Utilizing Problem Structure in Planning : A Local Search Approach

by Hoffmann, Jorg

• ISBN13: 9783540202592
• ISBN10: 3540202595
• Format: Paperback
• Copyright: 2004-01-01
• Publisher: Springer Verlag

Summary

Planning is a crucial skill for any autonomous agent, be it a physically embedded agent, such as a robot, or a purely simulated software agent. For this reason, planning, as a central research area of artificial intelligence from its beginnings, has gained even more attention and importance recently. After giving a general introduction to AI planning, the book describes and carefully evaluates the algorithmic techniques used in fast-forward planning systems (FF), demonstrating their excellent performance in many wellknown benchmark domains. In advance, an original and detailed investigation identifies the main patterns of structure which cause the performance of FF, categorizing planning domains in a taxonomy of different classes with respect to their aptitude for being solved by heuristic approaches, such as FF. As shown, the majority of the planning benchmark domains lie in classes which are easy to solve.

Table of Contents

Planning: Motivation, Definitions, Methodology
Introduction	p. 3
Planning	p. 11
A Local Search Approach
Base Architecture	p. 35
Dead Ends	p. 75
Goal Orderings	p. 89
The AIPS-2000 Competition	p. 107
Local Search Topology
Gathering Insights	p. 115
Verifying the h[superscript +] Hypotheses	p. 135
Supporting the h[superscript FF] Hypotheses	p. 181
Discussion	p. 199
Formalized Benchmark Domains	p. 203
Automated Instance Generation	p. 235
References	p. 243
Index	p. 249
Table of Contents provided by Blackwell. All Rights Reserved.