Modeling and Planning Robotic Manufacturing.
Abstract
The project has been primarily directed toward design and implementation of a basic planning system called EXPLAN. EXPLAN is a general-purpose system that uses domain-specific knowledge to guide its choices at every step. EXPLAN was debugged using and extended blocks world domain, which is a simplification of typical robot pick-and-place operations. EXPLAN is based on a new knowledge representation that associates a symbolic time interval and a truth value of true or false with every proposition about the world. As it builds a plan, EXPLAN uses these truth values and time intervals to discover all harmful interactions that could occur between parallel actions and to order these actions to remove these potential conflicts. EXPLAN uses this explicit representation of truth values and time intervals to describe preconditions for actions, as well as the actions themselves. This allows EXPLAN to reason efficiently about temporal relationships that occur during planning. This is a different approach from that used with state-change operator representations, which are the basis of many other artificial intelligence (A1) planning systems. These earlier planners must use protection mechanisms and fine-grained operators to represent information about actions and conditions taking place over time. These indirect mechanisms make the most frequently posed questions about temporal relationships difficult to answer efficiently. Keyword: Automatic planning.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1985
- Accession Number
- ADA161014
Entities
People
- Peter C. Cheeseman
- William T. Park
Organizations
- SRI International