12.02.2014 Views

Disjoint Pattern Database Heuristics

Disjoint Pattern Database Heuristics

Disjoint Pattern Database Heuristics

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

eratedfewernodesandincurredlessoverheadpernode.FortheTwenty-Four Puzzle,the2-and3-tiledatabasesusuallygeneratedfewernodesthanthe compute,itincurredalargerconstanttimepernode,andthustheactual disjointdatabase.Again,however,sincethisheuristicismorecomplexto<br />

foundin[2]. databaseheuristicsarebothsimplerandperformbetter,thedetailsofour experimentswiththe2-and3-tiledatabasesareomittedhere,butcanbe runningtimewasgreaterthanforthedisjointdatabases.Sincethedisjoint<br />

and3-tiledatabaseswasgreaterfortheFifteenPuzzlethantheTwenty- asingledatabase,butonlyaquarteroftheTwenty-FourPuzzletiles.For FourPuzzle,probablybecausewecouldstorehalftheFifteenPuzzletilesin Theperformancedierencebetweenthedisjointdatabasesandthe2-<br />

thesetwoproblems,thedisjointdatabaseheuristicsarebothsimplerand moreeective,buttheymaynotbeforlargerversionsoftheproblemor otherdomains.<br />

WehavefoundoptimalsolutionstoftyrandominstancesoftheTwenty- FourPuzzle,aproblemwithalmost1025states.Thebranchingfactorofthe 6 Summary,Conclusions,andFurtherWork<br />

problemis2.3676,andoptimalsolutionsaverageabout100moves.Wealso Thisisbyfarthebestperformanceontheseproblemstodate. ndoptimalsolutionstotheFifteenPuzzlein27millisecondsonaverage. functions,basedonpatterndatabases[1].Ratherthancomputingthecosts ofsolvingindividualsubgoalsindependently,apatterndatabaseheuristic considersthecostsofsolvingseveralsubgoalssimultaneously,takinginto Toachievethis,weimplementedIDA*withnewadmissibleheuristic<br />

admissibleheuristicsisalwaysanotheradmissibleheuristic. values.Thisisthemostgeneralapproach,sincethemaximumofanytwo accounttheinteractionsbetweenthem.CulbersonandSchaeer[1]combined heuristicsfromdierentpatterndatabasesbytakingthemaximumoftheir<br />

values.Adisjointpatterndatabasepartitionsthesetofsubgoalsintodisjoint ferentdatabasestobeaddedtogether,resultinginmoreaccurateheuristic groups,andthenaddstogetherthecostsofsolvingallthesubgoalsineach Weintroduceddisjointpatterndatabasestopermitthevaluesfromdif-<br />

group.Thisrequiresthatthegroupsbedisjoint,andthatasingleoperator 16

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!