>> General Information

Conformant-FF is a domain independent planning system developed by Joerg Hoffmann, in collaboration with Ronen Brafman. The system extends the classical FF planner with the ability to treat initial state uncertainty expressed in the form of a CNF formula. Contingent-FF is an extension to further treat partial observability (observation actions), finding tree-shaped plans with branches; the system also includes a preliminary treatment of a simple form of non-deterministic effects. Probabilistic-FF is a vastly modified version, developed in collaboration with Carmel Domshlak, that deals with probabilistic (Bayes Net) initial states as well as probabilistic effects (assigning probabilities to a list of possible outcomes).

>> Source Code Available

Contrary to my previous policy, I now also release source code for all of Conformant-FF, Contingent-FF, and Probabilistic-FF. Remarks:

• The Contingent-FF version here is a revised one we used in our ICAPS'16 work on undoability checking. It features: a cleaner AO* implementation; an additional simple heuristic function which just assumes that all unknown facts are true; a command line option to minimize the average cost of the plan tree; the option to add a give-up action, that allows the planner to achieve the goal in any state, at a give-up cost specified on the command line.

• The Probabilistic-FF version is the latest one, discussed in our JAIR article. It can deal with probabilistic effects, but it does not fully reproduce the results of the ICAPS version on domains without such effects.

DISCLAIMER: PROBABILISTIC-FF, AND IN SOME CONFIGURATIONS ALSO CONFORMANT-FF AND CONTINGENT-FF, MAKE USE OF ADDITIONAL PROGRAMS AND HENCE MAY NOT RUN OUT-OF-THE-BOX AS EASILY AS FF AND METRIC-FF DO. SINCE I AM LONG SINCE WORKING ON DIFFERENT STUFF, THERE WILL BE NO TECHNICAL SUPPORT FROM MY SIDE SO YOU MUST RESOLVE ANY DIFFICULTIES (INCLUDING ALSO THE PREVAILING DIFFICULTIES WITH THE BISON/FLEX PARSER VERSIONING) YOURSELF.

>> Executables Available

Here is a Linux executable of Conformant-FF. Here is a Linux executable of Contingent-FF. Here is a Linux executable of Probabilistic-FF.

Here is another executable of Conformant-FF, for use in the context of a project performed in the KnowledgeWeb framework. Via a PDDL translator, Conformant-FF can be used to compose functional level web service descriptions. Here, initial state uncertainty is used to model uncertainty about the precise form of input objects; in other words, uncertainty can be used to model certain kinds of ontologies. The executable differs from the one above in a number of minor optimizations for this particular task, and in that it outputs the plan to a file, for interfacing to the rest of the composition machinery.

Here are some benchmark problems for Conformant-FF, illustrating also the input syntax. Here are benchmark problems for Contingent-FF. Here are some benchmark problems for Probabilistic-FF.

>> Relevant Papers

• A conference paper on Conformant-FF is published at ICAPS'04, receiving the best paper award.

  R. Brafman, J. Hoffmann, Conformant Planning via Heuristic Forward Search: A New Approach, in: Proceedings of the 14th International Conference on Automated Planning and Scheduling, Whistler, Canada, June 2004. (gzip'ed postscript file) (bib entry)

• A long paper on Conformant-FF is published in Artificial Intelligence.

  J. Hoffmann, R. Brafman, Conformant Planning via Heuristic Forward Search: A New Approach, Artificial Intelligence, Volume 170 (6-7), 2006, pages 507 - 541. (pdf file) (bib entry)

• A conference paper on Contingent-FF is published at ICAPS'05.

  J. Hoffmann, R. Brafman Contingent Planning via Heuristic Forward Search with Implicit Belief States, accepted for: Proceedings of the 15th International Conference on Automated Planning and Scheduling, Monterey, CA, USA, June 2005. (gzip'ed postscript file) (bib entry)

• A conference paper on Probabilistic-FF is published at ICAPS'06.

  C. Domshlak and J. Hoffmann, Fast Probabilistic Planning Through Weighted Model Counting, Proceedings of the 16th International Conference on Automated Planning and Scheduling (ICAPS'06), The English Lake District, UK, June 2006. (pdf file) (bib entry)

• A long paper on Probabilistic-FF is published in JAIR.

  C. Domshlak and J. Hoffmann, Probabilistic Planning via Heuristic Forward Search and Weighted Model Counting, Journal of Artificial Intelligence Research, Volume 30, 2007, pages 565 - 620. (gzip'ed postscript file) (bib entry)