Towards Interactive RealTime Crowd Behavior Simulation

1
Towards Interactive Real-Time Crowd Behavior
Simulation

• Branislav Ulicny and Daniel Thalmann
• (Computer Graphics Forum 2002)

2
Outline

• Introduction
• Challenges
• Levels of Variety
• The Model of the World
• Behavior Model
• Implementation
• Case Study
• Conclusions Future work

3
Introduction

• Virtual crowds
• Common element of cinematic experience
• Rare in realm of real-time applications
• Real-time applications
• Interactivity needed
• Limited computational resources
• Real-time crowd simulation system
• Models a crowd as a multi-agent system with
  emphasis on individuals
• Defines a layered modular behavioral architecture

4
Challenges

• Multi-agent virtual human simulations
• able to run in real-time with 3D visualization
• allow user interaction
• Compared with single-agent simulations
• need the efficient variety management at every
  level
• need the methods easing the control of multiple
  characters
• Compared with non-real-time simulations
• need the fast and scalable methods to compute
  behaviors

5
Levels of Variety

• Variety
• The quality or state of having different forms
  or types
• Enhance the realism
• Levels of variety
• LV0 only a single solution
• LV1 a finite number of solutions
• LV1 the combinations of sub-solutions
• LV2 an infinite number of possible solutions
• Sometimes systems with a lower level of variety
  are justifiable
• Ex) emergency simulation

6
The Model of the World (1/3)

• The model is composed of the agents and the
  environment.

FSMs Finite State Machines
7
The Model of the World (2/3)

• The environment
• Static environments streets, buildings
• Dynamic objects fire, gas clouds
• Agents contain
• A set of internal attributes
• Various psychological or physiological states
• A set of higher-level complex behaviors
• A set of rules for selection among behaviors

8
The Model of the World (3/3)

• Interaction between the agents and the static
  environments
• Path-finder module
• Interaction is done via exchanges of events
• Interaction between the agents and dynamic
  objects
• simulating physical interactions or perceptual
  interactions
• Inter-agent interaction
• User-agent interaction

9
Behavior Model (1/4)

• Agents
• act in accordance with their surrounding
  environment
• react to environments changes, the other agents,
  actions of the user
• A behavior model should
• Simple enough to allow real-time animation of
  many agents
• Yet still Complex to provide interesting behaviors

10
Behavior model (2/4)

• Layered approach

11
Behavior model (3/4)

• The highest level
• Rules select high-level behaviors
• Three parts of the rules
• Selection part who
• Condition part when
• Consequent part what
• An example of a rule

FOR ALL WHEN EVENT in_danger_area AND
ATTRIBUTE fear gt 50 THEN BEHAVIOR FLEE
12
Behavior model (4/4)

• The middle level
• Hierarchical finite state machines
• select the low-level actions
• manage connections with the environment
• can call other FSMs
• The lowest level
• Path-finder module
• constructs a sequence of waypoints
• Collision-avoidance module
• corrects the trajectory in order to avoid
  collisions

13
System implementation

• Separating the model part from the visualization
  part
• allows different representations
• allows different update rate for different
  components
• allows platform independent
• Separating the action selecting part from the
  action executing part
• The animation variety

14
Case study Crowds in virtual reality Training
system (1/2)

• CROSSES (CROwd Simulation System for Emergency
  Situations)
• A training system for emergency situations
• Urban emergency situations involving crowds

15
Case study Crowds in virtual reality Training
system (2/2)

• An example
• The threat becomes active.
• The threat sends events to the affected agent.
• The rule become activated.
• Flee behavior is triggered.
• The agent selects the way out of danger.
• The agent return to normal behavior

FOR ALL WHEN EVENT in_danger_area AND
ATTRIBUTE fear gt 50 THEN BEHAVIOR FLEE
16
Case study Crowds in a virtual Heritage
reconstruction

• Reconstruction of a virtual heritage site
• Recreating life inside the architectural models
  for increment of the realism

17
Conclusions Future works

• Real-time crowd simulation system
• Introduce the notion of levels of variety
• Define a modular behavioral architecture
• Future works
• Enhance the levels of variety
• Improve the behavior model