3b3ee966e704fe6fbe5cd5d18d23ce83.ppt
- Количество слайдов: 54
Issues in Dynamic Fleet Management Talk at ROUTE 2000 - INTERNATIONAL WORKSHOP ON VEHICLE ROUTING SKODSBORG, DENMARK - AUGUST 16 -19, 2000 Geir Hasle Research Director, Department of Optimization SINTEF Applied Mathematics Oslo, Norway Geir. Hasle@math. sintef. no http: //www. oslo. sintef. no/am/ 1
My talk SINTEF Applied Mathematics (SAM) n Fleet Management n – – industrial potential, status, requirements technology research, science bridging the gap between science and industry Challenges n Routing etc. at SAM n Research Agenda n 2
SINTEF The Foundation for Scientific and Industrial Research at the Norwegian Institute of Technology The vision: Technology for a better society Business concept: SINTEF´s goal, in co-operation with NTNU and Ui. O, is to meet needs of the private and public sectors for research-based innovation and development Locations: The SINTEF Group have 1800 employees, 400 in Oslo and 1400 in Trondheim. 3
SINTEFs council SINTEFs board President/ Vice-president SINTEF Petroleum Research MARINTEK The Norwegian Marine Technology Research Institute SINTEF Energy Research SINTEF Applied Mathematics SINTEF Civil and Environmental Engineering SINTEF Electronics and Cybernetics SINTEF Applied Chemistry SINTEF Materials Technology SINTEF Fisheries and Aquaculture SINTEF Industrial Management SINTEF Telecom and Informatics SINTEF Unimed 4
SINTEF-Group turnover in 1999 Basic grants from The Research Council of Norway 3, 3% Strategic programs from The Research Council of Norway 4, 3% Contracts 92, 4% - Industrial and commercial enterprises 53, 0% Public sector 12, 5% International contracts 10, 5% The Research Council of Norway (project grants) 9, 9% Other sources 6, 5% 5
SINTEF Applied Mathematics http: //www. oslo. sintef. no/am A contract research institute in the SINTEF group Geometry Modeling Simulation Optimisation 6
SINTEF Applied Mathematics Department of Optimisation Focus: Applied research Planning Decision Support Application types: Resource optimisation Design/configuration Discrete Basis: Knowledge Based Systems Operations Research Computing science Main business areas: Transportation Area management Oil business Manufacturing Approach: Generic Tools Reuse Methodology 7
Transportation of goods in Norway and EU n 12. 000 companies (EU 1/2 mill. ) n Annual turnover 30 billion. (EU 1. 200 billion. ) n Many SMEs n 36 % empty driving n Capacity utilization with load: 60 % n Logistics costs 12% of product costs (EU 7%) n EU: 13 million trucks, 800 billion ton-kilometers (1990) n Germany: freight income some 60 billion DM (1990) 8
Industrial use of VRP Tools Excess travel, huge potential n Swedish report* 1999 (commercial road transport) n – – large end users, food & beverage generation of static routes vendors claim operational tools very high potential for savings * A. Henriksson, P. Liljevik: ”Dynamisk ruttplanlegging i verkligheten” Minirapport MR 123, TFK - Institutet för transportforskning, Stockholm October 1999 9
Increasing need for VRP Tools n focus on – – – time cost utilization customer service lead time reduction reactivity regulations, environmental concerns n e-commerce, home shopping n 10
Reasons for mismatch lack of awareness in industry n lack of data and infrastructure n price (SMEs) n organizational problems, resistance n practical constraints n – information availability – physical movement n tools not good enough – – functionality, modelling power user friendliness integration logistical performance 11
Existing tools - keywords • • Large variety: simple TSP - sophisticated VRP solvers focus: road transportation of goods, local distribution built for operative planning, used for generation of static routes packages primitive integration, but good import facilities inflexible and simple or heavy on consultancy Windows-platform good user interfaces, map visualization, manual changes • VRP algorithms? real-time planning? multiple users? continuous optimization? • priced at USD 40. 000 and above (high end) • • • 12
