11: characters & navigation

characters

• avatars
  • 'avatar' popularized as computer term by stephenson in snow crash (1992), originally from sanskrit: 'Earthly incarnation of a Hindu god or goddess'
  • h-anim specification
  • h-anim animation [boxman]
  • avatar me [»]
  • avatar-related links from web3d.org [»]
  • avatar-related stuff from paul flavin [»]
• working avatars
  • jack (anthropometrics guru)
  • mariela (soap saleswoman)
  • jackie strike (2000 presidential hopeful)
  • crowd simulators [»]
  • nude performers [www.somavision.de] [www.3dmodels.de]
• using mocap
  • examples [»]
  • scripts & info [»]
  • accad mocap lab [»]
• facegen [»]
• multi-user vrml worlds
  • typically communicate to vrml scene using java eai
  • Deep Matrix [»]
  • Blaxxun [»]
  • Cortona Islands [»]
  • Open Community [»]
• mpeg4
  • about [»]
  • the spec [»]
  • collaboration with vrml [»]

navigation

NavigationInfo{} node

• NavigationInfo{}
• avatarSize
  • [force field radius, eye height, step height, center of rotation]
  • should be scaled to match environment
  • can be parented by Transform{} node for scaling
  • affects collision detection and terrain following
• headlight: on by default
• speed: match with frame rate and environment size, to avoid overshoot / plodding
• type: ["EXAMINE", "WALK", "FLY", "NONE", "ANY"]
  test worlds: [examine] [walk] [fly]
• visibilityLimit: far clipping plane

viewpoints

• don't be lazy, 0 0 10 is rarely the optimal viewpoint
• use Info, it's easy
• can use Anchor{} nodes to bind viewpoints by DEFing viewpoint node and setting anchor url field to "#viewpointName"

collisions

• turning off:
  • put everything inside a Collision{} node
  • set collide to FALSE
• leaving collision on (default) results in a speed hit if scene is at all complicated: bad
• use proxies instead
  • guiderails (invisible or not) it's easy to fall off stairs, walkways, into valleys
  • moon [»]
  • stream [»]
• vrml specification only requires detection between avatar (camera) and scene
• cortona has an extension for object-object collsions [»]

environment authoring tips

• Authoring for Navigation in 3D Worlds, Susan Kropf
• Environmental Storytelling: Creating Immersive 3D Worlds Using Lessons Learned from the Theme Park industry, Don Carson
  • part 1
  • part 2

Anchor{} node

• Anchor{}
• to another url (webpage): url ["http://www.accad.ohio-state.edu"]
• to another world, optionally specifying viewpoint: url ["world.wrl#viewpoint"]
• to change viewpoint in current world: url ["#viewpoint"]
• example: [»]

depth cues

• most display technology is still lacking stereo cues, so we must exaggerate others:
  • overlap/occlusion: overlap background with foreground so positioning is unambiguous
  • relative size: don't compose a shot with a giant in the background and and expect it to seem large
  • relative height: don't put distant objects low in the frame and expect them to seem far away
  • atmospheric perspective: use a little fog when appropriate
  • familiar size: stick familiar objects next to unfamiliars to register size
  • linear perspective: use grids, rows, rivers etc. as depth cues
  • motion parallax: set viewpoints looking across motion paths create layers of left/right movement.
    

programming navigation interfaces

• great for HUDs
• will discuss more later

visible avatars

• "where does camera go?" problem
  • body can obstruct view
  • can't see character's face if behind
  • can't see from character's vantage point if in front
• immersion issue (1^st person vs 3^rd person)
• how many controls are necessary for expressive avatars? will using an avatar be like learning to type or drive, someday?