PLANET
|
As we develop larger and more complex intelligent systems in
knowledge-intensive domains, it becomes impractical to develop knowledge
bases from scratch. Recent research investigates how to develop
intelligent systems by drawing from libraries of reusable components
that include both ontologies and problem-solving methods. Planet is a
reusable ontology for representing plans that is designed to accomodate
a diverse range of real-world plans, both manually and automatically
created. We have drawn from our past experience in designing,
developing and integrating planning tools, and expect planet to ease
these tasks in the future in three ways. First, we have already found
it useful for knowledge modelling. By providing a structure that
formalizes useful distinctions for reasoning about states and actions, a
knowledge engineer can find the semantics of informal expressions of
plans (e.g., textual or domain-specific) through designing mappings to
the ontology. Second, a plan ontology can be a central vehicle for
knowledge reuse across planning applications. Planet contains general,
domain-independent definitions that are common and useful across
planning domains. Third, Planet can facilitate integration of
planning tools through knowledge sharing.
Currently, practical efforts to integrate planning tools are done by designing separate interchange formats for (almost) each pair of tools, since designing a more universal format is costly and often more difficult than designing the entire set of pairwise formats. These difficulties are in part because these systems include decision-support tools such as plan editors, plan evaluation tools, and plan critiquers, which represent plans in ways that are different from traditional AI plan generation systems. Planet can provide a shared plan representation for systems to communicate and exhange information about the plan, and can facilitate the creation of a common, overarching knowledge base for future integrations of planning tools. Planet is available in Loom, KIF, and CycL. Send email to Jim Blythe ([email protected]) or Yolanda Gil ([email protected]) if you would be interested in seeing the latest version. Planet is still under development, and we would appreciate your comments. We developed Planet as part of our work in the DARPA High Performance Knowledge Bases (HPKB) program. Planet is a middle-level theory that represents planning knowledge. We worked with personnel from Cycorp to integrate Planet into CYC, which resulted in an extension to the CYC IKB. A short note on the integration work is available as an internal project report here. We used Planet to develop a system for Course of Action (COA) Critiquing for the DARPA HPKB Battlespace Challenge Problem. We have demonstrated using Planet in this approach to integration in our work on the Rosetta project, an ISO Technology Integration Experiment between High Performance Knowledge Bases and Control of Agent-Based Systems exploring ontology-based agent communication. We also use Planet in our work on a problem-solving method for plan evaluation . HPKB project members can access the ontology here in Loom, CycL and KIF (password protected). There is more information in the following reports:
Contact: Jim Blythe |