AnovelstrategyforefficientnegotiationincomplexenvironmentsSiqiChenandGerhardWeissDepartmentofKnowledgeEngineering,MaastrichtUniversityMaastricht,TheNetherlands{siqi.chen,gerhard.weiss}@maastrichtuniversity.nlAbstract.Acomplexandchallengingbilateralnegotiationenvironmentforrationalautonomousagentsiswhereagentsnegotiatemulti-issuecon-tractsinunknownapplicationdomainsagainstunknownopponentsun-derreal-timeconstraints.InthispaperwepresentanovelnegotiationstrategycalledEMARforthiskindofenvironmentwhichisbasedonacombinationofEmpiricalModeDecomposition(EMD)andAutoregres-siveMovingAverage(ARMA).EMARenablesanegotiatingagenttoadjustitstargetutilityandconcessionrateadaptivelyinreal-timeac-cordingtothebehaviorofitsopponent.TheexperimentalresultsshowthatthisnewstrategyoutperformsthebestagentsfromthelatestAu-tomatedNegotiationAgents(ANAC)Competitioninawiderangeofapplicationdomains.1IntroductionAutomatednegotiationhasabroadspectrumofpotentialapplicationsindo-mainsandfieldssuchastaskandserviceallocation,webandgrid,electroniccommerceandelectronicmarkets,onlineinformationmarkets,andautomatedprocurement.Thispotentialhasledtorapidlyincreasingresearcheffortsonau-tomatednegotiationinrecentyears.Theworkdescribedinthispaperfocusesonautomatedbilateralmulti-issuenegotiation(e.g.,[16]).Akeyfeatureofthisnegotiationformisthattwoagentsnegotiatewiththeintentiontoagreeonaprofitablecontractforaproductorservice,wherethecontractconsistsofmul-tipleissueswhichareofconflictiveimportanceforthenegotiators.Examplesofsuchissuesarepriceandquality.Morespecifically,thepaperconcentratesonrealisticscenariosforbilateralmulti-issuenegotiationswhichareparticularlycomplexforthefollowingfourreasons.First,thenegotiatingagentsdonotknoweachother(i.e.,theyhavenotencounteredbefore)andthushavenoinforma-tionaboutthepreferencesorstrategiesoftheirrespectiveopponents.Second,thenegotiatorshavenopriorknowledgeaboutthenegotiationdomain(e.g.,aboutresourcelimitations)andthushavetocopewithuncertaintyaboutthedomain.Third,weconcentr...
