已阅读5页,还剩30页未读, 继续免费阅读
版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
Chapter18DeterministicDynamicProgramming toaccompanyOperationsResearch ApplicationsandAlgorithms4theditionbyWayneL Winston Copyright c 2004Brooks Cole adivisionofThomsonLearning Inc Description Dynamicprogrammingisatechniquethatcanbeusedtosolvemanyoptimizationproblems Inmostapplications dynamicprogrammingobtainssolutionsbyworkingbackwardfromtheendoftheproblemtowardthebeginning thusbreakingupalarge unwieldyproblemintoaseriesofsmaller moretractableproblems 18 1TwoPuzzlesExample Weshowhowworkingbackwardcanmakeaseeminglydifficultproblemalmosttrivialtosolve Supposethereare20matchesonatable Ibeginbypickingup1 2 or3matches Thenmyopponentmustpickup1 2 or3matches Wecontinueinthisfashionuntilthelastmatchispickedup Theplayerwhopicksupthelastmatchistheloser HowcanI thefirstplayer besureofwinningthegame IfIcanensurethatitwillbeopponent sturnwhen1matchremains Iwillcertainlywin Workingbackwardonestep ifIcanensurethatitwillbemyopponent sturnwhen5matchesremain Iwillwin IfIcanforcemyopponenttoplaywhen5 9 13 17 21 25 or29matchesremain Iamsureofvictory ThusIcannotloseifIpickup1matchonmyfirstturn 18 2ANetworkProblem Manyapplicationsofdynamicprogrammingreducetofindingtheshortest orlongest paththatjoinstwopointsinagivennetwork Forlargernetworksdynamicprogrammingismuchmoreefficientfordeterminingashortestpaththantheexplicitenumerationofallpaths CharacteristicsofDynamicProgrammingApplications Characteristic1Theproblemcanbedividedintostageswithadecisionrequiredateachstage Characteristic2Eachstagehasanumberofstatesassociatedwithit Byastate wemeantheinformationthatisneededatanystagetomakeanoptimaldecision Characteristic3Thedecisionchosenatanystagedescribeshowthestateatthecurrentstageistransformedintothestateatthenextstage Characteristic4Giventhecurrentstate theoptimaldecisionforeachoftheremainingstagesmustnotdependonpreviouslyreachedstatesorpreviouslychosendecisions Thisideaisknownastheprincipleofoptimality Characteristic5IfthestatesfortheproblemhavebeenclassifiedintoonofTstages theremustbearecursionthatrelatedthecostorrewardearnedduringstagest t 1 Ttothecostorrewardearnedfromstagest 1 t 2 T 18 3AnInventoryProblem Dynamicprogrammingcanbeusedtosolveaninventoryproblemwiththefollowingcharacteristics Timeisbrokenupintoperiods thepresentperiodbeingperiod1 thenextperiod2 andthefinalperiodT Atthebeginningofperiod1 thedemandduringeachperiodisknown Atthebeginningofeachperiod thefirmmustdeterminehowmanyunitsshouldbeproduced Productioncapacityduringeachperiodislimited Eachperiod sdemandmustbemetontimefrominventoryorcurrentproduction Duringanyperiodinwhichproductiontakesplace afixedcostofproductionaswellasavariableper unitcostisincurred Thefirmhaslimitedstoragecapacity Thisisreflectedbyalimitonend of periodinventory Aper unitholdingcostisincurredoneachperiod sendinginventory Thefirmsgoalistominimizethetotalcostofmeetingontimethedemandsforperiods1 2 T Inthismodel thefirm sinventorypositionisreviewedattheendofeachperiod andthentheproductiondecisionismade Suchamodeliscalledaperiodicreviewmodel Thismodelisincontrasttothecontinuousreviewmodelinwhichthefirmknowsitsinventorypositionatalltimesandmayplaceanorderorbeginproductionatanytime 18 4Resource AllocationProblems Resource allocationproblems inwhichlimitedresourcesmustbeallocatedamongseveralactivities areoftensolvedbydynamicprogramming Touselinearprogrammingtodoresourceallocation threeassumptionsmustbemade Assumption1 Theamountofaresourceassignedtoanactivitymaybeanynonnegativenumber Assumption2 