INFORMATION TO USERS

The most advanced technology has been used to photo­ graph and reproduce this manuscript from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are re­ produced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book. These are also available as one exposure on a standard 35mm slide or as a 17" x 23" black and white photographic print for an additional charge.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

University Microfilms International A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 313/761-4700 800/521-0600

Order Number 8913622

A physically-based simulation approach to three-dimensional computer animation

Caldwell, Craig Bemreuter, Ph.D.

The Ohio State University,1989

UMI 300 N. Zeeb Rd. Ann Arbor, MI 48106

A PHYSICALLY-BASED SIMULATION APPROACH TO

THREE-DIMENSIONAL COMPUTER ANIMATION

DISSERTATION

Presented in Partial Fulfillment of the Requirements for

the Degree Doctor of Philosophy in the Graduate

School of the Ohio State University

By

Craig Bernreuter Caldwell, B.A., M.F.A.

*****

The Ohio State University

1989

Dissertation Committee: Approved by

Thomas E. Linehan

"7 Charles A. Csuri

Advisor Richard E . Parent Department of Art Education ACKNOWLEDGEMENTS

I would like to express my sincere appreciation to Dr. Thomas E. Linehan for his guidance and insight throughout the research. I am indebted to him for his patience and encourage­ ment throughout this process. I am also deeply grateful to the the other members of my committee, Professors Charles A. Csuri and Rick E. Parent, for their invaluable suggestions and com­ ments .

None of this would be possible without the dedication and perseverence of Charles A. Csuri in starting and developing this graduate specialization with the considerable efforts of Thomas E. Linehan. Without both of them The Advanced Computing Center for the Arts & Design would not exist.

Gratitude is expressed to all the special people that have made up the Advanced Computing Center for the Arts & Design Team. I acknowledge the considerable assistance provided by John Chadwick, Dave Haumann, Ruedy Leeman, Michael Girard, Brian Guenter, Ron Piotrowski, and Mark Jansen.

To my wife, Amy, I have no sufficient way of thanking her for her unshakable faith in me and her willingness to endure with me the vicissitudes of my endeavors. To my children, Ryan and Joanna, I can only hope that now I can make up for the years of growing up they did while I was away, and thank them for their understanding of why I had to do it.

ii VITA

October 6, 1950...... Born - Chicago, Illinois

1972...... B.A., Florida Southern College, Lakeland, Florida

1974 ...... M.F.A., University of Florida, Gainesville, Florida

1974-1976...... Instructor, Columbus College of Art and Design, Columbus, Ohio

1976- Present...... Associate Professor, School of Art and Design, Northern Arizona University, Flagstaff, Arizona

PUBLICATIONS

Caldwell, C. (1989). "Looking In". In G. de Valois (Producer/Director), Computer Dreams. [Video Tape]. Hollywood, CA: Digital Visions.

Caldwell, C. (1989). "Looking In". In M. Kusahara (Producer/Director), COMPUTER GRAPHICS ANTHOLOGY. [Lazer Disk] Tokyo: AMAYAKAN.

Caldwell, C. (1988) . "Collaboration," ACM SIGGRAPH '88 Tu­ torial Notes, Educators Workshop.

Caldwell, C. (1988) . "Interaction Between Computer Science and Art in Computer Graphics," National Computer Graphic Association '88 Conference Proceedings, Volume III. pp. 653-663.

Caldwell, C. (1987). "Overlapping Images and Ideas," IEEE Computer Graphics and Applications, l_f (8) , pp. 3-6.

Caldwell, C. (1987). Edited by B. E. Brown in "SIGGRAPH '87 Technical Slide Set," Computer Graphics, 21, (5), p. 289.

iii Caldwell, C. (1987). Edited by J. P. Culver and B. E. Brown in "SIGGRAPH ' 87 Art Show Slide Set," Computer Graphics, 21, (5), p. 290.

Caldwell, C. (1987). Visual Works, PIXEL, 56, (5), pp. 38-39. Caldwell, C. (1987). Electronic Musician maga­ zine. May 1987, Vol. 3, No. 5.

Caldwell, C. (1987). Cover of PIXEL magazine. (Japanese Computer Graphics Magazine) April 1987, No. 55.

Caldwell, C. (1985). Computer Graphic Concepts and Tech­ niques for Artists and Designers. National Computer Graphic Association '85 Conference Proceedings, Volume I, pp. 393-413.

Wagner, P. and Caldwell, C. (1984). Digital Portfolio (Visual Work by Caldwell) Computer Graphics World, 7, (9), 62-63.

Caldwell, C. (1984). First Annual Computer Art Competi­ tion, Popular Computing magazine. September, 1984. (Work selected for reproduction in Special Graphics issue) p. 27.

FIELDS OF STUDY

Major Field: Computer Graphics and Animation (Department of Art Education)

Studies in Computer Graphics and Animation. Professors Thomas E. Linehan, Charles Csuri, J. Christian Wedge. TABLE OF CONTENTS

ACKNOWLEDGEMENTS...... ii

VITA...... iii

LIST OF FIGURES...... viii

CHAPTER

I. INTRODUCTION...... 1

1.1 Importance of Simulation as Animation 2 1.2 Purpose of Study...... 4 1.3 Overview of Study...... 7

II. METHODOLOGY...... 10

2.1 Inception of the Problem...... 12 2.2 Statement of the Problem...... 13 2.3 Criteria for Assessing the Model...... 14 2.3.1 Fidelity...... 15 2.3.2 Utility...... 15 2.4 Delimitation...... 16 2.5 Basic Assumptions...... 17 2.6 Definitions...... 18

III. REVIEW OF LITERATURE - ART, ANIMATION, AND SIMULATION...... 23

3.1 Traditional Art Media...... 24 3.2 Beginnings of animation...... 28 3.3 Animation - Early aesthetic directions.... 30 3.4 Media Artifacts...... 34 3.4.1 Media Transformations...... 35 3.5 Mimetic Tradition in Art...... 37 3.6 Animation - Timing...... 39 3.6.1 Principles of Animation...... 41 3.7 Rotoscoping...... 48 3.8 Animation - Computer...... 53 3.8.1 Computer Motion - Applications 56 3.8.2 Computer Animation - 2D beginnings.. 57 3.8.3 Computer Animation - System Hierarchy Levels...... 58 3.8.3.1 Guiding Level...... 59

v 3.8.3.2 Procedural Level...... 64 3.8.3.3 Task or Goal-Directed Level...... 66 3.8.3.4 Synergic Control...... 68 3.8.4 Computer Animation - Production Process...... 70 3.9 Simulation...... 72 3.9.1 Simulation Applications...... 73 3.9.2 Simulation Methods...... 74 3. 9.2.1 Robust...... 75 3.9.2.2 Stochastic...... 76 3. 9.2.2.1 Ad-Hoc...... 76 3.9.2.3 Deterministic versus Non-Deterministic Simulations.. 78 3.9.3 Simulation as Animation. 79 3.10 Physics - Mechanics...... 82 3.10.1 Laws of Motion...... 83 3.10.2 Kinematics...... 83 3.10.2.1 Forward Kinematics...... 85 3.10.2.2 Inverse Kinematics..... 86 3.10.3 Dynamics ...... 88 3.10.3.1 Forward Dynamics...... 89 3.10.3.2 Inverse Dynamics...... 89 3.10.3.3 Dynamic Analysis...... 90 3.11 Simulation of Aesthetic Criteria.,...... 91 3.11.1 Story Simulation...... 95

IV. REVIEW OF INTERDISCIPLINARY FACTORS AND SYSTEMS...... 110

4.1 Abstractions...... 113 4.2 Behavior - Physical...... 116 4.2.1 Behavior of Flexible Surfaces...... 117 4.2.1.1 Flexible Objects - Constrained Surfaces...... 117 4.2.1.2 Flexible Surface - Facial Animation...... 122 4.2.1.3 Deformation...... 126 4.2.2 Articulated Structures...... 128 4.2.2.1 Figurative Motion...... 128 4.2.2.2 Generative Modeling...... 137 4.2.2.3 Constraints...... 140 4.2.3 Particle Systems...... 144 4.3 Behavior - Environmental...... 146 4.3.1 Object Oriented Programming...... 147 4.3.2 Self-Scripting...... 149 4.3.2.1 Behavior Simulation - Flocking Behavior...... 150 4.4 Simulation/Animation Hybrid...... 152

V. THE MODEL...... 162

vi 5.1 Geometric Primitives...... 165 5.2 Mechanical Attributes...... 167 5.3 Functional Procedures...... 172 5.4 Behavioral Simulation...... 175 5.5 Interactivity of the Model...... 179 5.5.1 Usability...... 179 5.5.1.1 Usability - Resource Libr ar ie s ...... 179 5.5.1.2 Usability - Intuitive Modi­ fications, Visual Editor.. 180 5.5.1.3 Usability - Real-time Playback Facility...... 181 5.5.1.4 Usability - Weighting Facility...... 183 5.5.2 Flexibility...... 185 5.5.2.1 Flexibility - Aesthetic Criteria...... 192 5.5.3 Extensibility...... 194 5.5.4 Habitability...... 194 5.5.5 Sub-Model - Continuity Applications...... 195 5.6 Limitations of the Model...... 200 5.6.1 Limitations of the Model - Requirement upon the Artist... 202

VI. IMPLICATIONS OF THE MODEL...... 210

6.1 Fidelity...... 211 6.2 Utility - Creative Process...... 213 6.3 Model as Medium...... 217 6.4 Potential of the Model...... 220 6.5 Criticism of the Model...... 225 6.6 Parallels with Modern Art...... 228 6.7 Future Directions...... 234

VII. SUMMARY...... 247

BIBLIOGRAPHY...... 254

t 4 V l l LIST OF FIGURES

FIGURES

1. Kazimir Malevich, The Knife Grinder. (1912), oil on canvas, Yale University Art Gallery ......

2. Marcel Duchamp, Nude Descending a Staircase No. 2 . (1912). Oil on canvas, Philadelphia Museum of Art: Louise and Walter Arensberg Collection......

3. Umberto Boccioni, Unique Forms of Continuity in Space. (1913). Bronze, 43 7/8 x 34 7/8 x 15 3/4". Collection, The Museum of Modern Art, New York. Acquired through the Lillie P. Bliss Bequest......

4. Animator: Stuart Blackton, "Humorous Phases of Funny Faces". >From OF MICE AND MAGIC: A HISTORY OF AMERICAN ANIMATED CARTOONS by Leonard Maltin. Copyright 1980, 1981 by Leonard Maltin. Reprinted by Arrangement with New American Library, A Division of Penguin Books USA Inc......

5. Animator: Winsor McCay, "Gertie the Dinosaur". >From OF MICE AND MAGIC: A HISTORY OF AMERICAN ANIMATED CARTOONS by Leonard Maltin. Copyright 1980, 1981 by Leonard Maltin. Reprinted by Arrangement with New American Library, A Division of Penguin Books USA Inc......

6. Felix the Cat, new departure for animation. >From OF MICE AND MAGIC: A HISTORY OF AMERICAN ANIMATED CARTOONS by Leonard Maltin. Copyright 1980, 1981 by Leonard Maltin. Reprinted by Arrangement with New American Library, A Division of Penguin Books USA Inc.

7. Animator: Norm Ferguson, Mickey's Elephant, Principle of squash and stretch applied to Pluto's head Copyright 1988 The Walt Disney Company ......

viii 8. Animator: Ham Luske, "Elmer Elephant", Principles of anticipation, follow through, squash and stretch, and exaggeration. Copyright 1988 The Walt Disney Company ...... 44

9. Tracings from photostats of a swinging bird cage for the animation "Pinocchio", Copyright 1988 The Walt Disney Company ...... 50

10. Rotoscoping technique used in "101 Dalmatians". Copyright 1988 The Walt Disney Company ...... 51

11. Cumulative, sequential frames from deterministic simulation. A piece of paper falling. Dave Haumann, 1987 ...... 81

12. Springs, Hinges and Joints which represent the dynamic aspects of an object. Chris Wedge...... 118

13. Control points for "Ballon Guy" character created by dynamic control program. Chris Wedge...... 121

14. Key-frame facial animation. Daniel Langlois ...... 123

15. Facial animation generated by parameterization. Brian Guenter. 1987 ...... 124

16. Modeling elastic deformations. Demetri Terzoploulos...... 127

17. Example of figure used for generated motion tests. >From the TEMPUS program at the University of Pennsylvania. Low- resolution test figure. Norm Badler...... 129

18. Dynamic modeling of an articulated figure. "Man falling on surface with friction. Jane Wilhelms...... 133

19. Michael Girard. Example of PODA figure using inverse kinematic control...... 134

20. Time-speed graph. Chris Wedge ...... 137

21. Tree structure propagated by generative modeling technique...... 139

ix 22. Constraint force. Based on information from Ronen Barzel and Alan H. Barr. Figure reconstructed by author...... 141

23. Point-to-point constraint. Compound pendulum. Based on information from Ronen Barzel and Alan H. Barr. Figure reconstructed by author...... 142

24. Form of an explosion-like particle system, Bill Reeves...... 145

25. Richard Lundin's mechanical ant from "The Works"...... 153

26. The Model - Geometric primitives ...... 165

27. The Model - Mechanical Attributes ...... 168

28. The Model - Functional Procedures ..... 173

29. The Model - Behavioral Simulations ...... 178

30. Motion Modules of the Model ...... 187

31. Feedback Control Loop for the Model ...... 18 9

32. Jumping sequence of Luxo Jr. Lamp generated by key frames. John Lasseter...... 227

33. Jumping sequence of Luxo Jr. Lamp generated algorithmically, Andrew Witkin...... 227

x CHAPTER I

INTRODUCTION

Throughout the history of art, creative advances have

paralleled technological advances in media, permitting the

artist to illuminate new visual concepts. A unique feature of

today's technology is that the digital medium encourages the

synthesis of knowledge from different disciplines rather than

its fragmentation [1]. As the boundaries between disciplines blur, information previously restricted to the domain of one

discipline is available to other disciplines. Such a situation

has arisen in the merger of scientific simulation and artistic

animation. New simulation techniques which use properties of

kinematics and dynamics have application in animation. These

applications can prompt significant changes in the field of

animation.

Though implementation of physically-based motion simula­ tion as a routine animation tool is still in the future, gen­

eral structures and procedures can be anticipated.

It seems important to obtain an idea of this future in­ strumentation as early as possible since, seen from a future perspective, the activities occurring in this

1 2

field today can be regarded as paving the way for fu­ ture forms of expression [2].

Using physical laws to generate animated motion is not a new idea. Disney animators observed mechanical systems directly to obtain more life-like realism in their animations.

This study outlines a functional foundation for artists to use different simulation systems in the design of a computer anima­ tion model.

1.1 Importance of Simulation as Animation

Computer animation today is based on the traditional approach to creating imagery (drawing each frame by hand). All key-frame positioning is done individually by the animator on the computer. The computer assists in generating the needed frames "in between" these key-frames. This is not only tedious and complicated, but it is also limiting. It conditions the animator to think in terms of creating what a key-frame system can reasonably handle. This mind-set results in a tendency to dwell on 2D visual possibilities and character qualities that mimic traditional methods of generating animation.

Today, the physical, psychological and economic limita­ tions of time simply do not allow for more complexity in tradi­ tional animation. Attempts have been made to increase complex­ ity and subtlety through technological techniques such as 3

rotoscoping. However rotoscoping tends to inhibit the creative

initiative of the animator. Even with the computer the tern-

i poral problems of creating animation continue to exist: (a)

generation of realistic motion is difficult, and (b) animating

large collections of objects or figures that appear to interact

is very complex and generally avoided. As a result, animation

in general, and computer animation specifically, suffers from

an absence of intriguing and expressive motion. This stems

from the fact that certain phenomena are too visually or tem­

porally complex to be adequately reproduced by the artist's

visual skills alone. What is needed is a method for generating

motion which is analogous to how motion is generated in the

real world. "To achieve the degree of realism found in other

areas of computer graphics, the motion of objects must be simu­

lated by the physical principles of dynamics governing the

motion [3]."

The most viable alternative is the adaptation of computer

"simulation" techniques to an animation system. Computer simu­

lations offer the hope of creating significant complex motion

through the incorporation of behaviors of an object as it

responds to its environment. Simulation techniques expand the boundaries of the visual process so that the physical limita­

tions of an animator's time or the complexity of the idea does

not have an overbearing influence on the creative outcome. 4

Computer animations utilizing simulation techniques can

generate realistic motion through the use of dynamics laws or

kinematic descriptions. Newton's dynamic laws of motion permit

the simulation of large collections of objects that interact

without requiring the animator to explicitly specify the

motion. This in turn influences the creation of more subtle

and realistic animations for sophisticated viewers.

From these new simulation/animation hybrid systems new

operans will evolve that tap the unexplored realms of the com­

puter medium. With these new simulation techniques will come

parameters and constraints which, while allowing an animator to

exploit the system for new unimagined creative possibilities, may also encumber the artist.

The term "simulation", itself, rather than "animation", denotes a shift in control from the animator to the underlying physics of the environment. One would like a system for specifying motion which combines the real­ ism of dynamic simulation without removing control from the animator [4].

This study specifies a functional model that encompasses the

implementation of physically-based simulation as an animation

technique for the artist.

1.2 Purpose of Study

The study develops a functional model for a simulation- based animation system and examines its aesthetic implications. 5

The study analyzes the similarities and differences among motion simulation systems. The proposed model consists of methods for generating motion using temporal simulations of physically-based phenomena. Its use is proposed together with user modification. Specifically, this study demonstrates that simulation methods can be useful tools for animators in the management of complex motion. The aesthetic implications of the model's use are discussed and analyzed as well.

The study considers which elements of computer simulation/animation hold the greatest promise for expressive control by the animator. As Youngblood states, "Digital scene simulation is by far the most awesome and profound development in the history of symbolic discourse, its aesthetic implica­ tions are staggering [5]." The animation process and potential for motion generated through simulation methods receive atten­ tion throughout this study.

It is hypothesized that the re-creation of motion in the computer has far-reaching ramifications for the animator. In an effort to arrive at a functional mode this study addresses the following questions: a. How should such a physically-based simulation system be

structured for its use in animation? b. What interface features would be needed in such a

system? 6 c. What are the advantages and disadvantages of this

system? d. How does the proposed system extend the existing

means or create new ends for the animator?

The study then addresses the following questions: a. How can the computer be used to generate convincing motion

for the observer without sacrificing the expressive input

and control of the animator? b. What are the implications of physically-based processes of

generating motion for animation? c. Can a functional model be designed which allows for

heuristic techniques?

The new interactive capabilities of the computer actually encourages a trial and error, creative approach. This allows the artist to experiment and search for original, unique, and innovative possibilities [6]. This approach might best be exemplified by the question "What would happen if....?" In the future, combining heuristic techniques and common sense knowledge will change both the production and creative process.

As details of scene and motion execution are managed more and more by the computer, the creative efforts of the artist will focus more on the subtleties of design. The job of "animator" will shift to the job of "director" of animation. 7

1.3 Overview of Study

Chapter II of this study presents the research methodology employed. It includes the statement of the problem, delimita­ tions, basic assumptions, and definitions. Because this study relies upon work from several other fields, relevant informa­ tion from these divergent fields is presented in the next two chapters (Chapters III and IV). Chapter III reviews the issue of complexity and traditional attempts at resolving the prob­ lem. This includes relevant technical issues of process that have advanced the art form { e.g. animation, systems hierarchy levels, applications) and relevant background on simulation

(i.e. Laws of Motion).

Chapter IV reviews the specific issues of motion and their resultant implemented systems that are leading to this inter­ disciplinary crossover of simulation and animation in the com­ puter. This chapter includes physical behaviors (e.g. flexible objects, articulated structures) and behaviors as the results of external forces.

Chapter V contains the main result of the study, which is the functional model for motion simulation in animation. The model proposes a framework which links together physical attri­ butes and functionality. These attributes include structural primitives, mechanical attributes, functional procedures, and the necessary interactivity requirements associated with the 8 model.

Chapter VI assesses the aesthetic implications of this model in areas of fidelity, utility, and its emergent proper­ ties. The creative process is addressed under utility. Also covered are the creative implications in regard to criticism of the model and future directions. Chapter VII is the summary of the study. 9

NOTES TO CHAPTER I

1. Charles Csuri (1974). "Computer Graphics and Art," Proceedings of the IEEE, 62, (4), p. 514.

2. H. W. Franke (1986), In Lucas, R. E. (1986). Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study, Unpublished master's thesis, Ohio State University, Columbus, OH. p. 24.

3. Paul M. Issacs & M. F. Cohen (1987). "Controlling Dynamic Simulations with Kinematic Constraints, Behavior Functions and Inverse Dynamics," ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (3), p. 215.

4. Ibid., p. 216.

5. Rick E. Lucas (1986). Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study, Unpublished master's thesis, Ohio State University, Columbus, OH. p. 24.

6. Gary Demos, Maxine D. Brown, & R. A. Weinberg (1984). "Digital Scene Simulation: The Synergy of Computer Tech­ nology and Human Creativity," Proceedings of the IEEE, 72, (1), p. 23. CHAPTER II

METHODOLOGY

Pierce and Steiner Maccia [1] recognize that there are three main stages of inquiry: retroduction, deduction, and induction. Steiner Maccia outlines Pierce's distinctions between these three stages of inquiry. "Through retroduction, one devises characterizations — statements or theory [sic] about objects. Through deduction, one clarifies and completes such characterizations. Finally, through induction, one deter­ mines the objects falling within the range of the characteriza­ tion [2]

It is through retroduction that new concepts can arise.

Retroduction quite often results from a "surprising phenomenon, some experience which disappoints an expectation, or breaks in upon some habit of expectation... [3]."

The identification of the problem of this study was the result of retroductive inquiry through the author's interest in the culturally-accepted dichotomy between the subjective

(artistic, aesthetic) and objective (scientific visualization, analytical) points of view. These diverse points of view merge

10 11 in the computer, as all relevant data must be reduced to the same level of information(i.e. bits). Linehan [4] has indi­ cated that retroduction is the stage of inquiry most applicable to this type of research, as the discovery of a problem is not always attained by a direct route.

Overlapping areas of inquiry from the pursuit of two apparently dissimilar disciplines lead to "that stage which describes the characterization of the problem in a functional model form [5]." In his book Insight and Outlook, Arthur

Koestler refers to this process as "bi-sociation":

...the creative originality of this matchmaking bi- sociation is not apparent in the smooth syllogistic scheme. The scheme gives the impression that the mental achievement consisted of drawing the conclusions. In fact the achievement was to bring two premises under one roof, as it were. The conclusion is merely the offspring of the marriage, arrived at by routine ac­ tions. In other words, syllogism and deductive reason­ ing are not the method of creative thought, they merely serve as its formal justification after the act {and as a scheme for repeating the process by analogy after the original bi-sociation of the two fields in which the premises are representatively located). The solutions of problems are not "invented" or "deducted" — they are "found"; they "occur" [6].

In a desire to understand a problem from an interdisciplinary perspective a systems view of the problem provides a framework for the model. This systems view emphasizes the connections among the various parts that constitute a whole.

"A system may be defined as a collection of interacting elements that function together for some purpose [7]." In this study the model of a computer animation system is viewed as 12 encompassing the disciplines of animation, computer science, physics, and visual methodologies. This integration from vari­ ous disciplines is one of the major challenges of the systems approach [8], The essence of the study's approach requires components to work together to perform some function. This requires the capability for simplifying and ordering the infor­ mation, which is fundamental to the study's problem.

2.1 Inception of the Problem

An interest in the artistic as well as the computational process resulted in the author finding himself in computer graphic research environments. People working in these environments focus their efforts on problems of image and motion generation. In such environments artists and scientists coexist. In these computer animation research environments scientists are working on the problem of motion and its development direction as dictated by a strictly naturalistic point of reference. The "problematic situation" which provides the basis for this study grew out of the author's interest in the implications (creative and procedural) of these computer animation processes.

Following this initial path of inquiry, the author researches related bodies of literature concerning simulation of physical phenomenon and its principles, utilizing the creative process and creative potential as references for this investigation. Computer programs developed at The Ohio State 13

University, New York Institute of Technology, California Insti­

tute of Technology, University of Pennsylvania at Philadelphia,

PIXAR, and others have formed a starting point for a model.

(This work is described in Chapter IV). The simultaneous study

of literature on animation and simulation (traditional and com­ puter) and the aesthetic process set the stage for the incep­ tion of this study.

2.2 The Statement of the Problem

This study investigates the integration of simulation techniques in computer animation. An initial question was,

"How can naturalistic motion and complexity be created in 3D

computer animation?" This question, however, did not serve to establish a methodology which might attend to aesthetic expres­ siveness. In addition, this area is currently being researched and explored by computer scientists who do not have an interest in the use of simulation as an expressive animation tool

(Barr [9], BadlerflO]), as well as by scientists on the Ohio

State University campus who are interested in this one area alone (Chadwick, Parent[ll], Haumann[12], Girard[13]).

The more appropriate question for this study is: "Can a model be created which integrates the realism of motion simula­ tion with techniques for expressive control by the animator?"

The new development of computer techniques which make possible the integration of interactive real-time feedback, robotics techniques, and applicable mathematical techniques each provide 14 impetus for this timely study. The theoretical application of accurate physical simulation methods to the animator's manage­ ment of complexity and expressive control is the goal of this functional model.

The proposed model is not presented as the model for all of artistic computer animation, but is representative of the procedures needed for a new creative computer animation. The term model can be used in various ways. It is used in this study as specifying "certain relationships among elements of a process in such a way that new relationships or propositions can be generated from the model itself [14]." The model is not necessarily comprehensive, but rather is a starting point for problem identification, analysis, and synthesis of the informa­ tion uncovered.

2.3 Criteria for Assessing the Model

The chief criteria for assessing the functional model are its accuracy in producing what it is directed to do and its ability to increase both the complexity and credibility of motion for the viewer. The test for accuracy is in the area of fidelity. The model is held accountable for fidelity to the degree that it realistically reproduces motion. The second area for assessment of the model is utility (its usefulness).

The model's utility is assessed for its ability to provide a responsive interface for input and control by the animator in the management of complex animation. The use of fidelity and 15 utility as criteria for evaluating functional models is in keeping with standard practice as outlined in Communication

Theory and the Use of the New Media by Erwin P . Bettinghaus [15] .

2.3.1 Fidelity

The fidelity of the model is examined in relation to its ability to accurately emulate naturalistic motion. While the model does not claim a direct emulation of nature, it does require that a convincing correspondence be constructed within the computer medium. While the model should realistically reproduce motion it should not force the artist to use mechan­ isms or strategies that would be intrusive or inappropriate.

2.3.2 Utility

The utility of the model is assessed for its ability to increase the user's control of complexity using highly interac­ tive and expressive interface methods. Specifically, it must address those concerns of interactivity, extensibility, habita­ bility, and generality. An ideal interactive system would pro­ vide immediate real-time playback of full shaded images.

Extensibility includes facilities for reconfiguring existing mechanisms or including new ones as the need arises. Habita­ bility is based on the number of necessary features contribut­ ing to how friendly the system will be. Generality is con­ cerned with the levels of abstract control that can be assessed by the user. For example, can the user use both explicit 16

(guiding) and implicit (task-oriented) levels of control.

2.4 Delimitation

For the purpose of this investigation, the following limi­ tations are stated: This study is limited to the design of a functional model for motion simulation in computer animation.

The simulation of light propagation and its visual qualities

(e.g. color, surface properties) are considered to be another study in their own right. The functional model identifies potential application. A practicing level is presented by way of example.

This study is not concerned with individual techniques, but with the schematic notion of simulation as animation. The focus of this study is only on the new possibilities of expres­ sion provided by computer motion simulation. The model con­ structed in this study addresses motion complexities and credi­ bility and does not attempt to address other issues of image synthesis.

This model is tested using fidelity and utility as cri­ teria. The principal arguments offered in support of the model are drawn from art and computer literature. Accepted criteria from the fields of computer science and animation provide the basis for fidelity and utility assessment. This criteria is offered as evidence in support of selected features of the pro­ posed model. The rapidly evolving field of computer animation and the emerging nature of the arguments add reservations to 17 the study's findings.

The design of the model and its empirical validation are seen as two separate studies by this author. It must be noted that the integrated functional capabilities of the model do not currently exist. Separate isolated programs and components do exist. One of the study's main tasks is to describe how that integration of this component could evolve into a functional model. The identification of the problems, the generation of a functional model to address those problems, and the communica­ tion of that model to colleagues are seen to be significant and valuable research efforts in themselves.

2.5 Basic Assumptions

Animators generally strive to create original works that are many times identified as idiosyncratic to the artist/animator. Within this individual process there are universal communicable factors which viewers will respond to in a known way. Animators utilize these factors in their work, and computer artists will use similar simulation factors highlighted in this study to assist them in their work. It is assumed that the task of computer animation and its attendant problems have sufficient universality to promote visual commun­ ication. It is also assumed that the reader is familiar with three-dimensional computer graphics and its basic methods of representation. 18

2,6 Definitions

Animation: Individual images shown in rapid succession which produce the illusion of motion. Though there are many different types of animation, the term is identified in this culture's vernacular as "cartoons” . The essence of animation is "movement".

Behavior: Action or reaction that is observable. Behavior can apply to numerous levels of possible interaction. For instance, in computer graphics, behavior can be associated with a single force acting on a specific point or structure.

Behavior at its lowest level may be construed as a downward force applied to a mass {i.e. gravity).

Constraint: A restriction of the position, orientation, or physical state of an object [16].

Dynamics: The principles of motion of physical bodies based on Newton's laws and Euler's laws. Dynamics deals with the forces which cause motion, as opposed to kinematics, which deals with its geometric description.

Forward dynamics: The problem of computing the motion of a collection of objects, given the forces and torques acting on the objects. The solution is found by the application of the dynamic laws of motions [17].

Forward kinematics: The problem of computing the position of objects, given the rotational or translational 19 transformations. The solution is the straightforward applica­ tion of linear algebra [18].

Inverse dynamics: The problem of computing the forces and torques that would act on an object in order to produce the desired motion [19].

Inverse kinematics: The problem of computing the rota­ tional and translational transformations to apply to an object in order to produce a desired motion [20],

Key-frame: A key-frame is a frame in an animation that contains significant information about a change in the action or visuals. From two key-frames, inbetween frames can be cal­ culated or drawn which maintain the action or visual change specified by the key-frames.

Kinematics: The science of abstract motion without regard to forces or bodies of matter.

Functional Model: It is "not necessarily descriptive, but it specifies certain relationships among elements of a process in such a way that new relationships or propositions can be generated from the model itself [21]."

Newton's laws of motion: Basic laws of classical motion that are encapsulated in the expression F=ma (Force equals mass times acceleration). (Please see Chapter 3) [22]. 20

Physically-based modeling: "A mathematical representation

of an object (or of its behavior) which incorporates forces,

torques, energies, and other attributes of Newtonian physics

[23]

Primitive: A basic building block with which other objects

or entities are constructed.

Simulation: Simulation is the act of representing some

aspects of the real world by numbers or symbols which may be manipulated to facilitate their study [24]. It is the process of using numerical techniques to solve equations embodying natural law, so as to yield behavior of a computer model that is analogous to the behavior that the real-world objects being modeled would exhibit [25]. 21

NOTES FOR CHAPTER II

1. Elizabeth Steiner Maccia (1962). "Ways of Inquiring," Bureau of Educational Research and Service; Occasional Papers. Columbus, OH: The Ohio State University, p. 1.

2. Ibid., p. 12.

3. Ibid.

4. Thomas E. Linehan (1981) . A Computer-Mediated Model for Visual Preference Research and Implications for Instruc­ tion in Art Criticism, Doctoral dissertation, Ohio State University, p.50.

5. Arthur Koestler (1949). Insight and Outlook. New York: Macmillian & Company, pp. 254-255.

6. Ibid.

7. N. Roberts, D. Anderson, R. Deal, M. Garet & W. Shaffer (1983). Introduction to Computer Simulation, A System Dynamics Modeling Approach, New York: Addison-Wesley, p. 50.

8. Ibid., p. 51.

9. A. H. Barr (1988). Tutorial Notes, Physically-Based Modeling, SIGGRAPH '88,

10. Norm I. Badler (1984). "What is Required for Effective Human Figure Animation?,11 Proceedings Graphics Interface '-8_4, Ottawa, Ontario, pp. 119-120.

11. John Chadwick and Richard Parent (1988). "Critter Con­ struction: Developing Characters for Computer Animation," PIXIM Conference Proceedings. p. 1.

12. Dave R. Haumann (1987). "Modeling the Physical Behavior of Flexible Objects," SIGGRAPH ' 87 Tutorial Notes No. 17, Topics in Physically-Based Modeling.

13. Michael Girard (1986) . Interactive Design of 3^-D Computer-Animated Legged Animal Motion. University of North Carolina Motion Workshop, Chapel Hill, NC., p. 6.

14. Erwin P. Bettinghaus (1968). "Communication Theory and the Use of the New Media," Theory for the New Media in Educa­ tion . Educational Proceedings Series, #1, August 1968, Theory for the New Media in Education, Educational 22

Proceedings Series, Number 1, East Lansing: Michigan State University, p. 89-124.

15. Ibid.

16. Ibid.

17. A. H. Barr (1988). "Glossary," Tutorial Notes, Physically-Based Modeling, SIGGRAPH '88, p. E4.

18. Ibid.

19. Ibid.

20. Ibid.

21. Bettinghaus, op. cit., p. 90.

22. Barr, op. cit., p. E5.

23. Ibid., p. E6.

24. A. M. Collella, M. J. O'Sullivan and D. J. Carlino (1974). Systems Simulation, Washington, D.C.: Lexington Books. 25. Ibid. CHAPTER III

REVIEW OF LITERATURE - ART, ANIMATION, AND SIMULATION

There are several areas of literature that relate to this interdisciplinary topic of computer animation. It is essential to understand previous responses to temporal and scientific factors in art and animation. This chapter reviews art, anima­ tion, and simulation literature which has relevance to the problem stated in Chapter One.

There are several recent historical links which have relevance for this study. With the introduction of mechanical systems of visual reproduction - perspective and the camera obscura - the artist was now able to create an illusion compar­ able to the way our eyes perceive stimuli.

Since perspective had only untangled the problem of form and not of movement, naturalistic artists continued to search for some method of giving expression to the temporal dimension,

"time". Movement is a critical element in our perceptual iden­ tification, knowledge, and exploration of our world. With the

20th century came two avenues in art for this exploration into this dimension of motion: (a)explorations in representing

23 24

temporal effects in traditional art media {e.g. static paint­

ings [cubism, futurism] and sculpture), and (b)film (anima­

tion) .

3.1 Traditional art media

Cubism utilized temporal means that enabled the artist to

synthesize information about our perceptions in the environ­ ment . The focus of Cubism was directed toward a simultaneity,

a juxtaposition of several views of an object. Time, rather

than motion, was the directed focus of this simultaneity con­

cept. Cubistic time is structured to permit the viewer to form

an idea about an object or an environment's total dimensional­

ity. It is ironic that the cubist's concept of simultaneity

actually was first proposed in a Futurist's exhibition catalog in 1912 [1].

Futurist's theory moved away from the cubist's breaking up

of an object and the unfolding of an object's parts on a flat

surface (Figure 1) to symbolize the fourth dimension [2].

Futurist's theory embodied motion in time as essential elements of its interpretation. Some of their first efforts (e.g.

Kupka, Duchamp [Figure 2], Balia) entailed analysis of linear

sequential motion. Boccioni, however, probably distinguished himself in the Futurist movement for his scientific interest in the fourth dimension and his advocacy of the "concept of dynamic continuity" [3]. Figure 1. Kazimir Malevich, "The Knife Grinder". (1912). Oil on canvas, Yale University Art Gallery.

Boccioni rejected the concept of sharp differentiation of edges and searched for forms which would give a continuity in space. This is evidenced in his famous sculpture "Unique Forms of Continuity in Space" {Figure 3). Boccioni's goal in this piece was a synthetic depiction of motion, a "synthetic con­ tinuity" as opposed to the "analytical discontinuity" of Kupka and Duchamp, or the Cubist's "simultaneity". As the Futurist art movement wained, interest in motion and the fourth 26

Figure 2. Marcel Duchamp, "Nude Descending a Staircase No. 2". (1912). Oil on canvas. Philadelphia Museum of Art: Louise and Walter Arensberg Collection. dimension migrated into two separate directions. One direction focused on the widespread, early 20th century belief that the fourth dimension is a space dimension; this resulted in artifacts produced by artists such as Piet Mondrian and Theo van Doesburg [4]. The other direction focused on the temporal elements exemplified by the new medium of film and animation. 27

Figure 3. Umberto Boccioni, "Unique Forms of Continuity in Space". (1913). Bronze, 43 7/8 x 34 7/8 x 15 3/4". Collection, The Museum of Modern Art, New York. Acquired through the Lillie P. Bliss Bequest.

When static-artists desire to recreate the illusion of life or motion, they generally isolate a single momentary action which suggests a subject caught in the act of motion.

On the other hand the animator uses a series of drawings which actually appears to possess the quality of motion in time when projected through the film medium. The animator must see the whole, in its "continuity" [5]. PLEASE NOTE:

Copyrighted materials in this document have not been filmed at the request of the author. They are available for consultation, however, in the author's university library.

These consist of pages:

Walt Disney illustrations, figures 4-10 Pages: 28, 30, 32, 42, 44, 50, 51

UMI 28

3.2 Animation - the Beginnings

The first stop-motion picture made in America, The "Humpty

Dumpty Circus", approximately 1898, used wooden circus perform­ ers and animals. The wooden joints enabled James Blackton and

Albert Smith to place them in various positions. "It was a tedious process in as much as the movement could be achieved only by separately photographing each change of position [6]."

This "trickfilm" genre capitalized on the selective recording properties of the camera. In early 1906, Blackton made

"Humorous Phases of Funny Faces." It was during that film that

Blackton realized that if objects could be animated, why not 29

drawings [7]. In this film, caricatures of a man's and a

woman's face are drawn on a blackboard (Figure 3).

