Alfonso Gerevini
University of Brescia, Italy

Lenhart K. Schubert
University of Rochester, USA

General overview

DISCOPLAN (DIScovering State COnstraints for PLANning) is a collection of techniques for inferring state constraints (or invariants) and operator-parameter domains by inspecting the specification of a given planning problem.
A state constraint is a law that holds in every state that can be reached from the initial state by applying the operators of the domain. An operator-parameter domain is a (possibly empty) set including all the objects that can be used to instantiate the parameter.
State constraints and parameter domains can be used to speed up domain-independent planners by reducing their search space, as well as an aid to domain specification and debugging for knowledge engineers.
The input to DISCOPLAN can be either a UCPOP-like or PDDL-like specification of a planning problem. The output state constraints can be in either DISCOPLAN compact form or first order logic (F.O.L.)

Remote use of DISCOPLAN

DISCOPLAN on-line is a version of DISCOPLAN that can be remotely used through any web browser like Netscape or Microsoft Explorer. For the on-line use of the syststem please visit our demo page of DISCOPLAN 1.0 or of the newer version DISCOPLAN 2.0.

Available papers

A. Gerevini, Inferring State Constraints as Control Knowledge for Domain-Independent Planning (extended abstract) Proceedings of the Workshop on Planning as Combinatorial Search, AIPS-1998
A. Gerevini, L. K. Schubert, Discovering State Constraints in DISCOPLAN: Some New Results, Proceedings of the Seventeenth National Conference on Artificial Intelligence (AAAI-2000)
A. Gerevini, L. K. Schubert, Extending the Types of State Constraints Discovered by DISCOPLAN , Workshop on Analyzing and Exploiting Domain Knowledge for Efficent Planning, AIPS-2000
A. Gerevini, L. K. Schubert, Inferring Sate Constraints for Domain-Independent Planning, Proceedings of the Fifteenth National Conference on Artificial Intelligence (AAAI-1998)
A. Gerevini, L. K. Schubert, Computing Parameter Domains as an Aid to Planning, Proceedings of the Third International Conference on Artificial Intelligence Planning Systems (AIPS-1996)
A. Gerevini, L. K. Schubert, Accelerating Partial Orders Planners: Some Techniques for Effective Search Control and Pruning, Journal of Artificial Intelligence Research - JAIR, 5, 1996
A. Gerevini, L.K. Schubert, Discovering State Constaints for Planning: DISCOPLAN, Technical Report, forthcoming.

Software download

For the (Lisp) source code of Discoplan please contact Alfonso Gerevini
Domains and problems used in the Demo Page
PDDL2UCPOP & DISCO2FOL compilers

All the programs are written in Ansi Common Lisp and have been tested under Allegro Common Lisp 4.3 (Solaris & Linux), 5.0.1 (Linux)

Credits

PDDL2UCPOP and DISCO2FOL have been written by Devis Bianchini, Fabrizio Morbini and Maria Panteghini
Web engineering support: Fabrizio Morbini

Further information

For further information on DISCOPLAN on-line please contact Alfonso Gerevini

A. Gerevini, July 1 2003