1 / 59

Scheduling under Uncertainty : Solution Approaches

Scheduling under Uncertainty : Solution Approaches . Frank Werner Faculty of Mathematics. Outline of the talk. Introduction Stochastic approach Fuzzy approach Robust approach Stability approach Selection of a suitable approach. 1. Introduction. Notations.

leoma
Download Presentation

Scheduling under Uncertainty : Solution Approaches

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. SchedulingunderUncertainty:Solution Approaches Frank Werner FacultyofMathematics

  2. Outline ofthe talk • Introduction • Stochasticapproach • Fuzzyapproach • Robust approach • Stabilityapproach • Selectionof a suitableapproach

  3. 1. Introduction Notations

  4. Deterministicmodels:all dataaredeterministicallygiven in advance • Stochasticmodels:dataincluderandom variables In real-lifescheduling: manytypesofuncertainty(e.g. processingtimes not exactlyknown, machinebreakdowns, additionallyarivingjobswithhighpriorities, roundingerrors, etc.) Uncertain (interval) processingtimes:

  5. Relationshipbetweenstochasticanduncertainproblems: Distribution function Densityfunction

  6. Approaches forproblemswithinaccuratedata: • Stochasticapproach: useofrandom variables withknownprobabilitydistributions • Fuzzyapproach: fuzzynumbersasdata • Robust approach: determinationof a schedulehedgingagainsttheworst-casescenario • Stabilityapproach: combinationof a stabilityanalysis, a multi-stagedecisionframeworkandtheconceptof a minimal dominant setof semi-activeschedules → Thereisnouniquemethodfor all typesofuncertainties.

  7. Two-phasedecision-makingprocedure • Off-line (proactive) phaseconstructionof a setofpotentially optimal solutionsbeforetherealizationoftheactivities(staticschedulingenvironment, scheduleplanningphase) • On-line (reactive) phaseselectionof a solutionfromwhenmoreinformationisavailableand/or a partoftheschedulehasalreadybeenrealized→ useof fast algorithms(dynamicschedulingenvironment, scheduleexecutionphase)

  8. General literature (surveys) • Pinedo: Scheduling, Theory, Algorithmsand Systems, Prentice Hall, 1995, 2002, 2008, 2012 • SlowinskiandHapke: SchedulingunderFuzziness, Physica, 1999 • Kasperski: DiscreteOptimizationwithInterval Data, Springer, 2008 • Sotskov, Sotskova, Lai and Werner: SchedulingunderUncertainty; TheoryandAlgorithms, Belarusian Science, 2010 Forthe RCPSP underuncertainty, see e.g. • HerroelenandLeus, Int. J. Prod. Res.. 2004 • HerroelenandLeus, EJOR, 2005 • DemeulemeesterandHerroelen, Special Issue, J. Scheduling, 2007

  9. 2. Stochasticapproach • Distribution ofrandom variables(e.g. processingtimes, releasedates, due dates)known in advance • Often: minimizationofexpectationvalues(ofmakespan, total completion time, etc.) Classesofpolicies(seePinedo 1995) • Non-preemptivestaticlistpolicy (NSL) • Preemptivestaticlistpolicy (PSL) • Non-preemptivedynamicpolicy (ND) • Preemptivedynamicpolicy (PD)

  10. Someresultsforsingle-stageproblems(seePinedo 1995) Single machineproblems • ProblemWSEPT rule: order thejobsaccordingto non-increasingratiosTheorem 1: The WSEPT ruledetermines an optimal solution in theclassof NSL as well as ND policies. • Problem Theorem 2: The EDD ruledetermines an optimal solution in theclassof NSL, ND and PD policies.

  11. Problem Theorem 3: The WSEPT ruledetermines an optimal solution in theclassof NSL, ND and PD policies.Remark: The same resultholdsforgeometricallydistributed Parallel machineproblems • ProblemTheorem 4: The LEPT ruledetermines an optimal solution in theclassof NSL policies.

  12. ProblemTheorem 5: The non-preemptive LEPT policydetermines an optimal solution in theclassof PD policies. • ProblemTheorem 6: The non-preemptive SEPT policydetermines an optimal solution in theclassof PD policies.