When Blackton takes his hand away the faces roll their eyes, but there is little other actual movement. In­ stead, details are added on top of each other between exposures. When the man's cigar billows a cloud of smoke that obliterates the woman, Blackton erases the board and begins anew. This time the outline of a gen­ tleman with a bowler and umbrella draws itself, and the man doffs his hat [8].

Though Blackton's film was basically an assembly of isolated

effects, it was the first to demonstrate the concept of anima­

tion [9] .

Winsor McCay's animated feature "Gertie" in 1908 is

credited as the first animated cartoon, more so for the impact

it had than for it being the first "true" animation (Figure 5).

McCay wrote that "While these films made a big hit, the theatre

patrons suspected some trick with wires. Not until I drew

'Gertie, the Dinosaur' did the audience understand that I was

making the drawings move [10]." Maltin clarifies that most

movie audiences were naive and still trying to accustom them­

selves to the idea of motion pictures per se [11]. The tech­

nique of animation is obvious to us now in hindsight, but it

was an inconceivable concept to the general public of that

time. Similar new insights - which are at this time inconceiv­

able to us - should also evolve in the "magic" of computer ani­ mation. It is significant to note that there was approximately

a decade between the first stop-motion picture and the first

acknowledged animated feature which evolved from that stop- Figure 5. Winsor McCay, "Gertie the Dinosaur". From OF MICE AND MAGIC: A HISTORY OF AMERICAN ANIMATED CARTOONS by Leonard Maltin. Copyright 1980, 1981 by Leonard Maltin. Reprinted by Arrangement with New American Library, A Division of Penguin Books USA Inc. motion picture technique.

3.3 Animation - Early Artistic Directions

Early American animated films were influenced by the popu­ lar traditions of comic strips and the theater. These genres characterize the early narrative structures, gags and story material of animation [12]. Especially strong bonds with the 31 popular press were forged [13]. It is important to note that any new medium is influenced by the prevalent paradigms; anima­ tion was no exception.

Early subjects for European animation were derived from folklore, and drawing styles emerged from conventional graphic arts of the period [14] . The European animation community was unavoidably affected by the devastating effects of war. This condition colored their style and outlook. The effects were dramatically different and as a result their work differs from

America's style. The independent nature of European animation that evolved is exemplified by the early works of Wladyslaw

Starevitch and Lotte Reiniger whose efforts were concentrated in two fields of animation which were relatively unrealized in

American animation. Starevitch promoted social themes through the use of puppets which were corpses of bugs, birds and animals [15]. This lack of cuteness contrasts with the work of

American studios such as Disney. Lotte Reiniger developed

"two-dimensional shadow puppets to visualize fairy stories in the tradition of the 'ombres chinoises' of the eighteenth and nineteenth centuries [16] ."

Although the creators of the first animated films were not surrealists and were probably not even cognizant of that move­ ment, they inadvertently made films that demonstrated a similar disregard for everyday existence, normal logic, and causality.

They instead had a propensity for dreamlike action, which Andre

Breton (a noted film theoriest) and his followers admired. Most animated films begin by "establishing an alien universe" into which the spectator may project himself. This affinity between cartoons and the emerging antirealist aesthetic of the

1920's was first voiced in a 1925 article written by Gus Bofa, a French caricaturist [17]. "Bofa promised that the cartoon would succeed in the elusive quest of pure cinematic rhythm, and claimed that film gave man the power finally to create something new in the world (underline added) [18]This new departure was first exemplified by "Felix the Cat," whose films first appeared in France in 1927 (Figure 6). The cat was almost immediately linked with surrealism [19]. 33

Brion promoted Messmer's animal as a sur-chat, a super cat of mythology far removed from the reality of every­ day cats. Although Felix existed only on paper and celluloid, he paradoxically has more personality than "real" movie stars. His removable tail expressed the two creative faculties of the mind, surprise and cu­ riosity, by forming itself into punctuation marks! and ?. In Felix' s self-created world the lack of boun­ daries between real and imagined objects produced an oneirid state; Brion wrote that 'this creative power of the dream and this surrealist formation of the object give to these fantastic images the means by which the mind enjoys free play' [20].

Europeans enthusiastically responded to the surrealist poten­

tial in Winsor McCay's first experiments in which the animator

was transported into the picture and carried away by his

dinosaur, Gertie.

Films whose origin lay in the art world were continued by

graphic artists such as

Viking Eggerling, Fernand Leger, Francis Picabia, and Moholy-Nagy, who believed in the abstract, all attempt­ ed to make animated films. Oscar Fischinger began to make his abstract mobile patterns to music in 1931, which were followed by the experimental color abstract films that Len Lye initiated in Britain in 1933... [21] .

Though only sporadic films were made and their styles differed

sharply these artists derived their strength from the various

contemporary styles in modern art. Halas [22] asserts that their work has survived better than any but the very best

live-action films of the period. Halas is of the opinion that

Disney animators have all too often been prepared to sacrifice imaginative caricature (i.e. graphical qualities) for natural­ ism of style and movement in their characters [23]. 34

3.4 Media Artifacts

There are certain universal realities working within any medium. These can generally be found in the various manifestos

expressed by artists about their work. For example, Arnheim

[24] put forth the proposition that a medium can only be art as

it differs from a true rendering of reality. If there is no

opportunity for an artist to manipulate the medium, the artist

would be merely re-presenting reality, resulting in the focus­

ing of an audience's attention on what is represented, not on

the means of such a representation. Aesthetic values are

directed at the means rather than the object. Arnheim advo­

cates looking at the limitations of the medium. The limita­ tions (differences from reality) are the means by which the medium is recognized. Every change from the natural world is a

gain for potential aesthetics [25].

The inherent attributes of a medium are just those charac­ teristics that have direct aesthetic application. Examples of

"medium" attributes that have evolved in film and which have no real association in reality include: fast and slow motion, fades, dissolves, superimpositions, backward motion, use of a still in that sequence, and distortions through focus and filter [26].

For Arnheim, every medium, when used for artistic pur­ poses, draws attention away from the object which the medium conveys and focuses it on the characteristics of the medium itself. Furthermore, every medium proceeds by means of a central sensory nexus: music is the medi­ um of sound, dance of gesture, poetry of words, and so 35

on. The nexus becomes a symbolic language to be mani­ pulated by the artist, and the artist must learn to or­ ganize this physical material so that his vision or idea shines through. Even if he wants to duplicate as­ pects of the physical world, he must have command of his medium so that he can successfully translate his perception of the world into the proper codes of his medium [27].

This is no different than a painter who through the use of

learned techniques applies the proper blotches of paint in the

right order so people will marvel at the "realness" of the

work. An aesthetic idea evolves within the perceptual conven­

tions of a medium.

3.4.1 Media Transformations

This evolution within a medium results in a transforma­

tion . There are five basic transformations that can take place

in a representational image: graphic, perceptual, structural,

expressive, and symbolic [28] . This study focuses on the first

two transformations (graphic and perceptual). These transfor­ mations pertain to the veneer of the object as opposed to the

meaning behind the object's physical appearance, which is embo­

died in the other three transformations. Whether an image is

reproduced as a painting or displayed (digitized) on the com­ puter screen, a graphic transformation occurs.

An artistic image is never just a straight forward repro­

duction of the object. Even the computer and the digital encoding process itself ensures a basic transformation that no

artist can avoid; the image is translated into discrete points 36 called pixels. This is an artifact of the medium. In recreat­ ing the object by means of pixels, the concrete reality of our world is lost. The object in reality is a physical continuum occupying a discrete position in a contextual volume of space

[29]. It is not a collection of discrete points of light emu­ lating from a display monitor.

A three-dimensional, computer-generated image, employing many of the advances in image synthesis may embody a transfor­ mation of a higher order. The perceptual analysis that is foundational to the image synthesis capability of the computer can produce an image with a larger-than-life credibility. When this happens the image represents a perceptual transformation for the viewer.

Collier [30] makes the point that most of us perceive very general impressions when regarding an object in the ordinary course of our lives. An interrupted series of retinal sensa­ tions are received as we move or blink. The eyes and mind are rarely focused intently enough to consolidate these general impressions into a coherent, unified mental image or motion, allowing its formation to be indelibly imprinted on our mind in every particular. Consequently, when artists like Durer and

Leonardo da Vinci or scientists like Barr [31] and Girard [32] present us with such a completely "rendered" statement - with the phenomena now isolated within a defined display area and abstracted out of its natural (and distracting) surroundings - the image is seen as if for the first time [33] . This new 37

image overcomes the vagueness inherent in normal visual sensa­

tion to show us, so to speak, a new experience - a perceptual

transformation.

3.5 Mimetic Tradition in Art

There are those who would argue that the "mimetic" or

"perceptual" aspects of art are passe. This argument proposes

that although naturalism was once an artistic necessity - a

concession to its historical patrons (e.g. the state, the

church, the public) - this is no longer the case. The move to

abstraction away from mimetic preoccupation constitutes the

radical innovative essence of modern art [34]. However, it is

ironic that much of the significance of modern art is derived

from the artist's new emphasis on purely perceptual phenomena,

such as the use of color contrasts, ambiguities in figure-

ground relationships, illusions, and other expressions of a

profoundly sensed underlying perceptual logic [35,36]. A

predominant characteristic of modern art has been what can be

labeled its perceptual nature, its involvement in conceptual

issues, as well as perceptual ones.

Today, the outward manifestation of perceptual art is dis­

tinctly different from art in the past. These new manifesta­

tions have been nurtured by influences of modern science that have characterized much of the nineteenth and now the twentieth century. In fact, the whole era of modern art owes it existence to the technology of the camera, which relieved 38 artists from their role as recorders of visual history. The subject of science and its affect on the visual arts is easily a study in itself. There is a long and definitive influence of science on art [37], and this study extends that scientific influence within the realm of computationally derived motion and the mimetic tradition.

Art, even today, continues to be deeply involved in representation. This representation is of a new kind, however

[In this century] ...the artist has been representing either more specialized aspects of perception, such as figure-ground, illusions, cues to depth perception, color contrasts, or that the artist has been represent­ ing the nature of the elements, structures, and processes in the visual system with which we construct the perceptual world [38].

The act of perception and a work of art both depend on the organizing abilities of the mind. For Arnheim these abilities are supported by a world which seems to lend itself to certain kinds of organization. Arnheim [39] qualifies art as the organization not of a specific field of sensory data, but of a general pattern applicable beyond itself. In a representional painting artists see their subjects and then expressively organizes them. "A cubist painting of a building, for example, is a transformation not of the building but of the artist's particular mode of organizing certain kinds of perception

[40]." 39

3.6 Animation - Timing

Early animation began with visual jokes that moved along at an even tempo, and the German silhouette films ambled about at the normal tempo of stage marionettes [41].

Timing in early cartoons was limited mainly to fast moves and slow moves, with accents and thrusts calling for special handling. The personalities that were developing were defined more by their movements than their appearance, and the varying speed of those move­ ments determined whether the character was lethargic, excited, nervous, relaxed [42].

Max Fleischer (creator of KoKo the Clown) realized that an ani­ mator could originate special forms of timing suitable to the unreal world of the cartoon. When timing became a variable in itself, animated film took another step forward in its evolu­ tion as a medium. The new timing variations gave a unique comic punctuation to the story and substantially helped to dramatize it. The comic or climatic application of speed to the action, or the sudden arresting of movement altogether, became characteristic of the free treatment of the timing of movement in the cartoon film [43]. This freedom became exem­ plified by the comic actions of Felix the Cat. It was quite natural for Felix to be able to take off his tail, use it as a handle and then put it back on his body. This free and ima­ ginative approach to visual gags, as well as the new freedom of tempo, made the American cartoon a distinctive art form from the animated folklore of Europe [44]. 40

Timing gives meaning to movement and movement is the essence of the animation medium. Timing is actually two con­ cepts: (a) the correct timing for a movement as objectively known through physical experience and, (b)the communication of a temporal idea to an audience. This study is concerned pri­ marily with the first concept, that of achieving accurate move­ ment through physically-based techniques and the aesthetic ram­ ifications of new suggested possibilities of invention and con­ tinuity that arise on account of those techniques. The second concept is summarized by Halas [45] as a communication process which entails (l)time spent preparing the audience for some­ thing to happen (i.e. anticipation), (2)then on the action itself, and (3)finally on the reaction to the action (i.e. fol­ low through). Proper timing is critical to the communication of qualities of an object or character.

To interpret communicative timing factors correctly depends upon an awareness of how the minds of the audience react to the animation medium (e.g. How quickly or how slowly do they react? How long will they take to assimilate an idea?

How soon will they get bored?). This requires a firm knowledge of how the human consciousness reacts to verbal and visual stimuli and its role in maintaining continuity [4 6]. For exam­ ple, the perception of image detail decreases strongly with increasing eccentricity in the visual field. This type of change also holds for the perception of motion in the film medium. 41

Determining the appropriate timing for an audience's per­

ception starts with "cause and effect". All animators {tradi­ tional and computer) who strive for naturalism in their work

consider the mechanical forces on objects or characters. These mechanical forces (Newton's laws of motion) contain the "cause

and effect" information necessary to reproduce naturalistic motion. These mechanical forces represent a way of producing

animation that involves modeling the physical laws that actu­

ally affect objects. Since most animation encorporates natur­

alistic motion, it is logical to incorporate simulations of physical phenomena which can aid in the attainment of that

goal.

Newton's laws of motion aid the animator in portraying the

feeling of "size" or "scale" which is determined by timing. A giant will move more slowly because he has more weight, more mass, and more inertia than a normal man. He will take addi­ tional time to get moving and once started, he will take more time to stop. In contrast a small character or object will zip quickly around an environment because of less inertia [47].

Timing is the quintessence of animation. It embodies the necessary capability to convince and manipulate not only the objects on the screen but the viewer as well.

3.7 Principles of Animation

It was in the Disney animation studio that the commonly held consensus of definitive animation was created. From this studio narrative procedures were isolated and labeled. New animators learned these communicative practices as rules of the craft [48,49]. These basic principles have become the guide­ lines from which naturalistic and narrative animation can assess its technical prowness.

Squash and Stretch is considered to be the most important principle [50,51]. The rigidity and mass of an object is defined by its surface motion during movement. A fixed shape displays a marked rigidity when it is moved about. A living entity, no matter how bony, evidences considerable movement within its form during an action. This extreme change can be seen in Pluto's head in Figure 7.

Through the mid-thirties, everyone was making two draw­ ings for every conceivable action, and by working back and forth between the squash position (flattened by great pressure and the stretch position (extended con­ dition) we found we could make each position stronger 43

in both action and drawing [52].

This concept is also an integral part of computer generated facial animation (discussed later in Chapter IV). It provides the clues to the flexibility of the skin and muscle as well as showing the relationship between the parts of the face.

The Anticipation principle prepares the audience and the character for the action about to take place. Lasseter [53] points out that an action occurs in three parts: the prepara­ tion for the action, the action proper, and the termination of the action. Anticipation can be construed to represent an ana­ tomical provision for an action. Since muscles in the body function through contraction, each muscle must be extended first before it can contract; an arm must be swung back before it can catapult a ball into the air [54].

Follow Through and Overlapping Action is the natural ter­ mination of an action, just as the principle of "Anticipation" was the preparation of an action. In early animated films a character often came to an immediate and sudden stop after entering a scene. Disney pointed out that "Things don't come to a stop all at once; first there's one part and then another

[55]." This principle of "Follow Through and Overlapping

Action" evolved from such observations (Figure 8).

Two notable techniques grew from this principle, the lead and the drag. In the movement of any object or figure, the actions of the parts are not simultaneous: some part must initiate the move, like the engine of a train [56]. In walk­ ing, the action starts with the hips. As the hip swings for­ ward, it sets a leg in motion. The hip "leads", the leg "fol­ lows". The fingers will follow the wrist in a hand gesture [57] .

Appendages on an object or character will move at a slower pace and drag behind the skeletal parts. As the character comes to a stop these appendages will continue to move. Thomas

[58] makes the point that the movement of each part must be timed carefully so it will have the correct feeling of weight, 45 and it must continue to follow through in a pattern of action that is believable, no matter how broadly it is cartooned. It gives a looseness and solidity to the figure that is vital to the feeling of life.

The way in which an action is completed often tells us more about the person than the drawings of the movement itself. The anticipation sets up the action we ex­ pect..., the action whizzes past, and now we come to the "punch line" of the gag, the follow through, which tells us what happened— how it all turned out. Amaz­ ingly, the ending was hardly ever developed in early animation [59].

Slow In and Slow Out pertains to the timing of animation frames in-between the extreme key-frames. This results in a natural subtlety of movement. Thomas [60] indicated that this concept evolved because...

Walt continued to ask us to analyze the actions more carefully, and to understand how the body worked, since that was the only way to get the caricature of realism he wanted. "Our work must have a foundation of fact in order to have sincerity." One animator from outside the studio was "amazed that anyone would be that interested in the mechanics of motion" but this unique approach was the very heart of our work [61] .

Arcs have been found to describe the natural visual path of a movement.

... the movements of most living creatures will follow a slightly circular path. Perhaps this has to do with weight or maybe with the inner structure of the higher forms of life, but, whatever the reason, most movements will describe an arc of some kind [62].

Extensive use of arcs results in animation that is smoother and 46 less stiff than a straight path. Graham [63] goes even as far to point out that even if a path resolves itself into a straight line then usually the object rotates as it moves along that path.

Exaggeration is not an isolated distortion of an action or form. The purpose of exaggeration is to make things more real- istic by emphasizing the communicable essence of a thing. The following recollection gives an idea of the problem:

There was some confusion among the animators when Walt (Disney) first asked for more realism and then criti­ cized the result because it was not exaggerated enough. When Walt (Disney) asked for realism, he wanted a cari­ cature of realism. An artist analyzed it correctly when he said, "I don't think he (Walt Disney) meant 'realism'. I think he meant something that was more convincing, that made a bigger contact with people, and he just said 'realism'..." Walt Disney would not accept anything that destroyed believability, but he seldom asked an animator to tame down an action if the idea was right for the scene [64].

What is real for the Disney studios is what the audience will accept as being real-. This principle strives to utilize those preconceptions possessed by all human beings. This was evi­ denced by individual reactions to pencils tests shown at the

SIGGRAPH '88 Electronic Theatre of PIXAR's new animation of a human baby and a toy drummer. The baby was modeled very real­ istically, however the baby was perceived as a "monster" baby because it did not meet our conception of a baby. Realism is a relative concept dependent on more than perception. The mental constructs possessed by individuals plays an extremely impor­ tant role in the final reaction to a character. By 47

exaggerating physical attributes or motions the animator can manipulate the mental precepts possessed by the audience to

assist in their understanding of an idea.

Pose to Pose and Straight-Ahead action are two contrast­

ing approaches in traditional animation. "Straight-Ahead"

action starts with the first animation frame and works straight

forward, one drawing after another. New ideas are generated in

response to what has previously been done as the animator progresses through the scene. The animator knows what has to be accomplished but has little plan as to how to accomplish it.

The results have a spontaneity and a freshness because the pro­

cess contributes to the directions of the work [65].

"Pose-to-Pose" animation revolves around a definite plan of action. The animator carefully correlates all relationships of the scene such as size, scale, timing etc. The extreme key-frames are constructed. This enables the animator to turn the detail work over to the in-betweener. The "in-betweener" is a person whose main function is to generate the needed

frames between the key frames. The advantages of the Pose-to-

Pose approach are clarity and solid continuity in the work

[66] .

Appeal is that quality that attracts the viewer to the work. For an animator this includes a wide range of properties that comprise the total scene: design, simplicity, communica­ tion, drawing and painting techniques or style. A counterpart 48 to this concept in computer animation would be the "display algorithm". This algorithm determines how light propagates through the environment and produces the final rendered image.

This rendered image can be aesthetically pleasing with a sophisticated display algorithm.

Lasseter [67] indicates that another principle,

Secondary Action, is the action that results directly from another action. It adds a realistic complexity to the anima­ tion through the imitation of interacting forces. Obviously, the "secondary action" is subordinate to the "main" action.

The distinguishing factor is that this action must come either

"before" or "after" the main action, or it will not be seen.

Facial expressions fall under this set of actions which rein­ force the narrative.

These principles of animation are only guidelines. It is still the responsibility of the animator to make the necessary decisions and do the animation. This job becomes all the more difficult when complex objects interact. In these situations the animator turns to "live action", rotoscoping.

3.7.1 Rotoscoping

In the history of animation accomplished conventional ani­ mators have studied live models, scrutinized the movement of objects and characters on film, and directly used still frames of motion pictures (rotoscoping). Rotoscoping allows compli­ cated movement to be created by using live-action as a base. 49

Live-action film footage is transferred to photographic prints which are then traced over to create the actual animation frames. Thomas and Johnson [68] attribute their success in animation to the observation of humans and animals over many long years. The computer can directly parallel the traditional animators use of rotoscoping. This is accomplished by either simulating the mechanics of movement or by directly digitizing the actual movements of a figure (motion recording). This con­ cept is discussed more fully under the heading of Guiding sys­ tems later in this chapter.

Regular animation develops from the animator's experience and imagination while rotoscoping utilizes the filming of actors performing in scenes planned for the animated charac­ ters .

The direct use of live action film has been part of the animation industry for years— as an aid to animation, a companion to animation, and even as a replacement for animation. From time to time, almost every studio has fallen back on a strip of live film to perfect a specific action animators were not able to capture [69] .

In Disney's film "Pinocchio", Stromboli locked Pinocchio in a birdcage that swung and bounced as it moved about in the wagon.

"This intricate object would have been almost impossible to draw in the first place, let alone capture the weight and con­ vincing movement in the action [70]." The solution was to have the animator trace the movement from a photostate of the swing­ ing cage (Figure 9). 50

At the Disney studio, filmed action of humans and an­ imals was used in many ways to do many jobs, and it led to some important discoveries. Live action could dom­ inate the animator, or it could reach him. It could stifle imagination, or inspire great.new ideas. It all depended on how live action was conceived and shot and used [71].

Thus, live action came to be used in two distinctive ways: (1) as a resource, giving ideas that assist in the development of a caricature, and (2) as a motion model for learning and imple­ menting motion that was too subtle to be picked up in direct 51 observation [72]. From a careful scrutiny of film footage it was discovered that "Some actions were so complicated they were impossible to draw in caricature, and many of the moves that gave touches of personality were too subtle to capture at all

[73]." This is significant, because it indicates untapped areas of animation where computer simulation could demonstrate its unrivaled capability to aid in animation goals.

Stylistic integration problems materialized if animators stayed too close or copied from the action filmed. The results looked strange, didn't seem real, or, more appropriately 52

phrased, lost the "illusion of life" [74]. Thomas [75]

hypothesized that though the motion appeared real, it possessed

a certain authority and presence which just didn't fit in the

fantasy world of "animatable" forms. The solution to using

rotoscoping was to translate those realistic forms into the

"animatable" forms being employed in the film, using the same

timing but utilizing a change of proportions (Figure 10). "The

essential qualities of animation start at the point where

live-action film-making stops [76]."

In using computer simulations as an animation base a simi­

lar realization may evolve. The camera records everything that

is there, with an impartial lack of emphasis. An artist, on

the other hand can selectively delineate visual forms for

emphasis, especially to areas which might not otherwise be per­

ceived by viewers. Therefore, the animator's drawings can be

closer to the "perceived" realism of an object because he can

be selective and personal in what he chooses to show [77].

Notably, it is inanimate objects which do not suffer from

this incompatibility (i.e. live-action imagery versus animated

forms) in the animated environment. Falling bottles, automo­ biles, bird cages("Pinocchio") could be traced and the results

would give excellent results compared to hand methods, if not

better [78] . In retrospect Thomas [79] feels that the film

"Cinderella", which depended heavily on rotoscoping during its

creation, lacks the creativity and inventiveness that comes out

of the traditional animation process. This over-reliance was 53 for the most part overcome in the animated film "Peter Pan" which was also rotoscoped. Getting away from the actual use of the live-action scenes was accomplished by using film as a starting point from which to build, invent, and enrich [80].

The recognized advantages and disadvantages of live-action pro­ vide a valuable point of reference as computer simulation is incorporated into animation. An insightful observation - and a creative dilemma - by Thomas [81] is that the more realisti­ cally animals are drawn, the less real they will appear on the screen. It is hypothesized that this is because the viewer is distracted by the discrepancies from the known reality.

The Disney staff drew what people imagine a certain animal to be like, what the personality of that animal would be. Only then would the audience accept the animal character as being real. "The big point is that characters on the screen appear to be most real when they can be animated to have personali­ ties, and this can only be done when there is potential for movement in all parts of the body [82]."

3.8 Animation - Computer

The creative potential of computer animation resides in its ability to avoid the limitations that are inherent in the generation of hand-drawn animation. Computer animation is a subset of the global concept of "Animation". Though computer animation has by its nature properties in common (it generates distinctly created frames that are shown in rapid sequence to 54 create an illusion of motion or temporal change) with the other subsets of animation (e.g. claymation, traditional hand-drawn animation). Computer animation also has a number unique pro­ perties (i.e. generation of inbetween frames by other than individual human efforts) that differ from the previously known techniques of animation. Computer animation by its nature cir-

I cumvents the inevitable stylistic qualities (e.g. simplicity and economy) that are forced upon the traditional animation medium by its physical realities. The benefit of computer ani­ mation will be that the animator's energy will no longer be invested in the physical act of drawing, but instead the focus will be on designing and directing. This study is appropriate at this time in view of the fact that the emphasis in computer graphics is switching from visual complexity - after twenty years of concentration in various phases of modeling - to pro­ cedural complexity.

Initially, computer programmers thought that the task of generating single frames would be a trivial one, for they could easily perform the mathematical operations to produce a picture that represented an infinitely precise moment in time (83].

This attitude didn't account for the fact that it's not the visual information in one frame, but the visual information that is provided by the transition between the frames, that is important (84].

The problems inherent in this transition between frames include temporal undersampling, temporal aliasing, and frame- 55 time. The temporal undersampling problem occurs when the change or movement of an object has not been correctly recorded for last perception by the viewer. For example, when a wheel is spinning its spokes may spin at a rate which exactly corresponds to the frame rate at which the camera's shutter is open. Therefore the spin does not appear to take place because the second spoke occupies the same position the first spoke did in the previous frame. Next, temporal aliasing is the blurring effect of an object across a film frame during exposure of the film frame while the shutter is open. This effect does not occur naturally in computer animation. Thus motion blur must be artificially created if computer animation is to effectively simulate the viewers perception of motion in the film medium.

Lastly, frame-time deals with the difference between the actual

"real time" (i.e. days) an action takes and the relative time

(i.e. minutes) in which it is shown to the viewer. Time models have evolved that use temporal intervals such as disjointed frames, abutting sequential frames, and overlapping or nested time intervals.

The basic problem for a computer animation system reduces to computing key-frames or parameter values that are unambigu­ ous and specific. This requires a formal model of time that not only can be implemented in the computer but also is intui­ tive for the animator. Issues of animation control should be thought of as independent of the type of display system being used. The same animation techniques apply to any graphical 56 style ranging from

...2D line drawing and raster graphic on a 1-bit-per- pixel console of a workstation to 3D hidden surface, continuous tone, full-color graphics, at high resolu­ tion on 70 mm film. In fact the only restriction on the type of graphical system used is that it can be driven by a formal description of some type that the graphical system can be used under program control [85].

3.8.1 Computer Motion Applications

In this discussion of computer animation the focus is on motion control. There has developed a wide diversity of com­ puter systems for motion specification. Each system has been oriented towards a different concept and thus satisfies dif­ ferent needs. They range from 2D systems used by the enter­ tainment industry to generate Saturday morning cartoons more efficiently to 3D scientific applications which have quickly grown into a new identifiable area of computer graphics called

"scientific visualization".

New understandings of scientific processes are being real­ ized through new simulation techniques. The bonding of pro­ teins, the structure of viruses and DNA, and chemical processes are being simulated. Engineers are able to simulate previously unobservable dynamics events such as fluid flow or processes that unfold over large time spans [86]. Manufacturers experi­ ment with design variations before building expensive equip­ ment . The heaviest users have been the military and aviation industries. Simulators permit the training of personnel in 57

hazardous situations without risk to the participants.

Analysis of movement patterns is possible [87]. Medicine,

ergonomics, and sports [88] have all begun to utilize computer

visualization.

From these divergent applications a number of specific

application programs have arisen. Unfortunately, while these

programs often incorporate similar techniques, the actual

implementations have not often adapted well to other applica­

tions. The application field of most interest to this study is

robotics. This field has done extensive research into motion

control and the techniques that have evolved form the basis for

many of the recent developments in motion control.

3.8.2 Computer Animation - Two-Dimensional Beginnings

Computer-assisted animation first concentrated on the

automatic in-betweening of traditional two-dimensional(2D) ani­ mation. The film frames in an animation which contains the

extreme extensions of a movement or a distinct change in direc­

tion is referred to as key-frames. The in-between frames are

those drawn between key-frames to continue the action. These

in-between frames are basically slight modifications of the

key-frames.

The computer was first used to speed up the labor-

intensive aspects of the traditional animation process.

Burthyk [89] and Catmull [90] developed systems to generate the in-between frames (the most laborious step of animation) from 58 the key-frames specified by the animator. In-betweening for the computer is the same as for hand-drawn animation, "the pro­ cess of generating all the frames of a motion sequence given its first and last frames [91]." In these systems the animator sketches each key-frame, carefully drawing each line and end­ point of that line to the corresponding line in the previous key-frame. The endpoints of the lines are then mathematically interpolated to generate the lines for the in-between frames.

The essence of the in-betweening problem is determining the correspondence between the established key-frames.

Transformations (what type of motion will take place [i.e. translate, rotate, scale]) and trajectories (the paths these motions will follow) must be established. Success in this 2D process depends on the animator's ability to plan and draw the correct overall motion [92]. Explicit control is still in the hands of the animator. The majority of the work in this area has been done at the New York Institute of Technology. As with most systems there are limitations which result in discernible stylizations.

3.8.3 Computer Animation - System Hierarchy Levels

With the evolution of faster machines and improvements in display devices it became possible to work in three dimensions in computer graphics. The potential of working in three dimen­ sions opened up a new level of complexity. Some of the first animations were pioneered on 3D vector refresh displays through 59 interactive rotations and translations. Initially, the anima­ tor used a simplistic approach of "absolute positioning [93]."

This requires the animator to manually list a numerical value for each parameter for each frame. While absolute positioning is less prone to accumulating error ~ found in other systems - it is extremely time consuming and is prone to human error.

Furthermore, this approach does not utilize the capabilities of the computer and subverts the efficiency rational for using it in the first place. The intent of a computer animation system is to provide tools that allow the animator to specify motion with fewer but more powerful parameters.

The first 3D computer animation systems were logical extensions of 2D computer interpolation systems. Since then, a number of sophisticated approaches have evolved for 3D anima­ tion. These animation approaches can be viewed as a three- level hierarchy: guiding, procedural, and goal-directed [94].

The guiding mode requires the animator to describe or "guide" the behavior explicitly. This mode is the most prevalent in current interactive animation systems [95,96], The procedural mode requires the animator to describe the behavior algorithmi­ cally or through parameters using a programming notation. The goal-directed level requires the animator to describe the behavior implicitly {e.g. "get up and look out the window")

[97] . 60

3.8.3.1 Guiding systems

Guiding systems permit the animator to use the following straight-forward approach: interactively position an object on the screen and specify a frame number, then position the object a second time and specify another frame number, and finally interpolate between the frames and display the resulting anima­ tion [98]. This incremental positioning approach is less time consuming than "absolute positioning" and provides a more flex­ ible method for specifying motions [99]. Animations produced with guiding systems often have limitations in direct propor­ tion to constraints inherent within the system. There are four distinct types of guiding systems: 1) shape interpolation, 2) key-transformation, 3) notation-based systems, and 4) motion recording [100].

1. Shape Interpolation is analogous to 2D key-framing sys­ tems. In this system a direct (one-to-one) correspondence is established between the vertices and polygons. In-between frames are produced by interpolating between the corresponding vertices of an object in the different key-frames. The diffi­ culty in automatically determining the correct corresponding vertices and polygons of the objects is that the system has no understanding of what the objects really are [101]. Several types of interpolation (linear, cubic, spline, etc.) are usu­ ally provided in a key-frame system which allows the animator a measure of control over the in-betweening process. 61

2. Key-Trans formation systems - also labeled as event- driven by Gomez [102] or parameter-based [103] - provide a slightly higher control level than straight key-frame systems.

These systems have their own set of key parameters (e.g. rotate, translate), unlike key-frame systems which key a whole image. Objects are manipulated by transformations (e.g. rotate, scale, translate) about specified control points.

These transformations are accomplished by matrix multiplication of the object's vertices. This result's in a change in the object's orientation or position after the multiplication and re-display. The in-between frames are generated by calculating the difference between the previous parameter value and the current one and then moving the object by this incremental difference. This differs from shape interpolation where the actual points and lines that make up the drawings are actually moved.

The inbetween positions are incrementally calculated from the key-transformation specifications. For example, if the first key-transformation is specified at 30 degrees and the second at 70 degrees, the rotation of the object will be from

30 to 70 degrees with the program calculating the necessary increments. These incremental changes are less time consuming and provide a more flexible approach for specifying motions

[104]. At this level, objects are manipulated by using speci­ fied values and functional relationships to control motion

[105] . Most systems provide for the setting of hierarchies 62

(attachments) making possible the animation of articulated motion. These systems can be both visually interactive, so the animator can position in real time, or keyboard specified.

GRASS [106], TWIXT [107] and BBOP [108] are examples of systems based on this concept [109,110].

3. Motion Recording utilizes devices to acquire kinematic data from real objects. Equipment comprises a multiple camera system and small light-emitting diodes that are attached to the object or, in the case of a figure, to their joints (Watsmart system). The system records the position of the diodes in three-dimensional space. Although the system is usually lim­ ited to a laboratory setting, it can generate a large quantity of accurate data. This approach is appropriate for idiosyn­ cratic motions that an artist would frequently desire.

This system could be considered "computer rotoscoping" because it uses data from direct references of real-life objects as traditional rotoscoping does. Restrictions that the author encountered in working with such a system was that translational data must be converted to rotational angles for key-transformation system formats. In addition, a limited number of data input cameras resulted in a restricted range of possible recorded motion. These limitations can be overcome by additional software and a minimum of four input cameras

[111,112] . 63

4. Notation-Based systems provide yet another type of explicit motion control. Notation-based systems are generally text-mediated systems in which the user describes a movement in a choreographic notation or an alphanumeric equivalent [113].

LABAN is based on Labanotation, a dance notation language.

This notation is used by choreographers, anthropologists, phy­ siotherapists, etc. to represent movement [114].

In Labanotation the body is abstracted into a set of joints and extremities. Positions of the joints are oriented in terms of a 'cross of axes' coordinate frame associated with a more proximal joint. Movement can be described in five ways: direction; revolution; facing; contact; and shape. Direction signs translate a joint, giving either a desired orientation relative to a cross of axes or a path of motion. Revolution signs describe twisting or rotating. Pacing signs establish an orien­ tation for some body part. Contact signs indicate sur­ face contact with other body parts or the environment. Shape descriptions trace a path or describe a shape to be formed by some body part [115].

LABAN analyzes and interprets the notation from a list of move­ ments the figure is to execute. LABAN has been implemented in several systems (e.g. University of Pennsylvania [116], Digital

Productions [117]). These systems have limited capabilities due to problems of complexity, slowness or storage space [118].

The various guiding type systems all provide complete con­ trol but forces the user to specify motion explicitly. On one hand the animator is at liberty to create a complete motion sequence. On the other hand this process for complicated fig­ ures or intricate relationships may be so unwieldy as to be impossible, even with a well-designed device-mediated interface 64

[119] .

3.8.3.2 Procedural Level

The animation problem is also approached from the stand­ point of building an "expert system”.

The animator progressively teaches the animation system how to perform actions (usually with some kind of pro­ gramming language); eventually this process builds up a repertoire (or library) of actions that can be used as building blocks for more complex actions. Zeltzer de­ fines this as an "animator level" system because it al­ lows the animator to extend the animation system [120],

The procedural approach generally uses algorithmic control to generate motion or to simulate the actual physical properties of motion [121]. Gomez [122] states that typical procedural level systems require programming in some relatively "high level language", either a derivative of a common programming language (e.g. ASAS from LISP [123] or MIRA from Pascal [124]) or one developed specifically for the animation system (e.g.

GRAMPS [125], ZGRASS [126], sal [127], ANIMA-II [128], ANTS

[129).

GRAMPS [130] allows for the construction of motion macros.

Joints can be hierarchically linked together and constrained to remain within a set range of values. Though GRAMPS was ini­ tially designed for molecular modeling it did generate articu- f lated figure motion during its life span. ASAS [131] was adapted from LISP and provides low-level mechanisms for the definition of transformation hierarchies and motion. In addi­ 65

tion, a message-passing mechanism "makes it possible to imple­

ment adaptive motion, since animated entities can report

aspects of their physical attributes or their internal states

[132]." However, the animator must write in ASAS, a semi-Lisp

program language. Reynolds' program allows ASAS actors

(objects) to act in synchronization. It permits "tweaking" or

the process of progressive refinement. MIRA [133] is very

similar to ASAS; attributes of objects can be used to influence

the generation of motion.

In these languages the animator must write procedures in

the animation system's language which are then applied in

specific instances to produce the animation. Zeltzer [134]

points out that this method unfortunately imports the software

engineering problem into computer animation." An advantage of

using a programmable language is that the animator is required

to codify his algorithms, and therefore, as the system is used,

new capabilities and tools are added. This approach tends to be flexible because many different types of changes can - in principle - be controlled [135].

This level has been erroneously known as the "animator"

level. Gomez [136] points out that the name "animator" is not appropriate and that "procedural level" would be more applica­ ble . For the typical animator would not necessarily be someone with a background in high level language programming. Pro­ cedural level is therefore more appropriate because the "the animator must write procedures in the animation system's 66

language, which are then applied in specific instances to pro­

duce the animation [137].”

