Modeling Domain Independent Planning to Advance Application: Data

This project is sponsored by National Science Foundation under grant # IIS-0138690. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.

Available Data

This project focuses on the evaluation of planners on a wide range of problems. Expediting such a study requires access to both planners and problems.

To aid others with similar goals, we have collected together information on all of the problems and planners we have used or even considered using. Everything listed is freely available at other sites (which are listed as well in our repository); however, this site provides a single collection point.

• The Catalog of Planning Resources (CPR) contains the list of planners and problems we have examined for BUS inclusion. It attempts to situate in a single document the disparate collections of planners and benchmark problems along with their WWW homepages and other written material (conference proceedings, competition descriptions, journal papers, technical reports, etc.). Older versions of the CPR documents can be found in Planners, which describes classical planners in terms of their capabilities with information on the references and participation in planning competitions, and Planner components, which organizes the different modules (e.g., pre-processing or translation) that might be included in planners and references which planners do what.
• The Planning problem repository collects together benchmark problem sets from various sources
• Planning Data is the July 27,2006 copy of data for running 25 planners on 4297 problems to that date. This is the data used for the most recent publications of 2006. The tarball includes a single "all" file which is the cross-product of all planners and problems. There are a subset of 580 problems in the datafile called "medianOverOneSecond" whose median run-times across all planners were over one second. Finally, the "medianOverOneSecond" file is further split by planner name into 25 individual files. All data files are in the arff format of the Weka Machine Learning Package; ARFF format is easily converted CSV format by removing the file header before the "@data" line.
• We are constantly adding planners and problems and updating the entire suite of data. Please send us an email if you are interested in the most recent collection before our next publication.