Dynamo Dynamic, Data-driven Character Control with Adjustable Balance

1
Dynamo Dynamic, Data-driven Character Control
with Adjustable Balance

• Pawel Wrotek Electronic Arts
• Chad Jenkins Brown University
• Morgan McGuire Williams College

2
First, a video
3
Character Motion

• An integral part of modern video games

FIFA 2006 (EA)
San Andreas (Rockstar)
Antigrav (Harmonix)
4
Kinematic Character Motion

• Expressed by rigid body kinematics
• Rigid bodies connected by joints
• Character pose defined by rotation
• of joints
• Vector ?(t) represents pose
• at a given instant of time

5
Motion GenerationMocap and Keyframing

• Motion capture
• Manual keyframing
• () path of least resistance
• () absolute control, wyciwyg
• (-) not physically dynamic
• such motion is a static and partial snapshot of
  the dynamics occurred at the time of
  collection/creation Analogously, video is a
  snapshot of the physics of light
• Great for production animation, not so great for
  interactive virtual environments

6
Motion GenerationMocap and Keyframing

• () path of least resistance
• () absolute control wyciwyg
• (-) not physically dynamic
• static and partial snapshot of the dynamics
  occurred at the time of creation
• Production animation, not interactive games

God of War 2 (Sony)
7
Motion GenerationProcedural Animation

• Rules/algorithms to automatically generate motion
• Three categories of approaches
• Indirectly emulate physical plausibility
• Perlin,Goldberg 94 Popovic, Witkin 99 Kovar
  et al. 02
• Simulate physics only when necessary
• Shapiro et al. 03 Zordan et al. 05
• Simulate physics directly and persistently
• Hodgins et al. 95 Laszlo et al. 00

8
Procedural Animation

• Indirectly emulate physical plausibility
• Scripting Perlin,Goldberg 94
• Blending Rose et al. 98
• Optimization Liu et al. 05 Arikan et al. 03
• () creators retain control
• Creators define all rules for movement
• (-) violates the checks and balances of motion
• Motion control abuses its power over physics
• (-) limits emergent behavior

9
Procedural Animation

• Simulate physics directly
• Ragdolls
• Controllers to generate motor forces
• Zordan, Hodgins 02 Faloutsos et al.
  01 Popovic et al. 00
• () proper separation of powers
• Physics, control, AI
• Allows for emergent, natural interactions
• (-) inherit problems that plague robotics

Physics
Controller
10
Procedural Animation

• Simulate physics only when necessary
• Dynamic response
• Shapiro et al. 03 Zordan et al. 2005
  Natural Motion Endorphin
• Mocap for normal dynamics
• Simulation for disturbance dynamics
• () the best of mocap and simulation
• (-) limited to passive response
• Falling, getting hit, etc.
• No persistent interaction

11
Fundamental Question

• Can we have practical methods for physically
  simulated characters?
• Revisit the broader picture for autonomous
  control
• Decision making (AI) objectives, current state
  (xt) ? desired motion (xdt)
• Motion Control desired motion (xdt), current
  state (xt) ? motor forces (ut)
• Physics current state (xt) ? next state
  (xt1)

utMC(xdt-xt)
xt1P(xt,ut)
xdtAI(xt)
Physics
Motion Control
Decision Making
ut
xdt
objectives
xt1
12
The Autonomous Physical Motion Control Problem
utMC(xdt-xt)
xdtAI(xt)
xt1P(xt,ut)
Motion Control
Decision Making
ut
xdt
Physics
ut
objectives
xt1
13
The Autonomous Physical Motion Control Problem
utMC(xdt-xt)
xdtAI(xt)
Motion Control
Decision Making
ut
xdt
objectives
xt1

• Simulating physics
• Download ODE
• Buy Havoc
• Implement Guendelman et al. 03

14
The Autonomous Motion Control Problem
utMC(xdt-xt)
Motion Control
ut
xdt
Mocap data
xt1

• AI for autonomous decision making
• Someone elses problem
• Interface point for decision making
• Focus on motion control
• Motion capture as decision making placeholder

