Resource-constrained project scheduling_ Notation, classific(13)

项目进度管理

P.Bruckeretal./EuropeanJournalofOperationalResearch112(1999)3±4115

gorithmgproducesarealizationpwith p T andgorithmmax p TisdO jopt j3 elog.Thej runningtimeoftheal-makespanTheNP-completenessj .

showsof2canberealizedofdecidingwithwhetheraimprovedthattheperformanceratioofagivenbudgetcannotbetionalForprojects(unlesswithP NPp).

jT ,Skutellausesstronglypartitioningtechniquesandobtainsaddi-awithdi erentaperformancepolynomialvariantyieldsguaranteeapproximationalgorithmof2 log2 1 .AconsistsAnotherlog2 3.

getinrelaxingideaforapproximationalgorithmsbicriteriaordeadline),alsothetightconstraint(bud-similarapproximationthusarrivingattheso-calledvalue polynomialY (i.e.,0to`thosel` 1above, ingtime±costthat,ideasforpairaopt and fopt l ptimeconstructarealization ),ponesuchcanthatintwice ` 1a 1Àl andgmax p T 1al .For,realizationasthisexpensiveyieldsaandrealizationtwiceaslongwhichasanisatoptimalmostChoosing lforthegivendeadlineorbudget.tios1aeYexpectedof1 eoneamakespan. eÀobtainsuniformlyatrandomintheinterval1 %1Ximproved58fortheapproximationexpectedcostandra-5.Multi-modecase

rithmsThis®nedforsectionsolvingcoversw exactjpre andjgheuristicalgo-maxwhichisprojectaspologicallynetworkfollows.isLikeassumedinthede-toprevioussectionstheexecutedeachinsorted.EachactivitybeacyclicjP mayandto-beonceactivityonetivityselectedmayoutmaynotofasetofMjmodes.Also,notbebepreemptedchanged,thatandis,aanmodeac-completedjoncecessingisactivityinstartedmodeinmodemPMjhastobejinmodemwithoutmtakesinterruption.pPro-jmperiodsofsupportedbyasetRqofrenewableandasetandRm span,,nonrenewablethatis,anupperresources.boundonConsideringtheahorizonrenewablewehaveresourceanavailablekinperiodamountproject'st 1ofYFF qmake-FkY unitslikeofin

Sectionableprocessedresource3.ThekoverallPRmiscapacitygivenbyofthenonrenew-q

mk.Ifactivityjisnewableisunitsinprocess.resourceinmodemthenrjkmunitsofthere-Similarly,kareusedactivityeachperiodactivitym

j

theofthenonrenewableresourcejconsumesk.Ingeneral,rjkmWeparametersnondecreasingassumethemodesareassumedtobetobeinteger-valued.pprocessinglabeledtimes,withrespectthattois,mjmTpjYm 1forallactivitiesjP andmodesmakespan-minimalPf1YFFFYjMjjÀ1gconstraintsschedule.TheobjectiveSisto®ndaand formulatedjpre bylimitedimposedjgresourcebytheavailabilities.precedencethatmeetsSimilarrelationsthetomaxingeneralalsow optimizationmathematicallyintermsjpre ofajgbinarymaxisvariablesmodexmodelwhichmakesuseofbinaryjmt 1,ifshortness,minperiodt(0,activityotherwise).jisForcompletedinfeasibleIfjRmwedonotpresentaformalmodelthesakehere.ofever,solutionjP2andisjNP-completeMjjP2YjP (cf.Ythen[109]).®ndingariodupperavailabilitypresumingfeasibilityoftherenewableandaconstantper-pe-How-bytimes.theboundontheminimummakespanresources, isgivenanprecedenceGivensumofantheuppermaximumboundactivity weprocessingprocessingrelationstervalstimesandthemodesofcantheshortestusetheplained inigtoderivetimewindows,i.e.in-SectionjYvgj similar3.towhathasbeenex-5.1.Exactalgorithms

w Optimalinjpre jgproceduresforsolvingmaxgeneralizeproceduresdescribed jSectionpre jg3.1forsolvingthespecialcaseimprovedThePrecedencemax.WeTree:summarizeSprecherthreeandalgorithms.Drexl[186]ducedboundingbyPattersontheprecedenceetal.[155]treebyalgorithmintro-and-boundcriteria.quently,EachamodetreeanHere,onlevelgofincludingthebranch-newmeligibleactivityjgand,subse-jgrelatedcombinationofofanthiseligibleactivityactivityareselected.andacurrentmodenodeincorrespondsthebranch-and-boundtoadescendanttree.

ofthe