Roland Geraerts and Mark Overmars

1
Enhancing Corridor Maps for Real-Time Path
Planning in Virtual Environments

• Roland Geraerts and Mark Overmars
• CASA08

2
Criteria

• Fast and flexible path planner
• Real-time planning for thousands of characters
• Dealing with local hazards
• Natural paths
• Smooth
• Short
• Keeps some distance to obstacles
• Avoids other characters

3
The CMM Construction phase

• The Corridor Map
• A system of collision-free corridors for the
  static obstacles
• Corridor sequence of maximum clearance disks
• Data structure generalized VD clearance
  additional info

Corridor map
Corridor
4
The CMM Construction phase

• Computing the GVD
• Draw distance mesh for each obstacle with GPU
• Parallel projection of meshes
• Trace boundaries
• Prune the graph
• Re-sampling
• Increases efficiency
• Adding data
• Identify connected components
• For each corridor, store maximum clearance a
  character can have

5
Experiments Construction phase
McKenna MOUT environment
Footprint and Corridor Map 0.05s
6
Experiments Construction phase
City environment
Footprint and Corridor Map 0.64s
7
The CMM Query phase

• Extract corridor for start and goal ? global
  route
• Character follows attraction point ? local route
• Runs along backbone path toward goal
• Used to define a force function, applied to
  character
• Obtain path
• Integration over time, update velocity/position/at
  traction point
• Yields a smooth (C1-continuous) path
• Other behavior locally adjust path by adding
  forces

Query points
Corridorbackbone
Path
8
The CMM Query phase

• For start/goal, find closest disk enclosing the
  character
• kd-tree
• Find the shortest backbone path
• Dijkstra versus A
• Compute the corridor
• Compute the path
• Verlet integration

Corridorbackbone
Path
Query points
9
Experiments Query phase
McKenna MOUT environment
Corridor and path 0.2ms (average)
10
Experiments Query phase
City environment
Corridor and path 1.2ms (average)
11
Crowd Simulation

• Goal oriented behavior
• Each character has its own long term goal
• A start and goal fixes a corridor
• When a character has reached its goal, a new goal
  will be chosen
• Obstacle avoidance
• Helbing and Molnars social force model
• Efficient nearest neighbor computations
• 2D grid storing the characters

12
Crowd Simulation Experiments

• Performance (1 cpu)

13
Crowd Simulation

• Example

14
Conclusions

• The Corridor Map Method is fast
• 10,000 characters can be simulated in real-time
• The Corridor Map Method is flexible
• Collision avoidance
• Crowds
• The Corridor Map Method produces natural paths
• Smooth
• Short
• Keeps some distance to obstacles