The ideal approach combining both levels would be to allow the animator to design procedures visually. After all, there is no reason why an animation can't itself be considered a procedure; it simply happens to be in a different language [138].

3.8.3.3 Task or Goal-Directed Level

The goal in a task system would be to shift the chore of producing explicit motion descriptions to the animation

software. Badler, O'Rourke, Kaufmann, and Zeltzer have all

advocated the the importance of goal-directed motion. Zeltzer

[139] explains that at the task level the animation system

assigns the execution of motor programs to control characters,

and the motor programs generate the necessary pose vectors. To accomplish this requires the objects and figures to have a

knowledge base of the environment (containing information about their position, physical attributes, functionality, and environment). The animator specifies only broad outlines of a particular movement and the animation system fills in the details,

The task level attempts to permit the animation system to automate complex interactions of motions, while leaving impli­ cit control of the animation in the hands of the animator. The animator could instruct a figure to walk to the door, but need not "specify when each leg moves let alone details such as the 67

time/angle relationship for each joint of each leg [140]."

Although there is no task level system now developed for gen­

eral use, there have been several systems whose implementations

have had significant implications for fulfilling the ambition

of goal-oriented animation; sa [141], TEMPUS [142], Poda [143],

and Critter [144].

It is envisioned that a goal-oriented level would break

the task into "global functions". A global function "walk"

would call a procedural level gait controller with parameters

for speed and distance. This procedural level would invoke

"local motor programs" in the appropriate sequence. Local motor programs would include procedures such as "swing left

leg". These local motor programs would utilize dynamics and

kinematics or even add new explicitly defined sequences.

The motion of articulated figures is divided into two

issues: motor control {the coordination and control of joints to animate the figure) and motion planning (the blending of

sequential or concurrent motions and an autonomous interaction with the simulated environment. There is a need for integrat­

ing different animation levels to provide a complete yet economical process.

While a novice user may be satisfied with the "default" movements, an accomplished computer animator will desire total control to make a sequence as expressive as possible [145].

However, as Zeltzer [146] pointed out, the issue of control 68 should not be assumed to mean that the animator needs or even wants a an explicit system for motion specification. For the explicit (guiding) specification of complex motion soons becomes unwieldy - though it is well suited for specifying fine details. At the next level (procedural) a more implicit con­ trol mechanism exists that is powerful yet demanding of its user. The goal-oriented level (task) trade off specific con­ trol over details for broad control over complex complicated motions. The control solution lies in accessing each appropri­ ate level in a modular hierarchical organization. This organi­ zation "allows the user to identify the motion qualities that need to be adjusted, and at the same time it helps to localize the effect of such changes. This calls for a uniform represen­ tation of motor skills that incorporates, for each skill, a specification of the kinds of adjustments that are possible, and, in addition, a uniform set of mechanisms for interacting skills [147]."

3.8.3.4 Synergic control

Zeltzer [148] points out that animation software should incorporate sufficient knowledge about motions to allow anima­ tors to control a figure with a predetermined, yet flexible, set of movement commands. The organization of natural movement systems suggests that control programs should be distributed hierarchically throughout a motor control system in a top-down fashion. Biological motor control systems employ distributed problem solving (the motor control system is decomposed into a 69

coordinated set of smaller subsystems). For a figure these

could be modules consisting of a set of muscles and joints that

can effect a particular class of motions, each module being

under a set of local motor control. "Instead of storing expli­

cit movement descriptions, the biological motor system seems to maintain a repertoire of low-level motor control programs that

can be invoked in various combinations to suit the circumstance at hand [149]." The task of the top level command is to invoke that set of low-level subsystems necessary to perform a partic­ ular motion sequence. Greene, Tomovic and others have argued that this may be the only efficient way to control a large, complex system [150,151].

Zeltzer uses an example of the grasping motion of the hand to describe this kind of control.

The whole hand has over two dozen degrees of freedom, yet the grasping motion, which involves many small mus­ cles, can be thought of as being invoked by only a few parameters - for example, quickly or slowly, forcefully or gently, and so on- which entirely sufficient to characterize it. The hand muscles are programmed to perform this motion. Indeed, with most people, it takes an effort of will to flex and extend the fingers individually - that is, to "disable" the grasping pro­ gram [152] .

Zeltzer describes this "task" level in terms of some specific motion sequence (e.g. go to the door and open it, Run to the ball and kick it). At the top level of the hierarchy is the task manager. "It accepts task descriptions from the user and decomposes each of them into a list of composite skills 70

[153]." These skills represent some class of motions the struc­

ture can perform - walking, running, grasping and so on. This

skills are made up of primitive procedures previously referred

to as local motor control. These skills are then executed at

the appropriate point of the animation.

Specialized task systems consisting of flocks of birds and

fish have been recently constructed [154,155]. These simple behaviors have been very convincing. How well the success of

these simple systems translates up to more complex applications

remains to be determined.

3.8.4 Computer Animation - Production Process

Similar to traditional animation, computer animation begins as a concept or story. From this concept a storyboard

is created as an initial guide to the story and development of

key-frames. After the scene descriptions have been esta­ blished, the individual objects are isolated and its surface descriptions are encoded into polygons. The vertices of the polygons are converted into the computer as numerical coordi­ nates. This can be accomplished manually or with the assis­ tance of a digitizing device. The result is a data file from which the computer can display a three-dimensional object on a graphics screen. As in traditional animation the more compli­ cated an object is, the more time-intensive it is to produce it. The more polygons needed to describe an object results in more computer storage required (disk space) and more computa- 71

tion time (CPU cycles) for rendering.

Just as traditional animation works out its timing prob­ lems using simple drawings, so, too, is computer animation

driven by the need for simplicity.

Using wire-frame images, which require less data mani­ pulation and allow faster drawing speeds, a technical director would preview short motion segments in real time. The action commands being specified are visually checked for accuracy before the more costly and time- consuming final image rendering process [156].

In this way the animator is permitted indulgently experiment

with different variations in the timing with the computer's

assistance. This is possible because after specifying the beginning and ending key-frames for an action, the computer program will calculate the needed in-between frames. This is accomplished through a variety of techniques: guiding, motion recording, dynamics, etc.

The prevalent method for preserving the motion specified

in the computer is to use a formal script that can be inter­ preted unambiguously by the animation system. This script may be an explicit file with a description for each frame or a pro­ gram that describes the action or a variation of both [157].

The "script" combines the decisions of time, dynamics, and graphics. The animator can add one new aspect or modify an existing one without effecting the rest of the script. If, after testing, the modification is seen to be a mistake the animator can revert to a previous version of the script [158]. 72

The script provides a precise repeatability for the animator.

The goal of a scripted computer animation system is to allow the animator to write a compact, high-level description out into the more detailed, lower-level description required to fully specify and render each frame. This strict scripting is exactly the right thing in those situations where the entire animation is preplanned and where the desired appearance of essen­ tially every frame is known in advance [159].

In traditional animation the next step would be the tremendously time consuming process of rendering by the inkers and painters to produce the final images that the public will

see. In contrast, in computer animation the rendering is left to the computer. The computer is directed by the "script", which contains the artist's specifications on the object's illumination properties (lighting, color, texture, tran­ sparency, reflection, etc.) The rendering of an animation can even be subdivided and distributed to multiple computer facili­ ties .

3.9 Simulation

The ability to imitate is a defining characteristic of simulation, whether it be kids playing dress up in adult clothes or are playing like a pilot in a flight simulator.

Simulation generally involves some kind of model or simplified representation. During the course of a simulation, the model mimics important elements of what is being simulated. A simulation model may be a physi­ cal model, a mental conception, a mathematical model, a computer model, or some combination of all of these. For children playing house, their model is the toys they are using including the imaginary characters and 73

settings [160].

The previous efforts in simulation have been context dependent.

That is, simulation models have remained largely bound to discipline. Now, with the computer new simulations can be applied (with adaptations) to other areas of exploration. The potential exists for the cross-over of information from one discipline to another. This is possible because all data within the computer - whether it be physical, conceptual, or whatever form - is abstracted into the same homogeneous binary code.

3.9.1 Simulation Applications

Computer simulation generally comprises two "worlds".

There is the real world system (observed behavior or phenomenon) and then there is the mathematical world from which the computer simulation is produced. Simulation represents a procedure by which one can scientifically study or understand some behavior or phenomenon in the real world. Predictions about some behavior in the future can be forecast and then checked by setting up a simulation and running it [161]. Simu­ lation models help us to better understand "real world" sys­ tems .

Physical models comprise many of the current efforts in simulation. Wind tunnels, wave tanks, and weather chambers are examples of physical model simulations which imitate larger systems. Small replicas of transportation carriers are 74 constructed and placed in simulated environments to measure the effects of outside forces. For example, in a wind tunnel air is directed toward the replica so as to evaluate its aerodynam­ ics properties. The disadvantage of physical models are that they are expensive to build and maintain.

In a mathematical simulation numbers or symbols abstractly represent some aspect of the real world [162]. Early numerical simulations were done by hand. Many tedious years were spent performing numerical simulations for creating navigational tables in the 16th century. Simulations were later needed to design radar, gun turrets and other military equipment. From this need to perform simulation calculations the technology for the computer evolved. Computer simulation is currently used in a wide range of applications in the physical sciences, as well as in the social sciences and economics. This study is most concerned with those simulation applications of scientific visualization and robotic techniques now being developed.

3.9.2 Simulation Methods

Simulation uses several types of methods (e.g. Robust,

Stochastic, Deterministic.) to help bridge the discrepancies between the real world and the pseudo world created by the com­ puter. This methods are not exclusive and can be combined.

3.9.2.1 Robust simulation IS

Robust simulation generally means "insensitive to small departures from the idealized assumptions [163]." The word

"small" can have two different interpretations, both intrinsic to the concept: (1) fractionally small departures for all data points, or (2) proportionally large departures for a small number of data points [164]. Robust simulation methods would be applicable in computer animation when "we have a priori

[underline added] knowledge about the probable values and prob­ able uncertainties of some data [165]." In such cases a result that is desired takes this advanced information into account,

"neither completely freezing a parameter at a predetermined values nor completely leaving it to be determined by the data set. The formalism for doing this is called 'use of a priori covariances' [166]." This method can provide the animator with a degree of control over the outcome.

3.9.2.2 Stochastic simulation

In reality, the true set of parameters for a simulation can be hidden from the artist. It is frequently necessary to make an estimation of these parameters to get the desired results. The scientist may rely upon analytical formulas or numerical procedures to generate parameters for a simulation.

However, even the scientist must many times employ a method called Monte Carlo simulation or Stochastic Simulation. Accord­ ing to Press,

We can simulate our own sets of "synthetic" realiza­ tions of these parameters as "synthetic data sets". 76

The procedure is to draw random numbers from appropri­ ate distributions so as to mimic our best understanding of the measurement errors in our apparatus [167].

These random drawings construct data or parameters in the range that is desired. This method provides a way

...to characterize the errors of parameter estimation in a very precise way. One can also try out on the computer different methods of parameter estimation, or different data reduction methods, and seek to minimize the uncertainity of the result according to any desired criteria [168].

In addition, an overly simplified simulation could result in a animation with a lack of variety (same results each time the simulation is run). Due to the effects of randomizing fac­ tors in the real world this situation would be rare if not impossible. In an animation this sterile repeatability could be prevented by introducing stochastic processes into the model.

3.9,2.2.1 Ad hoc Approach

Many times an algorithm can be configured that is too com­ putationally expensive to run. One way around this is to

"fake11 the object (to find a method which reproduces the appearance of the object without actually simulating it) [169].

Ad hoc methods in simulation are necessary because either there are gaps in the knowledge known about the subject and it is necessary to proceed anyway, or because it would just be too computationally-intensive to proceed any other way. Ad hoc 77 simulations can provide a more realistic look than so called

"clean" simulations. Haumann explains that

In the first attempt to animate a sheet of paper, it was modeled as a perfectly flat sheet, placed in the air at a moderate angle and released. The result was that the paper slid sideways, edge first, looking like a flat toboggan traveling down a perfectly smooth, in­ visible slope. It has been placed in (an unrealisti- cally) perfect equilibrium situation, and there was nothing to cause it to deviate from this configuration: no upturned corners to catch the wind, no unbalanced aerodynamics forces. The solution was simple. No pa­ per is perfectly flat, so the positions of the mass particles representing the surface were randomly per­ turbed out of the plane of the surface, effectively wrinkling the paper [170].

Many Ad hoc simulations, though not entirely accurate, contain sufficient information to communicate the desired significance of the work. It is worth noting that acceptable levels of accuracy in a simulation can differ greatly from a scientific perspective to an artistic perspective.

If Ad hoc methods are used there is the question of suffi­ cient validity of the simulation. Is the computer facsimile really an accurate representation? This validity is dependent on the comprehensiveness of the mathematical model being used.

There are numerous instances in which all mathematical elements are known. From this data a non-computer simulation of physi­ cal phenomenon can be performed and compared to a computer simulation of the known mathematical variables involved. This has been done and has resulted in no discernible differences to the naked eye [171]. 78

3.9.2.3 Deterministic versus Non-Deterministic Simulations

For a simple physical system such as a frictionless pendu­

lum the equations of motion have a closed-form solution. A

closed-form solution uses a formula that can determine any

future state from an initial state. This is the case for periodic motion in which the solution to any state - in anima­ tion this would be an individual frame of animation - is deter­ ministic. A closed-form solution provides a short cut which

can predict the future without having to step through the

intermediate states [172]. Closed-form solutions do not exist

for a variety of mechanical systems (i.e. collisions between objects). There are no short cuts to predicting behavior for

systems that are chaotic.

"Chaos" challenges the ubiquitous reductionist theory (a system can be understood by breaking it down and studying each piece). While it is true that by using gravitational formulae the paths of the planets can be simulated for thousands of years, it is also true that there are other natural phenomenon that are not so predictable. "The weather, the flow of a stream, the roll of dice all have unpredictable aspects [173]."

It had been presumed that once a sufficient amount of data was accumulated and processed that predictability was inevitable.

The discovery that deterministic systems can generate seemingly random behavior has changed this past assumption. The implica­ tion is that there are fundamental limits on the ability to make predictions. Randomness generated in this way is called 79 chaos. This discovery has created a new paradigm in scientific modeling.

An offshoot of "chaos" research has been the parallel discovery that there is a determinism inherent in chaos theory; that many apparently random phenomena are more predictable than previously thought. Random-looking information in such diverse systems as the atmosphere, dripping faucets, and the mixing of fluids can now be explained in terms of simple laws. Chaos occurs in simple systems, where a small "change in the present causes a larger change in the future. The notion is clear if one thinks of a rock posied at the top of a hill. A tiny vari­ ation in a push one way or another is enough to send it tum­ bling down widely differing paths [174] .

3.9.3 Simulation as Animation

The first encounter with the term simulation in computer graphics is in the context of an illumination model - the simu­ lation of perspective, highlighting, shadowing, reflection, refraction, multiple light sources, volumetric transparency, translucency, surface texture, texture mapping and coloration techniques. In the context of this study simulation is a way of producing animation that involves modeling the physical laws that actually affect objects as opposed to positioning objects where they are supposed to be. For example, the specification of a figure walking by manually positioning the limbs is not a simulation. Though this hand method does model the desired 80

motion, it does so by declaring where things are to be posi­

tioned [175]. "A simulation system evaluates a physical system

to determine the actions of objects within the animation.

Rather than interpolating between values, a simulation system

simply uses its rules of behavior to compute the animation

[176]

Animation by simulation represents a different approach than the use of programming techniques or explicit descrip­ tions. These parameter controlled actions require no history.

Successive frames are not dependent upon previous frames because the sequence of frames may be calculated in any order since they are only dependent on explicit parameters. In con­ trast, a simulation may be thought of as a mathematical pro­ cedure that is solved incrementally. Rather than simply draw­ ing the result, a simulation "is used where it is necessary to evaluate the circumstances in order to determine the result

[177]." Before a simulation can solve the position of an object at frame two it must solve the position for frame one.

"Indeed, to get to an advanced state like frame 100, it is necessary to calculate all ninety-nine prior frames so 'it knows where it is'. Each successive frame is dependent upon the calculation of the previous frame [178]." (See Figure 11)

An engine simulation imitates the actual physics of engine operation. In the case of the engine it would calculate the rate of exploding gas in the top of the cylinder, the resis­ tance of the piston and the connecting rod, and determine how 81

Figure 11. Incremental animation sequence. Dave Haumann. far down the piston gets pushed. This interaction is displayed. The position of the piston is calculated for each

successive discrete frame in time, making animation [179]. The opposite of a scripted animation constructed by an animator would be just those type of simulations made by using such scientific and engineering information.

A structural engineer designing a bridge might use a finite-element model of the structure and its mechani­ cal properties in conjunction with a computer animation system in order to study how the structure would react to high winds or earthquakes [180] .

In these types of applications both methods of generating an animation can provide similar information to the viewer. At 82 this point, the quantitative and qualitative differences begin to blur, and animation and simulation begin to merge.

The demarcation between animation and simulation started to markedly blur with programs like TEMPUS [181], a program with sophisticated features that not only permit the definition and modification of human figures but also provides for resolved motion control. Generally, simulation-based programs like TEMPUS incorporate laws of motion to assist in the specif­ ication and accuracy of movement.

3.10 Physics - Mechanics

Physically-based simulations are derived from the field of mechanics in physics. Kinematics and Dynamics (subsets of mechanics) are the two pivotal disciplines that comprise the bulk of information necessary for the model in this study.

Kinematics is a mathematical descriptive account of motion without concern for its causes. Dynamics is concerned with the causes of motion and its mathematical laws and formulas.

Newton's Laws of Motion comprise the foundation for dynamics.

3.10.1 Laws of Motion

Galileo described motion, whereas Newton explained motion as the application of forces. Of Newton's three laws of motion, the first two laws replicate the effects of forces on an object and the resulting motion, while the third law repli­ cates how two or more bodies interact - momentum, gravity 83

[182] .

1. Newton's First Law of Motion (also the law of inertia).

If an object is at rest, it will remain at rest. If object is

in motion, it will remain in motion. The rate of change in the motion of an object is directly proportional to the unbalanced

force applied to the object. Inertia is the tendency of an

object to resist changes in velocity. Inertia may be defined as a constant which may be used as a measure of mass in a grav­

itational situation [183].

2. Second Law of Motion. Acceleration is directly propor­ tional to the unbalanced force applied to an object and is

inversely proportional to the mass of the object. Application of this law will determine how much force is needed to move

something at a certain speed, or if a measurable force is applied what is the resultant speed will be [184]. This law is

summed up in the equation, F = ma (Force equals mass times acceleration).

3. Third Law of Motion. "For every action there is an equal and opposite reaction [185]." Action and reaction should be thought of in terms of forces. If A exerts a force on B, then B must exert an equal and opposite force on A. This law applies to all physical substances whether they be gas, liquid or solid. If an "impulse" (force) is applied, "momentum" (the property of velocity or movement attached with an object) is the result. Conservation-of-momentum maintains that momentum 84

is spread from one mass to another (e.g. a boy jumping onto a

sled or a ball striking a bowling pin) [186].

The field of mechanics in physics is founded on the work of Galileo and Newton. From mechanics the interdisciplinary

subjects of kinematics and dynamics have evolved into a set of procedures which can be used to produce visual simulations.

3.10.2 Kinematics

Kinematics is a branch of mechanics within the field of physics concerned not with the causes of motion but instead with its description. Kinematics is a mathematically descrip­ tive account of motion without concern for its cause. Gen­ erally, kinematic motion specification is thought of in terms of positions and rotations. However, the field also encom­ passes velocities and accelerations. The forces and torques

(turning or twisting force) responsible for this motion are not taken into account.

3.10.2.1 Forward Kinematics

Forward kinematics is the specification of motion in terms of degrees of rotation in a joint-angle (i.e. the arm will rotate 90 degrees in the x-axis). Most key-transformation sys­ tems can be characterized as forward kinematic systems [187].

Forward kinematic systems are generally associated with articu­ lated figures in computer animation. This is because rigid segments, which make up an appendage, can be linked together in 85 a pseudo-joint for which key joint angles can be assigned by the animator. The animation is created by interpolating the resultant joint-angles from these specified key joint-angles.

The joint-angles determine the position and orientation of the final displayed appendages. Currently most forward kinematic descriptions requires the manual assignment of joint-angles in most all animation systems. This procedure is one of the underlying control mechanisms central to contemporary computer animation.

An articulated figure, hierarchically organized, would have an arm positioned using forward kinematics by requiring the user to first rotate the shoulder joint-angles into posi­ tion and orientation, then the elbow joint-angles, and so forth on down the hierarchical chain to each joint in the finger

[188]. A significant limitation of forward kinematics emerges when applied to the leg appendage.

Since the foot is hierarchically dependent on the an­ kle, knee, and hip joints, the hip cannot be moved without also moving the foot. This makes maintaining a foot hold on the ground very difficult [189].

Another drawback is that this procedure requires an immense amount of time to manually set all these key joint-angles.

Kinematic equations which relate acceleration, velocity and position would be a more effective way of determining the rota­ tional values needed for each degree of freedom. This would shift a large portion of the burden of maintaining ascribed joint-angle relationships to the computer. The computer would 86 maintain the ascribed joint-angle relationships for the dura­ tion of the desired motion.

The main advantage of forward kinematics is its usefulness in hierarchical structures, while its disadvantage is that it is not intuitive. We generally don't conceive in terms of joint-angles to place a hand, foot, or whatever into a desired position. People think and move in terms of target goals one wishes to achieve {i.e. move hand ten inches forward to get a piece of paper) [190].

3.10.2.2 Inverse Kinematics

The animator needs a method of generating motion that is analogous to how humans naturally move and think. We do not consciously think about how our joints move {i.e. forward kinematics), but instead we think in terms of how to reach tar­ get goals (i.e. grab, touch) [191]. Inverse kinematics pro­ vides the necessary joint-angles by way of specifying target goals (i.e. grab door knob). This is similar to how we ini­ tiate our own actions. Inverse kinematics could be considered a rudimentary goal-oriented approach.

Inverse kinematic control utilizes an end-effector (i.e. foot, hand) of a chain-of-joints (i.e. leg, arm) which is posi­ tioned and oriented in space [192]. Given an end-effector position, inverse kinematics solves for the degree of rotation required at each parent-joint (ankle to knee to hip, etc.)

These various joint-angles in the chain-of-joints in the leg 87 are then supplied to a key-transformation system (forward kinematics) for final input and subsequent display. Thus the foot on the ground problem described earlier is effectively solved using inverse kinematics [193] . Although actual specif­ ication can be done in either global or local space coordi­ nates, local end-effector coordinate space is more intuitive.

When a limb contains six degrees of freedom (three for position [x,y,z] and three for orientation [pitch, yah, roll], it is considered redundant. This redundancy makes multiple solutions to the inverse kinematic problem possible. For exam­ ple, we can "hold our hand in both position and orientation and still wave our elbow and shoulder [194]." The expressive poten­ tial of this redundancy is discussed in chapter VI. The strength of kinematic-based approach is the utilization of kinetic motion descriptions (i.e. running) which are available from biological literature [195].

The disadvantage of kinematic motion specification is the complexity of specifying motion for objects with many degrees of freedom (i.e. forward kinematics) and the difficulty of achieving naturalistic simulations of physical interactions with other objects or characters [196]. In addition, kinemat­ ics alone are inadequate for rapid or ballistic motion.

Kinematic models only describe motion. The appearance of weight (the force of gravity) has to be simulated manually in a kinematic system. By extending control to include dynamics, the animator is freed from manually simulating physical 88 attributes such as the illusion of weight of an object or char­ acter .

3.10.3 Dynamics

Dynamics uses mathematical equations derived from Newton's laws of motion. Dynamics refers to the explanation of motion in terms of forces (for translational motion) and torques (for rotary motion) in relation to position and orientation. Given the forces and torques acting on or from within a figure or object, the change in position and orientation which the entity undergoes can be calculated by dynamics. This assumes that the initial configuration, the current motion factors, motion con­ straints, and the physical properties such as mass and the dis­ tribution of mass are known (197]. Dynamics has been integrated into a variety of systems: flexible objects [198], rigid objects (199], soft tissue deformation [200], facial ani­ mation [201,202], and articulated structures [203, 204, 205].

3.10.3.1 Forward Dynamics

Forward dynamic motion is generated by applying forces to the known physical properties of an object which results in a change in the object's position or orientation. Simulating the motion of an inanimate-object is most easily accomplished using forward dynamics. Unfortunately, in forward dynamics, the specification of the various joint torques (rotational forces) is as laborious as the specification of joint-angle rotations in forward kinematics, and much less intuitive [206]. This 89

lack of intuitiveness stems from the fact that artists do not

think in terms of discrete, measurable forces.

3.10.3.2 Inverse Dynamics

Inverse dynamics provides a means of determining the

forces required to perform a specific motion [207] . If solving

an equation for position and orientation when the forces are

given is called "forward dynamics", then solving for forces when position and orientation are given is called "inverse

dynamics [208]." When working within a dynamics system it is often difficult to think of motion in terms of given forces.

Given a motion description (kinematic), inverse dynamics calcu­

lates the necessary forces and torques to simulate the motion described. This in turn, will be used by forward dynamics equations to generate the desired motion description. While certain forces and torques can be automatically computed (i.e. gravity) there are also a large number of other forces (muscles and other externally applied forces) that must be described by the animator to create the overall motion [209] . The final motion simulation is generated by forward dynamics, based on derived joint torques and external forces computed from inverse dynamics. Because both approaches (forward and inverse) have distinct advantages a dynamics system capable of executing both forward and inverse dynamics would be highly desirable. 90

3.10.3.3 Dynamic Analysis

"Dynamic analysis provides a means of accurately predict­

ing the behavior of a structure under the influence of forces

and torques [210]." Assuming the conditions specified are real­

istic a simulation model analyzed dynamically is constrained to move in a realistic fashion. According to Reynolds,

Given a physical (though simplified) description of a body and given a set of joint forces and torques that lie within the range that can be reasonably produced by its muscles or motors, a dynamic analysis program can produce a description of the actual motion such a body would undergo under these conditions in the real world. Such a motion prediction, consisting of the positions, velocities, and accelerations that occur at joints, can be used to drive a figure animation system such as those already developed to kinematically display the movement of articulated figures [211].

Dynamic analysis can automatically predict the effect of grav­

ity, collisions, reaction forces, and the joint-limiting forces and torques. The user specification problem is then reduced to choosing realistic values for controlling forces and torques that the animal or machine generates to move or stabilize itself.

Dynamic evaluation very naturally and realistically predicts the motion of the body models that are, after all, attempting to mimic physical reality, and appears to offer in the long run great advantages in increasing the realism of computer graphic images [212].

For cases, such as falling, or moving under the influence of large external forces and torques, dynamically-predicted motion may be far preferable [213]. 91

3.11 Simulation of Aesthetic Criteria

One of the most intriguing ideas to evolve from computer applications is that not only can the concept of simulation be applied to physically-based methods of creating art, but also to the creative process itself. This requires an analysis and formal description of the creative process as an algorithm.

Such a formal approach would appear to be the antithesis of the creative process. However, there is historical precedence for such formalism in art. For example, the ancient Greeks formal­ ized their knowledge and cultural standards of beauty into canons and their accompanying rules. From this formal aesthetic, an order of rightness evolved that is still recog­ nized today. At the very least an attempt to formalize infor­ mation leads to "a much deeper understanding than if we simply try to understand things in the traditional way" [214] .

The concept of Aesthetics has several meanings; (a)

Aesthetics may be an individual viewpoint which is used to interpret and evaluate works of art. Though our culture per­ petuates the belief in the validity of the individual aesthetic, it is often regarded as irrelevant if it does not link up in some way with the collective aesthetic of society.

(b) A collective, cultural aesthetic is the second meaning. It is probably safe to speculate that this collective aesthetic even evolved to "formalize" the art experience for society.

The acquisition of our society's collective aesthetic has gen­ erally not been a conscious introspective process as much as a 92

cultural and environmental indoctrination.

It appears that the aesthetic, in this case, is a con­

sensus of the culture, without which it would be difficult, at best, for us to share and appreciate art. If the aesthetic, within the creative process, is a consensus, then it follows that there must be a set of mutually acknowledged criteria.

Gips and Stiny [215] embrace this concept of an aesthetic consensus and define aesthetics as a "systemized” point of view, which can be extremely varied and complex. For example, in the Western culture originality is upheld as one of the pri­ mary, essential elements of a work of art. On the other hand, in an Eastern culture a continuity of tradition and thought is primary. Even though these viewpoints may vary considerably, their pervasiveness across cultural boundaries suggests they share a common underlying process. Properties that are not intrinsic to the object, but are relational between observer and object are the domain of this underlying process.

This process can be characterized by interpretative con­ ventions and evaluative criteria.

Interpretative conventions determine how an object is understood as a work of art, and they have two aspects: They decide (1) the types of interpretations that can be made for objects as works of art and (2) which in­ terpretations refer to which objects. Evaluative cri­ teria determine the judged quality of an object in terms of its interpretation as a work of art, and they have two aspects: They determine (1) the judged quality of an individual object and (2) the ordering of the judged qualities of two different objects [216]. 93

The (1) interpretative and (2) reference algorithms are classified under "interpretative conventions", and the (3) evaluative and (4) ordering algorithms are classified under

"evaluative criteria".

The "interpretative" algorithm consists of the description of an object or artwork in terms of its external evocations and internal coherence. External evocations and internal coherence are often discussed in terms of content, representation or expression. The internal coherence encompasses the underlying composition of an object. For an artwork this would involve qualities such as unity, variety, balance, etc. In the context of this study interpretation may pertain to rules for the building of motion or principles of organization. For example, what kind of structural arrangements are associated with rapid movements?

The "reference" algorithm links descriptions (interpreta­ tions) with specific objects or artworks. The algorithm impli­ citly determines which underlying structures acceptably match specific object descriptions based on the occurrence of speci­ fied criteria. For example, rapid movements can easily be associated with a corresponding structure (rabbit) that can actuate them. This algorithm may consist of a lookup table in its most basic implementation. In addition, this aspect of establishing correspondence alludes to a "generative" potential for structures with motion. 94

The "evaluation" algorithm assigns values to the interpre­ tations and its subsequent references. Continuing with the

"rapid movement and rabbit structure" scenario, the question is, "Was this a good match for the structural description?" If it was, it would receive a high evaluative rating. On the other hand, this match may be rated low aesthetically because of its obvious association and lack of expressive content.

Finally, an "ordering" algorithm would rank evaluative results in terms of a given aesthetic viewpoint. This ranking would be the result of the application of pre-specified factors of importance. If the expressive ranking of the correspondence match is lower than the structural ranking for that same match, and the expressive characteristic is predefined as more impor­ tant to the overall artwork, then the algorithm would continue to generate additional variations until a more suitable ranking resulted [217].

The potential of this approach is that the level of direc­ tion and qualitative output from the artist can be heightened.

The artist's time therefore is not consumed by explicit efforts of production. These aesthetic algorithms may also serve as a heuristic device capable of extending and focusing an artist's vision. The drawback of this approach would be in the actual specific determination of which relevant parameters need to be defined and how they will interrelate. Hopefully, the field of artificial intelligence will be able to provide insights into this problem of qualitative simulation in the future. 95

3.11.1 Story Simulation

Kenneth M. Kahn [218] has developed an animator language

called, "Director", that creates computer animations from story

descriptions. The required input includes the characters,

their personalities, the interactions between them, events, and

key scenes from the story. The action is produced by

"knowledge-directed aesthetic choices based upon the informa­

tion suggested by the story description, and built in knowledge

derived from common sense, perceptual, and ethological sources [219]

The Director language is based on the programming concept

of "actor" [220]. Each actor may be represented as a database

of knowledge in conjunction with its decision making strategy.

Decisions as to how the story is to proceed is handled by

"choice point" actors who are influenced by suggestions from

other actors.

These "suggestion" actors act as knowledge sources by comparing character personalities {such as "who is fas­ ter, stronger, etc."), or by culling information from the story itself. The choice point actors are defined as procedures for combining and resolving conflicts between suggestions as well as a data base for sugges­ tions received so far [221].

A rough animation is created by incrementally examining all

suggestions. While this work does achieve the desired outcome

of a story, it more significantly proposes that designing plot-based animation should be considered a knowledge-based 96 problem solving activity, whereas the generation of the actual

animation is accomplished by simulation- 97

NOTES TO CHAPTER III

1. L. D. Henderson (1983). The Fourth Dimension and Non- Euclidean Geometry. Princeton, N.J.: Princeton University Press, p. 110.

2. Ibid., p. 111.

3. Ibid., p. 112.

4. Ibid., p. 115.

5. J. Halas & R. Manvell (1976). The Technique of Film Animation. New York: Hastings House Publishers, p. 12.

6. A. E. Smith & P. A. Koury (1952). Two Reels and a Crank Garden City, N. Y.: DoubleDay, p. 51.

7. D. Crafton (1984). Before Mickey, The Animated Film, 1898-1926. Cambridge, MA: The MIT Press, p. 14.

8. Ibid., p. 22.

9. Ibid., p. 23.

10. Leonard Maltin (1980). Of mice and magic, A History of American Animated Cartoons. New York: Signet Books, p. 4.

11. Ibid., p. 5.

12. Crafton, 0£. cit., p. 7.

13. Ibid., p. 9.

14. J. Halas & R. Manvell (1976). The Technique of Film Animation. New York: Hastings House Publishers.

15. Crafton, ojo. cit., p. 244.

16. Ibid.

17. Ibid., p. 250.

18. Bofa, G. (1925). "Du Dessin anime," Les Cahiers du Mois, 16, p. 51.

19. D. Crafton (1984). Before Mickey, The animated Film, 1898-1926. Cambridge, MA: The MIT Press. 98

20. Maltin, ojo. cit., p. 4.

21. J. Halas & R. Manvell (1976). The Technique of Film Animation. New York: Hastings House Publishers, p. 14-15.

22. Ibid., p. 15.

23. Ibid.

24. J. D. Andrew (1976). The Major Film Theories. New York: Oxford University Press, p. 31.

25. Ibid., p. 32.

26. Ibid.

27. Ibid., p. 33.

28. Graham Collier (1972). Art and the Creative Consciousness. Englewood Cliffs, NJ: Prentice-Hall Inc., p. 77.

29. Ibid., p. 78.

30. Ibid., p. 79.

31. A. H. Barr (1987). "Introduction to Physically-Based Modeling," Tutorial Notes No. 1/7, SIGGRAPH r 87, Topic in Physically-Based Modeling.

32. Michael Girard (1986) . Interactive Design of 3^-D Computer-Animated Legged Animal Motion. University of North Carolina Motion Workshop, Chapel Hill, NC., p. 1-20.

33. Collier, o£. cit., p. 79.

34. P. C. Vitz & A. B. Glimcher (1984). Modern Art and Modern Science, The Parallel Analysis of Vision. New York: Praeger Publishers, p. 255.

35. Ibid.

36. J. D. Andrew (1976). The Major Film Theories. New York: Oxford University Press., p. 28.

37. Vitz, 0£. cit., p. 4.

38. Ibid., p. 255.

39. J. D. Andrew (1976). The Major Film Theories. New York: Oxford University Press., p. 31. 99

40. Rudolf Arnheim (1954). Art and Visual Perception. Berkeley, CA: University of California Press.

41. J. Halas & R. Manvell (1976). The Technique of Film Animation. New York: Hastings House Publishers, p. 29.

42. Frank Thomas & Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press., p. 32.

43. Halas, 0£. cit., p. 30.

44. Ibid.

45. H. Whitmaker & J. Halas (1981). Timing for Animation. London: Focal Press Limited., p. 9.

46. Ibid.

47. Ibid.

48. John Lasseter (1987). "Principles of Traditional Animation Applied to 3D Computer Animation," ACM SIGGRAPH * 87 Conference Proceedings, 21, (4), p. 35.

49. Frank Thomas & Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press., p. 32.

50. Lasseter, o£. cit,, p. 36.

51. Thomas, o£. cit., p. 42.

52. Ibid., p. 52.

53. Lasseter, o£. cit., p. 37.

54. Ibid.

55. Thomas, o£. cit., p. 27.

56. Lasseter, 0£. cit., p. 40.

57. Ibid.

58. Thomas, 0£. cit., p. 29.

59. Ibid.

60. Ibid., p. 30.

61. Ibid.

62. Ibid., p. 31. 100

63. John Lasseter (1987). "Principles of Traditional Animation Applied to 3D Computer Animation," ACM 5IGGRAPH ' 87 Conference Proceedings, 21, (4), p. 40.

64. Thomas, op. cit., p. 33-34.

65. Lasseter, op. cit., p. 40.

66. Thomas, op. cit., p. 25.

67. Lasseter, op. cit., p. 42.

68. Frank Thomas & Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press., p. 169.

69. Ibid., p . 169.

70. Ibid., p. 179.

71. Ibid., p. 169.

72. Thomas, pp. pit., p. 169.

73. Ibid., p. 172.

74. Ibid., p. 173.

75. Ibid.

76. J. Halas & R. Manvell (1976). The Technique of Film Animation. New York: Hastings HousePublishers, p. 25.

77. Frank Thomas & Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press., p. 173.

78. Ibid.

79. Ibid., p. 173.

80. Ibid.

81. Ibid., p. 183.

82. Ibid., p. 182.

83. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH 186, Tutorial Notes, Advanced Computer Animation, p. 89.

84. Ibid.

85. Ibid. 101

86. David Sturman (1986). "A Discussion on the Development of Motion Control Systems,11 Advanced Computer Animation, SIGGRAPH '86 Tutorial Notes. p. 5.

87. J. R. Wilhelms (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley, p. 1.

88. Ibid., p. 2.

89. N. Burtnyk & M. Wein (1976). "Interactive Skeleton Techniques for Enhancing Motion Dynamics in Key Animation," Communications of the ACM, 19, (10), pp. 564- 569.