Some Vendors • • • • • • Descartes Systems Caps Logistics -> Baan Micro. Analytics Roadnet Technologies (UPS) i 2 ESRI Kositzky and Associates Manugistics Carrier Logistics Inc Insight Inc. Caliper Corporation Trapeze Software Group Giro DPS International Paragon Software Systems Optrak (Andersen Consulting) Ilog Diagma PTV Alfaplan PLS Prologos USA USA/GB USA USA/The Netherlands/GB USA/Canada UK UK UK F F D D Typically claim 10 - 30% cost reductions - static routes 13
Few VRP Tools in Operation in Norway Coca-Cola n Taxi companies n Falken n NAS n NKL n Tollpost-Globe n Linjegods n Postal service n Hydro Agri n 14
Challenges - VRP Tools n n Functionality Modelling – – – n n n constraints criteria uncertainty dynamics supply-chain coordination Adaptability Power: speed vs. quality Large-size problems User Interface Integration Support etc. 15
Dynamic, real-time routing Success stories? n Paragon - Tesco – “… Home shoppers simply log onto the dedicated area of Tesco's website, select their purchases and identify a two hour time window for delivery to an address of their choosing”. . . n Truckstops – “… In some UK applications it is even used to recalculate routes during the day, modifying its original calculations to take account of new requirements and reflecting data transmitted back from vehicles by radio” … n Price. Waterhouse. Coopers 16
Goal - VRP Technology n real benefits for industry - logistical performance – solve right problem – plan quality vs. response time – user interaction, user-friendliness – configurability – reactivity – price 17
Future VRP technology GIS vendors n ERP vendors n ASP solutions, thin clients, Internet, www n better tools for strategic/tactical planning n supply-chain coordination, integrated solutions n n dynamic, real time fleet management 18
Dynamic Fleet Management Prerequisites n ICT infrastructure – order data – fleet data n access to high quality traffic data – speed predictions – “organic” electronic road network Better understanding of routing policies n Better VRP algorithms 19 n
Issues in VRP research n n Large, ill-structured problems rich models – uncertainty – dynamics – multiple criteria n reactivity – – – n n n disruption? slack policies plan quality vs. response time performance decomposition human issues 20
Stochastic and dynamic VRPs n what does “dynamic” mean? – problem changes dynamically – Psaraftis (1995): “. . . information on the problem is made known to the decision maker or is updated concurrently with the determination of the set of routes. ” – Baita, Ukovich, Pesenti, Favaretto (1998): “. . . releated decisions have to be taken at different times within some time horizon, and earlier decisions influence later decisions. ” – a. k. a. “real-time”, “on-line” n “organic” routing plans – challenges n n information flow physical goods – good idea? (talk of Carlos Daganzo) 21
Literature - dynamic VRPs 6 INFORMS sessions since 1995, some 20 papers n some 50 journal papers n 22
Some papers Psaraftis (1995): Dynamic vehicle routing: Status and prospects Bertsimas, DJ / Simchi. Levi, D (1996): A new generation of vehicle routing research: Robust algorithms, addressing uncertainty Crainic, TG / Laporte, G (1997): Planning models for freight transportation Baita, F / Ukovich, W / Pesenti, R / Favaretto, D (1998): Dynamic routing-and-inventory problems: A review Swihart, MR / Papastavrou, JD (1999): A stochastic and dynamic model for the single-vehicle pick-up and delivery problem Savelsbergh, M / Sol, M (1998): Drive: Dynamic routing of independent vehicles Ioachim, I / Desrosiers, J / Soumis, F / Belanger, N (1999): Fleet assignment and routing with schedule synchronization constraints Gans, N / Van. Ryzin, G (1999): Dynamic vehicle dispatching: Optimal heavy traffic performance and practical insights Reiman, MI (1999): Heavy traffic analysis of