Ribbon%20Networks%20for%20Modeling%20Navigable%20Paths%20of%20Autonomous%20Agents%20in%20Virtual%20Urban%20Environments

1
Ribbon Networks for Modeling Navigable Paths of
Autonomous Agents in Virtual Urban Environments

• Peter Willemsen Joseph Kearney Hongling
  Wang
• School of Computing Dept. Of Computer
  Science
• University of Utah
  University of Iowa

2
Research Overview

• Dynamic VE
• Vehicles, pedestrians, etc
• Lots of them!
• Behavior and scenario programming
• Environment Modeling
• Adapted to the needs of behavior and scenario
  programs

Environment
Scenario
Behavior
3
Motivation

• Virtual environments as laboratories for
  psychological study
• Child bicycle-riding behavior
• Dynamic virtual environments need activity
• Realistic behaviors
• Replicable scenarios
• Spatial awareness
• What routes are accessible?
• Where are other objects?
• What constraints are important?

4
Focus of Talk - Environment

• Geometric Information
• Shape and curve of pathways
• Topological Information
• Interconnections between pathways
• Logical information
• Rules governing behavior
• Occupancy Information
• Locations of objects

5
Related Research

• Farenc et al - Informed Environments
• Eurographics 1999
• Behavioral content connected to scene graph
• Donikian and Thomas - VUEMS
• Eurographics 2000, CGI 1997
• Comprehensive system for modeling urban networks
  of streets, sidewalks, and tramways
• Road modeling for driving simulation
• Civil engineering design of roadways
• Desirable properties for high speed roadways

6
Ribbon Networks

• Model urban streets, sidewalks, and navigable
  ways as ribbons in space
• Curvilinear coordinate system
• Arc-length parameterized spline
• Fast lookup based on distance
• Defines geometry and orientation of navigable
  surface
• Provides frame of reference for local spatial
  relationships
• Annotations
• (e.g. speed zones, passing regions,
• features)

7
Ribbon Network Computations

• Map (X,Y,Z) (D,O,L)
• Closest Point Computation
• Typical bottleneck
• Usual optimization techniques fail
• Slow convergence and divergence
• Hybrid Method
• Combines Newtons method and quadratic
  minimization
• fast and robust

(D, O, L)
8
Ribbon Network Attributes

• Ribbons capable of representing (all?) types of
  roadways
• Curvy, straight, engineering specified
• Cracks due to modeling errors fixed by
  imperceptible numerical nudges
• Attributes annotate ribbon structure
• Lanes (parallel streams channel traffic flow)
• Speed limit signs, passing zones, and behavior
  associated objects, such as flag men
• Ribbon establishes convenient local access to
  attributes

9
Intersections

• Splice together incoming and outgoing lanes
• Corridors (single lane ribbons)
• Dependencies
• Corridor relationships
• Crossing, merging, crosswalk, right of way,
  traffic control
• Zero-length intersections
• Change road cross-section
• Hierarchical intersections
• Pedestrians intersections
• Type and instance

Corridor C0 (crosses C1) (merges_with
C2) (right_of_way C1,C2)
10
Paths as Overlay Ribbons

• Single-lane ribbon overlayed on the road network
• Simplify behavior at road/intersection junctures
• Tracking
• Collision avoidance
• Dynamically composed by individual behaviors
• Provide egocentric, local coordinate system for
  environment queries

Insert picture of Washington and Clinton
intersection overlaid with a path ribbon.
11
Occupancy

• Spatial relationships based on ribbon structure
• (INSERT PICTURE HERE)
• Occupancy queries Two Forms
• Leader
• First object on ribbon segment
• List of leaders
• Ordered list of objects on ribbon segment

12
Results

• Ribbon network forms substrate for the Hank
  Simulator
• General purpose virtual environment
  software
• Simulator used hundreds of hours in production
  mode
• Database queries (almost) never fail
• Behavior code is greatly simplified
• Environment Description Format (EDF)
• Modeling language for road networks

13
The System in Action (movie)
14
Conclusions

• Solid conceptual model of urban landscape
• Ribbon networks
• Geometry matches shape and curve of pathways
• Topology describes interconnections
• Logical attributes provide constraints and
  socio-cultural aspects
• Occupancy sets up inter-object relationships
• Behaviors obtain spatial awareness
• Environment model is efficient, robust, and
  simple
• Supports behavior and scenario control programs

15
Future Work

• Automatic generation of models from GIS Data
• Terrain from roads
• Incorporating pedestrian models
• Integration into physical environment system
• VTerrain.org

16
Collaborators

• University of Iowa, Computer Science
• Jim Cremer
• Zhihong Wang
• Scott Davis
• Jill Secher
• University of Iowa, Psychology
• Jodie Plumert
• University of Iowa, Math
• Ken Atkinson

17
Acknowledgments

• Visual Database Expertise
• Joan Severson
• Shayne Gelo
• Kate Kearney
• NSF Support CDA-96-23614, INT-9724746,
  EIA-0130864, and IIS-0002535

18
(No Transcript)
19
(No Transcript)
20
Importance of Environment

• This slide was earlier, but didnt like it there
• Not sure if I like it in any case
• Behaviors are difficult to create!
• Time consuming
• Part of Environment!
• Behaviors and enviroment mesh together
• Provide sets of queries for behavior
• Where am I? (Spatial awareness)
• Whos near me? (Occupancy)
• Where can I go from here? (Geometry, topology)
• What rules should I obey here? (Logic)

21
Modeling Language

• System is built upon a language
• Describes the ribbon networks

• Example

22
Query Environment Database

• System provides run-time queries
• Where am I?
• Where can I go?
• Other examples
• Overview the general query structure of the
  run-time system
• Abstract behavioral queries