Organizing Committee |
| Yehuda Naveh, IBM Research Lab in Haifa naveh@il.ibm.com |
| |
| Andrea Roli, Universita "G. D'Annunzio" a.roli@unich.it |
|
|
Program Committee |
| Christian Blum, UPC, Spain |
| Yehuda Naveh, IBM Research, Israel |
| Justin Pearson, Uppsala U, Sweden |
| Gilles Pesant, Polytechnique Montreal, Canada |
| Steve Prestwich, 4C, Ireland |
| Andrea Roli, U G. D'Annunzio, Italy |
| Meinolf Sellmann, Brown U, USA |
| Bart Selman, Cornell, USA |
| Thomas Stuetzle, TU Darmstadt, Germany |
| Pascal Van Hentenryck, Brown U, USA |
| Toby Walsh, NICTA & UNSW, Australia |
| |
|
|
LSCS 2005 Workshop
Second International Workshop on Local Search Techniques in Constraint Satisfaction
in conjunction with CP 2005
Utilizing Problem Structure in Planning: A Local Search Approach
Joerg Hoffmann (Max Planck Institute, Germany)
Recently, a major breakthrough was achieved in terms of the scalability of automated planning systems. The talk describes one of the systems, FF, that brought about this breakthrough, and explains the reasons for its success.
FF uses a form of local search. I describe the main algorithms used in the system. Then an investigation of the properties of FF's heuristic function identifies the main patterns of problem structure, common to most of the benchmarks, that cause the system's performance. The domains are categorized in a taxonomy of different classes with respect to their aptitude for solution using heuristic planners such as FF. It is shown that the majority of the benchmark domains lie in classes that are "easy" to solve.
| |
|
|