Thebenefitobtainedfromeachactivityisproportionaltotheamountoftheresourceassignedtotheactivity Assumption3 Thebenefitobtainedfrommorethanoneactivityisthesumofthebenefitsobtainedfromtheindividualactivities Evenifassumptions1and2donothold dynamicprogrammingcanbeusedtosolveresource allocationproblemsefficientlywhenassumption3isvalidandwhentheamountoftheresourceallocatedtoeachactivityisamemberofafiniteset GeneralizedResourceAllocationProblem Theproblemofdeterminingtheallocationofresourcesthatmaximizestotalbenefitsubjecttothelimitedresourceavailabilitymaybewrittenaswherextmustbeamemberof 0 1 2 Tosolvethisbydynamicprogramming defineft d tobethemaximumbenefitthatcanbeobtainedfromactivitiest t 1 Tifdunitesoftheresourcemaybeallocatedtoactivitiest t 1 T WemaygeneralizetherecursionstothissituationbywritingfT 1 d 0foralldwherextmustbeanon negativeintegersatisfyinggt xt d ATurnpikeTheorem Turnpikeresultsaboundinthedynamicprogrammingliterature Whyaretheresultsreferredtoasaturnpiketheorem Thinkabouttakinganautomobiletripinwhichourgoalistominimizethetimeneededtocompletethetrip Foralongtripitmaybeadvantageoustogoslightlyoutofourwaysothatmostofthetripwillbespentonaturnpike onwhichwecantravelatthegreatestspeed 18 5EquipmentReplacementProblems Manycompaniesandcustomersfacetheproblemofdetermininghowlongamachineshouldbeutilizedbeforeitshouldbetradedinforanewone Problemsofthistypearecalledequipment replacementproblemsandcanbesolvedbydynamicprogramming AnequipmentreplacementmodelwasactuallyusedbyPhillipsPetroleumtoreducecostsassociatedwithmaintainingthecompany sstockoftrucks 18 6FormulatingDynamicProgrammingRecursions Inmanydynamicprogrammingproblems agivenstagesimplyconsistsofallthepossiblestatesthatthesystemcanoccupyatthatstage Ifthisisthecase thenthedynamicprogrammingrecursioncanoftenbewritteninthefollowingform Ft i min costduringstaget ft 1 newstageatstaget 1 wheretheminimumintheaboveequationisoveralldecisionsthatareallowable orfeasible whenthestateatstatetisi Correctformulationofarecursionoftheformrequiresthatweidentifythreeimportantaspectsoftheproblem Aspect1 Thesetofdecisionsthatisallowable orfeasible forthegivenstateandstage Aspect2 Wemustspecifyhowthecostduringthecurrenttimeperiods staget dependsonthevalueoft thecurrentstate andthedecisionchosenatstaget Aspect3 Wemustspecifyhowthestateatstaget 1dependsonthevalueoft thestatesatstaget andthedecisionchosenatstaget Notallrecursionsareoftheformshownbefore AFisheryExample Theownerofalakemustdecidehowmanybasstocatchandselleachyear Ifshesellsxbassduringyeart thenarevenuer x isearned Thecostofcatchingxbassduringayearisafunctionc x b ofthenumberofbasscaughtduringtheyearandofb thenumberofbassinthelakeatthebeginningoftheyear Ofcourse bassdoreproduce AFisheryExample Tomodelthis weassumethatthenumberofbassinthelakeatthebeginningofayearis20 morethanthenumberofbassleftinthelakeattheendofthepreviousyear Assumethatthereare10 000bassinthelakeatthebeginningofthefirstyear Developadynamicprogrammingrecursionthatcanbeusedtomaximizetheowner snetprofitoveraT yearhorizon AFisheryExample Inproblemswheredecisionsmustbemadeatseveralpointsintime thereisoftenatrade offofcurrentbenefitsagainstfuturebenefits AtthebeginningofyearT theownerofthelakeneednotworryabouttheeffectthatthecaptureofbasswillhaveonthefuturepopulationofthelake Sothebeginningoftheyearproblemisrelativelyeasytosolve Forthisreason welettimebethestage Ateachstage theownerofthelakemustdecidehowmanybasstocatch AFisheryExample Wedefinexttobethenumberofbasscaughtduringyeart Todetermineanoptimalvalueofxt theownerofthelakeneedonlyknowthenumberofbass callitbt inthelakeatthebeginningofyeart