1 / 36

A Genetic Algorithm for Period Vehicle Routing Problem with Practical Application

UNIVERSIDADE FEDERAL DO CEARÁ PRÓ-REITORIA DE PESQUISA E PÓS-GRADUAÇÃO PROGRAMA DE MESTRADO EM LOGÍSTICA E PESQUISA OPERACIONAL. A Genetic Algorithm for Period Vehicle Routing Problem with Practical Application. José Lassance de Castro Silva Felipe Pinheiro Bezerra CYTEDHAROSA 2012.

merrill
Télécharger la présentation

A Genetic Algorithm for Period Vehicle Routing Problem with Practical Application

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. UNIVERSIDADE FEDERAL DO CEARÁ PRÓ-REITORIA DE PESQUISA E PÓS-GRADUAÇÃO PROGRAMA DE MESTRADO EM LOGÍSTICA E PESQUISA OPERACIONAL A Genetic Algorithm for Period Vehicle Routing Problem with Practical Application José Lassance de Castro Silva Felipe Pinheiro Bezerra CYTEDHAROSA 2012

  2. Outline • MotivatingProblem • ProblemDefinition • SolutionMethodAproach • ComputationalExperiments • Conclusionsand Future ResearchDirections

  3. MotivatingProblem • WholesalerDistributor • Ice cream and ice pops division • Sales team • Marketing mix: • Product • Pricing • Promotion • Placement Practicalapplication:

  4. MotivatingProblem Practicalapplication: • SALES TEAM ROUTINE AT CUSTOMER STORE • Observe visibilityandpromotionelements • Inspectequipments (freezers) • Clean theequipmentsandrearrangetheproductsinsidethem • Remove strangeproducts • Analysesupply, assortmentandprices • Negotiateimprovementsandorders • Placeorder

  5. MotivatingProblem Practicalapplication: Currentsolutionmethod

  6. MotivatingProblem • Advantages: • Out of route serving • Intuitive inclusion of new customers • Sales representative´s familiarity with territory Practicalapplication: Currentsolutionmethod

  7. MotivatingProblem • Drawbacks: • No tour definition • Replanning cost (time) • Learning curve • Unable to handle customer with multiple service frequence demand Practicalapplication: Currentsolutionmethod

  8. MotivatingProblem • Predefinedfrequence a regularity • Routeoptimization • Savetravel time • Increasesalesoportunity • Minimize travelcostsandrisks • Fastandeasyto use • Operationalrestrictions • Team size • Daily workload Practicalapplication: Considerations

  9. The PeriodicVehicleRoutingProblem (PVRP) • Given: • a set ofcustomerswithknowndemandsandvisitfrequencies; • a set of schedule options for eachcustomer; • a planningperiodofmultipledays; • a homogeneousfleetofvehicleswithlimitedcapacity; • thelocationofthecustomersandthe central depot (wherealltrips must start andend); • the complete network wihtknownarccosts. • Find: • A set ofroutes over theplannigperiod. • Objective: • Minimize the global visiting cost.

  10. The Periodic Vehicle Routing Problem (PVRP) (BALDACCI et al., 2011) 1 vehicle 30 unitsofcapacity

  11. The Periodic Vehicle Routing Problem (PVRP) • Select a visit schedule for eachcustomer; • Define thecustomersthatshouldbevisitedbyeachvehicleoneachday; • Routethevehicles for eachday. Threesimultaneousdecisions: It´s a generalizationoftheVRP: NP-Hard.

  12. SolutionMethodAproach • Holland (1975) • Metaheuristic • Natural selection • Populationbased • Cromossomes/individuals • Recombinations • Fitness GeneticAlgorithms: Concepts

  13. SolutionMethodAproach Genetic Algorithms: Basic pseudocode Begin generateinitialpopulation evaluate fitness ofeach individual Whilestop criteriaisnottruedo proceed crossovers proceedmutations evaluate new individuals selectindividualstoreplaceandtheirreplacements update stop criteria End returnbestsolution End

  14. SolutionMethodAproach • Solutionrepresentation • Fitness function • Populationcontrol • Selectionmethod • Geneticoperators • Use ofhibridization • Stop criteria • Parametersdefinition GeneticAlgorithms: Key points