90. Ed Catmull (1978). "The Problems of Computer-assisted Animation," Computer Graphics:ACM-SIGGRAPH '78 Conference Proceedings. (4), pp. 348-353.

91. William T. Reeves (1981). "Inbetweening for Computer Animation Utilizing Moving Point Constraints," ACM SIGGRAPH '81 Conference Proceedings, Computer Graphics, 15, (3), p. 263.

92. David Sturman (1986). "A Discussion on the Development of Motion Control Systems" Advanced Computer Animation, SIGGRAPH '86 Tutorial Notes. p. 6.

93. Reynolds, op. cit., p. 76.

94. David Zeltzer (1984). Representation and Control of Three Dimensional Computer Animated Figures. (Unpublished doctoral dissertation, Ohio State University), p. 4.

95. Ibid.

96. Julian E. Gomez (1985). Computer Display of Time Variant Functions. (Unpublished doctoral dissertation, The Ohio State University), p. 35.

97. David Zeltzer (1985) "Towards an integrated view of 3-D computer animation," The Visual Computer, 1, (4), p. 251.

98. David Sturman (1987). " Discussion on the Development of Motion Control Systems," SIGGRAPH 1987 Tutorial N o . 10, Computer Animation: 3^-D Motion Specification and Control. p. 4 .

99. Reynolds, op. cit. 102

100. Zeltzer, op. cit.

101. William T. Reeves (1981). "Inbetweening for Computer Animation Utilizing Moving Point Constraints." ACM SIGGRAPH '81 Conference Proceedings, Computer Graphics, 15, (3), p. 264.

102. Gomez, op. cit.

103. P. Hanrahan and David Sturman (1985). "Interactive animation of parametric models." The Visual Computer, 1, 260-266.

104. Reynolds, op. cit.

105. Julian E. Gomez (1985). Computer Display of Time Variant Functions. Unpublished doctoral dissertation, The Ohio State University, p. 13.

106. Thomas A. Defanti (1973). The Graphics Symbiosis System — An Interactive Minicomputer Graphics Language Designed for Habitability and Extensibility. (Unpublished doctoral dissertation, The Ohio State University).

107. Julian E. Gomez (1985). Computer Display of Time Variant Functions. (Unpublished doctoral dissertation, The Ohio State University).

108. Lance Williams (1982). "BBOP," Tutorial Notes, Three- Dimensional Computer Animation, ACM SIGGRAPH '82.

109. R. Chuang & Glenn Entis (1983). "3-D Shaded Computer Animation-Step-by-Step," IEEE Computer Graphics and Applications, 3, pp. 18-25.

110. Williams, pp. cit.

111. T. W. Calvert, J. Chapman, & A. Patla (1980). "The Integration of Subjective and Objective Data in the Animation of Human Movement," Proceedings ACM SIGGRAPH '80, Computer Graphics, 14, pp. 198-203.

112. C. Ginsberg & D. Maxwell (1983). "Graphical Marionette," Proceedings ACM SIGGRAPH/SIGART Workshop on Motion, pp. 172-179.

113. L. Weber, S. W. Smoliar, & Norm Badler (1978). "An Architecture for the Simulation of Human Movement," Proceedings ACM Ann Arbor Conference. pp. 737-745.

114. Gary Demos, Maxine Brown, & R. A. Weinberg (1984). "Digital Scene Simulation: The Synergy of Computer Technology and Human Creativity." Proceedings of the IEEE, 103

72, (1), p. 28.

115. Jane P. Wilhelms (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley, p. 24.

116. Norm I. Badler (1982). "Human Body Models and Animation," IEEE Computer Graphics and Applications, 2, (11), p .7.

117. Demos, ojo. cit., p. 28.

118. Wilhelms, oja. cit.

119. Dick Lundin (1982). "3D Modeling, A Personal Orthodoxy," ACM SIGGRAPH '82, Course Notes, Three-Dimensional Computer Animation.

120. Julian E. Gomez (1985). Computer Display of Time Variant Functions. Unpublished doctoral dissertation, The Ohio State University, p. 13-14. 121. Ibid.

122. Ibid.

123. Craig W. Reynolds (1982). "Computer Animation with Scripts and Actors," ACM Computer Graphics, 16, (3), pp. 289-296.

124. N. Magnenat-Thalmann & D. Thalmann (1985). Computer Animation : Theory and Practice. Tokyo: Springer-Verlag, p. 76.

125. T. J. O'Donnel, and A. J. Olsen (1981). "GRAMPS - A Graphics Language Interpreter for Real-Time, Interactive, Three-Dimensional Picture Editing and Animation," Proceedings ACM SIGGRAPH '81, Computer Graphics, 3,.(15), pp. 133-142.

126. Thomas A. Defanti (1973). The Graphics Symbiosis System — An Interactive Minicomputer Graphics Language Designed for Habitability and Extensibility. Unpublished doctoral dissertation, The Ohio State University.

127. David Zeltzer (1982). "Motor Control Techniques for Figure Animation," IEEE Computer Graphics and Applications, 2, (11), p. 32-42.

128. R. Hackathorn (1977). "ANIMA II: A 3-D Color Animation System," Proceedings SIGGRAPH '11, Computer Graphics, 11, (2), 54-64. 104

129. Ronald Hanrahan, Rick Parent, Bob Marshall, & Mark Howard. (1981). "An Interactive Microcomputer Based 3-D Animation System," Proc. Conference Canadian Society for Man-Machine Interaction, pp. 181-191.

130. T. J. O'Donnel and Arthur J. Olson (1981). "GRAMPS -- A Graphics Language Interpreter for Real-Time, Interactive, Three-Dimensional Picture Editing and Animation,11 ACM SIGGRAPH '81 Conference Proceedings, Computer Graphic, 15, (3) .

131. Reynolds, op. cit., p. 290.

132. David Zeltzer (1985). "Towards an integrated view of 3-D computer animation." The Visual Computer, JL, (4), p. 255.

133. Thalmann, op. cit.

134. Zeltzer, op. cit.

135. Hanrahan, op. cit.

136. Julian E. Gomez (1985). Computer Display of Time Variant Functions. Unpublished doctoral dissertation, The Ohio State University, p. 17.

137. Ibid., p. 18.

138. Ibid.

139. Zeltzer, op. cit.

140. Craig Reynolds (1986), "Advanced Computer Animation," ACM SIGGRAPH '86, Tutorial Notes, Advanced Computer Animation, p. 205.

141. David Zeltzer (1984). Representation and Control of Three Dimensional Computer Animated Figures. Unpublished doctoral dissertation, Ohio State University, Columbus, OH.

142. J. Korein, G. Radack, and N. Badler. (1983). TEMPUS User Manual, Unpublished, Dept, of Computer and Information Science, University of Pennsylvania, Philadelphia, PA.

143. Michael Girard (1986). "Computer-Animated Legged Animal Motion," ACM Workshop on Interactive 3D Graphics, Chapel Hill, NC.~, p p 3,-20.

144. John Chadwick & Rick Parent (1988). "Controlling the Integration of Computational Models for Character Animation," National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, 105

pp. 423-432.

145. David Zeltzer (1985). "Towards an integrated view of 3-D computer animation." The Visual Computer, 1, (4), p. 255. 146. Ibid.

147. Ibid.

148. N. Magnenat-Thalmann & D. Thalmann (1985). "Three- Dimensional Computer Animation: More an Evolution Than a Motion Problem," IEEE Computer Graphics and Applications, J5, (11), p. 48.

149. Zeltzer, op. cit., p. 257.

150. Peter H. Greene (1972). "Problems of Organization of Motor Systems," Proqress in Theoretical Biology, 2, pp. 303-308.

151. R. Tomovic and R. B. McGhee (1966). "A Finite State Approach to the Synthesis of Bioengineering Control Systems," IEEE Transactions on Human Factors in Electronics, HFE-7, 2, (6), pp. 65-69.

152. Zeltzer, op. cit., p. 258.

153. David Zeltzer (1984). "Issues in 3-D Computer Character Animation," Introduction to Computer Animation, SIGGRAPH f 84 Course Notes, 8_, (1), p. 267.

154. Craig W, Reynolds (1987). "Flocks, Herds, and Schools: A Distributed Behavioral Model," ACM SIGGRAPH '87 Conference Proceedings, 21, p. 25-34.

155. Susan Amkraut, Michael Girard, & George Carl (1985). "Motion Studies for a work in progress entitled 'Eurythmy' in SIGGRAPH Video Review, Issue 21.

156. Gary Demos, Maxine Brown, & R. A. Weinberg (1984) . "Digital Scene Simulation: The Synergy of Computer Technology and Human Creativity." Proceedings of the IEEE, 72, (1), p. 23.

157. Julian E. Gomez (1985). Computer Display of Time Variant Functions. Unpublished doctoral dissertation, The Ohio State University, p. 28.

158. Craig W. Reynolds (1986) "Advanced Computer Animation," ACM SIGGRAPH '86, Tutorial Notes, Advanced Computer Animation, p. 203. 106

159. Ibid.

160. N. Roberts, D. Anderson, R. Deal, M. Garet, & W. Shaffer. (1983). Introduction to Computer Simulation, A System Dynamics Modelincr Approach. Mew York: Addison-Wesley. p. 3.

161. Ibid., p. 30.

162. A. M. Colella, M. J. O'Sullivan, & D. J. Carlino (1974). Systems Simulation. Washington, D.C.: Lexington Books, p. 1 .

163. W. H. Press, B. P. Flannery, S. A. Teukolsky, & W. T. Vetterling (1988) Numerical Recipes in Cf The Art of Scientific Computing. Cambridge, England: Cambridge University Press, p. 539.

164 . Ibid.

165. Ibid., P- 545.

166. Ibid., P. 546.

167. Ibid., P. 530.

168. Ibid., P- 531. 169. C. A. Csuri, J. Blinn, J. Gomez, N. Max, & W. Reeves (1983). "The Simulation of Natural Phenomena," ACM Computer Graphics, 17, (3), p. 138.

170. Dave H. Haumann, (1988). "Animating Using Behavioral Simulation," National Computer Graphics Association Proceedings, Anaheim, CA, p. 678.

171. See SIGGRAPH 1983 poster for example of actual test object and the computer generated test object.

172. J. P. Crutchfield, J. D. Farmer, N. H. Packard, fi R. S. Shaw. (1986). "Chaos," Scientific American, 38, (12), 49.

173. Ibid., p. 46.

174. Ibid., p. 48.

175. Craig Reynolds (1986) "Advanced Computer Animation," ACM SIGGRAPH '86, Tutorial Notes, Advanced Computer Animation, p. 200.

176. Julian E. Gomez (1985). Computer Display of Time Variant Functions. Unpublished doctoral dissertation, The Ohio State University, p. 14. 107

177. Reynolds, op. cit., p. 200.

178. Ibid.

179. Ibid.

180. Ibid.

181. Norm Badler and J. U. Korein (1982). "Techniques for Generating the Goal-Directed Motion of Articulated Structures," IEEE Computer Graphics and Applications, 2_, (11), pp. 71-81.

182. S. C. Frautschi, P. P. Olenick, T. M. Apostol, & D. L. Goodstein (1986). The Mechanical Universe, Mechanics and Heat. Cambridge, England: Cambridge University Press, p. 113.

183. S. R. Diamond (1970). Fundamental Concepts of Modern Physics. New York: Amsco Publications, p. 97.

184. Ibid., p. 98.

185. Frautschi, op. cit., p. 116,

186. Diamond, pp. cit., p. 103.

187. David Sturman (1986). "A Discussion on the Development of Motion Control Systems," Advanced Computer Animation, SIGGRAPH ' 86 Tutorial Notes. p. 3.

188. John Chadwick & Rick Parent (1988). "Controlling the Integration of Computational Models for Character Animation," National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, p. 424.

189. Ibid., p. 424.

190. Michael Girard (1986). Interactive Design of 3-D Computer-Animated Legged Animal Motion. University of North Carolina Motion Workshop, Chapel Hill, NC., p. 1-20.

191. Ibid.

192. Ibid.

193. Chadwick, op. cit., p. 425.

194. Ibid.

195. Girard, pp. cit., p. 6. 108

196. Jane R. Wilhelms (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley, p. 9.

197. Ibid., p. 3.

198. Dave R. Haumann (1987). "Modeling the Physical Behavior of Flexible Objects," pp. 1-13. SIGGRAPH '87 Tutorial Notes No. 3/7, Topics in Physically-Based Modeling.

199. James K. Hahn (1988). "Realistic Animation of Rigid Bodies," ACM SIGGRAPH '88 Conference Proceedings, Computer Graphics, 22, (3), pp. 1-8.

200. D. Terzopoulos, J. Platt, A. Barr, & K. Fleischer (1987). "Elastically Deformable Models," ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (4), pp. 205-212.

201. Keith Waters (1987). "A Muscle Model for Animating Three-Dimensional Facial Expression," ACM SIGGRAPH '87 Conference Proceedings, 21, (4), pp. 17-24.

202. Brian Guenter (1988). A System for Simulating Human Facial Expression. Unpublished manuscript.

203. W. W. Armstrong and M. Green (1985). "The Dynamics of Articulated Rigid Bodies for Purposes of Animation," Proceedings Graphics Interface '85, Montreal, pp. 407-416.

204. A. Witkin, K. Fleischer, & A. Barr (1987). "Energy Constraints on Parameterized Models," ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (4), pp. 225-229.

205. Michael Girard (1987). "Interactive Design of 3D Computer-Animated Legged Animal Motion," IEEE Computer Graphics and Applications, 1_, (6), pp. 39-51.

206. Chadwick, o£. cit., p. 425.

207. Issac, op. cit., p. 215.

208. Paul M. Issacs, & M. F. Cohen, (1987). "Controlling Dynamic Simulations with Kinematic Constraints, Behavior Functions and Inverse Dynamics," ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (3), p. 219.

209. Chadwick, 0£. cit., p. 426. 109

210. J. R. Wilhelms (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley, p. 3.

211. Ibid.

212. Ibid.

213. Ibid., p. 4.

214. D. E. Knuth (1973). "Computer Science and. Mathematics", American Scientist, 61, (6), p. 6.

215. J. Gips and G. Stiny (1979). "An investigation of algorithmic aesthetics", In F. Malina (Ed.), Visual art, mathematics and computers: selections from the journal Leonardo New York: Pergamon Press, p. 94.

216. Ibid., p. 96.

217. Ibid.

218. Kenneth M. Kahn (1979). The Director Animation Langugage. (Unpublished doctoral dissertation, Massachetts Institute of Technology).

219. Dave Haumann (1989). Unpublished manuscript.

220. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH ' 86, Tutorial Notes, Advanced Computer Animation, p. 89.

221. Haumann, op. cit. CHAPTER IV

REVIEW OF INTERDISCIPLINARY FACTORS AND SYSTEMS

This chapter reviews those factors leading to the integra­ tion of physically-based simulation as animation. The informa­ tion in this chapter explains how simulation systems work.

This chapter surveys the effectiveness of the new experimental simulations for generating visual imagery in contrast to current methods of generating computer animation. In looking at the results that "simulation'* programs have generated it is necessary to be cognizant of the new set of questions and con­ cepts that arise.

In the past twenty years the field of computer graphics has focused mainly on image synthesis. The problems of render­ ing and display have taken priority over deficiencies in motion generation. In the past several years a shift in priorities has taken place and problems of motion generation are receiving considerable attention. This is the result of several factors: image rendering has reached a high level of sophistication and the rediscovery that an object's characteristic movement con­ tributes greatly to its identification. In fact Thomas [1] believes that an object's facade has less bearing on its

110 I l l

identification than its motion. With this new emphasis being

placed on the problem of generating motion, "simulation" has

emerged as the most viable alternative to the explicit methods

of specifying motion in the computer.

Examining current examples of motion specification,

several areas emerge which would benefit from a simulation

approach. Human motion is difficult to simulate for two funda­

mental reasons. One reason is that the human body is extremely

complex, with several hundred joint-angle degrees of freedom

[2]. In a system with many degrees of freedom, there is a

staggering number of possible joint transformations that can

result in a satisfactory solution. The second reason is that movement patterns are very familiar to our eye. An observer

can easily recognize when a motion appears unnatural or mechanistic. For these reasons traditional animators have fre­

quently depended on the technique of rotoscoping to assist them with creating credible motion.

Traditional animators emulate the physical laws that con­ trol the behavior of objects. Computer animators continue in that tradition and are limited as well by the process of expli­ citly specifying positions, either through manual methods or through the use of motion recording devices. These methods are prone to errors of subjective judgement, and the tedium of trial and error. Accordingly, simplification and economy have 112 become a necessity for the sake of practicality.

The prevalent method of motion specification is to expli­ citly focus on "one" object or appendage at a time. What hap­ pens when there are a number of interacting objects or inter­ connected segments? The result is that not only does the task of animating a large collection objects become impossible using an explicit approach but also the subsequent modification of the animation turns into an overwhelming task because objects must appear to interact. As a result, most scenes exhibit a lack of interesting, meaningful motion. For the viewer, infor­ mation is lacking concerning the properties of the object as it responds to its environment [3].

Today the computer has the capacity to actually simulate

* physical laws. These "physical" laws (i.e. Newton's) manifest themselves in observable forces such as gravity and friction.

Motion can be initiated by a living organism as an act of will or as a response to its immediate physical surroundings.

To facilitate the animation of complicated data requires a procedural (implicit) approach in contrast to a guiding (expli­ cit) approach. Motion that is generated by procedural rule can produce complex realistic motion. Procedural methods rely on the computer to determine motion based on implicit instructions rather than by explicit positions. The procedural approach has the advantage of being able to generate complex motion from a small amount of user supplied information. For instance, simu­ 113

lation techniques (e.g. Robust, Stochastic) can also be used to produce variations from the "same" motion-generating procedure.

Motion procedures can also be parameterized so that the same procedure could be used to generate appropriate motion for dif­ ferent physical structures (i.e. rigid, deformable).

Polygon based computer systems require data structures which are too complex and detailed to be designed efficiently by an interactive user. Reeves [4] proposes the development of procedurally-based modeling domains, and, in particular, those based on stochastic processes [5]. "With such tools, the interactive user only specifies global constraints and parame­ ters and the procedural and stochastic elements are responsible for generating the actual images [6]." The rest of this chapter will describe procedural approaches that have recently shown promise for animation applications.

4.1 Abstractions

Using simulation as an animation technique requires an

"abstract" understanding of the process of generating motion by computer. In looking at different animation systems it is often confusing to see how they interrelate, to know what the commonalities are. The following abstractions, first proposed by David Zeltzer [7], will act as reference points for con­ structing a functional model for animators. There are five classes of abstraction to be considered in computer animation: structural, procedural, functional, character, and 114

world modeling.

Structural abstraction delineates the structural proper­ ties of an object (e.g. the transformation hierarchy, the con­

straints on joint motion, rigid or non-rigid links). Struc­ tural representations of 3D transformation hierarchies have been implemented in systems such as sa [8] and ASAS [9].

Procedural abstraction is the representation of a movement independent of an object's structure [10]. Kinematic and dynamic motion can be implemented as procedures and would be considered procedural abstractions. An example of procedural abstraction would be the computation of a trajectory for a fal­ ling object using a procedure based on dynamic motion. It is procedures that use processes such as dynamics or kinematics which effect "resolved motion" [11]. "Resolved motion" occurs when the position and the orientation of a target location is input and the computer automatically computes how to get to that target. Using procedures to solve the degrees-of-freedom

(i.e. inverse kinematics) problem for joint-angle specifica­ tions would likewise be resolved motion. Resolved motion is an important tool because humans are not able to accurately specify the necessary joint angles for even a few joints, much less the 200+ in a figure as complex as a human. Resolved motion is independent of a particular structure. This indepen­ dence is an important feature which is discussed further in

Chapter five and six. 115

In the case of a character with many links, it would be desirable to be able to group together both structural elements and procedures for a particular class of motions. Zeltzer [12] refers to this association as a functional abstraction. Func­ tional abstractions are valuable because they allow the anima­ tor to assemble "motor skills" [13]. If the general "shape" of the motion is known, then only a subregion of the "structural" hierarchy needs to be considered. For example, a hand "grasp­ ing" an object - a useful motion that would probably be needed often - can be constructed from knowing which joints need to move and approximately how they should move. This cluster of joint movements would be composed

...around the task "to grasp", and attach one or more procedures to implement it (perhaps resolved motion). Once this motor skill has been defined, the details of its execution can be suppressed. That is, only the ap­ propriate parameters need to be supplied, e.g., target location, fast or slow, hard or soft, to the motor pro­ gram for the grasping skill [14].

The animator is relieved from the burden of explicit specifica­ tion by defining functional abstractions for motor skills like

"grasping" and can instead conceive of figurative motion at a higher level.

Functional abstractions allow us to attach implicit goals to figure options. By decomposing a figure's po­ tential movements into a repertoire of skills the events and relationships can be associated with the specific skills (implemented as functional abstrac­ tions) that the figure controller "knows" about. More­ over, if functional abstractions refer to other func­ tional abstractions, it is possible to construct behaviors as compositions of simpler movements [15]. 116

4.2 Behavior - Physical

Until recently, motion specification in the computer has almost exclusively dealt with rigid monolithic objects organ­ ized in a hierarchical fashion [16]. Today, animation is branching out into new structural classes. The phenomena under study here can be classified into three structural categories:

(a) "structures" that contain rigid links (mammals, fish, trees, etc.), (b) "structures" that are bound together by sur­ face constraints (cloth, rubber, skin, clay, etc.), and (c)

"structures" that are not "true" structures but are actually processes. For example, the complex surface of a cloud is the result of local, temporal fluctuations of such conditions as temperature, humidity, and pressure. Under proper physical conditions diffuse water molecules condense into water vapor that is translucent, highly reflective, and in constant tur­ bulence. The perceived surface of a cloud is in reality the consequence of dynamics, not a fundamental property of clouds.

Static representations or fractal clouds miss the inherent dynamic aspect of clouds by only reproducing the visual epi- phenomenon [17].

Each structural class has inherent motion characteristics that we identify with the form. The advocated solution to the problem of specifying motion for these structures is to simu­ late the object's physical behavior. 117

4.2.1 Behavior of Flexible Surfaces

4.2.1.1 Flexible Objects - Constrained Surfaces

Haumann [18] has created a system, Dynaflex, based on the physical properties of flexible surfaces and their behaviors.

It uses dynamic simulation as the vehicle for motion. This approach differs from other dynamic simulations [19,20] in its emphasis on procedurally simulating the behaviors of animated objects as opposed to explicitly seeking a numerical solution to a system of equations. Dynaflex [21] also differs because it addresses the important need of generality for the animator.

A flexible object is defined as a clearly delineated surface constructed around interrelated elements (e.g. masses, springs, hinges). The resulting motion is controlled by manipulating the properties of these elements and their interrelationships.

In Dynaflex, the surface of an object is not only capable of bending or stretching when acted upon by an external force but can also oscillate and return to it previous form. Examples would include a soft rubber ball, cloth, string, etc.

Haumann [22] makes the case that interactions which often appear complex at the macro level are the result of more ele­ mentary interactions on the micro level. It is important to examine this level to understand what is actually taking place.

Initially, a flexible object is viewed 11 ...as exhibiting the following internal properties: resistance to changes in motion, resistance to bending, resistance to stretching [23]." In 118 addition, an object must respond appropriately to the external effects of gravity, the effects of the surrounding fluid (i.e. wind, water), and the effects of collisions with other objects

(such as ground contact). From this criteria Haumann concluded that flexible objects can be modeled by approximating the mass of an object and the forces which act on that mass.

A flexible object is synthetically represented in the com­ puter as masses assigned to the vertices. Initial conditions such as velocity and acceleration are assigned. Internal forces are represented as springs (resistance to stretching) and hinges (resistance to bending) between edges and masses.

This representation (Figure 12) maintains internal coherence which acts to hold the object together as well as maintain its shape. External forces are used to represent aerodynamic drag(vector fields which interact with the surface area of the

HBMBB

Figure 12. Spring and Mass representation. Chris Wedge. 119 triangular polygons that make up the object), collision detec­ tion (proximity of object and environment surfaces), and grav­ ity (a constant downward force). Haumann [24] has proposed that an animator should look at the situation as consisting of two levels of detail.

(1) at a coarse level for external constraints - for example:

a complex object is related to the air by drag and to the

ground by both gravity and contact.

(2) at a fine level for internal constraints - mass elements

interconnected by spring and hinge elements to maintain

internal object coherence. Explicit motion specification

is available in the form of control points.

Dynaflex operates through the following implementations:

(1) Resistance to changes in motion - "inertial", naive

"weight" This is accomplished through the assigning of a

point mass representation and directly related to Newton's

first two laws (inertia and F = ma).

(2) Resistance to being pulled apart - "stretchability" A

spring representation that constrains two masses is used.

This upholds Newton's third law (action-reaction).

(3) Resistance to bending - "stiffness". Masses are assigned

to the vertices of the default triangularized polygonal

surfaces that represent the object. Hinge constraints are

specified between the polygonal surfaces. The constraint

parameters control the degree of stiffness in the object

and preserve angular momentum. 120

(4) Resistance to motion through the surrounding fluid -

"damping". This is represented by surface area in direct

relationship with external velocity vectors. This force

(i.e. velocity and orientation) interaction is propor­

tionately propagated through the object. This charac­

teristic is used to simulate wind and water.

(5) Effects of gravity - Gravity is simulated by the attrac­

tion function between two masses.

(6) Collision Detection - Proximity of mass and surface is

computed to determine contact or collision with self­

intersection or external intersection (collision detection

between point mass and surface) [25].

While it is true that in a "simulation" the motion has to simply run its course; the course that it runs and the resul­ tant motion that is output depends on how the elements are hooked together and the magnitude of the parameter constraints.

Parameter values can be set at the global and local level.

This is important because an animator could, for example, assign stiffness or mass to one localized point of the object rather than to all the points. This results in a more idiosyn­ cratic and artistic reaction to the forces present. This flex­ ible object simulation program has been used to create an artistic animation.

The most novel use has been to help create the goofy characters created by Chris Wedge for the film "Balloon Guy". The motion of the hair, ears, eyes, tongues and attached strings were all controlled by a simulated response to the character's movements. The resulting 121

animation is a comical display of wind blown hair, flopping ears, bobbing eyes, dangling tongues, and trailing strings. All of this motion , painful and nearly impossible to specify manually, was easily gen­ erated within a few hours. The only information that was needed to drive the simulation for each character was a script of position and orientation for each frame of animation [26].

Control points function as constraints around which the dynam­ ics of the animation "revolved" (figuratively and literally).

(Figure 13) An epiphenomenon of this program is its inherent generation of the "Principles of Animation".

As each character spins or bobs the attached hair, ears, and strings, realistically reacts to the charac­ ters motion. In effect, the motion techniques demanded of the classical animators were automatically repro­ duced: squash and stretch, follow through, ease in and ease out [27].

Figure 13. Control points for "Ballon Guy" character. 122

The dynamic properties of this program provide an impor­ tant creative tool for exploration and experimentation. As

Haumann explains,

During the making of "Balloon Guy", we tried a variety of physical conditions. For example, increasing the mass of the hair would exaggerate the follow through of the hair, when the character stopped and started. In­ creasingly the aerodynamic drag on the string would cause it to act more like thread, floating on the air. Decreasing the drag caused it to look more like a dan­ gling chain being whipped about. All these experiments demonstrated to us that our single behavioral model of physical systems was capable of producing a wide variety of effects by simply changing a few physical parameters, and re-running the simulation [28].

4.2,1.2 Flexible Surface - Facial Animation

Facial animation is a special case application of a flexi­ ble surface. There are two basic approaches that have been implemented: key-frame [29] and parameterization [30,31,32].

The key-frame approach requires not only the construction of a complete facsimile of the face for each key-frame, but also requires the extremes of an expression from which to interpo­ late. Figure 14 shows key-frame expressions used in the anima­ tion "Tony DePeltri". Key-frames are labor-intensive and have a limited ability to vary from extremes. This limitation results in motion which may be perceived as mechanical or rigid when compared to reality.

The parameterization approach groups vertices together according to functional patterns (Figure 15). Key-nodes are established for the natural extremes of expression and maximum Figure 14. Key-frame facial animation. Daniel Langlois. and minimum zones of influence upon surrounding areas. Move­ ment is generated about these key-nodes (control points) which are embedded within the surrounding vertex groupings. When the control point is moved this change is propagated throughout the grouping. Individual vertices may be proportionally 124

Figure 15. Facial animation by parameterization. Brian Guenther. constrained to different vertex groupings. These parameteriza­ tion systems have primarily focused on straightforward kinematic descriptions of the skin. It is now becoming apparent that the inherent dynamics of the face must also be accounted for if realistic expressions are to be generated

[33]. Facial expressions are more than the kinematic transla­ tion of polygons.

Current implementation schemes rely on kinematic descrip­ tions based on the Facial Action Coding Scheme by Paul Ekman

[34]. "The Facial Action Coding System (FACS) is a notational-based environment that determines emotional states from the visible facial distortion. Individual muscles, or groups of muscles that operate in unison, are described as 125

Action Units (AU) that distort the skin tissue [35]." The six

basic expressions are anger, fear, disgust, happiness,

surprise, and sadness. From these six expressions almost all

other expressions can be derived. These other facial expres­

sions are composed of different combinations and magnitudes of

the basic action units that make up the original six expres­

sions. To generate different and more idiosyncratic expres­

sions the animator adjusts the magnitude of selected action

units.

The facial system contains a library of facial expressions

that are needed frequently and are foundational to other more

subtle expressions. On one hand this default feature lends

itself to global control, on the other hand the artist needs to

maintain explicit control over the facial parameters and under­

stand the parameters in order to prevent self-determination by

the system.

A facial animation system is a good example of the complex

nature of computer simulation. Not only must the program be

dynamic enough to produce the subtlety required, but to be able

to effectively use that subtlety the animator must have a knowledge of facial expressions and how the program is simulat­ ing them. For example, a facial expression generally consist of six to seven characteristics (i.e. wrinkled nose in an expression of disgust) [36]. An expression can either be called from a default lookup table or set specifically by the animator. The animator should know to a certain extent, how 126

this will affect the expression. Graphically, the animator

needs to be aware that the surface is a polygonal mesh with vertices and nodes that have a "finite degree of mobility".

The primary factors determining a node's mobility are:

1. tensile strength of the muscle and skin 2. Proximity to the muscle node of attachment 3. Depth of tissue at the node and the proximity to the bone. 4. The elastic bounds of the relaxed tissue, and the interaction of other muscles [37].

All these factors together will determine the degree of control and subsequent success a user will have.

4.2.1.3 Deformation

A deformation may be thought of as a flexible surface or volume that does not necessarily regain its original shape after a force has been applied. The form of an object changes shape in direct proportion to the external forces acting upon it (Figure 16). Surface deformation can be applied to a variety of objects: amorphous objects (fluids), organic objects

(trees, jello, muscles), and inorganic objects (plastics, soft metals, cloth).

Surfaces or volumes are deformed through the manipulation of control points which are mapped (i.e coordinated with the actual representation of the object) to the object. Shape manipulation techniques have been pioneered by Wein [38],

Parent [39], and more recently by Sederberg and Parry [40] with their free-form deformation (FFD) technique. Deformations can 127

G - G - G -

Figure 16. Elastic deformation of masses. Kurt Fleischer.

be applied both locally and globally, and maintain continuity

of surface and volume.

A significant advantage of applying deformations is that

it is highly intuitive for artists. Deformations transforms

the objects through a sculpture metaphor. This method is

geometrically independent from the objects it is applied upon

and can be used for a variety of objects such as skin or mus­

cles [41]. Terzopoulos [42] has created a physically-based

deformation technique which holds promise for future integra­ tion into a general physically-based animation system. 128

4.2.2 Articulated Structures

Articulated structures in computer animation are viewed as segmented parts connected by linkages. There are several modeling and motion generation techniques that can be applied to articulated structures. These include techniques specific to figurative motion, techniques specific to generative motion, and techniques that utilize structural constraints which effect the final motion. These procedural techniques provide a new level of productivity and facility not previously experienced.

4.2.2.1 Figurative Motion

Figurative {articulated human structure) motion distin­ guishes itself from the rigid monolithic motion demonstrated by most objects in two ways: (a) the figure is composed of articu­ lated links which can function individually or collectively and

(b) a figure can initiate motion in accordance with desired goals. This latter quality also distinguishes figurative motion from the reactionary motions to which most other struc­ tures are confined. There have been several systems which have attempted to produce figurative motion. TEMPUS [43] is a sys­ tem for the analysis and display of realistic human motion in a workspace. TEMPUS differs from general computer animation pro­ grams like MIRA [44] and ASAS [45] in that TEMPUS is restricted to the positioning and orientation of human figures and permits a large degree of device mediation. Movements available are the rotation and translation of the figure, rotations at Figure 17. Low-resolution figure from TEMPUS program. selected joints, and resolved motion[46]. Resolved motion per­ mits limbs to be positioned by the system from general specifi­ cations supplied by the user. Zeltzer believes that this implementation of a flexible resolved motion algorithm for positioning the limbs is an important step towards goal- directed motion [47].

Each system has generally based its solution to motion generation on the simulation of a particular class of motion

(e.g. dynamics or kinematics). The following describes several representative systems. Zeltzer [48] built a system (i.e.

"sa") based on forward kinematics. His end goal was a con­ venient, goal-directed, extensible system. Though he did not reach that goal many of the ideas that evolved from that effort helped pave the foundation for current efforts in that direc­ tion. 130

Zeltzer made significant progress in simulating actual physiological control mechanisms in conjunction with mechanical control strategies (forward kinematics) from robotics litera­ ture. Locomotion is resolved systematically from the lowest level (i.e. fingers) on up through each sequential level (i.e. hand to elbow to shoulder, etc.) of the control hierarchy. The system was designed for the general input of articulated struc­ tures (e.g. figures, machines, even imaginary creatures). Out­ put is in the form of skeletal structures which permits the user to preview (close to real time) the essence of the result­ ing movement [49]. The actual physical data (i.e. shape of limbs) is kept separately and used only for the final display of the image.

Figures are defined using a programming-language deriva­ tion which describes the joints, the range of movement at the joints, and joint connectivity [50]. This use of a derivative language makes a programming background highly desirable. The skeleton specification includes a "declarations" section and a

"descriptions" section [51]. The "declarations" section iden­ tifies each joint and its associated constraints with that joint (joint rotation limits). For example,

/* Skull and its Rotations */

skull: x -60 60 y -180 180 z -60 60

The description section, describes how two or more joints are structurally related. For example, a hand would be described 131

begin wrist (thumbl thumb2) (index 1 index 2 index 3) (middlel middle2 middle3) (ringl ring2 ring3) (littlel little2 little3) end /* wrist */ [52]

This example of a hand is composed of a wrist and five fingers.

The thumb description has two joints (thumbl and thumb2); the

rest of the hand description closely reflects the geometry of a

real hand. This information is used by the system to maintain

logical relationships between the parts. Because of the highly

specialized nature of particular structures and their associ­

ated motion it is likely that other similar grammatical

languages will evolve. In contrast, a key-frame system would

have required the user to maintain these relationships manu­

ally. This system, "sa", was designed at The Ohio State

University to operate with data in the form of procedural descriptions and output animations of articulated figures walk­

ing, running or doing somersaults.

Wilhelms [53] and Armstrong [54] have simulated articu­ lated limbs through the use of forward dynamics ("Deva"). They feel that most previous systems that have incorporated simula­ tion have specified motion kinematically (without consideration of the forces producing motion) which has produced jerky, unna­ tural motion. In the approach of Wilhelms and Armstrong, "one simulates the physical properties of the limb and solves its motion from the torques applied at each of the joints [55]." 132

(Figure 18) They believe a more natural motion is produced by

this method.

These representative examples of forward kinematics ("sa")

and forward dynamics ("Deva") both operate through interpola­

tion strategies. They have had limited success as animation

tools because joint-based strategy does not provide an "intui­

tive" interface for the user. The inherent problem with this

strategy is that the system operates solely in joint-space; the path, speed, and acceleration of the limb's end-effector (hand

or foot) can not be precisely controlled [56] . The most likely

strategy that will meet with success is based on both inverse

z z

y '

Figure 18. Sequence based on dynamic modeling. Jane Wilhelms 133 kinematics and inverse dynamics. Inverse kinematics is based

on end-effector goals which simulate the "conscious" component of how our limb movements are coordinated to grasp the a physi­ cal target (e.g. hand [end-effector] to doorknob [goal]) [57].

We intuitively conceive in terms of what we want to do, not how we do it.

Girard has constructed an integrated approach (PODA) that

incorporates not only kinematic motion but also limb and body dynamic control strategies for interactive limb positioning

(Figure 19). Its three main functional components are:

(1) According to Girard [58] movement of the end-effector (i.e. hand, foot) results when the animator positions or orients the end-effector specifically in space. In this mode, the program solves (by inverse-kinematics) for the joint angle rotations necessary to bring the end-effector to the new desired position. This process is essential for placing feet and hands at specific places in the environment, e.g., placing the feet on the ground, putting the hands on the hips, or cases involving reaching behavior.

(2) Secondly, in movements such as swinging of arms or kicking

of legs it is more "convenient to move a limb by changing

its joint angles rather than positioning its hand or foot

[59]. In such cases, the program generates output in the

form of forward kinematic rotational values to update the

joint angles which in turn result in new limb and new

end-effector positions [60],

(3) Third, to fine-tune the posture of a limb once its end-

effector has been constrained to a position(i.e. foot 134

0 ot 13 body s i 3 of 13 body ss 1 ol 11 body ss

4 of 13 body nun

Figure 19. Sequence based on inverse kinematics. Michael Girard.

anchored to the floor), the animator adjusts the selected

joint-angle, and the program solves (new joint-angles

through inverse kinematics) for the adjustment [61], The

end result is that the program, not the animator, calcu­

lates the translations needed for the joints; the animator

only has to determine goals and gaits.

