Plan

From TinyCog
Jump to: navigation, search
Figure 1: Plans are composted of scenes, with a script marking the actual execution path.

A "plan" is an internal data-structure of the Planner. A plan consists of a tree with one root task which is decomposed to a sequence of sub-tasks in the first level of the tree. When executing a plan, a Script is created with the actual states of the real world during plan execution. The reverse operation of Plan Recognition constructs a plan from a given Script.

A plan may be converted into a 2D Scene using a "pen on paper" representation, typically pruning task decompositions after the 1st level. This 2D Scene is then accessible for self-reference and meta-reasoning by the higher reasoning faculties, may become optimized or combined with other plans and then be converted again into an executable plan.

Implementation

TinyCog 0.0.1 implements deterministic plans (currently with a single outcomes per action). TinyCog 0.0.3 will introduce plans with alternative outcomes per task when elaborating on the Hunter Domain.