Lect. 1&2

Lectures 1&2

FIT4012: Artificial Life, Procedural Modelling & Animation
General Course References

Watt, Watt,
"Advanced Animation and Rendering Techniques, Theory and Practice"
Chapters 15, 16, 18 ACM Press, Addison-Wesley, 1992

For more pointers and as an excellent reference

Levy, S.,
"Artificial Life - The Quest For A New Creation"
Jonathan Cape 1992

A readable introduction to artificial life

Langton, Christopher G. (ed),
"Artificial Life"
Santa Fe Institute Studies in the Science of Complexity, Vol VI, Addison-Wesley 1989

Varela, F.J., Bourgine P. (eds),
"Toward a Practice of Autonomous Systems"
Proc's of the First European Conference on Artificial Life, MIT Press 1992

Some pioneering conference proceedings in artificial life

Adaptive systems
Animats
Artificial Life
Autonomy

Complex systems

Genetic Algorithm

Self-Organization

Synthetic Biology, etc.

A sample set of google search terms to try

Book reviews and a brief description of some of the above books are online.

Overview

This course is largely (but not exclusively) concerned with the modelling of natural phenomena.

Here are some characteristics of animate objects which will be of interest in this course...

Morphology (shape / form)
endo/exo-skeleton, musculature, membrane
Texture and surface characteristics
hair, fur, scales, feathers, flesh, bark, thorns, spines, veins, spots etc.
Nervous / sensory / perceptory systems
visual, aural, olfactory, touch
Growth / development
plants, animals
Locomotion
swim, walk, fly, roll, slither
Interactions with animate and inanimate objects
flee, pursue, communicate, consume, envelop, push, collide, mate, destroy

Here are some characteristics of in-animate objects which will be of interest in this course...

Shape
jagged, twisted, worn, rounded, rocky, amorphous, porous
Texture and surface characteristics
translucent, glistening, misty, glowing, stained, veined
Interactions with other objects
bounce, flex, break, adhere, stretch, cover, vibrate, absorb, transmit, reflect, refract, sink, repel, crumble, shatter, fold, score

Why do natural phenomena in particular require special modelling techniques?

Humans are very good at recognizing faulty attempts at visualizing the real world
Film industry demands realism
Some degree of accuracy is required for scientific purposes
Analysis of: bush fire behaviour, weather patterns, hurricane paths, tidal movement, success of pesticides, crop development in drought or areas of high salinity
Behaviour of 'simple' inanimate objects is in fact extremely complex
Dispersion of smoke in a breeze, crumpling of a piece of paper, water dripping from a tap, coin descending through a fluid
Animate objects are many orders of magnitude more complex than the most sophisticated artificial and inanimate objects
Interaction of cells in a human body to produce 'conscious thought'

Special techniques for modelling natural phenomena may exploit the computer's speed in order to handle complexity.

Manual modelling

Visual models of natural structures may be represented using any of the computer graphics modelling techniques.

Models may be produced by hand by manipulating:

polygons
control points

graphics primitives
splines... (the list goes on)

...via a GUI, keyboard and mouse, or from data obtained by measurement (using scanners and range-finders etc.) from the real world.

Here are some tools for interactive modelling using Specular Infini-D (mid 90's).
(3D modelling and animation software based on primitives)

Translating, rotating, scaling a primitive
Ray tracing the current view
Creating a new primitive of a specific type
Joining and breaking hierarchies of primitives
Adding lights and cameras to a scene
Moving and zooming the current camera view

...and some tools for interactive modelling using the old Hash Inc. Animation Master.
(3D modelling and animation software based on splines) (mid 90's)

Adding new splines
Adding control points to existing splines
Breaking splines
Peaking / smoothing splines at control points
Selecting control points to manipulate
Scaling, rotating a spline
Extruding a spline
Produce a surface of revolution from a spline
Apply texture / surface property to a patch
Flip or view normals of a patch
Create splines around text outlines

Interactive methods are tedious for the creation of complex structures. Even current software offers little improvement compared to these 10+ year old user interfaces.

Imagine trying to build a detailed model of a tree trunk and branches by manually positioning and shaping thousands of differently sized branch segments... and then having to add the leaves!

Manual animation

An animation of any kind may be produced:

Create a series of still images (taken at regular intervals across some time span) of objects whose characteristics change
Display these images at a suitable rate to allow persistence of vision to merge the stills into a continuous stream

Hand-drawn animation frames may be produced:

Draw an object at a set of key-frames
Draw the in-betweens

Computer animation frames may be produced:

Position a virtual object and set its attributes at key-frames (possibly specifying motion paths manually or by using data obtained from the real world)
Allow the computer to interpolate in-betweens
Render the series of images

Timeline from Inifini-D animation software

Manual computer animation is tedious when models are:

complex and have many degrees of freedom	human figure
required to interact with one another	tap-dancing human figure tossing a coin
expected to move in accordance with the laws of physics	snow-boarding, tap-dancing human figure rolling a coin down a ravine
of deformable or fluid bodies	snow-boarding, tap-dancing human figure rolling a coin down a ravine into a pond and falling in after it (Kids, don't try this trick at home unless supervised by an adult.)