  15. SolutionMethodAproach • Solutionrepresentation • Grand Tour • No tripdelimiters • Prins (2004), Chu et al. (2004) e Vidal et al. (2012) Proposedgeneticalgorithm: (VIDAL et al. 2012)

  16. SolutionMethodAproach • Individuals evaluation: Split algorithm (PRINS 2004) Proposedgeneticalgorithm: (PRINS, 2004)

  17. SolutionMethodAproach • Original crossover operator Proposed genetic algorithm:

  18. Computational experiments Benchmark instancestesting: Resultson benchmark instances. STATE-OF-THE-ART METHODS TanandBeasley (1984) - TB ChristofidesandBeasley (1984) - CB Chaoet al. (1995) - CGW Cordeau et al. (1997) - CGL Alegre et al. (2007) - ALP Hemmelmayret al. (2007) - HDR Baldacciet al.(2011) - BLD Vidal et al. (2012) - VDL

  19. Computational experiments Benchmark instancestesting: Averagecomputationalcost in minutes Source: Vidal et al. (2012) STATE-OF-THE-ART METHODS Cordeauet al. (1997) CGL Alegre et al. (2007) ALP Hemmelmayret al. (2007) HDR Chaoet al. (1995) CGW Vidal et al. (2012) VDL

  20. Computational experiments • Fair results • Lowcomputationalcosts Benchmark instancestesting:

  21. Computationalexperiments • Briefing • 629 Stores • 7 salesrepresentatives • Weeklyvisits, frommondaythroughfriday • 5 schedule options, except for 36 customers • Service time: 15 minutes • Maximumdailyworkload: 8 hours (480 minutes) • Travelspeed: 30km/h Practicalapplication: Solutionmethodapplied

  22. Computational experiments Practicalapplication: Solutionmethodapplied

  23. Computational experiments • Adjustments: • Demand = service time • Restrictions = dailyworkload in mimutes • Travel time • Penalties for notusingevery“vehicle” daily Practicalapplication: Solutionmethodapplied

  24. Computational experiments Practicalapplication: Solutionmethodapplied Distance savings over planning period Average daily workload composition per salesman (minutes).

  25. Computational experiments • Initialfindings: • Downtimeawareness • Trade-off betweensavingsandworkloadbalancing • “Howmuch does theworkloadbalancingcost?” Practicalapplication: Solutionmethodapplied

  26. Computational experiments Practicalapplication: Solutionmethodapplied .

  27. Computationalexperiments Practicalapplication: Solutionmethodapplied Comparisonsbetweencurrentsolutionmethodandproposedsolutionmethod

  28. Computational experiments Practicalapplication: Solutionmethodapplied MONDAY CURRENT PROPOSED

  29. Computational experiments Practicalapplication: Solutionmethodapplied TUESDAY CURRENT PROPOSED

  30. Computational experiments Practicalapplication: Solutionmethodapplied WEDNESDAY CURRENT PROPOSED

  31. Computational experiments Practicalapplication: Solutionmethodapplied THURSDAY CURRENT PROPOSED

  32. Computational experiments Practicalapplication: Solutionmethodapplied FRIDAY CURRENT PROPOSED

  33. Conclusions • Goodsolutionmethod for thePVRP • Goodresults for thepractical case: • Routeoptimization • Reliableprocedure • Service levelguaranteed • Costcontrol • Easyset-up • Decisionmaking tool

  34. Future ResearchDirections • Testinganotherinsertionmethods (i.e. GENI) • Populationdiversitycontrol • Apply more mutationoperators • Multicriteriaanalisysfor fitness evaluation • Automaticand/ordynamiccalibration • Meta-AGs • AI

  35. Future ResearchDirections • Directaproach for balancing • Spatialrouteclustering for eachvehicleduringplanningperiod

  36. UNIVERSIDADE FEDERAL DO CEARÁ PRÓ-REITORIA DE PESQUISA E PÓS-GRADUAÇÃO PROGRAMA DE MESTRADO EM LOGÍSTICA E PESQUISA OPERACIONAL THANK YOU! José Lassance de Castro Silva <lassance@lia.ufc.br> Felipe Pinheiro Bezerra <FELIPE@FORTALI.COM.BR> CYTEDHAROSA 2012

More Related