OntheInstantiationofADLOp eratorsInvolvingArbitraryFirst-OrderFormulasJanaKo ehlerSchindlerLiftsLtd.CH-6031EbikonSwitzerlandjanako ehler@ch.schindler.comJorgHo mannInstituteforComputerScienceAlb ertLudwigsUniversity79110Freiburg,Germanyho mann@informatik.uni-freiburg.deAbstractThegenerationofthesetallgroundactionsforagivensetofADLop erators,whichareallowedtohaveconditionale ectsandpreconditionsthatcanb erepre-sentedusingarbitrary rst-orderformulasisacomplexpro cesswhichheavilyinuencesthep erformanceofanyplannerorpre-planninganalysismetho d.Thepap erdescrib esasophisticatedinstantiationpro-cedurethatdeterminesso-calledinertiainagivenproblemrepresentationandusesthemtop erformsim-pli cationsofformulasduringtheinstantiationpro-cess.Asaresult,manyinapplicableactionsarede-tectedandruledoutfromthedomainrepresentationyieldingamuchsmallersearcspacefortheplanner.Intro ductionAplanningsystemthathandlesamoreexpressivelan-guagethanSTRIPSrequiressophisticatedalgorithmicsolutionstoquiteanumb erofproblems,whichhavenothingtodowiththeactualsearchpro cessforaplan.Oneoftheseproblemsconcernsthecomputationsetofactionsasallgroundinstancesagivenop erators.Theaimoftheinstantiationpro cessistogenerateallthosegroundinstancesoftheplanningop eratorsthatareapplicableinsomelegalworldstate.Thismeans,thatthepreconditionofop eratorshouldb esatis ableanditse ectsshouldconsistent.Ononehand,anaiveinstantiationpro cedurethatsimplyexpandslogicalquanti ersandenumeratesallp ossibleinstantiationsofop eratorparameterswillquicklyren-derevensimpleplanningproblemsunsolvable.Ontheotherhand,arathersophisticatedinstantiationpro ce-durecanruleoutmanyactions,whichwillneverb eapplicableinanyreachableworldstateorthatould|ifapplied|yieldaninconsistentstate.Itshouldalsoreturnthemostsimplesyntacticrepresentationofpre-conditionsande ects.Manyplanningsystemsdogeneratethecompletesetofactionsb eforeplanningactuallystarts.Theyusethisseteitherfortheenco dingofdomaininotherCopyrightc2000,AmericanAsso ciationforArti cialIn-telligence(www.aaai.org).Allrightsreserved.representationformalismssuchasSAT(Kautz&Sel-man1996)orforthederivationofusefulinformationthatcanhelpduringplanning,e.g.,distanceheuristics(Ge ner1999),symmetries(Fox&Longrelevantactions(Neb el,Dimop oulos,&Ko ehler1997),andgoalorderings(Ko ehler1998).WhenusingthePDDLlanguage(McDermott&oth-ers1998)torepresentADLop erators(Pednault1989),quitecomplexdescriptionsofpreconditionsande ectsarep ossible:arbitraryfunction-symb olfree rst-orderlogicformu-lasrepresentpreconditions,conditionale ectshaveform(whenantecedentconsequent)wheretheantecedentcanb earbitrarypreconditionandtheconsequentisaconjunctionofliterals,i.e.,anatomeithero ccursp ositivelyorneg-ativelyinit.conditionale ectcanalsob euniver-sallyquanti ed.Givensuchanop erator,theinstantiationhastore-placeallo ccurringvariables,whichareeitherquanti edoro ccurasparametersoftheop erator,bythosetyp e-consistentconstants,whichhaveb eendeclaredintheplanningproblem.Inordertoreplaceallvariablesbyconstants,theinstantiationpro cesspro ceedsinthreephases:1.Theexpansionofuniversalandexistentialquanti erso ccurringinthe rst-orderformulasrepresentingpre-conditionsorantecedentsofconditionale ectselim-inatesmostofthequanti edvariables,2.Theexpansionuniversallyquanti edconditionale ectseliminatesremainingquanti edvariables,3.theinstantiationofop eratorparameterseliminatesthevariableparameters.1Ineachphase,thefollowingatomicinstantiationtasko ccurs:Givenavariable?x,constantandanatomicformulap,determineresultinginstantiationp[?x=c].1InIPPtheassumptionismadethatdi erentop eratorparametersareinstantiatedwithdi erenconstants,i.e.,theplannernevergeneratesactionslikemove(a,a).
[1]
Jana Koehler,et al.
Solving Complex Planning Tasks Through Extraction of Subproblems
,
1998,
AIPS.
[2]
Craig A. Knoblock,et al.
PDDL-the planning domain definition language
,
1998
.
[3]
Jorg Homann.
A Heuristic for Domain Independent Planning and Its Use in an Enforced Hill-Climbing Algorithm
,
2000
.
[4]
Jj Org Hoomann.
A Heuristic for Domain Independent Planning and its Use in an Enforced Hill-climbing Algorithm
,
2000
.
[5]
Maria Fox,et al.
The Detection and Exploitation of Symmetry in Planning Problems
,
1999,
IJCAI.
[6]
Bernhard Nebel,et al.
Extending Planning Graphs to an ADL Subset
,
1997,
ECP.
[7]
Bernhard Nebel,et al.
Ignoring Irrelevant Facts and Operators in Plan Generation
,
1997,
ECP.
[8]
Edwin P. D. Pednault,et al.
ADL: Exploring the Middle Ground Between STRIPS and the Situation Calculus
,
1989,
KR.
[9]
P. Pandurang Nayak,et al.
A Reactive Planner for a Model-based Executive
,
1997,
IJCAI.
[10]
Jana Koehler,et al.
Handling of Inertia in a Planning System
,
1999
.
[11]
Bart Selman,et al.
Pushing the Envelope: Planning, Propositional Logic and Stochastic Search
,
1996,
AAAI/IAAI, Vol. 2.
[12]
Jana Koehler,et al.
Elevator Control as a Planning Problem
,
2000,
AIPS.