Kinematic programs do not necessarily conform to any natural laws or characteristic behaviors which govern motion. 135

Girard believes

...this is fine if we are not concerned with producing limb motion which appears to look natural. But if we are we must look for higher-level models of limb move­ ment which embody the dynamics and control strategies employed by real animals. The design of limb motion should be supported by a parametric model which cap­ tures the essential patterns of natural limb behavior [62] .

An important fact overlooked by those who advocate a strictly dynamic approach to articulated figures is that the

"control of coordinated actions by real animals frequently involves a planning stage wherein the animal anticipates its future possible positions [63]." This statement reinforces the previously defined difference between an "object" and an

"entity" that can initiate movement.

In Girard's system, PODA, trajectory look-up tables for limbs have been designed. These look-up tables provide the animator with the means to define a host of trajectories.

These trajectories may be used in many different types of leg- stepping motions, which in turn would establish the character of a figure.

Variations in leg motion occurs between gaits, between animals, and between legs (i.e. multi-legged creatures). For example, consider human locomotion. During walking the legs swing and touch down on the heel; running is often distinguished by a "kick" of the legs which brings the runner's foot high above the knee as his foot leaves the ground [64].

Girard's PODA program allows the animator to associate a given 136 gait with a trajectory from the trajectory table {i.e. leg motion look-up table, arm motion look-up table). "PODA pro­ vides an interface which allows the animator to associate a trajectory from a leg's trajectory table with a given gait

[65J

The user may freely alternate between the three modes

(trajectory specification, trajectory lookup table, gait specification) in order to converge upon some desired limb con­ figuration in PODA. Through this process the figure gradually accumulates constraints upon its degrees-of-freedom.

An animator may design a framework of motion through the use of interactively manipulated posture sequences and interactively accessible dynamic and temporal con­ straints . These parameters provide a control mechanism for establishing the set of postural and rhythmical goals which the anticipatory control strategies embed­ ded in the legged animal motion model are designed to meet [66].

The next step in the pipeline is to control the speed of the limbs.

Although speed may be controlled independent of path, in the context of limb trajectory motion, one typically wants to design the speed function in relation to pos­ tures which define the path. If speed were designed solely over time, changes in acceleration occurring about specific postures could be easily controlled. The animator requires a means of designing speed in re­ lation to specific distances associated with postures alongs the limb's path [67].

Girard's solution is to provide the animator with a dis­ tance time graph control mechanism, a graph which represents 137 distance as a function of time (Figure 20). Spline control in the graph provides continuity in the transitions between move­ ments .

4.2.2.2 Generative Modeling and Motion

New procedural techniques (i.e. stochastic processes) are being developed for generating highly complex objects (i.e. tree). With such synthetic objects comes the dilemma of how to specify realistic motion associated with these objects. This dilemma breaks down into two major concerns:

(1) Explicitly stating the motion transformations to a large

number of data primitives would be unfeasible. It would

be desirable to specify their motions in a general way.

For example, when the wind is blowing, most trees react in

Function*! < LUM J IfDO lU U nf f l m n I HIS I H'H I PHIHM BlNUMSPERn SiooWilngi M M H C«t*ul H*r»1ti (HUT01 0r»w f5HTH»PJTgn Potnti/Edgas 1 — — ' I X C 83 Fr«««i C 10BD 1 ■ I ,naatt^n ttodti add i?3i

Figure 20. Time-speed graph. 138

basically the same way. Armstrong points out that animat­

ing a tree in which transformations would have to be

applied to each leaf and branch "... is clearly not possi­

ble in a reasonable length of time [68]."

(2) If the complex structure of an object is generated pro-

cedurally using stochastic processes, then the exact

structure of the object is not known until after it has

been generated. How would an animator specify explicit

motion of an object whose form is generated by the system

during the rendering step?

Armstrong proposes that procedures should be designed to simultaneously generate motion with an object's structure. In the case of a tree blowing in the wind, a modeling procedure could be linked with motion parameters that indicate the strength and direction of the wind [69] . Motion specification is designed independent of the detailed modeling process. A generative tree structure was used by Armstrong [70] to suggest a method for incorporating motion and data generation (Figure 21) .

In this example, a tree is generated by applying a rule to the branches in each previous generation; "each branch splits into two child branches that diverge at a 45 degree angle

[71]," In this generative process a motion primitive is attached to the data primitive. Each primitive can receive information from the previous primitive and pass on information to the next primitive to be generated. Because the motion 139

Rule generated Tree Structure

Figure 21.

process can receive contextual information from neighboring

generations there exists a tremendous potential for automati­

cally generating motion. With an understanding of this process

the animator can anticipate what is going to take place and how

to creatively direct it.

A useful interactive feature Armstrong implemented was the

ability to choose between the amount of time an image would

take to render versus the degree of realism in the image. If

getting feedback on the generative motion was most important the animator would choose a lower resolution of the displayed

image so calculation would proceed more quickly. By selecting the computations to be performed on a group basis rather than individually the animator can control the amount of computation time required, and thus the realism of the motion [72]. 140

Armstrong also advocates that

...ideally the animator should only specify the values of one or two parameters in order to obtain the motion he needs. The values of these parameters should be in­ tuitively obvious, so the animator does not need to spend a considerable amount of time determining their values by trial and error [73].

4.2.2.3 Constraints

In constructing a simulation, setting the internal parame­

ters for a functional abstraction {structure and procedural

primitives) is difficult and time-consuming.

As the complexity of model increases, the number of parameters becomes large, and they tend to interact in ways that make the model difficult to control. The difficulty lies in finding settings of the parameters that achieve the desired effect. The utility of a sys­ tem would be greatly enhanced if this tedious process could be performed automatically, permitting the user to state in terms of constraints, [italics added] the properties a model is suppose to have, without the need to manually adjust parameters to give it those proper­ ties [74] .

The word "constraint", in the context of this study, is a restriction on the position, orientation, or physical state of the object. For example, a joint rotation may be constrained to 120 degrees of possible rotation. The joint is "con­ strained" not to go beyond that value. Systems may support various constraints. One of the first constraint based systems was Sutherland's Sketchpad. External constraints can dictate how an object reacts to its environment. A "point-to-nail" constraint specifies that an object may swivel about a 141 constrained point, while a "point-to-point" constraint forms a joint between two objects allowing the objects to move about freely, as long as the two constrained points stay in contact

[75]. Internal constraints can describe how the object holds together through the assigning of pseudo springs and hinges

[76]. Constraints provide an effective means of controlling parameterized models. They are amicable to user interaction and can possess a built-in generality.

In a three-dimensional cartesian coordinate system a ver­ tex or joint parameter can have six degrees of freedom (three translational and three rotational). In reality, joint parame­ ters rarely have unrestricted degrees of freedom. "Con-

1. Constraint to be met ball to point A ► 2. Introduce force, force pulls object

constraint force

3. Constraint is met

Figure 22. Constraint force. 142 straints" {limiting the number of degrees of freedom and/or its magnitude) can simulate the physical limitations in our world.

For example, a constraint specification would include a limita­ tion on how far a joint-angle parameter can rotate {i.e. the neck would be restricted to -90 to +90 degrees in the y-axis).

Other examples could include specifications of an attachment to a fixed point in space (Figure 22), surface-to-surface attach­ ment, floating attachment {i.e. a specific point on an object is attached to some point on a second object, allowing the point of contact to slide freely on the second object) (Figure 23) [77].

The animator should be cognizant that isolated constraints are relatively easy to maintain, while multiple and/or interacting constraints might be difficult or impossible to maintain. Solving for one constraint might violate another. \ \ I \ 7 Pendulum andEnds of rods Ends of rods Constraint is Gravityis introduced rod constrained come met compound pendulum together together svvin gs an d c onstraint stays met

Figure 23. Point-to-Point Constraint. 143

It is likely that a constraint system would be composed of a network of dependencies. This would typically result in cycli­ cal dependencies that would make it impossible for all con­ straints to be satisfied at the same time [78]. For every time one constraint is satisfied another is violated. However if underconstrained constraints are adopted a compromising process can be initiated. There are three levels of constraint specif­ ication :

(1) Fully constrained - If the number of independent con­

straints is equal to the number of degrees of freedom

("exactly specified: 'the ball is at the center of the box' ")

(2) Underconstrained - "only partially specified: 'the ball is

near the center of the box or the ball is inside the box'"

(3) Overconstrained - "a possibly contradictory situation:

'the ball is inside the box AND the ball is bigger that

the box' [79]."

Approximate solutions are derived when there are numerous interacting constraints. A methodology for approximate solu­ tions of interacting constraints would

...consist of making a small perturbation of the current state of the object in the "direction" implied by the constraint. Because these perturbations are small in magnitude, they tend to violate their co­ constraints only slightly, they tend to settle into configurations that satisfy nearly all of the active constraints. More energetic constraint systems might require several iterations for each frame, perhaps us­ ing a relaxation technique— the magnitude of the per­ 144

turbations starts relatively large and decreases asymp­ tomatically [80] .

Constraints in a computer animation program address such things as keeping the legs attached to the body. The user specifies high-level constraints, such as whether the body should remain a constant distance above the ground or follow an undulating motion [81].

4.2.3 Particle Systems

A simple representation of a behavioral simulation is a rule-generated, graphic technique known as "particle systems"

[82], A particle system is an assemblage of many individual, minute particles that can collectively represent - through highly complex geometry, motion and color - a "fuzzy" object.

Each particle has its own behavior. This behavior collectively results in a synergistic representation. "Over a period of time, particles are generated into a system, move and change from within the system, and die from the system [83]."

The shape, appearance, and dynamics of the particles are controlled by a set of parameters. These parameters constrain stochastic processes that randomly select each particle's appearance and movement. (Figure 24) "In general, each parame­ ter specifies a range in which a particle's value must lie [84] . » 145

at; par ini' speed & direction

ejection angle /

<-a, typical particle’s initial ______p osition

Figure 24. Particle System. Bill Reeves.

In the generation of a motion sequence the following steps are performed:

(1) New particles are generated in the system

(2) Each new particle is assigned its individual attributes

(3) Any particles that have existed within the system past

their prescribed lifetime are extinguished

(4) The remaining particles are moved and transformed accord­

ing to their dynamic attributes

(5) An image of the living particles is rendered [85].

An advantage to this method is that it is procedural and controlled by stochastic processes. Therefore, obtaining a highly detailed model does not necessarily require a great deal of human design time as is often the case with existing surface-based (i.e. polygonal) systems. Because this approach is procedural, it can "incorporate any computational model that 146 describes the appearance or dynamics of an object [86]." There­ fore, models can be used which have been developed in other scientific or engineering disciplines. They differ from the usual image synthesis methods in three ways:

(1) An object is not defined by a set of primitive surface

elements such as polygons but as "clouds of primitive par­

ticles that define its volume [87]."

(2) A particle system is dynamic. It changes with the passage

of time.

(3) An object "represented by a particle system is not deter­

ministic, since its shape and form are not completely

specified. Stochastic processes are used to create and

change an object's shape and appearance [88]." These

differences permit particle systems to more easily simu­

late objects that explode, flow, splatter, puff up, and

billow.

4.3 Behavior - Environmental

"Environmental" behavior in the context of this study encompasses the interaction between objects, characters, or both. Zeltzer [89] has defined this type of behavior as an

"adaptive motion" abstraction. It allows the animator to describe movement in terms of the interrelationships between objects (one object "adapting" to another). It is these rela­ tionships which determine the resultant motion that will take place in the animation. Behaviors lend generality to animation sequences through the ability of the characters to adjust their 147 motion in response to different environmental scenarios. A

distinct advantage of this approach is that it hides unneces­

sary detail from the animator. The burden of control specifi­

cations is left to animation software.

Zeltzer [90] and Reynolds [91] indicate that the future of computer animation will be in designing microworlds and popu­ lating them with interesting characters. The way animations are created will change, and, in turn, the energy of the anima­ tor will no longer be invested in the tedium of drawing in- between frames.

Objects and figures interact in the physical world in com­ plex ways at many different levels of detail. While "adaptive motion" requires techniques for collision testing and path planning; goal-directed animation control requires the addi­ tional implementation of sophisticated mechanisms for knowledge representation [92]. How is this possible? Initially in the computer, simple rules of behavior are used in the form of

"if-then" statements from programming grammer. More complex relationships necessitate an approach analogous to object- oriented programming.

4.3.1 Object-Oriented Programming

"Object-oriented programming" embodies the concept of functional abstraction. This programming environment consists of constructs (objects) which combine "state" with "behavior". 148

The first well-known object-oriented programming en­ vironment was SMALLTALK. The structure that encapsu­ lates the state (data) with the behavior (programs) is simply called an "object". Typically, objects are specific "instances" of a class— each instance has its own state, but the behavior is common to the whole class. (For example, there might be a class of "bounc­ ing balls," from which several separate instances could be created, each with its own properties such as size, velocity, position) [93].

Objects are transformed by message-passing (i.e. a message that lists the name of a requested action and the variables relevant to it are sent to the object). Message-passing is analogous to function calls in programming methodology, with the message name corresponding to the name of the function.

Object-oriented programming systems usually provide tools to allow various ways to combine and mix classes. (Hence the class "bouncing ball" could be mixed with the class "colored object" to produce a class of "colored bouncing balls") [94].

Reynolds [95] has created an "actor system", specifically for animation, based upon object-oriented programming. In this system the animator acts as a director and the "actors" are the objects being animated. "Actors" constitute a process attached to an object. An actor may be considered as an object or instance from a specific class (data structure with necessary procedures to manipulate that structure). The advantage of this actor/class definition is that it allows the artist to conveniently encapsulate both structure and operation in a sin­ gle definition. Once defined, the artist may then "direct" abstract notions of properties and behaviors which in turn 149 permit the artist to concentrate on a higher level ("what" an object does, not on "how" it does it) [96], These notions con­ tinued to evolve with the concept of "inheritance", where an object "inherits" the properties and behaviors of its parent or some other object with which it is linked.

The development of "actor" systems arose from a increasing realization that describing behavior as a single monolithic problem did not help the problem of complexity. Important aspects apparent in existing natural systems appear to be per­ formed in parallel and therefore information must be communi­ cated in a parallel rather than serial manner.

4.3.2 Self-Scripting

Some animation is made to match a preconceived image. Other times, animation is produced as an experiment, the answer to "what would happen if..." In the second type, which might be called "behavior simulation", the animator sets up a little world by defining the rules of behavior and selecting the cast of characters. When the behavior simulation is run we obtain images of what went on in the little world [97].

This behavioral scenario is accomplished through self­ scripting. One of the first self-scripting animations was done at Atari Systems Research under Ann Marion. The project sought to create an environmental simulation of sea creatures around a coral reef that would be "self-scripting". It is significant that these creatures were required to react in a variety of ways to each other. This reaction was structured such that a meeting of the "crab" and the "clown fish" did not lead to a 150

repetition of the same action. This was solved by assigning a

sufficient amount of information about possible internal states

to each character. This state might be hungry or full, alert

or tired, calm or frightened. The intent was not the quality

of the rendering but the development of the characters and

their interactions [98].

4.3.2.1 Behavior Simulation - Flocking Behavior

Reynolds [99] uses a behavioral simulation in the creation of an animated model of a flock of birds. This technique was derived from research on the actual flocking behavior of fish and birds [100]. The seemingly complex behavior of flocking is the result of simple following and collision avoiding behaviors between individual birds. The synchronized group behavior of birds is in reality discrete units that exhibit fluid motion.

Flock motion is the aggregate result of individual animals.

Problems with creating a computer graphic portrayal of flock behavior using explicit human methods include:

(1) Scripting would be tedious, labor intensive and prone to

error.

(2) If paths could be constructed collisions would probably

occur.

(3) Editing would be difficult.

(4) The result would not be efficient, robust or believable [101]. 151

Reynolds has created a flock model to simulate the way a real flock would operate. What is modeled is the behavior of an individual bird - not the behavior of the flock as a whole.

The flocking behavior is an epiphenomenon, a side ef­ fect of the way each bird behaves. The control of the flock must be distributed since there is no centralized structure that could be in control. This behavioral model would be implemented as an animation character, an object/actor of which any number of copies could be instantiated [102].

Success will be the result of the criteria selected for the simulation. This includes behavioral 'control structure' such as portions of the bird's perceptual mechanism and aspects of the physics of aerodynamic flight [103]. Rules of behaviors that are implemented in such a model include:

(1) Each bird tries to stay close to the main body of the

flock.

(2) Each bird tries to avoid colliding with its neighbors.

(3) Each bird tends to fly in generally the same direction as

its neighbors.

(4) The birds on the leading edge of the flock set the course

(specifically those birds are influenced less by the flock

behind them than by the attractions or dangers of the out­

side world, and the rest of the flock tend to follow them

due to the behaviors in 1, 2, and 3) [104].

The relationships between the birds will display realistic behavior but fall short of complete naturalism. Naturalistic results require environmental factors (i.e. prevailing winds) 152 that can alter a bird's behavior. The simulation of a bird in flight should be run with a non repeating "wind model" to prevent replays of the same action [105]. Reynolds points out that randomization from sources that are relevant {stochastic process) to the behavior would be far superior than a computer's randomization function.

Having created such a behavioral model, all that would be required would be to make many individual instances of the

"bird actor", then create the initial situation, and finally just allow the flock of simulated birds to go through its own simulation [106]. This type of behavioral simulation uses adaptive motion ability of a controller to use information about the environment and the figure itself [107] .

4 .4 Simulation/Animation Hybrid

"Of course, in practice, most real animation is a combina­ tion of various techniques— certain characters may be created via behavioral simulation, while others in the same scene might be fully prescripted [108]." There are hybrid methods of creat­ ing computer animation that utilize predetermined animation scripting and algorithmic simulation. Reynolds views this technique as a script/simulation hybrid. This hybrid provides a solution to the goal of automating complex motion interac­ tions while the animator maintains control of the overall ani­ mation. 153

The first notable hybrid animation was "The Works" from the New York Institute of Technology. A giant mechanical ant

in "The Works" is "a computer simulation of a large, six-legged mechanical vehicle that appears to be powered by a complex sys­ tem of hydraulic actuators, springs, and shock absorbers

[109]." (Figure 25) As the "Ant" moves forward its weight

shifts from leg to leg. This produces a realistic bounce and sway as the load is redistributed. Because these motions are based on physically accurate mathematical models they appear naturalistic which imparts a special believability to the work [110] .

While many aspects of the Ant's motion were generated with the help of mechanical simulation, other aspects were directed by the "script". This scripting/simulation distinction in the computer parallels the automatic "reflex" motion built into

Figure 25. Mechanical Ant. Richard Lundin. 154

this ant-vehicle and the "conscious" direction initiated by the

driver of the ant-vehicle. A similar distinction between

"conscious" motion and "reflex" motion can be found in most

legged creatures, including human locomotion [111]. Zeltzer

[112] has suggested human locomotion programs should be struc­

tured in a similar hierarchy with high-level 'conscious' goals

or tasks, middle-level skills, and low-level reflex motor con­

trols that are fully automatic.

This animation hybrid contrasts with scientific simulation

in which the animator gives up all direct control and the simu­

lation proceeds entirely on its own, relying entirely on the

initial conditions and behavior of the objects. "The intent is

not to automate the animator's job but rather to shift the

technique for directing the animation from one of 'scripting'

[quotes added] explicit motion to one of 'designing' [quotes

added] the behavior of animated characters [113]." Thus, the

animator becomes less the manual laborer doing the in-between

frames and assumes more the role of director and producer.

This concept has significant implications that could change the field of animation as we know it today. 155

NOTES TO CHAPTER IV

1. Frank Thomas and Olie Johnston (1984). Disney Animation: The Illusion of Life. New York: Abbeville Press. (8), p. 323.

2. Jane P. Wilheims (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley, p. 2.

3. Dave R. Haumann (1987). "Modeling the Physical Behavior of Flexible Objects," SIGGRAPH ' 87 Tutorial Notes No. 17, Topics in Physically-Based Modeling, p. 2.

4. W. T. Reeves (1983). "Particle Systems - A technique for Modeling a Class of Fuzzy Objects," ACM Computer Graphics, 17, (3), p. 329.

5. Ibid.

6. Ibid.

7. David Zeltzer (1984). Representation and Control of Three Dimensional Computer Animated Figures. Unpublished doctoral dissertation, Ohio State University, Columbus, OH.

8. Ibid.

9. Craig W. Reynolds (1982). "Computer Animation with Scripts and Actors," ACM Computer Graphics, 16, (3), pp. 289-296.

10. Zeltzer, D. (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, 1, (4), p. 252.

11. Ibid.

12. Ibid.

13. Ibid.

14. Ibid.

15. Ibid.

16. John Chadwick and Richard Parent (1988). "Critter Construction: Developing Characters for Computer 156

Animation," FIXIM Conference Proceedings. p. 1.

17. J. P. Crutchfield and J. Gomez, J. (1986). Chaos, Dynamics, and Natural Phenomena. Unpublished manuscript.

18. Dave R. Haumann (1987). "Modeling the Physical Behavior of Flexible Objects," SIGGRAPH '87 Tutorial Notes No. 17, Topics in Physically-Based Modeling.

19. Wilheims, op. cit.

20. R. Barzel & A. H. Barr (1987). "Modeling With Dynamic Constraints," ACM SIGGRAPH '88 Conference Proceedings, Computer Graphics, 22, (4), pp. 179-188.

21. Haumann, ojo. cit.

22. Ibid.

23. Dave R. Haumann (1987). The Animation of Flexible Objects. Manuscript submitted for publication, p. 1.

24. Dave R. Haumann (1987). "Modeling the Physical Behavior of Flexible Objects," SIGGRAPH '87 Tutorial Notes No. 17, Topics in Physically-Based Modeling, p. 5.

25. Ibid., p. 4.

26. Dave R. Haumann (1988). "Animation using Behavioral Simulation," National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, p. 679.

27. Ibid.

28. Ibid.

29. L. Bergerron (1986). "The Making of Tony De Peltri," SIGGRAPH '86 Tutorial Notes, Advanced Computer Animation, p. 251.

30. Fred I. Parke (1982). "Parameterized Models for Facial Animation," IEEE Computer Graphics and Applications, 2, (9), pp. 61-68.

31. Brian Guenter (1988). "A System for Simulating Human Facial Expression," Unpublished manuscript, pp. 1-15.

32. Keith Waters (1987). "A Muscle Model for Animating Three-Dimensional Facial Expression," ACM SIGGRAPH '87 Conference Proceedings, 21, (4), pp. 17-24. 157

33. Parke, op. cit., p. 68.

34. Paul Ekman and W. Friesen (1975) . Unmasking the Human Face. Englewood Cliffs, N.J.: Prentice Hall Inc.

35. Waters, op. cit., p. 18.

36. Ekman, op. cit.

37. Waters, op. cit., p. 19.

38. M. Wein and N. Burthnyk (1975). "Computer Animation of Free Form Images". Computer Graphics, 9^, (1), pp. 78-80.

39. Richard Parent (1977). A System for Generating Three- Dimensional Data for Computer Graphics. Unpublished doctoral dissertation, Ohio State University, Columbus, OH.

40. T. W. Sederberg and S. R. Parry (1986). "Free-Form Deformation of Solid Geometric Models," SIGGRAPH f86 Conference Proceedings, 20, (4), pp. 151.

41. Chadwick, op. cit., p. 8.

42. Demetri Terzopoulos, John Platt, Alan Barr and Kurt Fleischer (1987). "Elastically Deformable Models." ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (4), 205-212.

43. Norm I. Badler (1984). "What is Required for Effective Human Figure Animation?," Proceedings Graphics Interface '84, Ottawa, Ontario, pp. 119-120.

44. N. Magnenat-Thalmann & D. Thalmann (1985). Computer Animation : Theory and Practice. Tokyo: Springer-Verlag, p. 76.

45. Craig W. Reynolds (1982). "Computer Animation with Scripts and Actors," ACM Computer Graphics, 16, (3), pp. 289-296.

46. Badler, op. cit.

47. David Zeltzer (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, 1, (4), p. 256.

48. Zeltzer, D. (1984). "Issues in 3-D Computer Character Animation," Introduction to Computer Animation, SIGGRAPH '84 Course Notes, £, (1), p. 266.

49. David Zeltzer (1984). Representation and Control of Three Dimensional Computer Animated Figures. Unpublished doctoral dissertation, Ohio State University, p. 26. 158

50. Ibid., P- 35.

51. Ibid., P* 3B.

52. Ibid., P- 39. 53. Jane P. Wilheims (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley.

54. W. W. Armstrong and M. Green (1985). "The Dynamics of Articulated Rigid Bodies for Purposes of Animation". Proceedings Graphics Interface '85, Montreal, pp. 407-416,

55. Michael Girard (1986) . Interactive Design of 3^-D Computer-Animated Legged Animal Motion. University of North Carolina Motion Workshop, Chapel Hill, NC., p. 6.

56. Ibid., P* 2. 57. Ibid.

58. Ibid.

59. Ibid.

60. Ibid., P- 3.

61. Ibid., P- 4.

62. Ibid., P- 6. 63. Michael Girard (1986). Interactive Design of 3-D Computer-Animated Legged Animal Motion. University of North Carolina Motion Workshop, Chapel Hill, NC., p. 7

64. Ibid.

65. Ibid., P- 8.

66. Ibid., P* 18.

67. Ibid., P- 8.

CTl 00 M. Green (In press). "Animating Complex Objects." ACM Transactions on Graphics, p. 3,

69. Ibid.

70. W. W. Armstrong and M. Green (1985). "The Dynamics of Articulated Rigid Bodies for Purposes of Animation." 159

Proceedings Graphics Interface '85, Montreal, p. 409.

71. Green, o£. cit., p. 10.

72. Ibid., p. 15.

73. Ibid., p. 5.

74. A. Witkin, K. Fleischer & A. Barr (1987). "Energy Constraints on Parameterized Models," ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (4), 225.

75. A. Barr, (1988). "A Modeling System based on Dynamic Constraints" SIGGRAPH * 88 Tutorial Motes No. j5, Topics in Physically-Based Modeling, p. E-48.

76. Dave R. Haumann (1987). The Animation of Flexible Objects. Manuscript submitted for publication, p. 4.

77. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH '86, Tutorial Notes, Advanced Computer Animation. p. 126

78. Ibid.

79. Ibid., p. 127.

80. B. P. Zeigler, M. S. Elzas, G. J. Klir & T. I. Oren (1979). Methodology in Systems Modeling and Simulation. Amsterdam: North-Holland Publishing.

81. R. Garbutt, C. McPheeters, & B. Wyvill (1986). "University of Clagary 3-D Computer Animation System," SMPTE Journal, (5), p. 629.

82. W. T. Reeves (1983). "Particle Systems - A technique for Modeling a Class of Fuzzy Objects," ACM Computer Graphics, 17, (3), p. 359

83. Ibid., p. 360.

83. Ibid., p. 361.

85. Ibid.

86. Ibid.

87. Ibid., p . 359.

GO GO Ibid.

89. David Zeltzer (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, 1, (4), p. 251. 160

90. Ibid., p. 249.

91. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH f 86, Tutorial Notes, Advanced Computer Animation.

92. Zeltzer, o£. cit., p. 252.

93. Craig Reynolds (1986). Advanced Computer Animation. ACM SIGGRAPH '86, Tutorial Notes, Advanced Computer Animation, pp. 204-205.

94. Ibid.

95. Ibid., p. 205.

96. Dave Haumann (1989). Unpublished manuscript in preparation for dissertation document.

97. Craig W. Reynolds (1982). "Computer Animation with Scripts and Actors," ACM Computer Graphics, 16, (3), p. 294.

98. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH r 86, Tutorial Notes, Advanced Computer Animation, p. 206.

99. Ibid.

100. B. L, Partridge (1982). "The Structure and Function of Fish Schools." Scientific American, 6, pp. 114-123.

101. Craig W. Reynolds (1987). "Flocks, Herds, and Schools: A Distributed Behavioral Model," ACM SIGGRAPH '87 Conference Proceedings, 21, (4), p. 25.

102. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH '86, Tutorial Notes, Advanced Computer Animation, p. 206.

103. Craig W. Reynolds (1987). "Flocks, Herds, and Schools: A Distributed Behavioral Model," ACM SIGGRAPH '87 Conference Proceedings, 21, (4), p. 25.

104. Craig Reynolds (1986). "Advanced Computer Animation," ACM SIGGRAPH *86, Tutorial Notes, Advanced Computer Animation, p . 206.

105. Ibid.

106. Ibid.

107. Zeltzer, D. (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, (4), p. 251. 161

108. Reynolds, o£. cit., p. 205.

109. David Sturman (1986). "A Discussion on the Development of Motion Control Systems," Advanced Computer Animation, SIGGRAPH '86 Tutorial Notes. p. 4.

110. Reynolds, o£. cit., p. 203.

111. Ibid., p. 204.

112. Ibid.

113. Reynolds, o£. cit., p. 205. CHAPTER V

THE MODEL

This chapter proposes a functional model based on

computer-graphic simulations of physical relationships and phenomena that can be utilized for artistic expression. This model's structure and procedures are based on information reviewed in Chapter III and IV. The functional model seeks to bridge the qualitative - often idiosyncratic - conceptual orientation of the artist with the quantitative orientation of computer simulation.

It was determined that to create this functional model from the strictly subjective perspective of artists was not feasible. This would have dictated individual programs every time a new variation needed to be played out. It became obvi­ ous that the only way to structure the model was on a systems- oriented approach [1]. The fact that the model's structure is based on a system-oriented viewpoint rather than on a object- oriented viewpoint is appropriate from a utility standpoint.

The applicability of artistic initiatives differentiates this model from the current computer simulations and their applica­ tions .

162 163

This functional model is successively organized in concen­

tric layers. Each layer contains abstractions joined through a

logical interaction or interdependence. Layering provides a

powerful structure from which to create animations governed by

physical laws and driven by our imagination. Such a structure

permits the artist to transcend various levels of detail.

From this functional model the interrelationships of par­

ticular simulation components and the causes of changes in the

simulation can be understood. This, in turn, will permit the

animator to anticipate the effect of changes on the simulation.

It* should be noted that this model is based on a "hierarchical"

structure rather than a "linear" structure. The "hierarchical"

tree structure permits "branch" operations to be changed

without affecting the location and interrelationships of the

entire composition.

Commonalities are the basis around which this model is

organized, within the domain of all physically-based objects

and procedurally defined motion. It appears that what must be represented in the computer are the "structures", "attributes",

"functions" and "relationships" of objects in a scene. These factors represent the commonalities upon which this functional model can be built. This representation is to be uniform such that there is no distinction between agents (living entities that can initiate action) and objects (which can only react). 164

To reproduce the physical interactions of our world it is necessary to view the salient features of this model as a set of "geometric primitives" (how structure is represented in the computer), "mechanical attributes" (the characteristic features, such as weight or elasticity of the object), "motion functions" (how geometric primitives and mechanical attributes are interconnected), and "behaviors" (interrelationships between and among objects). This chapter also covers the issue of interactivity; this is critical in the successful implemen­ tation of the model.

Physically -based phenomena can be structurally classified in three categories: (a) Structures that contain rigid links

(mammals, fish, trees, etc.), (b) Structures that are bound together by surface constraints (cloth, rubber, skin, clay, etc.), and (c) Structures that are not "true" structures but are actually processes (i.e. clouds, water). Each of these phenomena have inherent motion characteristics that are identi­ fied with the form. From these structural categories

"geometric primitives" can be deduced that are applicable to not only physically-based modeling but also analogous to artis­ tic concepts of structural organization. It should be noted that parts of this functional model may not be immediately intuitive for the uninitiated artist. The mediated nature of the computer necessitates an expanded knowledge base on the part of the artist. This is discussed further in Chapter VI. 165

5.1 Geometric Primitives

The geometric description level is the foundation upon which each successive level is built. These descriptions comprise the bulk of information used to quantitatively clas­ sify a physically-based object. These geometric primitives are viewed as not only separate components, but also as the first level in a layered approach (Figure 26). The geometric descriptions of the functional model are classified as:

1. One-dimensional point primitives 2. Two-dimensional surface primitives (e.g. polygons, patches) 3. Three-dimensional volume primitives (rigid or flexible)

Geometric Primitives

Figure 26. Geometric Primitives. 166

These structures of physical properties represent a quan­ titative classification on the macro level of the physically- based world.

One dimensional points and lines would describe natural phenomena that are processes or composed of discrete elements

(e.g. clouds, water). Linear elements in our world (e.g. hair, string) can be represented as one-dimensional points linked together.

Two-dimensional primitives would be composed of two- dimensional planar surfaces connected together (e.g. paper, skin). Choices of linkages would determine the characteristic range of movement for an object. Hollow forms could be con­ structed from surfaces that are connected back on to themselves at their open edges (i.e. a flat surface curled into a cylinder shape).

Three-dimensional primitives (rigid or deformable). These would be composed of rigid (e.g. vases, bones, rocks) or deformable (e.g. muscles, jello) three-dimensional forms. The linking of these forms results in articulated structures which could be conceived of as a distinct sub-class within the domain of three-dimensional structures.

At this point there is no such thing as a flexible object until mechanical attributes are associated with the object.

The planning for such objects though can begin here. It is from a combination of geometric primitives that other 167

structures can be constructed (i.e. articulated structures,

faces). Articulated structures consist of rigid segments

linked together (e.g. figures, animals, insects, trees). A deformable surface like a face can be constructed in layers accounting for the underlying rigid or semi-rigid structures underneath (bone [skull] then mass [fat, muscle] and then sur­

face [skin]}. This model needs to be designed with an initial

state that permits these structures to be manipulated later with behavioral elements.

Objects can be structurally defined not only through these primitives but also from other objects themselves. One object

can be "part of" or a "movable part of" another [2]. For exam­ ple, a piston (i.e. movable-part-of) is part of a motor (i.e. part-of) which is part of a motorcycle. The functional model proposed here is based on the assumption that geometric primi­ tives are defined in a coordinate system, and the coordinate position of the object and its components are known locally or globally. This classification of one-dimensional, two- dimensional, and three-dimensional primitives constitute the basis upon which mechanical attributes can be bound.

5.2 Mechanical Attributes

The mobile character of an object or agent (i.e. an object that can initiate action) is defined by its mechanical attri­ butes. Without these mechanical qualities the potential for movement does not exist. Mechanical attributes associated with 168 an object includes joint linkages, mass, velocity, accelera­ tion, deformation, force, torque, and surface area as a func­ tion of damping and collision detection. All these attributes directly affect an objects characteristic movement. It is the combination of geometric primitives and mechanical attributes that permit a self-scripting or automatic simulation to proceed

(Figure 27). The animator can affect modifications in an ani­ mation by changing mechanical attributes at this local level.

It is necessary to represent the mechanical interaction of objects in terms of a well-defined set of relationships or dynamic attributes that is concise enough to be practical.

This will permit the logical assignment of mechanical attri-

Geometric Primitives

echanical Attribute:

Figure 27. Mechanical Attributes. 169 butes to the different geometries. These properties can be depicted as separate attributes.

Joint Linkages (connections between primitives) would be constrained to simulate a specified range of movement. Link­ ages in this model differ from an artist's concept of a physi­ cal structure. The linkage here determines the characteristic range of movement for the geometric primitives. Realistic joints serve as a reference for linkages. Such a listing would include:

(1) Ball-and-socket joints, as in the hip, which can have

three full rotational degrees of freedom. Generically,

this type of link would be unrestricted, similar to hip

and socket joints, but not able to rotate.

(2) Condyloid joints, as in the fingers, which move in various

planes but do not rotate.

(3) Gliding joints, as in the wrist, which slide or twist.

(4) Hinge joints, such as the elbow, which move in one plane.

(5) Pivot joints, such as the radius at the elbow, which

rotate

(6) Saddle joints, such as the base of the thumb, which are

capable of multiple motions but are more restricted by

their curved articulate surface than ball-and-socket

joints [3].

Applications of such linkages can be at the micro (i.e. polygons) as well as the macro (i.e. objects) level. Linkages are organized in either a hierarchical or lattice structure. 170

Specific linkages would be defined by range and type of con­

straints assigned to linkages. Under-constrained (Chapter IV)

linkages would lend itself to proximity attachments. This

would cover rubberband type (i.e muscle to skin) linkages.

Mass attributes represent matter or weight at a point.

This is the element that responds to environmental forces. It

is responsible for resistance to changes in motion (i.e. iner­

tia) . This attribute is the basis upon which most other

mechanical attributes interact. This primitive can be set glo­

bally for the object or locally at specific vertices on the

object.

Velocity attributes contain the initial state of velocity

an object has. Velocity is defined as the rate of change from

one position to another position. In the real world all

objects have a velocity relative to the environment, even if

that is a zero velocity. All "states" are important. They

represent the initial or current velocity which can be effected

by a change in velocity. An animation needs to start with

information about an object's initial motion. This velocity

primitive would contain whatever velocity the object possesses

for the duration of the animation. An object that is not mov­

ing has no change of position; it would be included in the set

of non-moving objects as defined by this primitive.

Acceleration attributes would be the rate of change in velocity, the change from one velocity (i.e. zero, no movement) 171

to another velocity (i.e. 5 MPH). This primitive would ini­

tiate changes in motion as the result of forces and masses

interacting. Barr [4] advocates including "impulse" attributes

(linear and angular) to account for the initial change when an

object begins to overcome inertia. At that point it takes a greater force to get it going than to maintain velocity. The

"impulse" attribute would be used to account for that differ­ ence .

Deformation attributes define the elasticity or stiffness of an object. This primitive contains information about an object's relative rigidity or flexibility. This would be in the context of resistance to being pulled apart and/or pushed together. This primitive would kinetically preserve linear and angular momentum. Haumann simulates this primitive as spring and hinge linkages between surfaces or masses [5]. This primi­ tive emulates Newton's third law (action-reaction). A neces­ sary variable in this attribute would be whether the object kinetically absorbs the external force and changes shape, or whether the force is released in the form of a reactive motion.