Therefore thestateatthebeginningofyeartisbt Wedefineft bt tobethemaximumnetprofitthatcanbeearnedfrombasscaughtduringyearst t 1 Tgiventhatbtbassareinthelakeatthebeginningofyeart AFisheryExample Wemaynowdisposeofaspects1 3oftherecursion Aspect1 Whataretheallowabledecisions Duringanyyearwecan tcatchmorebassthanthereareinthelake Thus ineachstateandforallt0 xt btmusthold Aspect2 Whatisthenetprofitearnedduringyeart Ifxtbassarecaughtduringayearthatbeginswithbtbassinthelake thenthenetprofitisr xt c xt bt Aspect3 Whatwillbethestateduringyeart 1 Theyeart 1statewillbe1 2 bt xt AFisheryExample AfteryearT therearenofutureprofitstoconsider soft bt max r xt c xt bt ft 1 1 2 bt xt where0 xt bt Weusethisequationtoworkbackwardsuntilf1 10 000 hasbeencomputed Thentodeterminetheoptimalfishingpolicy webeginbychoosingx1tobeanyvalueattainingthemaximumintheequationforf1 10 000 Thenyear2willbeginwith1 2 10 000 x1 bassinthelake AFisheryExample Thismeansthatx2shouldbechosentobeanyvalueattainingthemaximumintheequationforf2 1 2 10 000 x1 Continueinthisfashionuntiloptimalvaluesofx3 x4 xThavebeendetermined IncorporatingtheTimeValueofMoneyintoDynamicProgrammingFormulations Aweaknessofthecurrentformulationisthatprofitsreceivedduringthelateryearsareweightedthesameasprofitsreceivedduringtheearlieryears Supposethatforsome 1 1receivedatthebeginningofyeart 1isequivalentto dollarsreceivedatthebeginningofyeart Wecanincorporatethisideaintothedynamicprogrammingrecursionbyreplacingthepreviousequationwithwhere0 xt bt Thenweredefineft bt tobethemaximumnetprofitthatcanbeearnedduringyearst t 1 T Thisapproachcanbeusedtoaccountforthetimevalueofmoneyinanydynamicprogrammingformulation ComputationalDifficultiesinUsingDynamicProgramming Thereisaproblemthatlimitsthepracticalapplicationofdynamicprogramming Inmanyproblems thestatespacebecomessolargethatexcessivecomputationaltimeisrequiredtosolvetheproblembydynamicprogramming 18 7TheWagner WhitinAlgorithmandtheSilver MealHeuristic TheInventoryExampleinthischapterisaspecialcaseofthedynamiclot sizemodel DescriptionofDynamicLot SizeModelDemanddtduringperiodst t 1 2 T isknownatthebeginningofperiod1 Demandforperiodtmustbemetontimefrominventoryorfromperiodtproduction Thecostc x ofproducingxunitsduringanyperiodisgivenbyc 0 0 andforx 0 c x K cx whereKisafixedcostforsettingupproductionduringaperiod andcisavariableper unitcostofproduction Attheendofperiodt theinventorylevelitisobserved andaholdingcosthitisincurred Weleti0denotestheinventorylevelbeforeperiod1productionoccurs Thegoalistodetermineaproductionlevelxiforeachperiodtthatminimizesthetotalcostofmeeting ontime thedemandsforperiods1 2 T Thereisalimitctplacedonperiodt sendinginventory Thereisalimitrtplacedonperiodt sproduction WagnerandWhitinhavedevelopedamethodthatgreatlysimplifiesthecomputationofoptimalproductionschedulesfordynamiclot sizemodels Lemmas1and2arenecessaryforthedevelopmentoftheWagner Whitinalgorithm Lemma1 Supposeitisoptimaltoproduceapositivequantityduringaperiodt Thenforsomej 0 1 T t theamountproducedduringperiodtmustbesuchthatafterperiodt sproduction aquantitydt dt 1 dt jwillbeinstock Inotherwords ifproductionoccursduringp
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 新建建筑消防验收合格报告模板
- 少先队活动策划及会议记录范本
- 企业印章使用风险防控与管理规范
- 项目管理团队激励方案模板
- 餐饮行业标准化操作流程
- 餐厅物资采购合同范本
- 餐饮采购协议合同范本
- 企业安全管理体系审查标准化模板事故预防版
- 个人承诺服务保障书(8篇)
- 社会组织项目申报及资金管理指南
- XX集团董事会2025年度工作报告
- 多重耐药菌的课件
- 交安设施冬季施工方案
- 行业的客户信息管理表格模板
- 航天员工知识培训内容课件
- 民航招飞英语试题及答案
- 风电场安全检查表
- GB/T 19981.1-2014纺织品织物和服装的专业维护、干洗和湿洗第1部分:清洗和整烫后性能的评价
- GB 5903-2011工业闭式齿轮油
- 国开经济学(本)1-14章练习试题及答案
- 斯巴达勇士赛合作方案(精彩策划文案)课件
评论
0/150
提交评论