AI Planning and Scheduling Systems: Domain Repository

Note: this page is no longer being maintained. For recent collections of planning domains, we direct you to the 2000, 2002, and 2004 International Planning Competitions.

ADL/"Strips" Operator domains

Ucpop Domains

The UCPOP Planner page includes a large set of domain theories & search functions for testing.

TlPlan domains

The TlPlan planner is a planning system designed to utilize domain dependent control information specified in a temporal logic. A version coded in C is available for download. The system allows actions expressed in the full ADL language which includes first-order preconditions and conditional action effect. The system comes with over 30 implemented planning domains that illustrate the nature of the domain specific control information accepted by the system and how this information speeds up planning. Included in the set of domains are traditional planning domains, various search domains, e.g., Rubik's Cube and MicroSoft's game of FreeCell, and a few optimization domains, e.g., job-shop scheduling.

Schema (HTN) based domains

Domains for the COLLAGE Planner

COLLAGE is a domain independent planning system developed by Dr. Amy Lansky and her colleagues at NASA Ames Research Center. As well as information about the planner, the web page contains domain knowledge for two domains:

HTN Schemas for UM Translog Domain

The UM Translog Domain contains an English description of a domain for comparing planning systems (see below). Linked from this page are operator specifications for this domain for both the University of Maryland implementation of the Nonlin planner and the more recent UMCP HTN planner.

O-Plan domains

O-Plan is an agenda-based planner, developed at the University of Edinburgh. This page has information about the planner and using it. To see some domains used by O-Plan, the standard O-Plan examples document describes the domains and the tasks and provides pointers to the TF (task formalism) model descriptions. (To obtain solutions to the sample problems, O-Plan can be invoked from the web page.)

SIPE-2 Domains

SIPE-2 is a performance-oriented, general-purpose software system for generating and monitoring the execution of plans. It plans hierarchically, using different levels of abstraction, and provides a formalism for describing actions as operators. Given an arbitrary initial situation and a set of goals, SIPE-2 either automatically or under interactive control combines operators to generate plans to achieve the prescribed goals in the given world. In contrast to most AI planning research, heuristic adequacy (efficiency) has been one of the primary goals in the design of SIPE-2, which includes many heuristics for reducing computational complexity. Unlike expert systems, the SIPE-2 architecture is capable of generating a novel sequence of actions that responds precisely to the situation at hand. More information about SIPE-2 and some example domains can be found on the SIPE web page.

Domains specified in English or other natural language.

UM Translog Domain Specification

The UM Translog Domain was inspired by the CMU Transport Logistics domain developed by Manuela Veloso. UM Translog is an order of magnitude larger in size (41 actions versus 6), number of features and types interactions. It provides a rich set of entities, attributes, actions and conditions, which can be used to specify rather complex planning problems with a variety of plan interactions. The detailed specification provides for long plans (~40 steps) with many possible solutions to the same problem, and thus this domain can also be used to evaluate the solution quality of planning systems.

Dynamic domains, simulators, etc.

The Tileworld

The Tileworld is an abstract testbed system designed to support experimentation with agent architectures in dynamic and unpredictable environments. The system includes a simulated environment, an embedded agent, and a set of routines to facilitate experimentation. The environment is a two-dimensional grid on which are located different kinds of objects, notably tiles, holes, obstacles and a ``gas station.'' Exogenous events can occur in the TileWorld: specifically, objects can appear and disappear during a simulation. The experimenter can control of variety of characteristics associated with the objects in the environment, such as the average rate at which they appear or disappear.


There are a number of pages on the web with information about specific planners and various aspects of planning research. This repository is aimed specifically at domain specifications for comparing planning systems, not for information about systems themselves. A good resource for the latter is the AI Planning Resources Page .