Ioannis Karamouzas, Roland Geraerts, Mark Overmars

1
Indicative Routes for Path Planning and Crowd
Simulation

• Ioannis Karamouzas, Roland Geraerts, Mark Overmars

2
Path Planning

• What is path planning
• Steer character from A to B
• Computer games and path planning
• Fast and flexible
• Real-time planning for thousands of characters
• Flexibility to avoid other characters and local
  hazards
• Individuals and groups
• Natural Paths
• Smooth
• Collision-free
• Short
• Keep a preferred amount of clearance from
  obstacles
• Flexible
• ..

3
Path Planning

• Maintain suspense of disbelief
• Realistic graphics and physics
• Still though, the path choices that characters
  make are poor

4
Errors in Path Planning
5
Existing Algorithms

• Grid-based A Algorithms
• Computational expensive
• Aesthetically unpleasant paths
• Waypoint graphs
• Hand designed
• Do not adapt to changes in the environment
• Navigation Meshes
• Automatic construction is slow
• Paths need to be smoothed

6
Existing Algorithms

• Local Methods
• Flocking Reynolds, 1987 1999,
• Helbings Social Force Model Helbing et al,
  2000
• Reactive style planners
• Local methods fail to find a route
• Suffer from local minima problems
• Lead to repeated motion

7
The Indicative Route Method

• In real life, people
• Do not plan an exact path, but
• A preferred/desired global route
• A path planning algorithm should produce
• An Indicative Route
• Guides character to its goal
• A corridor
• Allows for flexibility

8
The Indicative Route Method

• The IRM method in action
• A collision-free indicative route determines the
  characters preferred route
• A corridor around this route defines the walkable
  area for the character
• A smooth path is generated using a force-field
  approach

9
Computing Corridors

• The Corridor Map
• Introduced by Geraerts and Overmars, 2007
• Provides a system of collision-free corridors
• Corridor sequence of maximum clearance disks
• The Corridor Map is computed as follows
• The Generalized Voronoi Diagram is approximated
  using GPU Hoff et al, 1999
• Clearance additional info is stored

3D Environment
Skeleton of the map
Corridor
10
Computing Corridors

• Computing the Corridor Map
• Only required during preprocessing
• Very fast (50 ms, NVIDIA GeForce 8800 GTX)
• Compute a corridor
• Retract the indicative route to the Generalized
  Voronoi Diagram
• Find corresponding path in diagram
• Use clearance information as a representation of
  the corridor

11
Local Navigation in IRM

• Boundary force
• Find closest point on corridor boundary
• Increases to infinity when close to boundary
• 0 when clearance is large enough (or when on GVD)
• Steering force
• An attraction point moves along the indicative
  route
• Attracts the character with a constant steering
  force
• Noise force
• Create variation in paths
• Perlin noise is used

12
Local Navigation in IRM

• Collision Avoidance Force
• Avoid collision with other characters and moving
  entities
• Helbings model can be used
• Additional models can be easily incorporated
• Obtain the final path
• Force leads to an acceleration term
• Integration over time, update velocity/position/at
  traction point
• Results in a smooth (C1-continuous) path

13
IRM method

• Resulting vector field
• Indicative Route is the medial axis

14
Creating Indicative Routes

• Use the Generalized Voronoi Diagram
• Retract start and goal
• Find shortest path (using A)
• The corridor is obtained immediately
• Use a network of Indicative Routes
• Created by level designer
• Voronoi-Visibility Complex Wein et al, 2005
• A on coarse grid
• Additional information can be incorporated
• For example flow into account
• Use the notion of Influence Regions

15
Crowd Simulation

• Method can plan paths of a large number of
  characters
• Goal oriented behavior
• Each character has its own long term goal
• When a character reaches its goal, a new goal is
  chosen
• Wandering behavior
• Attraction points do a random walk on the
  indicative network
• Experiments
• Goal-oriented behavior
• Simulation ran for 1000 steps
• Each step calculates 0.1s of simulation time

16
Crowd Simulation - Experiments

• Test Environment

City environment
2D footprint (640 ms)
17
Crowd Simulation - Experiments

• Performance
• 2.4 GHz Intel Core2 Duo, 2 GB memory
• One CPU core used
• 3000 characters, CPU usage 26, FPS 33

18
Crowd Simulation Video
19
Current Research

• Global behavior
• Incorporating influence regions
• Types of behavior (shopping, tourists, )
• Further improving the local methods
• Take mood and personality into account
• Dealing with small groups
• Observing and modeling paths of real humans
• Motion capture data
• Tracking pedestrians
• Evaluation of the results
• Projects Website
• http//people.cs.uu.nl/ioannis/irm