For example, an aluminum can will deform in proportion to the magnitude of a force exerted on it while, on the other hand, a rubber ball will only momentarily deform before kinetically releasing the absorbed force in an observable reaction. The ball retains its original shape after a force, while the alumi­ num can deforms from its original shape. 172

Force attributes simulate external forces such as gravity and wind, or as internal force's resulting from muscle contrac­ tion. For example, gravity effects a downward force on the mass attributes. These forces are used by the dynamic motion procedures later.

Torque attributes would contain the magnitude of a rota­ tional force being applied at a joint. This would result in the desired rotation of an appendage from the joint. Again, this would be used by procedures at a higher level for dynamic motion simulation.

Surface-area attributes are a requisite for damping and collision detection. Damping is the motion of an object as the result of contact forces propagated by the surrounding fluid

(i.e. wind or water). Damping can be computed through a for­ mula that relates surface-area orientation to velocity vectors.

Haumann has employed an effective ad-hoc technique to simulate damping through the use of hinge joints at the polygonal level

[6] .

5.3 Functional Procedures

For an object or agent with many links, it is desirable to be able to combine "geometric primitives and mechanical attri­ butes" with "motion procedures" (e.g. dynamics, kinematics) into functional procedures which are necessary to effect a par­ ticular set of motions. Figure 28 illustrates the combining of geometric primitives and mechanical attributes into functional 173 procedures. Functional procedures permit the animator to create motor skills [7]. A prototypical physical object, for example, might obey some subset of the laws of Newtonian mechanics which can be assigned at the functional level. Arti­ culated figures can build a repertoire of behaviors from these functions, such as walking and grasping [8].

Movements that would be repeated frequently in an anima­ tion would be assembled into a library of motor skills. For example, the grasping movement of the hand can be factored into a functional procedure. The necessary structure primitives and mechanical attributes are selected from the known joint-angle

Functional Procedures

Geometric Primitives

Mechanical Attributes

Figure 28. Functional Procedures. 174 rotations and hand movements and assembled into a functional primitive for "grasping". Once defined, the lower level details do not need to be redefined again later. This allows the animator "to direct" the motion through parameters (e.g. target location, fast or slow, hard or soft) [9]. This func­ tional capability is fundamental to the behavioral level, described in the next section, "Behavioral Simulation".

A functional procedure, like grasping, would be con­ structed as a kinematic or dynamic motion. It is at this func­ tional level that the actual specification of mechanical attri­ butes to specific geometric primitives would be assigned in conjunction with the desired type of motion (i.e. kinematic, dynamic). How these elements are hooked together will directly affect the subsequent motion of the objects. Haumann has sug­ gested two conceptual levels for functional procedures that should be useful: (a) "at a coarse level for external con­ straints - for example: a complex object is related to the air by drag and to the ground by both gravity and contact," and (b)

"at a fine level for internal constraints - mass elements interconnected by spring and hinge elements to maintain inter­ nal object coherence [10]."

User-designed functions can generate motion in virtually any manner. Functions can range from the simple specification of gravity to the complex motion of a scripted sequence (i.e. a dog getting a newspaper). Most important, these functional procedures may be nested together resulting in meta-behavioral 175

functions. One functional procedure may invoke others, allow­

ing for higher level movements such as walking.

5.4 Behavioral Simulation

Behavior is the result of many complex factors and

interactions. Though physical behavior can be simulated

through Newtonian dynamics, behavior is more than the numerical

solutions output from dynamic equations of motion. Girard [11]

believes the difficulty in creating intelligent behavioral

motion is that given some desired behavior (or property of

behavior), we must find the forces which will produce it [12].

Motion behaviors are significant because of the informa­

tion that can be associated within the interrelationships

between objects. These behaviors contain behavioral elements

such as (a) programmable behaviors, (b) properties (mechanical,

logical, social), and local memory (event history, current

state). These behavioral elements may be built from a library

of functional procedures and their interconnections. In the

model's hierarchy motion behaviors can be composed of several

functional procedures combined to form a more powerful func­

tional procedure. This "functions within a function" concept permits the simulation of low-level behavior (Figure 29).

Nevertheless, behaviors are also more than nested functions.

True behavioral situations require objects to have local or

global knowledge of their environments. That is, objects need the capacity to obtain information from their environment and 176

Behavioral Simulation

Geometric Frinfttives

Mechanical Attributes

Figure 29. Behavioral Simulation. also from other objects. If we wish to utilize goal-directed characters capable of achieving "non-trivial tasks then the character must take into account the geometry and mechanics of physical environments [13]." This capability is similar to the message-passing facility designed by Reynolds.

Collision detection is a quality of the physical environ­ ment that must also be taken into account. The physical re­ action between objects or between an object and the environment is in the domain of collision detection. The general concept of collision detection as a reaction is too narrow to account for all types of collisions such as self-intersection (i.e. an extremely flexible object which can turn in on itself, like 177 cloth in the wind). Therefore collision detection must be

"robust" so that not only contact with objects can be detected and simulated but also self-intersection within an object

[14,15]." In addition, the animator can specify parameter values for "contact adhesion" and "sliding friction".

In examining the problem of how to simulate the "grasping" of an object it is evident that the previous grasping function primitive will suffice for only one of the large number of pos­ sible grasps that result from the number of finger units and inherent links. Girard [16] suggests that grasping must also involve the analysis of an object's geometry. This analysis is for the purpose of locating possible "handles" which are suit­ able for finger grasping or insertion strategies, which in turn would be used in the establishment of "pre-complied" grasping strategies for specific objects [17]. It is from this juncture that procedures for grasping can proceed.

(1) Classify the object to be grasped in terms of one of the

known geometric primitives.

(2) Determine the grasping shape of the hand in terms of the

dimensions of the object to be grasped.

(3) Generate the determined pre-shape during approach. Upon

contact, adjust the grip about the object by sliding the

fingers onto the (possibly irregular) surface until secure

contact is reached [18]. 178

This grasping behavior may be generated through either

kinematic or dynamic means. It should be noted that in this

functional model behaviors are not inherently locked to

specific structures. This ability to arbitrarily interconnect

primitives leads to creative associations.

While humans are in some sense active agents, they also obey the laws of Newtonian mechanics; a person falls just like a rock when pushed off a cliff. On the other hand, an animator may want the chairs and tables to dance around the room when the villain leaves. It should be easy to ascribe such behaviors to otherwise inanimate objects [19] .

Such creative associations will require motor problem

solving within the animation system. In order to do simple

motor problem solving, it will be necessary to embed common

sense "knowledge1' in object descriptions. That is, we want to

be able to encode such default knowledge as one usually leaves

a room by finding and opening the door. From our surrounding

environment we have absorbed knowledge of naive physics, common

sense, and mechanisms that are built on very non-conscious

movements. Not only does this knowledge need to be accessible

for use, but it also needs the option of being overridden once

implemented so a character can leave by the window.

This cognizance of the environment leads to intelligent motion behavior. Such behavior consists of tasks such as: (a)

executing highly coordinated gymnastic maneuvers, (b) assem­

bling and manipulating objects, and

cluttered environments [20]. At the highest level, "task 179 planning" {the overall sequence of the task's individual steps or components) must be determined to perform tasks which involve such complicated maneuvers. "Path planning" is the determination of collision-free paths between an initial and final position in an environment. "Trajectory planning" solves for the speed of movement.

In this functional model there must be libraries of prede­ fined functions and behaviors. These permit the artist to not only add to the available motion repertoire but to also modify default behaviors that would be used often. Zeltzer in

"Interacting in Animated Microworlds" states that for each behavior or skill to be simulated there must be: preconditions and triggering stimuli, an execution schema (i.e. scripts or finite state network), associated potentiated skills, and finally subgoal generation.

5.5 Interactivity of the Model

To be able to weave all these primitives into a cohesive operational system will require a responsive interface. Inter­ face issues of usability, flexibility, extensibility and habi­ tability [21] as each relates to this functional model need to be addressed.

5.5.1. Usability

Usability cover the relative ease and effectiveness of the needed interactive control parameters for a simulation. 180

Whether artists must preconceive their idea and design the for­ mal elements analytically prior to the application in question, or whether they can design and conceptualize their idea while interacting with the system stems largely from its interactive capabilities. Valuable and integral editor components should include resource libraries, intuitive modification designs, real-time playback facility, and a weighting facility. These elements can mean the difference between a system that merely functions to one that is expressively responsive.

5.5.1.1 Usability - Resource Library

A library of primitives and procedural functions (i.e. motion primitives) would be an invaluable resource. This library would not only be graphically interactive but also be expandable and flexible. This library facility would contain a variety of predefined geometric, mechanical, functional, and behavioral primitives. This would alleviate the syndrome of

"re-inventing the wheel" which is so pervasive in the computer graphics and simulation field today.

5.5.1.2 Usability - Intuitive Modifications, a Visual Editor

The editor is to be visually interactive on three levels:

(a) a motion graph of primitives over time,

(c) the complete visual representation. Though the motion-time graph may not be initially intuitive, it does establish a usable correspondence between the often incomprehensible 181

numerical specifications and the complete pictorial representa­

tion. This graph may be thought of as a "parameter path" for

defining the motion [22].

A parameter path is composed of a sequence of values for a

given parameter over time. While a path can be defined as

either a list of numbers, an actual pictorial animation, or as

a graph of parameter values versus time, it is the latter which

will be most useful as a general interactive tool in control­

ling discontinuities. Discontinuities occur in conventional

key-framing because there is no effective means for specifying

and controlling interpolation over a sequence of key-frames.

With a motion-graph editor an animator should be able to con­

trol such discontinuities by means of moving points [23]. This

should result in straight-forward changes being easily accom­

plished, provided the animator can change individual parameters

rather than being limited to a global parameters. In addition,

Reynolds [24] points out that motion-graph manipulation

encourages the animator to comprehend the mathematical aspects

of his motion specification.

5.5,1.3 Usability - Real-time Playback Facility

Real-time playback should be available after animators

have finishing making adjustments in their scripts. Real-time playback permits the animator to see the motion just as it will be shown in the final animation, with no perceptual delay between the display of frames. The playback system is very 182

important, because it enables animators to pencil-test the

current motion sequence. An important point to note here is

that animators work by successive refinement - adding new con­

straints, seeing how the sequence looks, then going back and

adjusting constraints until the sequence is just right [25].

Some systems generate pencil tests in real-time. Others

require a few seconds for compilation, and then playback in

real-time. The latter are often limited by the amount of memory available to store computer pencil test frames.

As long as the compilation procedure involves LITERALLY just a few seconds of idle time on the animator's part, then the playback is useful during the design process and is online. Otherwise, we categorize the playback facility as offline [26].

A couple of seconds of delay for precalculation is acceptable

(it takes that long to prepare mentally for feedback). Such a system may be considered an "online" system. An "offline" sys­ tem would be one in which the precalculation would be equivalent to the time it takes an animator to lose their train of thought. Motion cannot be accurately and efficiently adjusted unless it is in real-time. Multiuser systems with intermittent display updates {because of input/output swapping) will result in an unacceptable jerk in the playback [27].

It is rather difficult to work out the desired timing of an animation with no steps between storyboard and final foo­ tage. The result generally lacks refinement due to the lack of feedback. It is mandatory that the animator get immediate 183

feedback on what his work will look like, regardless of whether

the playback is either offline or online. Any animation even­

tually provides offline playback once the work has been

recorded and projected. This delay between design and viewing

is, however, unacceptable because it would make many of the

continuity decisions at best "guesswork".

5.5.1.4 Usability - Weighting Facility

Physical properties should be designed to be very inter­

dependent. When one physical property is changed the effect is

propagated to the other physical properties within the anima­

tion. Maintaining this correlation when one of the parameters

changes can be achieved through a weighting facility. For

example, the problem with early color specification in the com­

puter can be used to demonstrate this effect. Color is com­

posed of red, green, and blue primaries in the computer. The

final color can vary greatly by the adjustment of just one of

these primary components. Changing just the intensity of the

desired color becomes a difficult and trying experience,

because all three primary colors must be changed in varying proportion. Eventually, more intuitive parameters {hue,

saturation and value) have emerged which changed the red, green and blue components proportionally to achieve the desired change. Similar "weighting" programs will also need to evolve

so that the interactive quality of the physical properties will be more predictable and controllable. In effect, a "weighting facility" provides the system with a proportional "correcting" 184

of parameters.

Implementing constraints is one method of "weighting" parameters. Under-constrained specifications {i.e. the ball is

near the box) permits an efficient weighting control specifica­

tion. The animator might prefer to specify several "con­

straints" and declare that they must continue to be "main­ tained" . It is then up to the animation system to determine the appropriate parameter values for each frame such that the constraints are satisfied.

The animation system would permit the straight-forward specification of constraints. This would include a library of built-in constraint relationships. These geometric relation­ ships may include the distance between two objects, the align­ ment of one object with another, or the coincidence of a selected point on each of two objects [28] . Animators would probably find themselves leaving objects underconstrained (i.e. the ball is near the box) and specify the basic motion through a script. The motion from the script would be modified by the constraints that are in effect during the duration of the ani­ mation. "If a model is known to be fully constrained it is often possible to derive an explicit 'closed-form solution', a formula based on a system Of simultaneous solutions. Such a formula allows all relevant parameters to be computed directly

[29]." 185

5.5.2 Flexibility

The goal of "flexibility" is to have the necessary con­ straints put on the system, not on the animator. A system should not force a way of working on the animator. Though a system should not impose physical laws on the animator, it should have them available when needed. As Gomez [30] pointed out, "although Wily Coyote falls in a fashion that may be related to d = l/2at(squared), it usually does not happen until he has been walking on air for a few seconds." Sometimes an animator may want a function that in the programmer's opinion is not as appropriate as another function that has been pro­ vided. Gomez [31] makes the argument that an animator may want a linear interpolation, rather than a spline, even with its inherent problem of discontinuity at times.

Therefore, accessibility to the needed control level will be critical for success. An integration of the three modes of control (guiding, procedural and task) can be projected through this functional model's structure. This need to access dif­ ferent control levels stems from the inability of just one mode to provide the animator with complete yet reasonable control.

Reynolds believes,

...in practice, most real animation is a combination of various techniques— certain characters may be creat­ ed via behavioral simulation, while others in the same scene might be fully prescripted [32]. 186

The current prevalent guiding mode provides excellent

refinement of explicit details but is too unweildy for control­

ling complex motion. Heavy reliance on this explicit level

results in discontinuities in the motion. It is within this

guiding mode that explicit geometric structures and mechanical

attributes would be assigned.

The specification of functional procedures is located

within the procedural mode. This mode is very powerful but

requires a thorough understanding of procedures to utilize its

potential. To simplify the interaction an animator may prefer

to specify only a small set of the global parameters and rely

on the default values.

The animator would have access to the following four pro­

cedural methods of motion control: forward kinematics, forward

dynamics, inverse kinematics, and inverse dynamics. "Forward

kinematics" permits the animator to manipulate an object or

articulated figure by degrees of rotation. "Forward dynamics" also permits explicit placement but by means of forces and torques. "Inverse kinematics" and "inverse dynamics" permit the input of the position and orientation of a target location.

From this transformation data the intermediate positions or torques and forces (necessary to reach the desired position or orientation) are computed. This "inverse" procedure automati­ cally resolves the motion specifications needed. These pro­ cedures should be viewed as operating in a pipeline, with dif­ ferent motion procedures interacting with each other (See 187

Figure 30) .

This pipeline permits access to different methods of motion specification. These motion specification levels are organized as modules. By linking the different modules together through a feedback control loop the artist has access to the different specification levels (guiding, procedural, task) when needed. These different levels permit artists to interject their desires either implicitly, explicitly, or algo­ rithmically.

General motion planning schemes such as gait specification and path-planning will constitute high-level, implicit control.

For a number of animations no other specifications may be

■t Motion Inverse Inverse Forward orward Planning Kinematics Dynamics Dynamics Kinematics ' Feedback {Key*fnmH >: etc.) _ ControlContr Loop

Figure 30. Motion Pipeline and its Modules. 188

necessary. If "directed" or predictable control is desired the

animator may choose the appropriate motion control module. For

example, the inverse kinematics module permits the intuitive

specification of objects by constraints. Kinematic constraints

may be assigned in several ways. The most obvious is the pre­

specification of position. Constraints may also come into

effect when some inequality is satisfied, such as when one

object attempts to occupy the same global position already

occupied by another object. Constraints also "... may be

invoked by a behavior based on current criteria in the system,

(e.g., a ball stays in the hand after being caught until

thrown) [33]." The inverse dynamic module can take the data

from inverse kinematic specifications and determine force mag­

nitudes that result from dynamic analysis. These forces can in

turn be supplied to the forward dynamics module, which in turn

can output rotational and translational values to the forward

kinematic module. The feedback control loop provides a method

of linking modules together: whether it be a straight sequence,

an individual preference different modules, or repetitive loops of single or mixed modules. Not all the modules need to be operative at the same time; this will be determined by the

immediate goals. This ability to mix different modules can permit a keyframing animation system to be connected within a dynamic system. In addition, the knowledge of forces permits the development of behavioral functions. 189

In the past, each person who has created a procedural

motion system has generally championed one of the four methods

at the expense of the others. The individual limitations of

each system, however, serves to point out the need for alter­

nating control strategies to fit the situation. This is funda­ mental to the concept of flexibility.

The feedback control loop (Figure 31) is the mechanism

which provides for how these modules can be linked together and

controlled. This process determines how well the results ful­

fill the artist's expectations. This control mechanism may be

iifflmtitli

U ti

Interpretative Guiding Conventions

Procedural (Browsers) Eval uati ve C riteria Goal-Oriented

Figure 31. Feedback Control Loop. 190 operated through explicit manipulation, the coupling of modules as procedures, implicit goal-oriented direction, or by the predefined aesthetic-interpretative conventions in conjunction with evaluative criteria. In the feedback control loop we can­ not only implement the explicit guiding control needed for fine tuning but also implement aesthetic controls of a higher order.

The significance of this loop is that animators are not handed one module but a collection of modules and possible con­ nections from which to tailor the motion simulation to their vision. Rather than being confined to the specification of parameter values artists can now construct their own aesthetic algorithm from this model. It is the up to the artist to select which motion generating modules are to activated and in what order. This combination approach is conceptually similar to recent surface description languages such as Perlin's Image

Synthesizer [34].

It is anticipated that control will be initially focused at the guiding and procedural levels. It is here that the ani­ mator will specifically alter individual values or link (e.g. sequentially, intermixed, repeatedly) the different motion modules. The artist in the role of selective agent initiates action, scrutinizes the results, and either accepts or rejects the outcome with the option to continue the process.

The flexibility to interact on different levels with dif­ ferent modules broadens the creative potential of the animation 191 process and provides a base for the concept of "browsers" as an interactive, procedural "what if" tool.

The notion of browsers as implemented in Smalltalk (Tesler, 1981) or Loops (Stefik, 1983) suggests a powerful method for attaching guiding controls to motor skills. Suppose I have on my RGB monitor a shaded display of a human character. On my terminal screen is a representation of the structure of the character and its skills. Now suppose I trace a curve on the graph­ ics tablet. If I specify that that curve represents a particular joint rotation, - i.e., I point to the node for the little finger on my terminal, I should immedi­ ately see on the display the little finger of my char­ acter wiggling. Suppose now I point to the node for "grasping with the left hand" - I should see the figure's left hand open and close with the velocity I have specified. Lastly, if I pick the node labeled "walk", the figure should begin to walk across the screen, and this time, the curve I have drawn could determine, say, the speed of the gait [35].

Nevertheless there is a good chance that "...the easiest way to specify a motion might be to specify the goal rather than 'how' to achieve the goal [36]." Such a goal-oriented mode is appropriate when there exists either a limited understanding of the concepts involved, a desire to begin a rough sketching out of an animation idea, or where higher level control is needed. Up to this point it has been assumed that specifica­ tion will be by motion types and parameter values. Thus, any sort of motion is explicitly specified "a prior" and is often burdensome. By trading off explicit command of the details an implicit control over complex motions can be achieved. By not explicitly specifying motion prior to its execution there is the added possibility of avoiding unintentional preconceptions and prejudices entering the work. This goal-oriented mode is 192 composed of the previous two - guiding and procedural.

5.5.2.1 Flexibility - Aesthetic Criteria

Another alternative - aesthetic specific - is to simulate aspects of the creative process about which the artist already has some notion. This aesthetic strategy could take the form of interpretative conventions and evaluative criteria [37].

These aspects address the concern that there is no new art form if the artist continues to manually control the medium. A sys­ tem derived from the functional model of this study should be able to handle not only visual complexities but also creative complexities on a conceptual level.

Gips and Stiny [38] have looked at the creative process as one in which external relationships and internal coherence can be codified into an aesthetic algorithm. It is worth noting that this approach is similar to the framework of this study's model.

These strategies permit formalized aesthetic viewpoints to be used to select and link motion modules according to prede­ fined criteria. One of the first bodies of information likely for this type of integration would be "Principles of Anima­ tion", {e.g. squash and stretch, anticipation, etc.) {See Con­ tinuity Applications later in this chapter). The aesthetic continuity of motion in animation is directly proportional to the successful application of these principles. 193

For example, squash and stretch can be simulated by dynam­ ics. However, the correct application of dynamics to a charac­ ter would not necessarily produce the extreme - and many times unnaturalistic - changes needed to be convincing. This dispar­ ity between naturalism and animation exists because the medium of film is, of course, not reality. Aesthetic changes are necessary to fill in the discontinuities inherent in the medium and the mind of the audience. Arnason [39] makes the point that the more artists attempt to reproduce the world as they see it; the more they realize that reality rests not so much in the simple object nature of natural phenomena, as in the eye of the spectator. The effective communication of squash and stretch requires an exaggeration of dynamics in the figure.

The interpretative conventions module would match "squash and stretch" with the dynamic modules, and the evaluative module would use a ranking algorithm to determine if it sufficiently matched criteria specified - possibly exaggerated - by the artist.

This example of aesthetic input is one of countless orien­ tations this structure provides {See Future Implications in next chapter). Overall, these modules provide a flexible hierarchy for interacting within the system at the appropriate points of the animation. 194

5.5.3 Extensibility

No matter what capabilities the system provides, a desire for additional modules will arise. For artistic uses this will be especially true because of the nature of the artistic pro­ cess. A system based on this functional model would include facilities for including new mechanisms or reconfiguring exist­ ing ones. This would mean that animators could define their own movement criteria and implement them into the animation system library. Effectively animators are reconfiguring the system to meet their own needs [40].

Experienced computer users can utilize their expertise. They can use mathematical methods to analyze and mani­ pulate paths, while the editor can provide algorithmic tools for path manipulation. A programmer can generate the simple path database with his own program, if desired. The new path can then be read into the editor [41] .

5.5.4 Habitability.

Necessary features in an animation system contributing to its habitability (how user-friendly it is) would include the following: "guarded exit" {do not exit unless the script is saved or the user is sure), "interactive exception handling"

("file exists - do you want to overwrite it??"), "help" facili­ ties . Habitability would include plug-in facilities receptive to divergent - or format ir-ompatible - types of systems and motions (e.g. particle systems, polygons, procedurally defined objects). It's one thing to construct a system that can accept 195

outside information; it's another it be able to integrate that

facility into the system for later usability.

Another necessary feature would be the integration of this procedural level with a visual interface. Gomez [42] states

that "...there is no reason why an animation can't itself be considered a procedure; it simply happens to be in a different

language. There are many functions (e.g. picking, locating,

sketching) for which the graphical gesture is clearly the pre­ ferred mode of interaction [43]." As desirable as a visual interface would be, language will probably remain the medium of choice for specifying complex spatial, temporal, and behavioral relationships. Zeltzer [44] recognizes that there is a contin­ ued resistance to the text interface which probably has more to do with the chore of typing rather than with text-mediated issues. This resistance will be greatly alleviated by develop­ ments in speech recognition in the future.

5.5.5 Submodel - Continuity Applications

Continuity in an animation can be achieved through the application of the known successful techniques (e.g. Principles of Animation). These techniques developed at the Disney studio guide the animator in maintaining continuity in an animation.

A system implemented from the functional model described in this study will be successful in direct relation to how its elements are applied. Indiscriminate use of this model's features without consideration of the "principles of animation" 196 will result in continuity problems within an animation. Anima­ tion is a visual facsimile of reality in need of creative dev­ ices to fill in the discontinuities inherent in the medium.

Simulating the "squash and stretch" principle of animation in computer animation can be accomplished primarily through techniques of "surface deformation" and secondarily through

"motion blur" techniques. Motion blur alleviates the disturb­ ing effects of temporal strobing. Temporal strobing is the disruption of the sequential perception of an image as it moves. Because there is no blurring effect the sequential position of an object becomes spaced far apart. This problem does not exist in live-action film because while the shutter is open the object's motion is recorded as a "smear" across the frame. This smearing contributes to the communication of con­ tinuity and in its own way contributes to the perception of

"squash and stretch". In the computer this "smear" can be simulated by elongating an object and applying successive degrees of transparency in the direction of its motion. Thus the stretch factor of "squash and stretch" is an elongation or the appearance of elasticity relative to the direction of motion. Stretch does not necessarily follow physical laws but can help to alleviate the problem of temporal aliasing in com­ puter graphics and film. Squash directly follows the laws of physics. Simulating the shape changes that objects undergo while accelerating and decelerating improves the perceptual clarity of the motion [45]. 197

An alternate method for simulating "squash and stretch" is to use a sophisticated deformation technique. Deformation cap­ tures the fluid, elastic behavior of a changing form. By building this elastic behavior into the deformation of a form as a relationship between the kinematic and dynamic attributes of the skeleton, squash and stretch, follow through and over­ lapping action, and exaggeration can be automated [46].

Chadwick and Parent [47] have suggested that prismatic joints, functioning as springs or shock absorbers can be used to form the foundation for exaggerated squash and stretch where needed.

The principle of "anticipation" can be viewed as the ana­ tomical provision for an action. It is a counter balance to the action impulse; the body stance that permits the action to be launched. The principle of "follow through and overlapping action" would be the natural dynamic consequence of an action.

Inputting dynamics logically results in this kind of motion.

"Slow in and slow out" deals with the spacing of the in-between drawings between the extreme poses. "Cel animators felt that one of the most objectionable traits of early computer anima­ tion motion was its lack of easing [48]." In most 3D key-frame computer animation systems the "in-betweening is done automati­ cally using spline interpolation [49]." "Slow in and slow out" is achieved by adjusting the tension and continuity of the splines. "This works well to give the affect of slow in and out, but a graphical representation of the spline is required to see the effect of tension, direction, and continuity have on 1

198

its shape [50]." However, this concept can most easily be han­

dled by the acceleration attributes associated with an object

and its mass.

"Arcs" (curvilinear paths of movement) are accomplished in

key-frame systems through splines. An important consideration

for a system should be that the

...spline that defines the path of action should be separate from the spline that defines the timing or spacing of the in-between frames for several reasons: so that the arc of a fast action doesn't flatten out; so that you can adjust the timing of the in-between frames without effecting the path of action; so that you can use different splines to define the path of ac­ tion. .. [51] .

If information from biological descriptions of the kinematics of organisms and planning strategies is incorporated arcs will be naturally output.

Another principle of animation, "secondary action" is the reaction that results from an action. Secondary action can be accomplished through a collision detection mechanism or behavioral simulation. As an object collides or interacts with another, a force is transmitted which results in movement being propagated through the scene. Facial expressions are secondary actions in the context that they evidence an "emotional" reac­ tion. When the main idea of an action is communicated through the movement of the body the facial expression becomes subordi­ nate to the main idea. 199

"Appeal" (the attraction or aesthetic quality of the work) might be considered one of the strongest points of this model.

In animation, awkward, inconsistent, jerky, or unnatural motion results in a breakdown of the continuity and a lost of that illusion of reality created by the film medium. The use of simulation techniques results in a much tighter control of the motion continuity. From this new control more expressive mani­ pulations will evolve. The "exaggeration" principle may be thought of as the consistent change of a quality or charac­ teristic for communicative emphasis. This factor is discussed in more depth in Chapter VI.

Methodology in animation has generally centered around the two approaches of "pose to pose" and "straight-ahead". Both of these methods are applicable to this model. However, the straight-ahead method has more far-reaching implications for the future. These ramifications are also discussed in Chapter

VI.

Whether the animator uses a key-frame method or a simula­ tion approach, it will be necessary to assemble the components in some responsive or interactive manner. The mediated aspect of computer animation makes it difficult to control key-frames and in-between frames. Lasseter [52] points out that in work­ ing with a complex character, creating one complete pose at a time (all characteristics defined together) would make the in- between frames too unpredictable. Unexpected changes would materialize between pose extremes requiring numerous revisions 200 of in-between frames,

In the context of hierarchical modeling there is a much better approach which works "layer by layer" down the hierar­ chy. Lasseter describes the process:

Instead of animating one complete pose to another, one transformation is animated at a time, starting with the trunk of the hierarchical tree structure, working transformation by transformation down the branches to the end. Fewer extremes are used. Not all translates, rotates, and scales have extremes on the same frames; some have many extremes and others very few. With fewer extremes, the importance of the in-between frames increases. Tension and direction controls on the in­ terpolating splines are helpful in controlling the spacing of the in-between and to achieve slow in and out [53].

Chadwick makes the point that

In order to effectively fine tune each degree of free­ dom precisely, each parameter is worked individually for a sequence of motion. Parameters are added and layered to build the desired motion. This effectively allows the user to isolate parameters which require fine-tuned adjustment [54].

Organizing the parameters (e.g. translation, rotation, dynamics etc.) into a hierarchical system and having the animation proceed "layer by layer" down the hierarchy should prove to be a unique and likely paradigm in computer animation.

5.6 Limitations of the Model

An animation system based on this functional model faces the dilemma of how much should be ready-made for the animator and how much should be constructed by the artist in the system 201 itself. On one hand, ready-made procedures would contribute greatly to ease of control. On the other hand, this ease of control is entirely dependent upon how the motion functions are linked. The specific modules that are supplied in a resultant system is determined by the assumptions made about an artist's working process -and desired output.

For the uninitiated user there should be no "major11 prob­ lems if there is a small number of global parameter values for the motion functions and they are intuitively obvious. How­ ever, these ready-made procedures will unintentionally guide an animator to sets of preconceived forms reminiscent of tradi­ tional art work. Using a ready-made system would negate the primary artistic use of a medium - to discover, create, and produce original imagery.

If the system relies heavily on the artist to specify a large quantity of parameters, some of which have non-intuitive values, then the motion may be very hard to control. Such a cumbersome situation could easily materialize if too many options are integrated in the system.

The animator is faced with a tradeoff of powerful options against efficiency. The compromise will likely be a system that allows for basic motion procedures that are easy to con­ trol, but at the same time permit the construction of complex yet powerful procedures when desired. 202

5.6.1 Limitations of the Mode1-Requirements upon the Artist

This model introduces technical levels of complexity gen­

erally not found in recognized contemporary artistic methods.

In using such a model the artist must acquire a renaissance

perspective of considering all the factors involved (e.g.

goals, limitations, structures, and related activity inside and

outside the system) [55]. As Csuri states, "Hopefully, as we

handle the complexities better, and understand the relation­

ships of the individual parts of an image or motion, then the potential for realizing computer-generated pictures will be

imminent and the justification for new criteria will be forth­

coming [56]."

There is a tremendous amount of information about the

field and the medium itself of which the artist needs to be

aware. Sometimes this reduces to the fundamental obstacle of

finding special information - which is often available from

scientific literature - and deciphering it. Csuri [57] advo­

cates working with other researchers from the diverse fields in question. This is often helpful in the initial stages of

focusing in on the correct problem. For example, there is the problem of domain dependence; what you need to know to animate a human is different from what you need to know to animate a tree. Not only is the model allowing for a changing medium that is active rather than passive; it also prompts the emer­ gence of a new type of artist. 203

Such an artist must be equipped to deal with the technical as well as the aesthetic. A new type of artist will be needed, an interdisciplinary artist. This new artist will have access to tools (i.e. artificial intelligence, virtual realities, robotics) that embody the potential of this medium. Csuri [58] states that it will require "knowledge, skill, perseverance, ingenuity and understanding backed by a sense of order, thought, purpose, and insight" on the part of this new artist.

For example, creative facial animation requires going beyond default expressions (e.g. a pre-packaged smile, a surprised look); the animator needs to be able to define the desired cranium, mandible, and skin elasticity from built-in proto­ types. Artists-users who only allow themselves the depth of information that a list of instructions provides will not be

"computer artists", just as the possession of a camera does not necessarily make one a photographer.

This leads to the question of what type of artist is a computer artist? A computer artist does fit today's stereotype of an "inspired ignoramus" reacting to standardization and industrialization in society. The concept of the interdisci­ plinary Renaissance artist more closely aligns with the artist involved with this digital medium. This interdisciplinary artist will have to be able to traverse the new paths being opened with the computer's assistance - expressive alternatives which emerge from logical processes and choices one makes in the process. 204

In that regard, computer animation can only be mastered to the degree that the artist has knowledge of the system. This is exemplified by Miller's [59] definition between "wiggle" and

"wobble". These qualitative movements are readily known by artists but describing the difference in quantitative terms to the computer can be difficult - if not impossible - if artists do not understand the difference themselves. This is demon­ strated by the following quantitative definitions:

wobble - resonant oscillation in response to external

forces.

wiggle - deformation of shape due to internal forces.

The artist must understand the difference if they are to simu­ late the subtle difference that can exist between movements such as "wiggle" and "wobble".

As to what the future will bring, it is difficult to predict what the ULTIMATE computer animation system will b e .

Reynolds believes it would be hard to visualize a system which allows arbitrary extensions into unexpected realms without being fully programmable even though programming and aesthetic judgement-making seem to be disjointed in most people's think­ ing processes. Therefore the user of such a programmable sys­ tem must always be on guard against compromising aesthetic judgements to simplify the programming [60]. A practical solu­ tion may be a joint situation such as instituted at the

Advanced Computing Center for the Arts and Design at the Ohio 205

State University, a common laboratory where people from dif­ ferent disciplines and expertise work together. 206

NOTES TO CHAPTER V

1. Jack Burnham (1978). "System Esthetics," In, R. Kostelanetz (Ed.), Esthetics Contemporary, New York: Promethus Books, p. 160.

2. David Zeltzer (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, 1, (4), p. 254.

3. Wilhelms, J. P. (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley, p. 30.

4. Alan H. Barr (1987). "Introduction to Physically-Based Modeling," Tutorial Notes No. 3/7, SIGGRAPH * 87, Topic in Physically-Based Modeling, p . 2.

5. David R. Haumann (1987). "Modeling the Physical Behavior of Flexible Objects," SIGGRAPH '87 Tutorial Notes No. 17, Topics in Physically-Based Modeling.

6. Ibid.

7. Zeltzer, ojo. cit., p. 252.

8. Ibid., p. 253.

9. Ibid.

10. Haumann, op. cit., p. 5.

11. Michael Girard (1988). An Overview of Motion Planning Problems in Robotics and Artificial Intelligence. Unpublished manuscript, p. 1.

12. Ibid.

13. Ibid.

14. Michael Girard, (1986). "Computer-Animated Legged Animal Motion," ACM Workshop on Interactive 3D Graphics, Chapel Hill, NC., £. 8_.

15. Haumann, op. cit., p. 62.

16. Michael Girard (1987). "Interactive Design of 3D Computer-Animated Legged Animal Motion," IEEE Computer 207

Graphics and Applications, 1_, (6), p.. 49.

17. R. Tomovic and R. B. McGhee (1966). "A Finite State Approach to the Synthesis of Bioengineering Control Systems," IEEE Transactions on Human Factors in Electronics, HFE-7, 2, (6), pp. 65-69.

18. Michael Girard (1988). An Overview of Motion Planning Problems in Robotics and Artificial Intelligence. Unpublished manuscript, p. 5.

19. Zeltzer, op. cit., p. 253.

20. Girard. op. cit., p. 6.

21. Julian Gomez (1987). "Comments on Event Driven Computer Animation," Notes for SIGGRAPH 87 Tutorial Computer Animation: 3D Motion Specification and Control, pp. 77-90.

22. Craig Reynolds (1986). "Advanced Computer Animation,11 ACM Siggraph ' 86, Tutorial Notes, Advanced Computer Animation, p. 90.

23. W. T. Reeves (1981). "Inbetweening for Computer Animation Utilizing Moving Point Constraints," "ACM SIGGRAPH '81 Conference Proceedings, Computer Graphics, 15" (3), p. 264.

24. Reynolds, op. cit., p. 90.

25. Ibid.

26. Julian E. Gomez (1985). Computer Display of Time Variant Functions. (Doctoral disseratation, The Ohio State University). p. 16.

27. Ibid.

28.( Reynolds, op. cit., p. 126.

29. Ibid., p. 127.

30. Julian Gomez (1987). "Comments on Event Driven Computer Animation," Notes for SIGGRAPH 87 Tutorial Computer Animation: 3D Motion Specification and Control, p. 85.

31. Ibid., p. 78.

32. Reynolds, op. cit., p. 205.

33. Paul M. Issacs, & M. F. Cohen, (1987). "Controlling Dynamic Simulations with Kinematic Constraints, Behavior Functions and Inverse Dynamics," ACM SIGGRAPH '87 208

Conference Proceedings, Computer Graphics, 21, (3), p. 219.

34. Ken Perlin, (1985). "An Image Synthesizer," ACM SIGGRAPH '85 Conference Proceedings, Computer Graphics, 19, (3), p. 287-296.

35. Issac, oj>. cit., p. 126.

36. Zeltzer, o£. cit., p. 257.

37. J. Gips and G. Stiny (1979). "An investigation of algorithmic aesthetics". In F. Malina (Ed.), Visual art, mathematics and computers: selections from the journal Leonardo New York: Pergamon Press, pp. 93-100. 38. Ibid.

39. H. H. Arnason, (1976). History of Modern Art. Englewood-Cliffs, N. J.: Prentice-Hall Inc., p. 22.

40. Julian Gomez (1987). "Comments on Event Driven Computer Animation," Notes for SIGGRAPH 87 Tutorial Computer Animation: 3D Motion Specification and Control, p. 78.

