PDDL is the standard language for the encoding of the planning domains. The earlier version of the language was developed by Drew McDermott, with the help of the 1998 Planning Competition committee. We have extended this to allow expression of timed actions and to describe the effects that time has on actions. The language has been developed in consultation with a committee (particular thanks go to David E. Smith, Fahiem Bacchus, Drew McDermott and Hector Geffner for their contributions to the discussions).

A PDDL2.1 language manual can be found here, describing the syntax and semantics of the extensions. We have also prepared a second document that discusses the semantics of the most compex end of the language and raises further possible directions for extension (see below).

Features of PDDL2.1:

It was originally intended to cover planning with uncertainty, but it has proved hard to establish a common basis for temporal planning and we are keen to make our proposals known, rather than embarking on a further round of contraversial extensions. However, a group of interested parties are exploring this and are considering making a proposal to be aired as an informal part of the competition this year: we see this as a challenges track and encourage others with ideas for the way in which the competition should move forward beyond 2002 to contact us with their ideas.


This document describes an extended language (not to be used in the competition), based on PDDL2.1, but introducing events and processes, allowing its use in modelling complex discrete-continuous real-time systems. Although we do not intend this language to be used in the competition, we believe it represents an important direction in which to pursue the further extension of planning domain expressiveness and offer it as a foundation for future developments.

In the document is included a semantics for level 4 PDDL2.1 actions, via a mapping to the extended expressive power. We do not claim that this is the only plausible semantics for level 4 (indeed, we have explored a version based on the level 3 semantics we provide in the PDDL2.1 document), but consider that this approach has various advantages.