1
Department of Computer Science,
University of Maryland,
College Park, MD 20742, USA,
2
MINDSWAP,
University of Maryland,
8400 Baltimore Ave., College Park, MD 20742, USA
Hierarchical Task-Network (HTN) based planning techniques have been applied to the problem of composing Web Services, especially when described using the OWL-S service ontologies. Many of the existing Web Services are either exclusively information providing or crucially depend on information-providing services. Thus, many interesting service compositions involve collecting information either during execution or during the composition process itself. In this paper, we focus on the latter issue. In particular, we present ENQUIRER, an HTN-planning algorithm designed for planning domains in which the information about the initial state of the world may not be complete, but it is discoverable through plan-time information-gathering queries. We have shown that ENQUIRER is sound and complete, and derived several mathematical relationships among the amount of available information, the likelihood of the planner finding a plan, and the quality of the plan found. We have confirmed our theoretical results by an experimental evaluation. We have also performed experimental tests that demonstrated how our approach can be employed for the problem of Web Service composition.
In Proceedings of the Third International Semantic Web Conference 2004 (ISWC2004), 7-11 November 2004, Hiroshima, Japan.