(MIT Media Lab Master's Thesis)

Disney Meets Darwin

An Evolution-based Interface
for Exploration and Design
of Expressive Animated Behavior

Jeffrey Ventrella

Visible Language Workshop
Media Lab, MIT
May, 1994



Abstract

In this thesis I describe a computer animation system called the Character Evolution Tool. It was developed both as a prototype system for animators and graphic designers, and as a testbed for the applicability of genetic algorithms in the design process. Although the focus is primarily on physically based articulated characters, these are considered as only a sub-class in the class of all graphical objects which can exhibit expressive motion behavior, termed, "behavior objects." The Character Evolution Tool employs a genetic algorithm for the automatic evolution of goal-oriented behavior in animated graphics, with an overlay of interactive evolution. This overlay affords the user of the system the ability to encourage expressivity and communicative behaviors in the animated characters, as they evolve towards some otherwise pre-defined objective goal. Central to this system is an experimental technique for guiding the direction of evolution by way of a gesture drawn into the scene by the user. This gesture constitutes a novel approach to defining the genetic algorithm's objective fitness function, in that the characters are encouraged to emulate properties of the gestured motion, thereby assuming some of the expressive qualities that the user has specified.


Thesis Supervisor: Ronald MacNeil
Title: Principal Research Associate, Program in Media Arts and Sciences

This work was supported in part by Paws, Inc., USDOT, and NIF.




TABLE OF CONTENTS


ACKNOWLEDGEMENTS

1. INTRODUCTION
Expressivity Through Guided Evolution
Computer Animation Not Alive Enough
"Move Like This" - Survival of the Fittest
Evolution and Design

2. BACKGROUND AND RELATED WORK
Autonomous Character Animation
Spacetime Constraints
Motion Control
The Genetic Algorithm
Artificial Life
Interactive Evolution
Gesturing for Animation

3. APPROACH
Two Levels of Evolution
Gesturing

4. NUTS AND BOLTS
Behavior Objects
The Articulated Figures
Morphology
Motor Control
Periodicity
Physics
Architecture
The Interface
Gesturing
Saving and Loading Genetic Data

5. RESULTS
Subject's Responses
Lag Time Learning
Aperiodic Expressions
Prototyping and Phenotyping
Artificial Life

6. CONCLUSIONS
An Animation Tool
A Research Environment for Exploring Evolution in Design
A Tool for Exploration of Evolutionary Principles
Design and Evolution
User Interfaces to the Genetic Algorithm
Conclusion

7. FUTURE WORK
Clip Behaviors
Family Trees
Flesh and Bones
Interacting Characters

GLOSSARY

Appendix A: The Morphology Scheme for the Articulated Figures
Appendix B: The Motor Scheme for the 3D Articulated Figures
Appendix C: The Physics for the Articulated Figures

REFERENCES