CAOS Presentation: SIGGRAPH 2000

ACM STUDENT CHAPTER MEETING!
RESEARCH DIRECTIONS IN
COMPUTER & VIDEO GAMES
Bill White will present an overview of several tutorials and panel discussions
that took place at SIGGRAPH 2000 concerning research directions for the
computer and video gaming industry.
TERRAIN
RENDERING
OBSTACLE
AVOIDANCE
Wednesday
September 20, 2000
2:00 – 3:00 PM
Engineering Building 0012
COMPOSITE
TEXTURES
The video gaming industry is booming!
This year, for the first time, computer and
video game revenue in the U.S. will exceed
feature film revenue!
Cinema-quality graphics have
been achieved in affordable
platforms, driven by advances
in processor and memory
technology.
Where does computer graphics fit in?
Unfortunately, like the film industry, the
computer game industry has limited
resources (particularly time) for R&D.
It relies instead on
academia and game
platform developers
to provide the
research that results
in game advances.
What’s so different about games research?
The momentum for graphics research in the past has
come from applications in the film industry and the
scientific community.
Prioritized:
 High quality images
 Massive data sets
Deprioritized:
 Real-time processing
 Inexpensive platform
Graphics research is being increasingly propelled by
the video gaming and virtual reality industries.
Prioritized:
 High quality images
Reprioritized:
 Real-time processing
 Massive data sets
 Inexpensive platform
Problem Area #1: Scene Complexity
Unlike the simple games of yesteryear, modern
games require multiple characters interacting in
elaborate environments, with independent motion
and (hopefully) realistic levels of detail.
Approach A:
Brute force
modeling of
vast numbers
of polygons.
Approach B:
Fast-processing,
low-memory
NURBS models
of surfaces.
Approach C:
Progressive
meshes for
continuous
level-of-detail.
Problem:
Serious
processing.
Problem:
Problem:
Counterintuitive Smooth
modeling.
interpolation.
Brute Force Large Poly-Count Display
Tens of thousands of
polygons rendered at
once.
The vertices of distant
polygons might be
merged to make
larger, less detailed
polygons.
Active polygon list
must be continually
updated as player
navigates playing
field.
NURBS Surface Modeling
Large models can be effectively stored in a fraction of
the space and processed in a fraction of the time as
polygonal models.
Developing the vertex sets for such models
is rather elaborate and counterintuitive.
70,000
triangles
30,000
vertices
153
patches
Continuous Level-of-Detail
Rather than using a single polygonal model for each
object, use multiple models, displaying the more
detailed one when the object is closer to the viewer.
Artists are encouraged to
composite multiple
texture maps, to improve
run-time performance by
reducing retrieval time.
This intensifies the
problem of texture seams
where the texture has
been stitched together.
Problem Area #2:
Realistic Behavior
Having characters and objects
within the game environment
behave “appropriately” is a
particular challenge to game
developers.
Physics in Games
 How to efficiently program the
laws of physics into a game?
Steering Behavior
 How to get large numbers of
characters to behave like a group
of individuals?
Emotion Synthesis
How to mimic human behavior
so characters not controlled by
the player react in a seemingly
emotional fashion?
Physics in Games
Years of research in the field of
scientific visualization have
yielded several efficient
methods for rendering rigid
bodies and fluid dynamics.
Everything gets a bit more complex when the game
player is allowed to interact with the environment at will.
Steering Behavior
Some game objects (enemy starfighters, attacking dinosaurs, etc.)
tend to behave in “flocks” or “herds”, yielding behavior patterns that
aren’t that difficult to program or that time-consuming to render.
Pursuit
Obstacle Avoidance
Formation Cornering
Neighbor Detection
Alignment
Emotion Synthesis
Having game characters react “emotionally” to the environment
and the developments in the game’s storyline can greatly
enhance the gaming experience.
Programming the character to react to interactive (i.e., noncanned) stimuli can be incredibly difficult.
One solution: Network multiple players together so each
character is controlled by a separate emotion-driven player!