  13. Selected references (1) • PinedoandWeiss, Nav. Res. Log. Quart., 1979 • Glazebrook, J. Appl. Prob., 1979 • WeissandPinedo, J. Appl. Prob., 1980 • Weber, J. Appl. Prob., 1982 • Pinedo, Oper. Res., 1982; 1983 • Pinedo, EJOR, 1984 • PinedoandWeiss, Oper. Res., 1984 • Möhring, Radermacher andWeiss, ZOR, 1984; 1985 • Pinedo, Management Sci., 1985 • Wie andPinedo, Math. Oper. Res., 1986 • Weber, VaraiyaandWalrand, J. Appl. Prob., 1986 • Righter, System andControl Letters, 1988 • Weiss, Ann. Oper. Res., 1990

  14. Selected references (2) • Weiss, Math. Oper. Res., 1992 • Righter, Stochastic Orders, 1994 • Cai and Tu, Nav. Res. Log., 1996 • Cai and Zhou, Oper. Res., 1999 • Möhring, Schulz andUetz, J. ACM, 1999 • Nino-Mora, Encyclop. Optimiz., 2001 • Cai, Sun and Zhou, Prob. Eng. Inform. Sci., 2003 • Ebben, Hans andOldeWeghuis, OR Spectrum, 2005 • Ivanescu, Fransooand Bertrand, OR Spectrum, 2005 • Cai, Wu and Zhou, IEEE Transactions Autom. Sci. Eng., 2007 • Cai, Wu and Zhou, J. Scheduling, 2007; 2011 • Cai, Wu and Zhou, Oper. Res., 2009 • Tam, Ehrgott, Ryan andZakeri, OR Spectrum, 2011

  15. 3. Fuzzyapproach • Fuzzyschedulingtechniqueseitherfuzzifyexistingschedulingrulesorsolvemathematicalprogrammingproblems • Often: fuzzyprocessingtimes , fuzzy due dates • Examplestriangularfuzzyprocessingtimestrapezoidalfuzzyprocessingtimes

  16. Often: possibilisticapproach(Dubois andPrade 1988) Chanasand Kasperski (2001) Problem Objective: Assumption: → adaptionofLawler‘salgorithmforproblem

  17. Special cases: Alternative goalapproach - fuzzygoal, Objective: Chanasand Kasperski (2003) Problem Objective: → adaptionofLawler‘salgorithmforproblem

  18. Selected references (1) • Dumitru andLuban, Fuzzy Sets and Systems, 1982 • Tada, Ishii andNishida, APORS, 1990 • Ishii, TadaandMasuda, Fuzzy Sets and Systems, 1992 • GrabotandGeneste, Int. J. Prod. Res., 1994 • Han, Ishii and Fuji, EJOR, 1994 • Ishii andTada, EJOR, 1995 • Stanfield, King andJoines, EJOR, 1996 • Kuroda and Wang, Int. J. Prod. Econ., 1996 • Özelkanand Duckstein, EJOR, 1999 • SakawaandKubota, EJOR, 2000

  19. Selected references (2) • Chanasand Kasperski, Eng. Appl. Artif. Intell., 2001 • Chanasand Kasperski, EJOR, 2003 • Chanasand Kasperski, Fuzzy Sets and Systems, 2004 • Itohand Ishii, FuzzyOptim. andDec. Mak., 2005 • Kasperski, Fuzzy Sets and Systems, 2005 • Inuiguchi, LNCS, 2007 • Petrovic, Fayad, Petrovic, Burke and Kendall, Ann. Oper. Res., 2008

  20. 4. Robust approach Objective: Find a solution, whichminimizesthe „worst-case“ performanceover all scenarios. Notations (singlemachineproblems) maximal regretof Minmaxregretproblem (MRP): Find a sequence such that

  21. Somepolynomiallysolvable MRP (Kasperski 2005) (VolgenantandDuin 2010) (Averbakh 2006) (Kasperski 2008) Some NP-hard MRP (LebedevandAverbakh 2006) (for a 2-approximation algorithm, see Kasperski and Zielinski 2008) (Kasperski, Kurpiszand Zielinski 2012)

  22. Kasperski andZielinski (2011) ConsiderationofMRP‘susingfuzzyintervals Deviation interval Known: deviation Applicationofpossibilitytheory(Dubois andPrade 1988) possibly optimal if necessarily optimal if

  23. Fuzzyproblem orequivalently whereis a fuzzyintervalandisthecomplementofwithmembershipfunction The fuzzyproblemcanbeefficientlysolvedif a polynomialalgorithmforthecorresponding MRP exists.

