Knowledge-Based Logistics Planning: Its Application in Manufacturing and Logistics Planning.
Abstract
This document summarizes research in CORTES, a project in constraint-based planning, scheduling, and control for complex large-scale domains such as military transportation and manufacturing. The CORTES approach keeps the planning/scheduling combinatories in check, using quantitative problem space metrics called texture measures to: (1) identify critical decisions that require early attention; and (2) steer search toward promising solutions. This basic approach has been applied and validated in several key contexts, including (1) micro-opportunistic search, which focuses on efficient generation and dynamic maintenance of complex large-scale Just-In-Time schedules; (2) simulated annealing search, where texture measures have been developed to focus search and learn to recognize (un)promising runs; (3) iterative constraint posting, which combines flexible schedule representation and dynamic identification of conflicts requiring further arbitration; (4) integration of predictive planning scheduling and execution control, where texture measures and flexible schedule representations are combined to coordinate multiple planning/scheduling and control agents; and (5) Interactive schedule repair, where adaptive similarity metrics direct re-use of previous repair histories and help select repair focus and actions. Visualization issues have been addressed by developing comprehensive (see reverse) languages for characterizing: (1) domain concepts to be displayed; (2) user analysis tasks; and (3) graphical presentation techniques that can be assembled to create displays. (MM)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1994
- Accession Number
- ADA291617
Entities
People
- Katia Sycara
- Nicola Muscettola
- Norman Sadeh
- Steve Roth
Organizations
- Carnegie Mellon University