Animation Wrapup

1
Animation Wrap-up

• Summary of Animation
• Collision Detection

2
Generating Motion

• What matters?
• Quality of motion appropriate for rendering style
  and frame rate
  
• Controllable from UI
• Controllable from AI
• Skills of the animated character
• Personality of the animated character

3
Keyframing

• Fine level of control
• Quality of motion depends on skill of animator

4
Motion Capture

• Natural-looking motion
• Hard to generalize motions
• Registration is difficult
• Often seems weightless Bill Kroyer, Rhythm
  Hues

5
Simulation (Broadly Defined)

• Physics is hard to simulate
• Pseudo-physics is somewhat hard
• Control is very hard
• Gives Generalization Interactivity

Desired Behavior
Forces and Torques
Model Numerical Integrator
User/AI
Control
State
Graphics
6
When to Use What Method?

• Keyframing
• Sprites and other simple animations
• Non-human characters
• Coarse collision detection
• Motion Capture
• Human figures
• Subtle motions, long motions
• Simulation
• Passive simulations
• When interactivity w/ motion is important

7
Integration of Technologies

• Layering
• Add hand/finger motion later
• Facial animation
• Use keyframing to modify data
• Fix holes in data
• Use motion capture to drive simulation

8
Collision Detection

• Essential for many games
• Shooting
• Kicking, punching, beating, whacking, smashing,
  hitting, chopping, dicing, slicing, julienne
  fries
  
• Car crashes
• Expensive
• tests!!

9
Collision Detection

• For each object i containing polygons p
• Test for intersection with object j with polygons
  q
  
• (j i)
• For polyhedral objects, test if object i
  penetrates surface of j
  
• Test if vertices of i straddle polygon q of j
• If straddle, then test intersection of polygon q
  with polygon p of object i

10
Collision Detection

• Efficiency hacks/cheats
• Fewer tests Exploit spatial coherence
• Use bounding boxes/spheres
• Hierarchies of bounding boxes/spheres

11
Bounding Boxes

• Axis-aligned vs. Object-aligned
• Axis-aligned BBox change as object moves
• Approximate by rotating BBox

Swept volume
12
Collision Detection

• Convex objects
• Look for separating plane
• Test all faces
• Test each edge from obj 1 against vertex of
  obj 2
  
• Save separating plane for next animation frame

13
Collision Detection

• Concave Objects
• Break apart
• Convex hull
• Automatic or artist-created

14
Collision Detection

• To go faster
• Sort on one dimension
• Bucket sort (i.e. discretize in 1 dimension)
• Exploit temporal coherence
• Maintain a list of object pairs that are close to
  each other
  
• Use current speeds to estimate likely collisions
• Use cheaper tests

15
Collision Detection
16
Collision Detection

• Cheaper distance calculation
• Compare against
• Approximation
• Manhattan distance - Shortest side/2

17
Collision Detection Sprites

• AND for each pixel in sprites

18
Collision Detection

• Discretization in 3D
• Create a voxel array
• Store an ID in each voxel where an object is
• Collision where voxel has an ID already
• Difficult to determine good voxel size
• Huge memory
• Hash table -- gives constant time point queries

19
Inter-Object Distance

• A related problem
• Collision Avoidance
• Part of motion planning is to avoid collisions
• Many collision detection algorithms take time
  into account
  
• Estimate Collision Time, distance
• www.cs.unc.edu/geom/collision.html

20
AI in video games

• 5-10 of CPU for Realtime
• 25-50 of CPU for Turn-based
• Chase/Escape behaviors
• Group behaviors
• Finite State machines
• Adaptation/Learning

21
Questions

• How good should the AI be?
• Why are people more fun than NPCs?
• Will networked games reduce AI?
• New directions for AI in games?

22
Chase/Evade

• Algorithm for the predator?

23
Enhancements to Chase

• Speed Control
• Velocity, Acceleration max/min
• Limited turning Radius
• Randomness
• Moves
• Patterns

24
Enhancements to Chase

• Anticipation
• Build a model of user behavior

McCain
Bush
25
Enhancements to Chase

• Anticipation
• Build a model of user behavior

26
Group Behaviors

• Lots of background characters to create a feeling
  of motion
  
• Make area appear interesting, rich, populated

27
Formations
Turning corners Obstacles
28
Flocking -- (HalfLife, Unreal)
Simple version Compute trajectory to head toward
s centroid

• What might go wrong?

29
Group Behaviors
Craig Reynolds SIGGRAPH 1987

• Reaction to neighbors

30
Steering Behaviors

• Pursue
• Evade
• Wander
• Obstacle Avoidance
• Wall/Path following
• Queuing
• Combine behaviors with weights
• What could go wrong?

31
What could go wrong?
Forces balance out in dead end
Exactly aligned

• Does not handle changes in strategy

32
Perceptual Models
33
Production Rules

• If( enemy in sight ) fire
• If( big enemy in sight ) run away
• If( --- ) ----
• Selecting among multiple valid rules
• Priority weighting for rules or sensor events
• Random selection
• No state (in pure form)