41. Craig Reynolds (1986). "Advanced Computer Animation," ACM Siggraph '86, Tutorial Notes, Advanced Computer Animation, p. 91.

42. Julian E. Gomez (1985). Computer Display of Time Variant Functions. (Doctoral disseratation, The Ohio State University). p. 18.

43. David Zeltzer (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, If (4), p. 254. 44. Ibid.

45. Craig Reynolds (1986). "Advanced Computer Animation," ACM Siggraph '86, Tutorial Notes, Advanced Computer Animation, p. 71.

46. John Chadwick and Richard Parent (1988). "Critter Construction: Developing Characters for Computer Animation," PIXIM Conference Proceedings, 1_, (1), p. 3.

47. Ibid.

48 Reynolds, o£. cit., p. 93.

49. John Lasseter (1987). "Principles of Traditional Animation Applied to 3D Computer Animation," ACM SIGGRAPH '87 Conference Proceedings, 21, (4), p. 41. 209

50. Ibid.

51. Ibid.

52. Ibid., p. 40.

53. Ibid.

54. Chadwick, o£. cit., p.

55. Jack Burnham (1978). "System Esthetics," In, R. Kostelanetz (Ed.), Esthetics Contemporary. New York: Promethus Books, p. 164.

56. Rick E. Lucas (1986). Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study. Unpublished master's thesis, Ohio State University, Columbus, OH. p. 13.

57. C. A. Csuri, J. Blinn, J. Gomez, N. Max & W. Reeves (1983). "The Simulation of Natural Phenomena," ACM Computer Graphics, 17, (3), p. 139.

58. Charles Csuri (1974). "Computer Graphics and Art," Proceedings of the IEEE, 62, (4), p. 514.

59. Presented at discussion of Galvin Miller's Paper, "The Motion Dynamics of Snakes and Worms," at the SIGGRAPH '88 Conference, Aug 4, 1988.

60. Craig W. Reynolds (1982). "Computer Animation with Scripts and Actors," ACM Computer Graphics, 16, (3), pp. 289-296. CHAPTER VI

IMPLICATIONS OF THE MODEL

This study presents a model which is designed to provide a structure for the specification of aesthetic criteria in con­ junction with complex physical phenomena. The aesthetic impli­ cations of this model lie in the 20th century interest in the nature of illusion rather than appearances. It is on the prem­ ise of separating the mode of representation from the objects represented that the modern art movement began. This separa­ tion is inherent in the process of computer graphics and this model.

Like contemporary art, the creative potential of this simulation/animation model resides in the relationships between the parts. The actual creative ramifications are dictated by its mechanisms of control and how they are linked together. As the relationships of the individual parts of an image and its motion are better understood a better handling of the complexi­ ties involved will emerge. From this process creative works will evolve, and it is hoped that the “justification for new criteria will be forthcoming [1].“

210 211

6.1 Fidelity

The model's fidelity is based on its ability to reproduce

and control complex realistic motion for creative applications.

The judgement of fidelity is concerned with the realism and

efficiency of the properties generated from the model. This

judgement contrasts with the simulation of natural phenomena in

the descriptive sciences. The descriptive sciences use a more

comprehensive standard of correctness for judgement.

This model provides a structure for recreating realistic

motion phenomena that are generally too complex to be ade­

quately reproduced by the artist's skills alone. Complexity is

accomplished through a concentrically organized structure which

permits a layered approach. This "layered" approach permits

the artist to focus on selected areas for refinement. For

example, the artist may first refine the horizontal translation

information, then separately select and refine the vertical

translation, then refine the rotation around the x axis and so

forth until the sequence is satisfactory. This approach dis­ tinguishes this model from its manual predecessors in which all qualities of motion had to be worked out upon the same series of drawings.

This procedural approach provides an automatic assistance through the functional and behavioral procedures which incor­ porate kinematics and dynamics. This, in turn, leads to a 212

greater efficiency and accuracy in producing realistic motion.

The resulting realistic motion produced is within the domain of

the mimetic tradition in art.

Art, in the context of imitation, adheres to the mimetic

tradition; the quality of the work is judged by its faithful­

ness to the reality imitated. This is the foundation from

which the majority of computer graphics/animation is viewed

today. The artist who desires to duplicate the physical world must have at his command tools that permit this realistic

reproduction. This is particularly important whenever two or more characters are to be animated. "Technically, it is diffi­ cult to animate two characters sharing a space, moving them around, without their stepping on each other, while keeping a general feeling of dimension and volume in the scene [2]." In this respect, the fidelity of the model excels. Through the behavioral mechanisms of the model, complex motion can be real­ istically accomplished in a straightforward manner.

Technically, the mere reproduction of known reality through simulation methods does not require special artistic activity. However, artistic creativity will become increasing important as imaginary scenarios present themselves needing visualization. The resulting visualization techniques would lend themselves to surrealistic animations. Through the use of sophisticated illusion technology, the spectator, being part of the scene, would partake in a genuine experience, an "event".

As these new options present themselves the fidelity 213

demonstrated by this model will become mandatory if realistic

motion is to be satisfactorily reproduced.

6.2 Utility - Creative Process

The utility of the model can be assessed by examining how

well the creative process works as it unfolds through the use

of the simulation model. The "creative process" may be defined

as the transformation or emergence of new ideas or artifacts

which result from a series of interconnected actions or events

[3]. There is a close similarity between that definition and

the model because of the model's interconnected organization.

Wallas [4] further delineates the creative process as composed

of four stages: preparation, incubation, illumination and

verification.

The preparation stage of the creative process is identi­

fied by a high degree of divergent mental activity. In this model divergent exploration is accomplished by the interre­

lated, concentric organization (e.g. the linking of structures,

attributes, and functions) coupled with real-time interac­ tivity. This exploration enables artists to systematically

investigate their intellectual, emotional, and intuitive reac­ tions through contemplation and inquiry. Flexibility is a major factor in the preparation stage. The flexibility of the model contributes to how versatile the process is towards applying variations to the problem domain. Browsers (i.e. fast software-controlled switching patterns) provide the ability to 214 facilitate this exploration in the model.

While interactivity, as a function of art, is not in itself unique, the concentric organization and "real-time" response of the model qualifies as a distinct change from pre­ vious endeavors in the art process. The coupling of mechanical and structural complexities with the real-time interactivity of the computer permits numerous variations to be worked out in rapid succession. This divergent, interconnected ability pro­ vided by the model's structure is critical in the search for new creative solutions and insights.

The incubation stage of the creative process - the result of the highly-divergent, discovery-oriented preparation stage - permits a transition to more convergent activity. This is pos­ sible because in the model the computer preserves motion descriptions within the form of a script. Scripting can be in the form of a fully-specified script with exact key-frame specifications or as a set of initial variables and constraints to be imposed during the process, as in the case of self- scripting. A variety of creative solutions can be explored because the script permits the artist to preserve the original information for later manipulation.

The stage of illumination generally overlaps with the stage of preparation in the creative process. As a direct result of interaction with this model, "inspirational" leaps of association can take place in which new or novel solutions sur­ 215 face. This may happen in a state of play, but most likely will occur as a result of linking structures, attributes, and func­ tions together, resulting in motion behaviors for which the viewer does not hold "preconceived" notions. The most per­ vasive limitations to overcome in computer animation are the assimilated rules and conventions that have evolved from our familiarity with physical media and society's feedback as to what is significant and what is not. Computer art/animation is tightly bound by these acquired aesthetic conventions.

Scientific artifacts inherent in the model may be viewed as positive elements which promote "illumination" and stifle prejudicial preconceptions. For example, in the inverse- kinematic program PODA [5], articulated figures are referred to as bi-peds or multi-peds, depending on the number of support limbs. Though this nomenclature originated from a scientific orientation, it is useful in avoiding prejudicial expectations.

This orientation can prevent the artist from conceiving subjec­ tively (dogs, horses, humans versus multilegged characters).

This orientation permits new associative insights which will be needed if the artist is to truly exploit this medium. Accord­ ing to Burnham, "Disengagement with preconceived enduring forms and orders of things is a positive assertion [6]." New complex associations not previously conceivable are to be realized through the model.

Another important factor in the utility aspect of the model is the extended capability of the computational process 216 in using variables. People can generally retain a limited number variables (i.e. 7 to 10) at any one time in their cons­ cious memory. The computer is not limited in the same way. and can retain and manipulate many variables simultaneously. The computer extends the capabilities of what the artist can do.

After illumination the individual must at this point work through a number of convergent associations in order to make the new insights a practical working proposition. This verifi­ cation phase of activity generally consists of a convergent effort often accompanied by much trial and error. The output from the model is in the form of synthetic data (i.e. polygons and vertices) which is then piped to the rendering algorithm for display.

After working through the model and its relationship to the creative process, it is found that the model meets the necessary criteria set aside for assessing its utility - usa­ bility, flexibility, extensibility, and habitability. Artists working within the structure of this model are able to either accept the tool structure provided or expand the structure to satisfy their creative needs. Artists creates their own work­ ing algorithm from the model's different levels and components.

Upon further development of the algorithm the output is removed further and further from the application process and drawn closer to the art process. Finally, the process will combine into "higher order tools leading up to the creation of the resulting art object [7]." 217

This higher order tool can take the form of interpretative conventions and evaluative criteria. The creative process con­ tinues to follow a similar path but takes on a different form in the computer. For example, preparation would be in the form of codifying evaluative criteria. This criteria would be applied by the computer as a type of screening process in the incubation. The illumination and verification steps still rely on the judgement of the artist.

6.3 Model as Medium

Jerome Stolnitz, in his book Aesthetics and Philosophy of

Art Criticism, concluded that what can be said about the creative artist "with relative certainty, if anything, is prob­ ably this: the artist is one who 'thinks' in terms of some artistic medium [8]." Stolnitz [9] concluded that a "medium" is comprised of certain sensory elements - colors, tones, etc.- which are ordered and interrelated. Just as painting or sculp­ ture embody specific elements inherent to their vehicle, so does the digital medium. Arnheim supports this premise in his analogy, "Poetry returns us to words, not to messages, and painting focuses our sensibilities on line, color, composition, and other formal properties rather than on representation

[10]." Similarly, computer simulation has the capacity to return us to the essence of movement rather than just emphasiz­ ing the world it imitates. 218

A medium is not simply "passive", not just a respondent, a

recipient of action from the artist. A medium has constraints,

physical or otherwise, that influence its manipulator. The

medium, in fact, assumes a character of its own, which permits

some things to be done with it, but not others. The computer

as an active medium gives direction to the creative process by

suggesting to the artist unanticipated images and ideas [11].

These suggested new areas of exploration motivate the artist to

deviate into unexplored realms.

Simulation as animation, however, will not constitute a

new and unique form until it is characterized by "unique" pro­

perties of the medium, unique properties such as interactivity

(direct, immediate response), simulation (virtual space and

time), and intelligence (the capability to incorporate

decision-making strategies from embedded rules and con­

straints) . It is these qualities - conveyed by the structure

and elements of the model - that carry the artistic meaning.

Content, in the formalist tradition of modern art is embodied

in the elements of the medium. The important new elements of

this computer animation medium are its accompanying abstrac­

tions (e.g. functional, procedural, resolved motion, etc.). It

is through this model that artists can layer abstractions into

a powerful medium through which to fulfill their creative

impulses.

Zeltzer [12] hypothesizes that interacting with the com­ puter in this manner is analogous to the "straight-ahead" 219

technique of 2-D animation. In straight-ahead animation the

animator's ideas are stimulated by what has taken place previ­

ously in the animation. The animator receives stimulus from

the piece itself. By automating this process, with the

computer's assistance, an animator could establish the rules of

a simulated environment and "let the animation system generate

the sequence [13]." The animator could then choose from the

numerous variations that could be generated in this manner.

Because the possible variations can quickly increase into a

staggering number of choices, it is unlikely someone unfamiliar

with the structure of the system would be able to generate any­

thing more than novelty sequences.

To truly utilize the computer to its full potential, it is

necessary that the artist be not only proficient with the

medium but that he understand and assimilate the new conceptual

framework. In this way the artist may continue to selectively

emphasize (e.g. grouping, closure, simplicity) those elements

which have a direct bearing on the expressive meaning. "It is

no contradiction that an artist like Caravaggio became more

imaginative and inventive by not trying to imagine or invent,

but by selectively viewing the variety that nature presents

[14]

In the case of animation, continuity of the motion is the primary creative element, and it must be either credible, prob­ able, or necessary to have relevance in the work. Ironically,

computer animation is assessed in a large part by noting how 220

"convincing" the motion is because it is no longer being gen­

erated by hand-drawn methods - resulting in the "simplified and

stylized imagery" generally associated with traditional anima­ tion. Computer-generated animation can be made to look more realistic or more convincing in its other-worldliness; it can take on added complexity and detail. The credibility of the motion generated by simulation techniques conveys a convincing and consistent sense of reality.

The discovery and communication of new realities are characteristic of the artist in our culture. Therefore, the animator should not only have the ability to generate realistic movement, but also be able to create an "ordered" foundation away from the accepted "natural" by varying these same physical laws.

6.4 Potential of the Model

The functional model in this study proposes a structure to seamlessly integrate pseudo live-action - dynamics and kinematic control - within the digital medium. The potential is for new animation techniques which transcend the stiff and unnaturalistic movement associated with traditional rotoscoping techniques. Thomas [15] hypothesized that the problem with traditional rotoscoping is that there is no reality to the copy and that the camera records everything with an impartial lack of emphasis. 221

This model provides a solution to both of these deficien­ cies. First, procedurally-based motion provides a reality of sorts that can be manipulated (tweaked) by the artist, a way to take it beyond the copy stage. This procedural motion is accessible and manipulatable, which is not the case for live- action unless the scene is reshot. Secondly, the artist, by selectively altering parameter variables, can effectively direct a viewer's attention within the animation. For example, an inanimate object such as a chair could be assigned kinematic motion that ordinarily originates from a live character result­ ing in the viewer's attention being drawn to this contrast.

As technology continues to become more sophisticated new creative forms, involving interactivity will play an ever greater role. This interactive ability of computer systems holds the key to radical changes within the art-making process as spectator becomes an integral part of the environment.

These interactive transformations occur in "real time"; that is, the processing happens as soon as the stimuli is received, and the results are visible immediately. The relevant question will soon be not whether this interactivity can provide dramatic new perceptions, but whether it becomes the new per­ ception. Early in the history of computer graphics, Csuri pro­ posed the concept of interactive, real-time art.

Real-time computer art objects are an intellectual con­ cept which can be visually experienced rather than a finalized materialized object. This kind of computer art exists for the time the participant and the comput­ er with the CRT display are interacting as a process. 222

The art object is not the computer or the display, but the interactivity of both interacting with the partici­ pant... Real-time computer art objects are a unique art form [16].

Nake [17] believes that this interactivity is an issue of amount (a level of sophistication); that the interactivity issue can influence the audience-artist relationship beyond traditional art forms. This logically leads to the switching of the "artist and audience" relationship to the relationship of "author and participant". This new interactive environment becomes conversational; its laws change as a result of its interactions. "This is the ultimate case of Marcel Duchamp's dictum that the artist begins the artwork and the witness com­ pletes it [18]."

The unique aspect of simulation by the computer can also have an equally dramatic effect on the role of "animator" which may be exchanged for the role of "director". As director, the artist does not use the computer to create a specific image or sequence, but rather supplies input that permits the derivation of image variants and sequences by the character itself as an active character [19]. While "character animation" normally refers to the likes of Mickey Mouse, in this context the term

"character" is meant to refer to any object to which a viewer would tend to attach a personality. According to Reynolds,

Such an object should have a "characteristic" appear­ ance and style of moving. If they are well-presented, the viewer can quickly absorb these traits and start to anticipate how a character will react to a given 223

situation... Once that level of sophistication has been reached the important issues become the same as with any sort of dramatic art: story, characters, believa- bility and pacing [20].

This automata is possible through the use of generative model­

ing. The simulated world is not previously defined in detail

but guided by built-in computer instructions. These instruc­

tions on the classes of objects that exist and their relation­

ship to other objects are actived as the user approaches them.

This enables the user to venture as far as desired, all the

time moving in environments that are different from the ones

previously "walked" through.

Simulation can also provide a foundation for evolutionary

forms that could be thought of as "artificial life". This

artificial life is accomplished through the simulation of the

ecology and evolution of organisms (however primitive) in terms

of their motion control skills [21] . These motor skills might

include walking, climbing, grasping, throwing, etc. Computer­

generated organisms would use these skills - in conjunction

with path-planning algorithms - to navigate through their

environment by making adaptive adjustments to external forces.

The "animator" would specify the initial conditions of the

system's ecology to be simulated. According to Girard [22] this might be accomplished by building organisms from libraries

of predefined physical and behavioral skills. These libraries would classify motor and planning skills in terms of behavioral traits. 224

In addition an organism's biological system could be simu­ lated by superimposing traits (e.g. life span, interaction strategies with various organisms, impact thresholds, reproduc­ tive cycles, viewing distance) [23]. If an environment popu­ lated with such organisms were given the capacity to reproduce, along with the potential to introduce subtle structural changes as they do so, a physically mutating population would emerge

[24] .

This mutation capacity could be creatively extended to psychological forces which in turn could influence behavior.

One such scenario could be based on psychological research.

For example, there is research that suggests that the "self" is a product of an individual's relationship with other people

[25]. This "self" trait sometimes manifests itself within a high-monitoring individual who changes behavior to promote smooth interaction with others. Woody Allen, in the film

"Selig", portrayed such a character whose adaptive and change­ able behavior was so extreme as to cause physical changes as well. A similar computer-generated character might be created whose physical form would be mutated by not only physical con­ straints but also by behavioral and stylistic constraints [26].

Csuri [27] has suggested that this natural selection pro­ cess simulated by the computer could be extended to embrace the creative process. The creative choices an artist makes represent the artist's view of reality and eventually determine the aesthetic content of his work. These choices are made from 225

a primary set of working constructs that an artist has at any

one time. It should be possible to codify this finite set of

constructs into the computer as elements which aid in the final determination of the aesthetic context, representation, or expression.

The symbolic issues of art and this new medium are beyond the scope of this study. Theory is not dismissed at this time but requires the further development of computer resources to be fully explored and their ramifications understood.

6.5 Criticism of the Model

Nake [28] states that a work of art should be judged from the message it tries to communicate in relation to the means it employs for achieving this. It is difficult not to immediately compare computer-generated imagery with the prevalent art para­ digm. This comparison is perpetuated by the fact that "we", as artists, inevitably bring our current constraints, preconcep­ tions, and prejudices to the computer, however these biases are not inherent in the computer medium. Artists must look at options outside of their current realm of experience. Computer graphics/animation has new emerging features which make current guideposts of art criticism irrelevant at times.

The significant question that should be asked is,

Where within this struggle to produce art by the com­ puter lies that elusive new breath of life meriting the long awaited computer art movement? Are we mistaking a new medium, which merits a new approach not necessarily 226

a movement? [29].

This study'' s model integrates algorithmic aesthetics as a solu­ tion to incorporating aesthetic criteria into the process. It may be that attention should turn to the potential of the aesthetic quality of algorithms instead, not for the purpose of creating more elegant algorithms but for how qualities of algo­ rithms (e.g. efficiency, flexibility, robustness, and interac­ tivity) might reflect on issues of aesthetics [30].

Another major concern is a dependency on the motion gen­ erated by the model's modules which would result in a deference of judgement by the user to the "correctness" of the motion generated. Paradoxically, this deference to the computer is misplaced if the sequence illustrated in Figure 32 (generated by a keyframe animation system) and a similar sequence in Fig­ ure 33 (generated by a procedural methods without artistic input) is any indication.

In the example by Witkin (Figure 33) the Lamp moves con­ vincingly but with an impartial lack of emphasis on any one aspect. In contrast the example by Lasseter (Figure 32) demon­ strates an emphasis on aspects of anticipation (a more extreme compression of the lamp before springing from its position) and an exaggeration of the struggle of the Lamp to overcome weight in its base while jumping. These changes by Lasseter more closely align with our anthropomorphic expectations of the character. Most users of a computer animation system will not Figure 33. Procedural generated motion Andrew Witkin.

Figure 32. Key-framed motion. John Lasseter. 228 possess the technical skill necessary to generate such an effective sequence by keyframe methods. However with the motion modules of this model a close proximity can be generated and then refined to fulfill artistic goals.

6.6 Parallels with Modern Art

Most artists are not involved with the computer and gen­ erally regard it as a foreign entity with no applicability to themselves or aesthetic issues. While the computer does not necessarily address artistic issues, it is indicative of signi­ ficant developments, just as the camera was. With its use artists are forced to examine their own basic beliefs about art. This at the very least parallels the introduction of modern art which challenged artists "to refine and extend our conception of art [31]." It is probably safe to say that modern art was not born by choice but by necessity. Technological advances (e.g. introduction of the camera, new scientific theories) rendered previous justifications for art obsolete.

Today the computer's capabilities to graphically simulate phy­ sical phenomena will have an equally profound, though dif­ ferent, effect on art. There are numerous parallels between the various introductions of new descriptive systems (e.g. per­ spective, Impressionism).

Similarities exist between the state of art after the introduction of the camera (modern art)and art after the intro­ duction of the computer (computer art). A review of key areas 229

will lead .to a a better understanding of the similarities that

exist for traditional artists. Today there exists a repressive

environment for computer art similar to modern art in its

beginnings. As Kuhn [32] explains, facts which arise through

experimentation and which don't fit the prevailing paradigm are

invariably labeled as "bogus' or trivial until the emergence of

a new and more encompassing general theory. This paradigm con­

tains a body of knowledge and beliefs so pervasive that it dom­

inates all ensuing discovery. Such a paradigm incited the

labeling of modern art as crude and childish. Similarly, com­

puter art is often dismissed as being too cold and impersonal

to be taken seriously as an art form. While such labels are

unerroneous, they overshadow the priorities of originality and

creativity which are identified with art in our culture. This

situation is further exasperated today by the rapid flux of

current technological and artistic shifts.

The early movements of modern art - like computer explora­

tions today - were characterized by a search for fundamental

visual qualities which could be considered universal rather

than specific. The Impressionists reexamined nature in rela­

tion to the media with which they worked with and in the pro­

cess discovered a new range and intensity of color peculiar to pigments as opposed to nature. Van Gogh and Gauguin strove to bring to fuller consciousness their concerns with problems of

structural organization, irrespective of natural realistic

appearances. Cubism stimulated a creative seesaw in which a 230

bottle could be turned into a cylinder and then later, through

analytical cubism, a particular bottle could be made from a

cylinder [33]. This approach resulted in new dynamic arrange­

ments based on universal “plastic" elements of art. These

early explorations in modern art resulted not only in new

creative works, but also in an acquired knowledge of the

characteristics, limitations, and appropriate methodologies

which would set the stage for future endeavors.

The Futurist movement (approximately 1912) holds special

interest in relation to this study. Futurism emphasized move­ ment as a viable expressive subject matter and stressed process

over product. For Futurists, the concept of movement changes the meaning of forms, as movement continues to alter relation­

ships between the visual forms of the work. In this regard,

Futurists broke with the past tradition of object-orientation

in favor of more creative avenues. As Boccocini states,

There is no fear more stupid than that which makes us afraid to go beyond the bounds of the art we are prac­ ticing. These is no such thing as painting, sculpture, music, or poetry; there is only creation! Therefore if a composition i3 in need of a special rhythmical move­ ment to aid or contrast with the static rhythm. ... one can superimpose any structure whatsoever that is capa­ ble of giving the required movements to the planes or lines [34].

It is the ability of the computer to realistically simu­ late the physical properties of nature which imparts a surreal­ istic quality to resultant images. Surrealist Rene Magritte would probably have assimilated these surrealistic simulation 231

capabilities of the computer in his work if they had been

available. The meta-reality created by Magritte was accom­

plished through techniques now associated with the computer

(e.g. numerous variations, derivations, cross-references, com­

binations, transformations and synthesis) [35]. Magritte gen­

erated his magic realism in a precise and realistic manner; the

magic arises from the juxtaposition of elements that do not

normally belong together. This style of realism is evidenced

in numerous computer-generated works today. It is interesting

to note Magritte's fascination with the surrealism of early

efforts in film technology.

This surreal expressive change from the norm - accom­

plished by computer animation/simulation - is demonstrated by

the New York Institute of Technology's film "The Works". In this film the character of the ant is clearly not "real", yet the viewer can feel its mass and power and marvel at its agil­

ity (Figure 26). "The correctness - the absolute surreal

correctness - of the mechanical simulation gives a unique and captivating action [36]." The surrealistic quality of computer-generated imagery - of unflawed surfaces and con­ sistent movements - has an aesthetic appeal of its own.

Motion, in the real world is seldom perfectly smooth or geometrically precise in its movement. The surreal nature of simulated mechanical motion has a special appeal for the viewer. It has been suggested that our eyes may "like" to look at these surreal forms because they provide just the sort of 232

cues that our vision is best attuned to perceive [37].

Interest in the conceptual framework of art and its objec­

tive realities characterized the work of such artists as Piet

Mondrian and Marcel Duchamp. Their ideas are quoted here because they correspond to ideas that are being re-explored in

computer art today. For Mondrian,

Art makes us realize that there are fixed laws which govern and point to the use of the constructive ele­ ments of composition and of the inherent interrelation­ ships between them. These laws may be regarded as sub­ sidiary laws to fundamental laws which create dynamic equilibrium and reveal the true content of reality [38] .

This idea reflects the motivation this author had for research­ ing simulation and its artistic potential. Marcel Duchamp pro­

fessed a personal movement away from the physicality of paint­

ing. "I was interested in ideas— not merely in visual products

[39]." This movement resulted in explorations on the part of

Duchamp, near the-latter part of his life, in mechanics as a potential medium.

More recently there has.been a disparity between contem­ porary artists that has a direct bearing on the use of the com­ puter in art. On one hand, there are established artists such as David Hockney, Philip Pearlstein, Kenneth Noland and Larry

Rivers who have worked with digital paint systems to produce art work. However these artists used the digital process to produce work derivative of methods they already employed. On the other hand, there are contemporary art forms (e.g. 233 environmental art and performance art) which embrace the pro­ cess over product and utilize the computer for the unique qual­ ities it can bring to the work. To account for this dichotomy

Jack Burnham [40] has proposed that an artist's philosophical orientation has a significant bearing on his working process and final product. It may be that the most significant indica­ tion of whether an artist might lean towards incorporating a computer into his working process, rather than using it as just another tool, is whether the artist is object-oriented or system-oriented.

For the system-oriented artist, art does not reside in material entities but in relationships. "For these people sig­ nificance does not emanate from things but from the way things are done [41]. Jack Burnham defines the system artist as a perspectivist considering goals, boundaries, structure, input and output. The continuity of a system may be attended to in time and space, and its behavior determined by its mechanisms of control and interconnection. Systems resist functioning as an applied aesthetic but reveal principles underlying the ongo­ ing reorganization of art. The system-oriented artist strives to reduce technical distance between his artistic sensibilities and the medium, gradually transforming artistic and technologi­ cal decision-making into a single activity [42]. It appears that object-oriented artists have re-integrated the mystic tradition of craft and the value of private creation into their work, traits which do not easily fit within the domain of the 234 computer? It is the system-oriented artists who will success­ fully exploit the computer and this study's model. This system-orientation has been evidenced in the collaborative working process of such contemporary artists such as Robert

Rauschenberg and Andy Warhol.

Other noticeable parallels include: the noticeable divi­ sion between the art establishment and computer art as there was in the beginning of modern art, the circumventing of art value wholly within finite objects is evident in early modern art (e.g. Cubism, Dadaism) as it is within computer art (elec­ tronic screen), the fear that the technology - whether it be camera or computer - will usurp the artist's creativity, and that the resistance to both modern art and computer art ori­ ginates primarily from the established art community not from the general public. The truth of these parallels remains to be determined. It is up to the computer artist to exploit the distinct characteristics of the medium (i.e. simulation) that will make it a distinct art form.

6.7 Future directions

In the future, artists will have to come to terms with the escalating interaction between man and machine and the inevit­ able changes this will bring to art. Developments in interac­ tive computer graphics, simulation techniques, and intelligence applications will profoundly transform this cybernetic interac­ tion. In turn, our efforts will be focused towards finding the 235 proper repertory of ideas and techniques by which to communi­ cate new concepts which emerge from these developments. Robert

Mallory [43] proposes a frame of reference for this cybernetic encroachment which will serve as a guide for the computer's progressive influence.

Mallory [44] proposes six distinct levels. At the first level, the computer simply executes mathematical calculations.

In the second, the computer becomes an indispensable part of the art-making process. By the third level, the computer per­ forms "not only routine discriminations but decides on alterna­ tive courses of action governing the whole system." The options and their functions have been strictly defined and remain deterministic during operation. In the fourth level, courses of action unanticipated (non-deterministic) by the programmer are determined by the computer "...the computer has arrogated to itself both human and machine functions." The computer at the fifth level surpasses the performance of the artist to such an extent that he or she "like a child, can only get in the way." If the sixth level is ever attained the very term

"artist" may no longer have meaning. The machine may "have achieved some kind of organic, self-replicating mode of existence, or will have evolved into a state of pure, disembo­ died energy or spirit [45,46]."

First Level 236

The computer, at this stage, is used primarily to make

already developed techniques more efficient. For example,

artists may use computers to control light, objects, or for

special effects. The light sculptures of Bruce Naumann are not

noticeably different from his non-computerized work [47], Used

in this manner the computer replaces electromechanical devices

that are not as sophisticated.

The computer does provide a higher degree of complexity

and new artists (i.e. Eric Stalher) are discovering the unique aesthetic qualities that have a direct bearing on their work.

This leads the next level in the framework.

Second Level

The concept of ,,computer,, animation assumes the computer is already an integral part of the art-making process and has

incorporated the first cybernetic level. Interpolation tech­ niques within computer animation relieve the animator from the tedium of drawing the inbetween frames from keyframe to key­ frame. However, the limitation of this approach is that all motion must still be explicitly described by the animator.

This places an inescapable limit on the motion complexity that can be introduced. It is at the next level that a means to automate motion generation is undertaken.

At this second level the artist must come to grips with a shift from a material orientation to an electronic orientation caused by the computer's integration in the art process. In 237

this context, art shifts from objects to events [48] . For

instance, while a painting is a "material object" when it is in

our presence, it is an event when it is viewed on a display

screen. This contextual shift dramatically affects the defini­

tion of art and its identifying characteristics that the gen­

eral public relys on to help them in their determination of

what is art.

Third Level

In this level the expectation is for more convincing

images because the computer can be programmed to lighten the burden of motion specification. Images previously simplified and stylized because of limitations inherent in a more explicit approach can take on an added complexity and detail, which in turn permits the artist a wider range of naturalism from which to select. A more directorial role can be assumed by the ani­ mator as the motion, once defined, can be generated by the com­ puter and the artist can concentrate on "what" a character does rather than "how" it does it. Simple kinematic and dynamic descriptions can form a foundation layer on top of which more sophisticated actor-based objects can be implicitly defined.

At this point, unique features (e.g. interactivity, simu­ lation, intelligence) of the computer medium that can be exploited become evident. Besides being able to automate motion by new simulation techniques, intelligent control over animated objects can be introduced. By means of defined 238 abstractions {e.g. functional, structural, etc.) and embedded knowledge, a degree of flexible intelligence can be attained.

The use of goals and multi-directional constraints to achieve knowledge-based control is applicable to both physically-based simulations as well as the "creative intentions" of the artist.

This is evidenced in this study's model where the "Principles of Animation" provided a ready made set of rudimentary aesthetic criteria. This study's model - as far as it remains deterministic - would be initially classified with this level.

In addition, the implementation of interactivity in con­ junction with basic embedded information, in the form of multi-directional constraints, alludes to the potential of interactivity as a new art form. Levels of sophistication

{e.g. speed of response, deterministic versus non-deterministic response, intelligence) appear to be the qualifying factor.

Fourth Level

While the third level will be an important area of inves­ tigation for the near future, it is the fourth level, whose implementation is now being proposed in this study's model, that holds special interest for the artist. At this level the computer becomes capable of coming up with unanticipated courses of action. This is the primary difference between level three and level four. The man-machine interaction pro­ duces non-deterministic (unanticipated) responses as the result of an increased number of options being made available within 239 the system and increased computer power. Some of the same scenarios involving interactivity, simulation, and intelligence evidenced at the third level become fourth level when a suffi­ cient number of variables have been specified and are interre­ lated. Chaos research had indicated that a deterministic

(anticipatory) system will unavoidably become non-deterministic

(un-anticipatory) because of fluctuations (e.g. options chosen, numerical accuracy) which can rapidly amplify through the sys­ tem. Chaos through the amplification of small fluctuations which propogate into larger unanticipated fluctuations in another part can provide natural systems with access to novelty.

There is a point, consequently, where the computer will inevitably generate unanticipated changes. This leads to a

"what would happen if" strategy for the artist to employ. For instance, an animator could procedurally define the behaviors of a set of "actors" and populate a "micro-world" with these actors. By running a simulation of this micro-world, a "what would happen if" scenario would take place. On an elementary level, a character within this micro-world could use chaotic path control as an element of surprise to evade a predator's attack. This is analogous to the random experiments of the

Dadaists early in the history of Modern Art.

The artist does not specifically know what will happen but does control the environment and the known parameters. This process of interactive, knowledge-based simulation could be 240

elevated to its own art form if it could be run in real time.

Today many have turned their attention to the potential of

superconductivity. Yet,, in the future it is the parallel

nature of physical simulation and graphics that could have the

greatest effect on speed, productivity, and new emergent options.

With the advance in parallel computing this fourth level

of the framework will become a functioning reality. The

increase in computation will power the broad application of

complex variables which will reveal unimagined realms in need

of exploration.

Fifth Level

At this fifth level, technological advances in parallel

computing will permit the incorporation of previously incon­

ceivable amounts of aesthetic criteria (interpretative conven­

tions, evaluative criteria). This study's model provides a

basic structure for this futuristic scenario. Advances in

artificial intelligence, qualitative simulation, and chaos

theory will fuel this superior function of the computer.

While only a few years ago such a possibility was still

largely unimaginable, new research in the field of "chaos" sug­

gests that a system - with as many variables that we would need

- would begin to enter into an unavoidable randomness [49].

Chaos provides a means of structuring random changes, thereby 241 providing the possibility of evolutionary, variability control

in the model (50].

It has even been suggested that innate creativity may have an underlying chaotic process that amplifies small fluctuations into new ideas or new ways of connecting old ideas [51] . These fluctuations experienced as macroscopic thoughts may be per­ ceived as decisions based on free will. This prospect, if true, would open up unimagined realms and challenge us to reex­ amine the inalienable assumptions we hold about creativity,

Chaos provides a mechanism that allows for creativity within an environment governed by deterministic laws [52].

At this level the concept of environment may be extended to a truly responsive environment. This is an environment where human behavior is perceived by the computer, which inter­ prets what it observes and responds through intelligent visual and auditory displays [53]. Previously such an environment would be controlled by a preexisting program, however, at this level the artist may choose from a network of artificial intel­ ligence which could be geared towards a specific philosophical orientation. The art form would be a composed interaction between man and computer, mediated by the artist [54] .

As exciting and attractive as this environment and fifth level appear there is the real danger of accepting reality "in a box" [55] . If this level is attained questions of validity must be raised. Today, we partake of life's experiences 242 through numerous media. Books, television etc. all extend to us vicarious experiences which in many ways are more immediate and fulfilling than if we were actually there.

Sixth Level

At level six, if it is ever attained, art will not be in question as much as ourselves. If the current evolutionary rate of machine compared to man is any indication, there are inescapable issues of biological versus "silicon’1 life forms as the inheritors of the future. While such issues are beyond the scope of this study, it is important to consider the inevitable consequences of the course we are on.

One point that must not be lost in the technology is the purpose of art. A statement concerning the function of art by the German philosopher Max Scheler refers to three sorts of images an artist can produce. His remarks are quoted here because they affirm the goals of this study.

The purpose of art is not to reproduce what is already given {which would be superfluous), nor to create some­ thing in the pure play of subjective fancy (which can only be transitory and must necessarily be a matter of complete indifference to other people), but to press forward into the whole of the external world and the soul, to see and communicate those realities within it which rule and conception have hitherto concealed [underlining added] [56]. 243

NOTES TO CHAPTER VI

1. Rick E. Lucas (1986). Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study, Unpublished master's thesis, Ohio State University, Columbus, OH. p. 13.

2. Frank Thomas and Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press, p. 179.

3. J. Wallas (1926). Creativity. New York: Macmillian, p. 34.

4. Ibid.

5. Michael Girard (1987). "Interactive Design of 3D Computer-Animated Legged Animal Motion," IEEE Computer Graphics and Applications, 1_, (6), pp. 39-51.

6. Jack Burnham (1978). "System Esthetics," In, R. Kostelanetz (Ed.), "Esthetics Contemporary" New York: Promethus Books, pp. 160-171.

7. John Chadwick (1988) . "Getting the Aesthetics of the Computational Process Out of the Process and into the Product." Unpublished paper, p. 11.

8. Jerome Stolnitz (I960). Aesthetics and Philosophy of Art Criticism, Boston: Houghton Mifflin Company, p. 103.

9. Ibid.

10. J. D. Andrew (1976). "The Major Film Theories" New York, Oxford University Press, p. 32.

11. Stolnitz, o£. cit., p. 104.

12. David Zeltzer, D. (1985). "Towards an integrated view of 3-D computer animation," The Visual Computer, 1, (4), p. 250.

13. Zeltzer, o£. cit., p. 251.

14. B. Martinez and J. Block (1988). "Visual Forces" Englewood Cliffs, NJ: Prentice Hall Inc., p. 102.

15. Frank Thomas and Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press, p. 173. 244

16. Charles Csuri (1974). "Computer Graphics and Art," Proceedings of the IEEE, 62, (4), 503-515.

17. Frederick Nake (1986). In Rick E. Lucas (Ed.). Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study, Unpublished master's thesis, Ohio State University, Columbus, OH. p. 70.