15
Motion Control Impediments
utMC(xdt-xt)
Motion Control
ut
xdt
Mocap data
xt1

• Gain tuning for motion control
• Balance for upright motion

16
Motion Control Impediments
utMC(xdt-xt)
Motion Control
ut
xdt
Mocap data
xt1

• Gain tuning for motion control
• Balance for upright motion

Problem parent space control?
17
Motion Control Impediments
utMC(xdt-xt)
Motion Control
ut
xdt
Mocap data
xt1

• Gain tuning for motion control
• Balance for upright motion

Problem parent space control?
Solution world space control?
18
Segway Analogy
19
Segway Analogy
20
Segway Analogy
21
Feedback Motion Control

• Parent PD-servo
• Torque u about an axis
• Appropriate kp and kd values are necessary for
  stable control
• Tedious and difficult
• Holdover from robot rotational sensing

u
D. Brogan
22
Parent Space Control

• Moving reference frame
• Interferes with stability
• Lacks consideration of global orientation

23
World Space Control

• Fixed global reference frame
• Steady target desireds
• Implicit balance

24
World Space PD-Servo

• t kp (v ?) kd (?d ?a)
• Wd desired world space rotation matrix
• Wa actual world space rotation matrix
• T Wd Wa-1 (transformation from Wa to Wd)
• v, ? rotation axis, angle derived from T
• ?d desired world space angular velocity
• ?a actual world space angular velocity

?
Wa
v
Wd
25
A Note about Axis-Angle(Source code in the paper)

• Torques determined by desired angular
  acceleration
• i.e., Proportional to 2nd derivative of rotation
• 1D Hinge Hodgins95 t ? ?2q/?t2
• 3D Ball joint t ? ?2rotation/?t2
• but Matrix/Quat derivatives produce denormalized
  results under ODEs Euler integration and are
  awkward to convert to torques
• Rotation axis is fixed anyway during the Euler
  timestep, so reduce to a 1D problem
• 3D Ball joint

26
Early Results

• Gain Tuning
• Cartwheel with object

27
Super-balancing

• An artifact of world space control
• Retain separation of powers
• Desired pose relative to character root (Person
  space)
• Desired root orientation specified by AI
• Actual position and orientation determined by
  physics

28
Root-spring control

• Spring only opposes gravity (no rotation about
  FG)
• Torque-limited and breaks under excessive strain

t maximum
Applied Torque t
t tbalance
t 0
tbalance
Torque limit
Breaking point
29
Results

• Obstacle course
• Parent space
• Person space
• User interaction
• Balance comparison
• Ballistic
• Person space (meathook)
• Person space (root spring), Parent space
• In-game boxing

Parent space
Dynamo
30
Results

• Obstacle course
• Parent space
• Person space
• User interaction
• Balance comparison
• Ballistic
• Person space (meathook)
• Person space (root spring), Parent space
• In-game boxing

31
Results

• Obstacle course
• Parent space
• Person space
• User interaction
• Balance comparison
• Ballistic
• Person space (meathook)
• Person space (root spring), Parent space
• In-game boxing

32
Results

• Obstacle course
• Parent space
• Person space
• User interaction
• Balance comparison
• Ballistic
• Person space (meathook)
• Person space (root spring), Parent space
• In-game boxing

33
Results

• Obstacle course
• Parent space
• Person space
• User interaction
• Balance comparison
• Ballistic
• Person space (meathook)
• Person space (root spring), Parent space
• In-game boxing

34
Future Work

• AI for goal-oriented motion generation
• Experimental parent vs. world analysis
• Biomechanical character constraints
• Embodied perception

35
Conclusion

• Physically dynamic characters are practical
• World-space control yields
• Implicit character balance
• Easier gain tuning
• Path to emergent behavior for interactive
  characters

36
Acknowledgements

• NSF Award IIS-0534858
• Dan Byers
• Sam Howell
• mocap.cs.cmu.edu
• G3D and ODE user communities
• Innovating Game Development
• Guest Lecturers
• A-Lab

37
RoboCup Dynamical Soccer

• cjenkins_at_cs.brown.edu
• morgan_at_cs.williams.edu