From TinyCog
Jump to: navigation, search
Figure 1: Plans are composted of scenes, with a script marking the actual execution path.
Figure 2: A man stands in front of a table
Figure 3: The scene graph for figure 1

We define a "scene" as a 3 dimensional representation of real-world objects represented by a scene graph ([Strauss 1993]). A scene graph consists of a number of Objects, together with their position, orientation and spatial relationship, so that it can be rendered into a 2D image by a Render Engine and processed using symbolic AI.

A "scene" is the basic data-structures of SBR, and can be thought to be a union of the following concepts:

  • A real-world 3D scene derived from sensor information,
  • a "mental image" as defined in psychology,
  • a Scene Graph as known from computer gaming,
  • a "Frame" or "semantic network" from good ol' fashioned Artificial Intelligence (GOFAI),
  • an "ABox" ("assertional box") object network from Description Logics,
  • a "start scene" of a Planner, which is the initial state of affairs,
  • a "goal" of a Planner, which is the desired end state of a Plan,
  • a intermediate state of a Planner and
  • a series of object parameters to be handled by statistical AI.

Scenes can also represent internal TinyCog data-structures as a "pen on paper" 2D diagram. This way, scenes provide the base for self referencing and Meta Reasoning. Scenes are implemented on the lowest level as 3D scene graphs (as in computer gaming), while behaving more like semantic networks on higher levels capturing objects and their relationships. Scenes are used in all Subsystems in the same way and provide common semantics and a kind of integration backbone.