[SOLVED] CS GPU Crowd Animation

Crowd Animation
Michael Andereck CSE 888 March 7, 2008

Particle systems
Flocking systems
Behavioral systems
History

Particle Systems
Tens of Thousands of members
No intelligence
Respond to global forces, including collision reaction

Flocking Systems
Thousands of members
Limited intelligence
Some physical basis
Collision avoidance
Local control system

Flocks, Herds, and Schools:
A Distributed Behavioral Model
Boids given simple behavioral constraints:
Separation Alignment Cohesion
Often cited as one of the first examples of flocking and emergent behavior

Behavioral Systems
Tens to thousands of members
Each member is highly intelligent
Collision avoidance
Behavior based on rules

Applications of Crowd Models
Computer animation
Games, interactive graphics, and virtual reality
Robotics
Aerospace
Education
Artificial life
Art
Biology
Physics
Other areas

Hierarchical Model for Real Time Simulation of Virtual Human Crowds
Crowds are modeled at varying hierarchical levels:
Crowd Groups Agents
Behavior is controlled by external control, scripted control, and at a low level, the innate behavior

Soraia Raupp Musse and Daniel Thalmann, 2001

Densely Populated Urban Environments
Focus is on rendering large crowds with robust collision detection, realistic shading, culling, and pedestrian movement
Yiorgos Chrysanthou and Franco Tecchia, 2001
Figure 2. Particles avoid walls but climb over the smaller steps

Scalable Behaviors for Crowd Simulation
Allows increase in crowd behavioral complexity without a corresponding increase in agent complexity
Users can dynamically specify crowd behaviors
Agents have a situation-based control structure
Mankyu Sung, Michael Gleicher, Stephen Chenney, 2004
Figure 5: Left: Spatial situations can be easily set by drawing
directly on the environment. Situation composition can
be specied by overlaying regions. Right: Non- spatial situation
can be set on the crowd by grouping participants.

Autonomous Pedestrians
Tackles the open problem of emulating real pedestrians in urban environments
Motions controlled at different levels
Reactive behaviors
Navigationalandmotivationalbehaviors Otherinterestingbehaviors
Information stored in mental states
Wei Shao and Demetri Terzopoulos, 2005

Fast and Accurate Goal-Directed Motion Synthesis for Crowds
Limit characters to specific postitions
Generate a space of possible actions with a motion graph
which is searched to find the appropriate action
Mankyu Sung, Lucas Kovar and Michael Gleicher, 2005

Modeling Behavior in a School of Fish by Fast Synthetic Distributed Vision
Uses distributed vision rays to lower the cost of computing the fishs perception at each step.
Adriano Rinaldi, Alessio Malizia, Rick Parent, 2006

Continuum Crowds
Environment creates a potential field based on density and topographical speed
Individual direction based on goals, speed, and discomfort
New characters can be added at low cost

Imposters and pseudo-instancing for the GPU
Goal: to use the GPU to render as many characters as possible at interactive speed
Imposters and pseudo- instancing are used to simplify rendering where applicable
2^20 characters at 5 fps

Crowd Creation Pipeline for Games
Realistic crowds are still too expensive for realistic gaming
Combine geometry and imposter approach for faster rendering
Also include pre-simulated cloth sequences
Model prepping
Exporting the data
Rendering
R. McDonnell, S. Dobbyn and C. OSullivan, 2006

A Conceptual Framework for Modeling Crowd Behavior
Priya Dey and David Roberts in SIGSIM 2007

A context-aware agent which is able to dynamically change its desires and needs

Controlling Individual Agents in High- Density Crowd Simulation
Focus on global wayfinding and local motion
Take into account agent personalities
Line forming and character interaction, including pushing N. Pelechano, J.M. Allbeck and N.I. Badler, 2007

Massive Software
The most widely used package for large-scale crowd simulations
Used in The Lord of the Rings, The Chronicles of Narnia, I, Robot, Happy Feet and others
Developed by Stephen Regelous

Future Work
Larger scale crowds at interactive speeds
Add a wider variety of behaviors for the crowd
Add more dynamic control
Balance between agent autonomy and animator control