In all but the most trivial of cases, manual computer animation is a difficult, time-consuming process.

Procedural modelling

Model geometry and surface specified algorithmically or mathematically
The inclusion of time-varying parameters facilitates procedural animation

void setGeometryAndColour (double time)
{
    for (long i=-2; i<20; i++)
    {
      for (long j=-2; j<20; j++)
      {
        controlPt[i][j]-> x = i;
        controlPt[i][j]-> y = sin(i*time);
        controlPt[i][j]-> z = j;
			
        cel[i][j]-> red = (1+ sin(i*time)) / 0.5;
        cel[i][j]-> green = 0.0;
        cel[i][j]-> blue = 0.0;
      }
    }
}

Image from Macintosh Graphing Calculator

Sample code to create a plane containing sin waves in colour and geometry


Image created using the interactive "Artmatic", procedural texture generator by U&I Software	Image showing structure of the function used to create the Artmatic image (left)

Procedural Animation

Computer code is used to determine / control how the model changes over time
(Just add a parameter 't' to your procedural models for a simple example of this)

General methods include but are not limited to:

Script-based systems

Work with actors (objects in an animation) which

Have variables to describe their current state

Have instructions to be carried out at each frame

May pass messages about their state to other actors

Special-purpose animation scripting languages have been developed which offer high-level descriptions of what an actor is to do.

E.g. Define a "car" actor and four "wheel" actors. Each frame, the car may be moved some distance. The car actor passes the distance it moves on to the wheel actors, that respond to these messages by rotating the correct amount so that they appear to have rolled along a surface without slipping.

Physically-based modelling

Specify a set of bodies and their physical properties
Specify the manner in which forces will be applied to bodies

Model some aspect of physics in order to simulate the behaviour of bodies acting under forces as applied by/during:
collision detection and response

fluid dynamics

rigid and articulated body dynamics

deformation
...and run the simulation for some time-span, rendering the objects at regular intervals.
Behavioural animation

Specify a set of agents and their properties

tendencies to mate,

eat,

sleep or

change colour

visual acuity... etc.

Specify a set of behavioural rules for each agent

eg. An agent simulation for an undergraduate student:
If (Need (food))
    Move-Towards (Union Building)

Else If ((Weather == SUNNY) AND (Visible (mate))
{
    Set-FacialExpression (cheerful)
    Move-Towards (mate)
}

Else
    Sleep(5)
and run the simulation for some time-span, rendering the objects at regular intervals

Stochastic animation

Specify probabilities of certain events occurring
eg. tendencies for spontaneous combustion / emission of particles or shoots of grass

Seed / constrain the simulation
eg. specify boundaries within which fireworks or plants are to emerge

and run the simulation for some time-span, rendering the objects at regular intervals

General principles of procedural modelling and animation

None of the above methods are set in concrete
All of the above techniques may be combined
Improvise

Dorin's Razor for procedural modelling

Decide on your purpose...
Science Model exactly the aspects of the real system...	?	Art Model (or ignore) any aspects of any system...
...necessary to achieve the desired result

Failure to observe this simple rule is guaranteed to result in much wasted effort and heartache.

An Introduction to Artificial Life

Some themes (not 'truths') of a-life:

The study of "life as it could be"
Life as a property of form, independent of matter
Synthesis and bottom-up approaches
Complex, global behaviour, emergent from simple, local rules

Some opinions (not 'truths') on a-life:

Langton et. al. 1991, Alife (USA) II...

"Artificial Life is a field of study devoted to understanding life by attempting to abstract the fundamental dynamical principles underlying biological phenomena, and recreating these dynamics in other media - such as computers - making them accessible to new kinds of experimental manipulation and testing"

A living system can be viewed as having an input (stimulus), an output (response) and some internal control acting on the input to produce the output.

Bourgine & Varela 1992, Alife (Europe) I...

"Artificial Life can be better defined as a research program concerned with autonomous systems, their characterization and specific modes of viability."

A living system can be viewed as an autonomous device with its autonomy / viability arising through internal self-organizing processes which are responsible for its behaviour.

Historical "Artificial Life"

Hero of Alexandria's pneumatic animals	1st C. AD
The Golem of Prague	~1400->
Jacks, clockwork armoured figures	~1400->
Vaucanson's duck	1735
Mary Shelley's "Frankenstein"	1818
Von Neumann's self-reproducing automata	1929->
Penrose's machines	1959
Conway's "Game of Life"	1970
Braitenberg's vehicles	1984
Dawkin's biomorphs	1986
Reynold's boids	1987
Marionettes, Automata etc.	~early ->
Artificial Life Art	recent

Artificial Life Techniques

Software

Genetic Algorithms (with explicit, implicit or interactive fitness functions)
Physical simulations (with abstract or real-world-modelled Physics and Chemistry)
Multiple agent / ecosystem simulations
Cellular Automata

Hardware

Autonomous robot construction

Wetware

Experimenting with chemical processes in a beaker

Artificial Life, Procedural Modelling, Procedural Animation

Artificial Life as a field of endeavour is so loosely defined that almost any animated procedural model, vaguely reminiscent of a living thing, will be called 'Artificial Life'.

FIT4012: Artificial Life, Procedural Modelling & Animation General Course References