Artificial Intelligence
BIDA: an improved perimeter search algorithm
Artificial Intelligence
Disjoint pattern database heuristics
Artificial Intelligence - Chips challenging champions: games, computers and Artificial Intelligence
Ranking and unranking permutations in linear time
Information Processing Letters
Maximizing over multiple pattern databases speeds up heuristic search
Artificial Intelligence
Additive pattern database heuristics
Journal of Artificial Intelligence Research
Journal of Artificial Intelligence Research
Recent progress in heuristic search: a case study of the four-peg towers of Hanoi problem
IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence
Dual lookups in pattern databases
IJCAI'05 Proceedings of the 19th international joint conference on Artificial intelligence
Combining perimeter search and pattern database abstractions
SARA'07 Proceedings of the 7th International conference on Abstraction, reformulation, and approximation
Symbolic merge-and-shrink for cost-optimal planning
IJCAI'13 Proceedings of the Twenty-Third international joint conference on Artificial Intelligence
Hi-index | 0.00 |
Pattern Databases were a major breakthrough in heuristic search by solving hard combinatorial problems various orders of magnitude faster than state-of-the-art techniques at that time. Since then, they have received a lot of attention. Moreover, pattern databases are also researched in conjunction with other domain-independent techniques for solving planning tasks. However, they are not the only technique for improving heuristic estimates. Although more modest, perimeter search can also lead to significant improvements in the number of generated nodes and overall running time. Therefore, whether they can be combined or not is a natural and interesting issue. While other researchers have recently proven that a joint application of both ideas (termed as multiple goal) leads to no progress at all, it is shown here that there are other alternatives for putting both techniques together---denoted here as multi-valued. This paper shows that multi-valued pattern databases can still improve the performance of standard (or single-valued) pattern databases in practice. It also examines how to enhance memory usage when comparing multi-valued pattern databases in contraposition to various single-valued standard pattern databases.