An Agent-Based Approach to Optimal Configuration Design with Application to Manufacturing Process Planning
Abstract
This project explored an agent-based strategy for configuration design. A bulk manufacturing process planning problem served as the focus, but the methods and algorithms are readily extendable to any serial configuration and parameter instantiation problem. The project focused on three goals: process instantiation, configuration optimization, and effective management of the agent process. To meet the first goal, a probabilistic approach was used to determine which agents to select to instantiate a configuration through parameter selection. To meet the second goal, the parameter optimization algorithm was modified to select and optimize sequences as well as parameters. The advantage of this approach is that it is open-ended; there is no preconceived assumption of what sequence might be valid or optimal. Configuration agents were developed and integrated with the parameter agents to effectively optimize both process selection and parameters, concurrently. To meet the third goal, two approaches were taken. First, a strategy for learning across problems using findings in cognitive science was applied to the author's agent-based method. Second, using organizational models, a collaborative approach was formalized and implemented, allowing the agents to interact and make group decisions as to how to best solve the configuration and parameter optimization problem. A listing of 10 publications supported by this grant is included.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 26, 2003
- Accession Number
- ADA419532
Entities
People
- Jonathan Cagan
Organizations
- Carnegie Mellon University