Using dynamic memory structures in planning and its application to manufacturing
Abstract
The work presented in this thesis covers the development of a methodology for case-based, expert domain planning that is based on episodic memories, is capable of clustering these memories with the help of constraints, failure prediction knowledge and premise importance values, creates a solution tree by adding memory chunks to the unfinished solution, adapts its memory preferences to the individual user and uses memories of failures in the form of TOPs to predict future failures and to give advise for correcting them. The use of episodic memory structures allows reorganization of knowledge according to new experiences and failures of predictions. An intelligent CAD module and feature extraction system that is integrated with the intelligent planner developed is also presented. The application of this intelligent planning system is demonstrated on numerous examples taken from the manufacturing domain of process planning.
Degree
Ph.D.
Advisors
Kashyap, Purdue University.
Subject Area
Electrical engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.