
animation. They included squash and stretch, anticipation,
staging, straight ahead action and pose to pose, follow
through and overlapping action, slow in and slow out, arcs,
secondary action, timing, exaggeration, and appeal. Some
of these principles are purely artistic (such as appeal), but
many others describe the dynamic and behavioral
properties of real materials and creatures.
The primary method for animating 3-D characters is by
key framing
. The technique takes its name from the
practice in 2-D cel animation of a senior animator drawing
the figure at particular frames in which significant events
or extrema of motion occur. More junior animators then
use the key frame drawings as a guide for the intermediate
frames (doing a kind of human-powered interpolation of
the poses). These techniques were extended into the realm
of 3-D computer animation by specifying and interpolating
positions and joint angles of the characters. Extrema for
each degree of freedom (DOF) are specified by human
animators, and the intermediate values are interpolated
(frequently by cubic spline) by the animation software.
The motion resulting from interpolating the key frames
is purely kinematic. If the motion is to be physically
realistic (exhibiting inertia, obeying force laws, etc.),
biologically plausible (obeying joint limits or other
constraints of a particular creature), follow any of the
principles of animation described earlier, or be
entertaining,
it is entirely up to the human animator to
choose key times and values which meet these
constraints.
Virtually all production character animation
is done using this method, in spite of the large amount of
research into automating character animation. It is the
principles mentioned earlier which make animation (of any