Mini-project 2: Making Your Own Planning Domain

Assigned [Aug 30, 2004]. Due [September 22, 2004]

In this project, you are required to come up with a planning domain of your own choosing (hopefully different from the existing ones), and experiment with it on one of the existing planners (say AltAlt). Here are the individual tasks.

  1. Provide an english descrption of your domain
  2. Provide a PDDL description of it.
  3. Explain (in an accompanying report) what aspects of the domain were hard to convert to PDDL, and how you abstracted/simplified them. (You may have a chance to bring these issues back for latter projects).
  4. Generate at least 5 problems of varying complexity for your domain.
  5. Test these problems on an existing planner that can eat PDDL (AltAlt, GP-CSP, Blackbox, and/or FF will be fine).
  6. Write a report on the results, with focus on whether this domain is different or similar to the ones you considered in mini-project 2.
  7. Turn in the report as well as the domain/problem descriptions (you will be asked to send the electronic version of the domain too--so that in the next project, you will exchange it with other people's domains.

I would encourage you to think of domains quite different from the ones you have seen. Even if all aspects of your domain cannot be modelled with PDDL, it would be worth finding some abstraction of the domain. Latter you can consider developing a version of the domain with PDDL 2.1.

If your domain requires ADL effects/preconditions (quantified and conditional effects), you may want to develop one version with ADL and one without ADL. (This is standard practice in the planning competitions now--where they give out multiple versions of the same domain).


Here are some papers that describe the efforts of the various authors in converting complex domains into PDDL-like syntax (you don't have to read these at all. Please note that these may or may not be the best ways to model or solve the respective problems.