  24. Solution approaches • Binary searchmethod- repeatedexactsolutionofthe MRP - applications: : binarysearchsubroutine in B&B algorithm

  25. Mixed integer programmingformulation- useof a MIP solver- application: • Parametricapproach- solutionof a parametricversionof a MRP(often time-consuming)- application:

  26. Selected references (1) • Daniels andKouvelis, Management Sci., 1995 • KouvelisandYu, Kluwer, 1997 • Kouvelis, Daniels andVairaktarakis, IEEE Transactions, 2000 • Averbakh, OR Letters, 2001 • Yang andYu, J. Comb. Optimiz., 2002 • Kasperski, OR Letters, 2005 • Kasperski and Zielinski, Inf. Proc. Letters, 2006 • LebedevandAverbakh, DAM, 2006 • Averbakh, EJOR, 2006 • Montemanni, JMMA, 2007

  27. Selected references (2) • Kasperski and Zielinski, OR Letters, 2008 • Sabuncuogluand Goren, Int. J. Comp. Integr. Manufact., 2009 • Aissi, BazganandVanderpooten, EJOR, 2009 • VolgenantandDuin, COR, 2010 • Kasperski and Zielinski, FUZZ-IEEE, 2011 • Kasperski, Kurpiszand Zielinski, EJOR, 2012

  28. 5. Stabilityapproach 5.1. Foundations 5.2. General shopproblem 5.3. Two-machineflowandjobshopproblems 5.4. Problem

  29. 5.1. Foundations Mixed Graph Example: 11 12 13 00 ** 21 22 23

  30. Example (continued) 11 12 13 ** 00 21 22 23

  31. Stabilityanalysisof an optimal digraph Definition 1 The closed ball iscalled a stability ball ofifforanyremains optimal. The maximal value iscalledthestabilityradiusofdigraph Known: • Characterizationofthe extreme valuesof • Formulasforcalculating • Computationalresultsforjobshopproblemswith(seeSotskov, Sotskovaand Werner, Omega, 1997)

  32. 5.2. General shopproblem Definition 2 iscalled a G-solutionforproblemifforanyfixedcontains an optimal digraph. Ifanyis not a G-solution, iscalled a minimal G-solutiondenotedas Introductionofthe relative stabilityradius:

  33. Definition 3 Letbe such thatforanyThe maximal valueofof such a stability ball iscalledtherelative stabilityradius Known: • Dominancerelationsamongpathsandsetsofpaths • Characterizationofthe extreme valuesof

  34. Characterizationof a G-solutionforproblem Definition 4 (strongly) dominatesin → dominancerelation Theorem 7: is a G-solution. Thereexists a finite coveringofpolytopebyclosedconvexsetswith such thatforanyandanythereexists a forwhich Corollary:

  35. Theorem 8: Letbe a G-solutionwith Then: is a minimal G-solution. Foranythereexists a vector such that Algorithmsforproblem

  36. Several 3-phase schemes: • B&B: implicit (or explicit) enumerationschemeforgenerating a G-solution • SOL: reductionofbygenerating a sequencewiththe same and different • MINSOL: generationof a minimal G-solutionby a repeatedapplicationofalgorithm SOL

  37. Somecomputationalresults: Exact sol.: , Heuristic sol.:

  38. 5.3. Two-machineproblemswithintervalprocessingtimes • ProblemJohnson permutation:Partition ofthejobset

  39. Theorem 9: • foranyeither (either ) and • andifsatisfies

  40. Theorem 10: If then Percentageofinstanceswith , where

  41. General caseofproblem Theorem 11: Thereexists an Theorem 12:

  42. Example: without transitive arcs: J2 J1 J3 J5 J6 J4

  43. Propertiesof in thecaseofseeMatsveichuk, Sotskovand Werner, Optimization, 2011 Schedule executionphase:seeSotskov, Sotskova, Lai and Werner, Schedulingunderuncertainty, 2010 (Section 3.5) Computationalresultsforandfor • Problem→ Reductiontotwoproblems:seeSotskov, Sotskova, Lai and Werner, Schedulingunderuncertainty, 2010 (Section 3.6)

  44. 5.4. Problem Notations:

  45. Definition ofthestability box:

  46. Definition 5 The maximal closedrectangular box is a stability box ofpermutation , ifpermu-tationbeing optimal forinstancewith a scenarioremains optimal fortheinstancewith a scenarioforeachIftheredoes not exist a scenario such thatpermutationis optimal forinstance , then Remark: The stability box is a subsetofthestabilityregion. However, thestability box isusedsinceitcaneasilybecomputed.

  47. Theorem 13: Fortheproblem , jobdominatesifandonlyifthefollowinginequalityholds: Lower (upper) boundon therangeofpreservingtheoptimalityof :

  48. Theorem 14: Ifthereisnojob , in permutation such thatinequality holdsforat least onejob , thenthestability box iscalculatedasfollows: otherwise

  49. Example: Data forcalculating

  50. Stability box for Relative volumeof a stability box Maximal rangesofpossiblevariationsoftheprocessingtimes , withinthestability box aredashed.

More Related