18. Lucus, op. cit., p. 25.

19. John Chadwick & Rick Parent (1988). "Controlling the Integration of Computational Models for Character Animation," National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, p. 428.

20. Reynolds, op. cit., p. 204.

21. Michael Girard (May 1988). Unpublished manuscript from PhD General Exam, p. 9.

22. Michael Girard (May 1988). Unpublished manuscript from PhD General Exam, p. 11.

23. Chuck Csuri (June 1988). Private Conversation.

24. Ibid.

25. M. Synder (1980). "The Many Me's of the Self-Monitor", Psychology Today, 4_, (13), pp. 33-40.

26. Mark Jansen, (June, 1987). Private Conversation.

27. Csuri, op. cit.

28. Nake, op. cit.

29. John Chadwick (1988). Getting the Aesthetics of the Computational Process Out of the Process and into the Product, Unpublished manuscript., p. 10.

30. Ibid.

31. T. S. Kuhn (1970). The Structure of Scientific Revolutions. Chicago, University of Chicago Press, p. 36.

32. Charles Csuri (1988). Computers and Art: A Medium in Search of a Movement. Unpublished Manuscript.

33. Charles Biederman (1948). Art as the Evolution of Visual Knowledge, Redwing Minnesota, Charles Biederman, p. 283. 245

34. Umberto Boccioni (1968). Technical Manifesto of Futurist Sculpture, 1912. In Herschel B. Chipp (Ed.) Theories of Modern Art, A Source Book by Artists and Critics. Berkeley, CA: University of California, p. 303.

35. Gablik, Suzi (1970) . Magritte. Greenwich, Connecticut: New York Graphics Society Ltd, p. 14.

36. Craig Reynolds (1986). "Advanced Computer Animation," ACM Siggraph '86, Tutorial Notes, Advanced Computer Animation. p. 200.

37. Reynolds, op. cit., p. 203.

38. Mondrian, Piet (1945). Plastic Art and Pure Plastic Art. New York: Wittenborn, p. 54,

39. James Johnson Sweeney (1946) . Eleven Europeans in America, Bulletin of the Museum of Modern Art, 13. (4-5), p. 20.

40. Jack Burnham (1978). "System Esthetics," In, R. Kostelanetz (Ed.), "Esthetics Contemporary" New York: Promethus Books, pp. 160-171.

41. Burnham, o£. cit., p. 160.

42. Ibid.

43. Robert Mallory (1969). "Computer Sculpture: Six Levels of Cybernetics" Art Forum Magazine.

44. Ibid.

45. Cynthia Goodman (1987). Digital Vision. Englewood Cliffs, N.J.: Prentice-Hall, p. 160.

46. Mallory, 0£. cit.

47. Goodman, o£. cit.

48. Roy H. Hill (1983). "Conceiving Art in the 80's," Paper delivered at Digicon 83, International Conference on the Digital Arts. Vancouver, B. C., p. 1.

49. J. P. Crutchfield, J. D. Farmer, N. H. Packard, & R. S. Shaw (1986). "Chaos," Scientific American, 38, (12), 46- 57.

50. Crutchfield, o£. cit., p. 57.

51. Ibid. 246

52. Ibid.

53. Robin King (1988). "Aesthetic Experiences, Personal Constructs and the Evolution of Computers in Art" National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, p. 444.

54. Myron Krueger (1983). Artificial Reality Reading, Massachusetts: Addison Wesley. 55. Ibid.

56. Graham Collier (1972). Art and the Creative Consciousness. Englewood Cliffs, N.J.: Prentice-Hall, Inc., p. 11. CHAPTER VII

SUMMARY

This study proposes a model designed to integrate the simulation of physically-based motion with expressive control strategies for the animator/artist. The model presented in

Chapter V presents the elements, their sequence and relation­ ship, and the procedures necessary to control new levels of visual complexity. The elements of the model are: Geometric

Primitives, Mechanical Attributes, Functional Procedures, and

Behavioral Simulation. The levels of abstract control for these elements include guiding, procedural, and goal-oriented modes. These levels encompass motion generation modules (for­ ward & inverse kinematics and forward and inverse dynamics) that can be linked together inside a feedback control loop with aesthetic control modules of interpretative conventions and evaluative criteria.

The structure and the elements of the model permit the use of complex systems which simulate the appearance of natural processes of movement. This model promotes the reality of

"simulation as animation". The use of simulation methods per­ mits the use of a large body of actions that are not only so

247 248

complicated that they have been impossible to replicate manu­ ally, but can also contribute touches of personality too subtle to capture previously.

These actions indicate untapped areas of animation where com­ puter simulation could demonstrate its unrivaled capability to aid in animation goals.

Simulations can open up new possibilities through the use of automatically generated motion {an object's behavior deter­ mines its next move), and environmental responsiveness (objects are affected by their environment). The results are extremely credible, even when implemented on a elementary level. Simula­ tion techniques that show promise for new innovative capabili­ ties include robust simulation, non-deterministic simulation, and ad-hoc tactics for simulation processes.

The implications are that future animation systems will not only know about surface properties, dimensions, and the configuration of objects in an environment, but also the object's dynamic properties of mass, elasticity, and internal cohesiveness. It is predicted that future systems will embody many of the features of this study's model and lead to the evo­ lution of new paradigms of creative work.

The model's structure parallels the contextual organiza­ tion of contemporary art. In essence, art does not reside in 249 material entities, but in relations between people, and between people and their environment [1]. This is evidenced by the fact that the physical arts of painting and sculpture are not the powerful media they once were; media influence is now gen­ erated by print, cinema, and television. This reality gives rise to the question of whether art as a concept is in a pro­ cess of disintegration or in a process of reformation. The answer lies in the response to the changing context of creativity. Non-deterministic generative techniques, aesthetic algorithms, and interactive capacities of the computer have the power to introduce significant changes to both the creative process and the role of the artist. As Dietrich points out,

The real advances of this art form are still ahead of us, when modes of human intellectual endeavor such as emotion, association and intuition, which are main in­ gredients of art can be successfully simulated by com­ puters. The act of creation itself will become automat­ ed by learning machines which will self-modify their programs, thus interacting intelligently and auto­ nomously with the world. Art-making machines will in­ still new meaning to the process of bridging mind and matter [2].

To bring this potential to fruition requires an artist to truly understand how the medium operates. The motivation for this study was that desire to understand.

This search into the simulation of temporal, physical phenomena has revealed that the field is fragmented, a cohesive body of literature does not now exist, and that there sharply differing opinions on the direction that of future research.

The field has been split between advocates for dynamics and 250 those for kinematics. This devisive attitude is changing.

Prompting this change is the realisation that no matter how good the solution is for kinematic control, it cannot account for rapid or boisterous motions and no matter how realistic dynamic motion is, it cannot account for goal-directed actions.

It is encouraging to find researchers who are now advocating the search for a more holistic structure that might take into account the differing animation goals of artists [3].

As the problem of fragmentation is overcome it is then in the hands of the artist to assimilate a new working procedure.

This study's model introduces a new modeling problem {proper­ ties and behaviors) which requires new control methodologies

{e.g. procedural, goal-directed, aesthetic algorithms). For many this will mean a radical departure from their usual method of working. It is anticipated that motion and continuity can become as plastic an art element for the artist as traditional art elements (e.g. line, shape, value, etc.) have been. >From this perspective, the study suggests a model that is an exten­ sion of our knowledge base, rather than a departure from it.

An argument could be leveled that the elements of this model are naturally evolving regardless of a formal proposal for such a model. Historically such a logical scenario has not occurred, it has generally taken an individual or group initia­ tive to break the prejudicial barriers of the time. Quite often information arises which doesn't fit the prevailing para­ digm and is invariably labeled as "bogus" or trivial until a 251 new and more encompassing general theory emerges [4]. King [5] has proposed that we are currently in the midst of moving from a paradigm based on physical art forms toward another - com­ posed of events and utilizing computers - which has yet to materialize.

Another concern that should be addressed is that the model has not been sufficiently analyzed in terms of the implementa­ tion of its various components. The purpose of this study has not been to analyze, but to establish the extent to which this simulation model could be used and propose a structure for its function. This has not been done previously, and without such a model artists would continue to be limited in their knowledge as to how the computer can profoundly change the course of ani­ mation and possibly the form of art itself.

It is apparent to the researcher, after reviewing the information collected for this study, that this model will eventually emerge in various forms. No one program can assimi­ late all of the "bells and whistles" (options) needed to accom­ modate the idiosyncrasies of the creative and inquisitive mind of the artist. At this time the proposed model can be quali­ fied as following in the footsteps of significant art forms in which the work is distinguished more by its scope and energies than by its originality [6].

Research into the use of simulation as an animation tool has in many ways just begun, and the subsequent recognition of 252 the computer as a modern fine art medium is just beginning to take place. The affects that these efforts in simulation and animation have will be a natural area for further research.

In summary, the implication of this study is that the impact of simulation techniques as animation will be in the release of the animator's energy from the physical act of draw­ ing, and re-focused into designing and directing. As artists aspire to create new original works of art, computer simulation will help them break new ground previously barred by con­ straints of time and complexity. It is the desire of this researcher that this model will provide artists desiring to explore the medium of computer simulation with a heuristic guide. This model has the potential of providing insights into creative possibilities that have yet to be conceived. 253

NOTES TO CHAPTER VII

1. J. Burnham (1978). System Esthetics. In R. Kostelanetz (Ed.), "Esthetics Contemporary" New York: Promethus Books, p. 162.

2. Frank Dietrich (1986). In R. Lucus (Ed.), Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study, Unpublished master's thesis, Ohio State University, Columbus, OH. p. 15.

3. J. Chadwick and M. Girard (1988). Private Conversation.

4. T. S. Kuhn (1970). The Structure of Scientific Revolutions. Chicago: University of Chicago Press, p. 12.

5. Robin King (1988). Private Conversation.

6. Frank Thomas and Olie Johnston (1984). Disney Animation: The illusion of Life. New York: Abbeville Press, p. 173. BIBLIOGRAPHY

Abas, S. J. (1983). "On Didactic Implications in the Teaching of Applied Mathematics from a Coupling of Interactive Graphics and Waves." Eurographics '83 P. J. W. ten Hagen (editor), Elsevier Science Publishers B.V. (North Holland), 225-233.

Adelsberger, H H., Pooch, U. W., Shannon, R. E., Williams, G. N. (1986) "Rule based object oriented simulation systems", SIGPLAN NOTICES, 21, '‘(6) , 107-112.

Amkraut, S. & Girard, M. (1986) Pencil test of Eurthemy shown at the SIGGRAPH '86 Electronic Theatre.

Andrew, J. D. (1976). The Major Film Theories. New York: Oxford University Press.

Aono, M. & Kunii, T. L., (1984). "Botanical Tree Image Generation." IEEE Computer Graphics and Applications, 4, (5), 10-34.

Armstrong, W. W. and Green, M. (1985). "The Dynamics of Articulated Rigid Bodies for Purposes of Animation." Proceedings Graphics Interface '85, Montreal, 407-416.

Arnheim, R. (1954). Art and Visual Perception: A Psychology of the Creative Eye. Berkeley: University of California Press.

Arnheim, R. (1957). Film as Art. Berkeley: University of California Press.

Badler, N. I., O'Rourke, J., & Kaufmann, B. (1980). "Special Problems in Human Movement Simulation." Siggraph f 80 Conference Proceedings, 14, (3), 189-197.

Badler, N. I., & Platt, S. M. (1981). "Animating Facial Expressions." ACM Computer Graphics, 15, (3), 245-251.

254 255

Badler, N. i.r & j. u. Korein (1982). "Techniques for Generating the Goal-Directed Motion of Articulated Structures." IEEE Computer Graphics and Applications, 2, (11), 71-81.

Badler, N. I. (1982). "Human Body Models and Animation." IEEE Computer Graphics and Applications, 2, (11), 6-7.

Badler, N. I. (1984). "What is Required for Effective Human Figure Animation?" Proceedings Graphics Interface ' 84, Ottawa, Ontario, 119-120.

Badler, N. I., Mannoochehri, K. H. & Baraff, D. (1986). "Multi-Dimensional Input Techniques and Articulated Figure Positioning by Multiple Constraints." Proceedings of Computer Graphics Workshop, University of North Carolina Chapel Hill, NC., 1-18.

Baerle, S. V. (1986). "Character Animation: Combining Computer Graphics and Tradition Animation," Advanced Computer Animation, Siggraph ' 86 Tutorial Notes.

Barr, A. H. (1987). "Introduction to Physically-Based Modeling," Tutorial Notes No. 17, SIGGRAPH ' 87, Topic in Physically-Based Modeling, 1-5.

Barr, A. H. (1988). "Glossary," Physically-Based Modeling, ACM SIGGRAPH '88 Tutorial Notes. E1-E7. Barzel, R. & Barr, A. H. (1987). "Modeling With Dynamic Constraints," ACM SIGGRAPH '88 Conference Proceedings, Computer Graphics, 22, (4), 179-188.

Barzel, R. & Barr, A. H. (1988). "A Modeling System Based on Dynamic Constraints," Tutorial Notes No. 3/7, SIGGRAPH '87, Topic in Physically-Based Modeling.

Barzun, J. and Graff, H. F. (1977). The Modern Researcher. New York: Harcourt, Brace Jovanovich.

Bazin, A. (1974). What is Cinema? Berkeley: University of California Press.

Benthall, J. (1972). Science and Technology in Art Today. London: Thames and Hudson.

Berkum, A. V. and Blekkenhorst, T. (1986). Science * Art. Utracht, Holland: Fentener van Vlissingren.

Berlyne, D. E. (1965). Structure and Direction in Thinking. Wiley, N. Y., 30-44. 256

Bettinghaus, E. P. (1968). Communication Theory and the Use of the New Media in Educational Proceedings Series, #1, August 1968 Theory for the New Media in Education, Educational Proceedings Series, Number 1, East Lansing: Michigan State University, 1968. 89-121.

Biederman, C. (1948). Art as the Evolution of Visual Knowledge. Redwing, Minnesota. Charles Biedermann.

Bofa, G. (1925). "Du Dessin anime," Les Cahiers du Mois, 16, 51-57.

Brockman, R. A. (1982). "Numerical Solution of Large Deformation Problems involving Sufrace Contact and Impact," Mechanical Engineers, 123-136.

Broudy, H. S. (1966). "Aesthetic Education in a Technological Society: The Other Excuses for Art," The Journal of Aesthetic Education, 1, (1), 28-37.

Burnham, J. (1978). "System Esthetics," In, R. Kostelanetz (Ed.), Esthetics Contemporary (pp. 160-171). New York: Promethus Books.

Burtnyk, N. & Wein, M. (1976). "Interactive skeleton techniques for Enhancing Motion Dynamics in Key Frame Animation," Communications of the ACM, 19, (10), 564-569.

Caldwell, C. (1988). "Interaction Between Computer Science and Art in Computer Graphics," National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, 653-663.

Calvert, T. W., Chapman, J. & Patla, A. (1980). "The Integration of Subjective and Objective Data in the Animation of Human Movement," Proceedings ACM SIGGRAPH f80. Computer Graphics, 14, 198-203.

Calvert, T. W., & Chapman, J. (1982). "Aspects of the Kinematic Simulation of Human Movement," IEEE Computer Graphics and Applications, 2, (11), 41-50.

Catmull, E. (1978) "The Problems of Computer-assisted Animation," Computer Graphics:ACM-SIGGRAPH '78 Conference Proceedings, ±, 348-353.

Chadwick, J., & Parent, R. (1988) "Controlling the Integration of Computational Models for Character Animation," National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, 423-432.

Chadwick, J. (1988). Getting the Aesthetics of the Computational Process Out of the Process and into the 257

Product, Unpublished manuscript.

Chadwick, J. & Parent, R. (1988). "Critter Construction: Developing Characters for Computer Animation," PIXIM Conference Proceedings, 1_, (1), 1-14,

Chipp H. B. (1968). Theories of Modern Art, A Source Book by Artists and Critics. Berkeley, CA., University of California Press.

Chuang, R. & Entis, G. (1983). "3-D Shaded Computer Animation-Step-by-Step," IEEE Computer Graphics and Applications, 3^ 18-25.

Colella, A. M., O'Sullivan, M. J. £ Carlino, D. J. (1974). Systems Simulation Washington, D.C.: Lexington Books.

Collier, G. (1972). Art and the Creative Consciousness. Englewood Cliffs, N.J.: Prentice-Hall, Inc.

Collins, G. C. (1978). The Sex-Appropriateness of Art Activity for the Female. Unpublished doctoral dissertation, The Ohio State University, Columbus, Ohio.

Crafton, D. (1984). Before Mickey, The animated Film, 1898- 1926. Cambridge, Mass, The MIT Press.

Crow, F. C. (1982). "A More Flexible Image Generation Environment," ACM Computer Graphics, 16, (3), 9-16.

Crutchfield, J. P. £ Gomez, J. (1986). Chaos, Dynamics, and Natural Phenomena. Unpublished manuscript.

Crutchfield, J. P., Farmer, J. D., Packard, N. H. & Shaw, R. S. (1986). "Chaos," Scientific American, 38, (12), 46-57.

Csuri, C. (1974). "Computer Graphics and Art," Proceedings of the IEEE, 62, (4),*503-515.

Csuri, C. (1978). "3-D Computer Animation," In M. Rubinoff and M. C. Yovits (Eds.), Advances in Computers, Volume 16, (pp. 1-55). New York: Academic Press.

Csuri, C., Carlson, W., Hackathorn, R., Parent, R. £ Howard, M. (1979). "Towards an Interactive High Visual Complexity Animation System," Proceedings SIGGRAPH 79, 13, (2), 289- 299.

Csuri, C. A., Blinn, J., Gomez, J., Max, N. £ Reeves, W. (1983). "The Simulation of Natural Phenomena," ACM Computer Graphics, 17, (3), 137-138. 258

Defanti, T. A. (1973). The Graphics Symbiosis System — An INteractive Minicomputer Graphics Language Designed for Habitability and Extensibility. (Unpublished doctoral dissertation, The Ohio State University).

Demos, G. Brown, M. D. & Weinberg, R. A. (1984) "Digital Scene Simulation: The Synergy of Computer Technology and Human Creativity." Proceedings of the IEEE, 72, (1), 22-31.

Diamond, S. R. (1970). Fundamental Concepts of Modern Physics, New York: Amsco Publications.

Dietrich, F. (1987), "The Computer: A Tool for Thought- Experiments." Leonardo, 20, (1), 1-11.

Ecker, D. W. (1962). "Teaching Machines and Aesthetic Values." Philosophy of Education Society Symposium of Teaching Machines, National Society of College Teachers of Education, Chicago, Feb. 15, 1962. 1-17.

Ekman, P. & Friesen, W. (1975). Unmasking the Human Face. Englewood Cliffs, NJ: Prentice Hall Inc.

Efland, A. D. (1979) . "Conceptions of Teaching in Art Education." Art Education, 79, (32), 21-26.

Efron, A. (1966). "Technology and Future of Art." The Massachusetts Review, Autumn, 1_, (4), 1-21.

Feldman, E. B. (1970). In Becoming Human through Art: Aesthetic Experience in the School. (pp. 348-377). Englewood Cliffs, NJ: Prentice-Hall, Inc.

Fetter, W. A. (1982). "A Progression of Human Figures Simulated by Computer Graphics." IEEE Computer Graphics and Applications, 2, (11), 9-13.

Forest, L., Magnenat-Thalman, N., Rambaud, D. & Thalmann, D. (1986). Proceedings of Graphics Interface '86, Vancouver B.C., Vancouver B.C.: Canadian Information Processing Society.

Franke, H. W. (1987). "The Expanding Medium: The Future of Computer Art." Leonardo, 20, (1), 1-4.

Frautschi, S. C., Olenick, P. P., Apostol, T. M., & Goodstein, D. L. (1986). The Mechanical Universe, Mechanics and Heat. Cambridge, England: Cambridge University Press.

Gablik, Suzi (1970). Magritte. Greenwich, Connecticut: New York Graphics Society Ltd. 259

Garbutt, R., McPheeters, C., & Wyvill, B. (1986). "University of Clagary 3-D Computer Animation System." SMPTE Journal, 5, 629-636.

Ginsberg, C. M. & Maxwell, D. (1983). "Graphical Marionette." ACM-SIGGRAPH/SIGART Interdisciplinary Workshop, Motion: Representation and Perception, 172-179.

Gips, J. and Stiny, G. (1979). An investigation of algorithmic aesthetics. In F. Malina (Ed.), Visual art, mathematics and computers: selections from the journal Leonardo (93- 100), New York: Pergamon Press

Girard, M. and Maciejewski, A. A. (1985). "Computational Modeling for the Computer Animation of Legged Figures." SIGGRAPH f85 Proceedings, Computer Graphics, 19, (3), 263-270.

Girard, M. (1986). "Computer-Animated Legged Animal Motion." ACM Workshop on Interactive 3D Graphics, Chapel Hill, N C .

Girard, M. (1987). "Interactive Design of 3D Computer-Animated Legged Animal Motion." IEEE Computer Graphics and Applications, 1_, (6), 39-51.

Girard, M. (1988). Private conversation. May 29, 1988.

Girard, M. (1988) . An Overview of Motion Planning Problems in Robotics and Artificial Intelligence, Unpublished Manuscript.

Gomez, J. E. (1984). "Twixt: A 3-D Animation System." In K. Bo and H. A. Tucker (Eds.), Proceedings Eurographics ' 84, (pp. 279-288) North Holland: Elsevier Science Publishers B.V.

Gomez, J., MacDougal, P. & Zeltzer D. (1984). "A Tool Set for 3-D Computer Animation.' ACM Siggraph Proceedings 1984, 4, 125-137.

Gomez, J. E. (1985). "Computer Display of Time Variant Functions." (Doctoral disseratation, The Ohio State University, 1985).

Gomez, J. (1987). "Comments on Event Driven Computer Animation." Notes for SIGGRAPH 87 Tutorial Computer Animation: 3D Motion Specification and Control, 77-90

Green, M. (In press). "Animating Complex Objects." ACM Transactions on Graphics.

Greene, P. H. (1972). "Problems of Organization of Motor Systems." Progress in Theoretical Biology, 2, pp. 303-308. 260

Guenter, B. (1988). A System for Simulating Human Facial Expression. Unpublished manuscript.

Hackathorn, R. (1977). "ANIMA II: A 3-D Color Animation System." Proceedings SIGGRAPH '77, Computer Graphics, 11, (2), 54-64.

Hanrahan, R., Parent, R., Marshall, B. and Howard, M. (1981) "An Interactive Microcomputer Based 3-d Animation System." Proc. Conference Canadian Society for Man-Machine Interaction. pp. 181-191.

Haeberli, P. E. (1988). "ConMan: A Visual Programming Language for Interactive Graphics." SIGGRAPH '88 Proceedings, Computer Graphics, 22, (4), 103-110.

Hahn, J. K. (1988). "Introduction to Issues in Motion Control." SIGGRAPH '88 Tutorial Notes for 'Introduction to Computer Animation'

Hahn, J. K. (1988). "Realistic Animation of Rigid Bodies." ACM SIGGRAPH '88 Conference Proceedings, Computer Graphics, 22, (3), 1-8.

Halas, J. & Manvell, R. (1976). The Technigue of Film Animation. New York: Hastings House Publishers.

Hanrahan, P. & Sturman, D. (1985). "Interactive Animation of Parametric Models." The Visual Computer, 1, 260-266.

Haumann, D. R. (1987). The Animation of Flexible Objects. Manuscript submitted for publication.

Haumann, D. R. (1987). "Modeling the Physical Behavior of Flexible Objects." SIGGRAPH '87 Tutorial Notes N o . 17, Topics in Physically-Based Modeling.

Haumann, D. R. (1988) "Animation using Behavioral Simulation." National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III Anaheim, CA, 672-679.

Henderson, L. D. (1983). The Fourth Dimension and Non- Euclidean Geometry Princeton, NJ: Princeton University Press.

Hill, R. H. (1983, August). "Conceiving Art in the 80's." Paper delivered at Digicon 83, International Conference on the Digital Arts Vancouver, B. C., August 1983, 1-7

Hofstadter, D. R . (1979). Godel, Escher, Bach: New York: Vintage Books. 261

Issacs, P. M. & Cohen, M. F. (1987). "Controlling Dynamic Simulations with Kinematic Constraints, Behavior Functions and Inverse Dynamics." ACM SIGGRAPH f 87 Conference Proceeding's, Computer Graphics, 21 (3), 215-224.

Kaelin, E. K. (1963). Aesthetics and the Teaching of Art Unpublished Manuscript, Madison, WI: University of Wisconsin.

Kawaguchi, Y. (1982). "A Morphological Study of the Form of Nature." ACM Computer Graphics, 16, (3), 223-230.

Kern, E. J. (1969). Descriptive Aesthetics, Three Dimensions of Experience: A Curiculum Model for Art Education (Doctoral dissertation, The Ohio State University, 1978). 1-15.

King, R. G. (1981). "A General Systems Model of the Creative Process, Proceedings of the Twenty-Fifth Annual North American Meeting of the Society for General Systems Research with the American Association for the Advancement of Science, January 6.-9.. 23-29.

King, R. (1988) . "Aesthetic Experiences, Personal Constructs and the Evolution of Computers in Art" National Computer Graphics Association Conference Proceedings, Technical Sessions, Volume III, 437-445.

Kochanek, D. & Bartels, R. (1984). "Interpolating Splines with Local Tension, Continuity and Bias Control." SIGGRAPH 84, Computer Graphics, 18, (3), 33-41.

Koestler, A. (1949) . Insight and Outlook. New York: Macmillan & Company.

Korein, J. & Badler, N. (1983). "Temporal Anti-Aliasing in Computer Generated Animation." ACM Computer Graphics, 17, (3), 377-388.

Krueger, M. (1983). Artificial Reality Reading, Massachusetts: Addison Wesley.

Kuhn, T. S. (1970), The Structure of Scientific Revolutions. Chicago, University of Chicago Press.

Lasseter, J. (1987). "Principles of Traditional Animation Applied to 3D Computer Animation." ACM SIGGRAPH r87 Conference Proceedings, 21, (4), 35-44.

Linehan, T. E. (1976). "An Investigation of Criteria for Evaluating Computer Art." Computer Graphics and Art, 4, (1), 6-9. 262

Linehan, T. E. (1981). A Computer-Mediated Model for Visual Preference Research and Implications for Instruction in Art Criticism. (Doctoral dissertation, Ohio State University, 1981).

Luenberger, D. G. (1979). Introduction to Dynamic Systems. New York: John Wiley & Sons.

Lucas, R. E. (1986). Evolving Aesthetic Criteria for Computer Generated Art: A Delphi Study. Unpublished master's thesis, Ohio State University, Columbus, OH.

Lundin, D. (1982). "3D Modeling, A Personal Orthodoxy." ACM Siqgraph 82, Course Notes, Three-Dimensional Computer Animation.

Lundin, R. V. (1986). "Motion Simulation." Advanced Computer Animation, Siqgraph '86 Tutorial Notes.

Maccia, E. S. (1962). "Ways of Inquiring." Bureau of Educational Research and Service; Occasional Papers. Columbus: The Ohio State University. 1-13.

Magnenat-Thalmann, N. & Thalmann, D. (1985) . Computer Animation : Theory and Practice. Tokyo: Springer-Verlag,

Magnenat-Thalmann, N. & Thalmann, D. (1985) . "Three-Dimensional Computer Animation: More an Evolution Than a Motion Problem." IEEE Computer Graphics and Applications, 5, (11), 47-57.

Maltin, L. (1980). Of mice and magic, A History of American Animated Cartoons. New York: Signet Books.

Marcel, B. (1928) "Felix le chat ou la poesie creatrice." Le Rouge et le Noir (7), 21-26.

Martinez, B. & Block, J. (1988) . Visual Forces. Englewood Cliffs, NJ: Prentice Hall Inc.

McGovern, T. (1987). "The use of Live-Action Footage as a Tool for the Animator". Tutorial Notes SIGGRAPH '87, 3^-D Character Animation by Computer, (5), 1-26.

Miller, G. S. P. (1988). "The Motion Dynamics of Snakes and Worms." SIGGRAPH '88 Proceedings, Computer Graphics, 22, (4), 169-173.

Mondrian, P. (1945) Plastic Art and Pure Plastic Art. New York: Wittenborn, 1945.

Nielan, P. E. & Kane, T. R. (1985). "Symbolic Generation of Efficient Simulation/Control Routines for Multibody 263

Systems." In G. Bianch and W. Schiehein (Eds.), Dynamics of Multibody Systems, IUTAM Symposium, Berlin: Springer.

O'Donnel, T. J. & Olsen, A. J. (1981). GRAMPS - A Graphics "Language Interpreter for Real-Time, Interactive, Three- Dimesional Picture Editing and Animation." Proceedings ACM SIGGRAPH 81. Computer Graphics, (15), 133-142.

Ostby, E. (1986). "Describing Free-Form 3D Surfaces for Animation." Proceedings of Computer Graphics Workshop, University of North Carolina Chapel Hill, NC., 1-8.

Parent, R. E. (1977). A System for Generating Three- Dimensional Data for Computer Graphics. Unpublished doctoral dissertation, Ohio State University, Columbus, OH.

Parke, F. I. (1982). "Parameterized Models for Facial Animation." IEEE Computer Graphics and Applications, 2, (9), 61-68.

Parke, F. I. (1984). "Parameterized Models for Facial Animation." Tutorial Notes Siggraph '84, Intro to Computer Animation, 296-310.

Parke, F. I. (1986). "Interactive Tools to Support Animation Tasks." Proceedings from ACM Computer Graphics Workshop on Interactive 3D Graphics, University of North Carolina, Chapel Hill, NC, 1-3.

Partridge, B. L. (1982). "The Structure and Function of Fish Schools." Scientific American, j3, 114-123.

Pearce, A. Wyvill, B. Wyvill, G. fi Hill, D. (1986). "Speech and Expression: A Computer Solution to Face Animation." Proceedings Graphics Interface '86, 136-140.

Pique, M. E. (1986). "Manipulations within Rotations." Proceedings of Computer Graphics Workshop, University of North Carolina. Chapel Hill, NC., 1-10.

Platt, J. C. (1988). "Constraint Methods for Flexible Models." ACM SIGGRAPH '88 Proceedings, Computer Graphics, 22, (4), 279-285.

Pollock, M. (Ed.). (1985). Common Denominators in Art and Science Edinburgh: Aberden University Press.

Press, W. H., Flannery, B. P., Teukolsky, S. A. & Vetterling, W. T. (1988). Numerical Recipes in C, The Art of Scientific Computing. Cambridge, England: Cambridge University Press. 264

Raibert, M. H. (1988) . Biped Gymnastics submitted for publication.

Reeves, W. T. (1981). "Inbetweening for Computer Animation Utilizing Moving Point Constraints." ACM SIGGRAPH '81 Conference Proceedings, Computer Graphics, 15, (3), 263- 270.

Reeves, W. T. (1983). "Particle Systems - A technique for Modeling a Class of Fuzzy Objects." ACM Computer Graphics, 17, (3), 359-375.

Reynolds, C. W. (1982). "Computer Animation with Scripts and Actors." ACM Computer Graphics, 16, (3), 289-296.

Reynolds, C. (1986). "Advanced Computer Animation." ACM Siggraph '86, Tutorial Notes, Advanced Computer Animation

Reynolds, C. W. (1987). "Flocks, Herds, and Schools: A Distributed Behavioral Model." ACM SIGGRAPH '87 Conference Proceedings, 21, (4), 25-34.

Ribble, E. (1982). Design of a Skeletal Motion Processor. Unpublished manuscript.

Richardson, J. A. (1971). Modern Art and Scientific Thought. Urbana, IL: University of Illinois Press.

Roberts, N., Anderson, D., Deal, R., Garet, M. & Shaffer, W. (1983). Introduction to Computer Simulation, A System Dynamics Modeling Approach. New York: Addison-Wesley.

Roger, E. & Siskel, G. (1986). "Animation Special." At the Movies, July 6, 1986.

Sachter, J. (1984). Three-Dimensional Concepts of Computer Graphics for Artists. Unpublished master's thesis, Ohio State University, Columbus, OH.

Scheler, M. (1954). The Nature of Sympathy. London: Routledge & Kegan Paul Publishers, 253.

Schillinger, J. (1948). The Mathematical Basis of the Arts. New York: Philosophical Library.

Schlag, J. F. (1985). "Eliminating the Dichotomy Between Scripting and Interaction." Proceedings Graphics Interface '85, Vancover, B.C.

Schweppe, M. K. (1984). Timing and Movement in Three- Dimensional Computer Animation. Unpublished master's thesis, Ohio State University, Columbus, OH. 265

Sederberg, T. W. and Parry, S. R. (1986). "Free-Form Deformation of Solid Geometric Models." SIGGRAPH '86 Conference Proceedings, 20, (4), 151-159.

Smith, A. E. & Koury, P. A. (1952). Two Reels and a Crank. Garden City, NY: DoubleDay.

Stefik, M., Bobrow, D., Mittal, S. & Conway, L. (1983). "Knowledge Programming in Loops: Report on an Experimental Course." AI Magazine, _4, 3-13.

Stern, G. (1983). "Bbop - A Program for 3-Dimensional Animation." Proceedings Nicographic '83, Japan, 236-295.

Stolnitz, J. (1960). Aesthetics and Philosophy of Art Criticism. Boston: Houghton Mifflin Company, 102-107.

Sturman, D. (1986). "A Discussion on the Development of Motion Control Systems." Advanced Computer Animation, Siggraph /86 Tutorial Notes. 3-15.

Sturman, D. (1987). "Interactive Keyframe Animation of 3-D Articulated Models." SIGGRAPH 1987 Tutorial No. 10, Computer Animation: 3-D Motion Specification and Contol, 17-26.

Sturman, D. (1987). "A Discussion on the Development of Motion Control Systems." SIGGRAPH 1987 Tutorial N o . 10, Computer Animation: 3^-D Motion Specification and Contol. 3-15.

James Johnson Sweeney (1946). Eleven Europeans in America, Bulletin of the Museum of Modern Art, 13. (4-5), 20-22.

Terzopoulos, D., Platt, J., Barr, A. & Fleischer, K. (1987). "Elastically Deformable Models." ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (4), 205- 212.

Terzopoulos, D., Fleischer, K. (1988). "Modeling Inelastic Deformation: Viscoelasticity, Plasticity, Fracture." ACM SIGGRAPH r88 Conference Proceedings, Computer Graphics, 22, (4), 269-278.

Thomas, F. & Johnston, 0. (1984). Disney Animation: The illusion of Life. New York: Abbeville Press. (8), 90-147.

Thomas, F. (1984). "Can Classic Disney Animation be Duplicated on the Computer." Computer Pictures, (Jul./Aug.), 20-26.

Thomas, F. (1987). "The Future of Character Animation by Computer." Tutorial Notes SIGGRAPH ' 8T_, 3-D Character Animation by Computer, (5), 1-16. 266

Van de Grind, W. A. (1986) . ’’Vision and the Graphical Simulation of Spatial Structure." Proceedings from ACM Computer Graphics Workshop on Interactive 3D Graphics, University of North Carolina, Chapel Hill, NC, 1-4.

Tomovic, P.. and McGhee, R. B. (1966) . "A Finite State Approach to the Synthesis of Bioengineering Control Systems." IEEE Transactions on Human Factors in Electronics, HFE-7, 2, (6), pp. 65-69.

Vitz, P. C. & Glimcher, A. B. (1984). Modern Art and Modern Science, The Parallel Analysis of Vision. New York: Praeger Publishers.

Waters, K. (1987) . "A Muscle Model for Animating Three- Dimensional Facial Expression." ACM SIGGRAPH '87 Conference Proceedings, 21, (4), 17-24.

Weber, L., Smoliar, S. W., S Badler, N. I. (1978). "An Architecture for the Simulation of Human Movement." Proceedings ACM Ann Arbor Conference, 737-745.

Wedge, C. (1987). ACM SIGGRAPH '87 Tutorial Notes: 3-D Character Animation by Computer. (5), 1-25.

Whitmaker, H. & Halas, J. (1981). Timing for Animation. London: Focal Press Limited.

Wilhelms, J. P. (1985). Graphical Simulation of the Motion of Articulated Bodies such as Humans and Robots, with Special Emphasis on the use of Dynamic Analysis. Unpublished doctoral dissertation, University of California, Berkeley.

Williams, L. (1982). "BBOP." Tutorial Notes, Three-Dimensional Computer Animation, ACM SIGGRAPH '82.

Willmert, K. D. (1982). Visualizing Human Body Motion Simulations. IEEE Computer Graphics and Applications, 2, (11), 71-81.

Witkin, A., Fleischer, K. & Barr, A. (1987). "Energy Constraints on Parameterized Models." ACM SIGGRAPH '87 Conference Proceedings, Computer Graphics, 21, (4), 225- 229.

Witkin, A. and Kass, M. (1988). "Spacetime Constraints." ACM SIGGRAPH '88 Conference Proceedings, Computer Graphics, 22, (4), 159-168.

Zeigler, B. P., Elzas, M. S., Klir, G. J. & Oren, T. I. (1979). Methodology in Systems Modeling and Simulation. Amsterdam: North-Holland Publishing. 267

Zeltzer, D. (1982). "Motor Control Techniques for Figure Animation." IEEE Computer Graphics and Applications, 2, (11), 32-42.

Zeltzer, D. (1983). "Knowledge-based Animation." ACM- SIGGRAPH/SIGART Interdisciplinary Workshop, Motion: Representation and Perception (pp. 172-179). Toronto, Quebec, Canada.

Zeltzer, D. (1984). Representation and Control of Three Dimensional Computer Animated Figures Unpublished doctoral dissertation, Ohio State University, Columbus, OH.

Zeltzer, D. (1984). "Issues in 3-D Computer Character Animation." Introduction to Computer Animation, SIGGRAPH '84 Course Notes, 8., (1), 261-275.

Zeltzer, D. (1985). "Towards an integrated view of 3-D computer animation." The Visual Computer, 1, (4), 249-259.