Wilensky, Robert (1980). Meta-Planning: Representing and Using Knowledge About Planning in Problem Solving and Natural Language Understanding.

Author of the summary: Jake Auxier

Motivation and Goals

Higher level plans and goals, i.e., meta-plans and meta-goals can be used by the same sort of mechanisms that are needed for processing ordinary goals and plans.

The meta-planning knowledge will enable the mechanisms needed for processing to handle much more complicated situations involving multiplicity of goals in a uniform manner.

Issue

Development of meta-plans and meta-goals that can be used both in problem solvers and story understanders.

Design

Meta-Plans

Meta-planning is the body of knowledge about the planning process.

• Set of goals for the planning process: meta-goals
• Set of plans to achieve the meta-goals: meta-plans

Meta-planning vs. Planning w/ Constraints and Critics

Supply the understander with the description of the situations in which meta-themes cause meta-goals to arise, and with knowledge about he meta-plans associated with each meta-goal.

The understander then must detect goal conflicts within the story it is trying to understand. When it does, it infers the meta-goals of resolving the goal conflicts, and subsequent actions are interpreted as plans fulfilling the resolution.

Meta-planning is involved in the component of the Planner called the Goal Detector. Goal detection is apparent when an autonomous system is trying to solve a problem. The autonomous system must contend with other autonomous systems that need its resources, it must know when it needs certain resources, it must know if it needs help from other systems, and it must deal with internally generated goals. All these goals can interact and conflict with each other. This is where meta-planning comes into play.

Implementation

1. Goal Conflict: an adverse interaction between goals of the planner
2. Goal Competition: an adverse interaction between goals of a different planner
3. Goal Overlap: a positive interaction between the goals of the planner
4. Goal Concord: a positive interaction between the goals of different planners
5. Goal Subsumption: establishing a state that makes it easier to fulfill a recurring goal

To help in determining which of a number of goals to use in goal conflicts, goals are given values and costs.

Both ordinary goals and meta-goals can be compared against each other in case of conflicts.

Meta-planning has been implemented in the systems, PAM and PANDORA.