From TinyCog
Jump to: navigation, search

An "Action" in Scene Based Reasoning (SBR) is a change in the state of the world. Technically, an action results in the modification of Object parameters in a Scene. There are two types of actions:

  • Actions initiated by an Agent (active behavior):
    Example: "The hunter runs towards the deer." or
  • Actions due to the laws of physics as simulated in SBR using Physics Simulation (passive behavior):
    Example: "The stone falls to the ground, because it's support was removed."

Actions are usually part of a Plan data-structure where an action connects two Scenes.

Agents are equipped in TinyCog with a small set of "primitive actions" that the agents "know" to execute directly after the start of the system. The Agents have to learn all other actions during the course of TinyCog execution using various types of Learning processes.