Warning: Most of the links on this page are not yet connected.

Planning is an important activity for autonomous agents. It is used for more than just moving objects around. Examples include the following:

• Planning to achieve some goal such as move an object from one location to another,
• Planning a discourse that will cause another agent to do something,
• Understanding the plans of another agent in order to be able to assist (thwart) that agent.

We concentrate on the first of these. We remark here that if one uses top down plan generation then a bottom up analysis of the plan tree can be used for pan understanding.

We use a state space model for the planning process. We must be able to represent the current state of the world and the effects of the agent's actions upon it. T this end we associate to each action three things:

1. The preconditions: The conditions that must be satisfied if the action is to be possible;
2. The add-list: list of relations that are made true as a result of taking the action;
3. The delete-list: the list of realtionships that the action destroys;

This raises the (in)famous frame problem which for the moment we deal with by ignoring it.

There are various approaches to planning. We will consider three examples of planning.

1. STRIPS
2. Goal regression
3. Nonlinear planning

To return to the AI home page click AIHOME.

Jonathan Hodgson
Last Changed 23/11/94