Plan Debugging Using Approximate Domain Theories.
Abstract
The overall goal of this project was to investigate the idea that planning in the presence of uncertain information could be easier, not harder, than planning in its absence. The basic reason proposed was that, since it is impossible to control planning search with precision, uncertain domain information had a potential for use in this area that more accurate information might lack. This suggestion turned out to be correct. During the course of the project, we developed compelling theoretical arguments showing that reasoning can be controlled more generally by using uncertain domain information than by using existing techniques such as hierarchy. Doing so, however, involves surmounting substantial practical difficulties in terms of both the underlying planning technology and our ability to manipulate uncertain information itself. This report describes these theoretical arguments and the problems (and potential solutions) that an implementation exploiting our ideas must face. (AN)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1995
- Accession Number
- ADA294862
Entities
People
- Matthew L. Ginsberg
Organizations
- Stanford University