the dynamic stochastic inventory-routing problem Gendreau, M / Guertin, F / Potvin, JY / Taillard, E (1999): Parallel tabu search for real-time vehicle routing and dispatching Powell, WB / Towns, MT / Marar, A (2000): On the value of optimal myopic solutions for dynamic routing and scheduling problems in the presence of user noncompliance Cheung, RK / Muralidharan, B (2000): Dynamic routing for priority shipments in LTL service networks Gendreau, M / Laporte, G / Seguin, R (1996): Stochastic vehicle routing Gendreau, M / Laporte, G / Seguin, R (1996): A tabu search heuristic for the vehicle routing problem with stochastic demands and customers Haughton, MA (1998): The performance of route modification and demand stabilization strategies in stochastic vehicle routing Yang, WH / Mathur, K / Ballou, RH (2000): Stochastic vehicle routing problem with restocking Haughton, MA (2000): Quantifying the benefits of route reoptimisation under stochastic customer demands Secomandi, N (2000): Comparing neuro-dynamic programming algorithms for the vehicle routing problem with stochastic demands Shieh, HM / May, MD (1998): On-line vehicle routing with time windows - Optimization-based heuristics approach for freight demands requested in real-time 23 Kilby / Prosser / Shaw: Dynamic VRPs: A Study of Scenarios (forthcoming)
Approaches - uncertainty, dynamics ignore n deterministic model - and repair n – – – n crisp, optimized plans are brittle is disruption costly? add slack, how? stochastic model – investigation of policies – still need dynamic decision-making n lessons to be learnt from factory scheduling 24
Dynamic VRP DSS n dependent on high quality updated information – fleet status – order status n “organic” route planning – – – concept of current plan when do we commit? when do we include changes? locking parts of plan do we need to worry about disruption? dependence on type of operation / business rules n delivery vs. pickup – applicable algorithms – (how much) do we save by taking a dynamic approach? 25
Approaches insertion heuristics + iterative improvement n constraint propagation n n MP formulations? n Minimal disruption possibly an additional goal criterion component 26
Routing at SAM SPIDER n Green. Trip n HAMMER - vessel routing with inventory constraints n Bus scheduling n e. CSPlain, EU FP V n Distributed problem solving n Proposals n 27
SPIDER n a VRP Solver C++ program library – – – n UNIX Windows COM component instantiates to a module for optimised transport management – plan-administrasjon – VRP optimisation – cheapest path calculations n n adaptable to wide variety of applications distribution through sw vendors 28
29
Green. Trip n n Esprit 20603, January 1996 -March 1999, > 40 person-years Consortium – Tollpost-Globe (N) – Pirelli (I) – Ilog (F) – University of Strathclyde (GB) – SINTEF (N) n RTD effort in methods, algorithms, and generic sw for optimised fleet management 30
The goal of Green. Trip n Produce a cost-effective tool to optimise routing of vehicles that – is generic – takes into account multiple business constraints – permits efficient (re)configuration – integrates easily in existing IT infrastructure 31
Green. Trip Technical Approach OO Programming n Constraint Programming n Iterative Improvement Techniques n Applications Modelling n Automated Systems (Re)Configuration n 32
The Green. Trip Consortium Tollpost. Globe Pirelli SINTEF ILOG Uo. S 33
CASE : TOLLPOST-GLOBE n n n n Pick up orders : 600 Regular and non-regular customers Deliveries : 2. 400 Time windows - Customer service Two days are not the same some 100 vehicles Different vehicles (size, volume, equipment) Depot with automatic sorting / registration 34
CASE : TOLLPOST-GLOBE n n n Electronic road and address data are available via the GIS Transportation Demonstrator Mobile communication installed in 15 vehicles GPS installed in 5 vehicles some 100. 000 customers in the Oslo region goal: dynamic fleet management system 35
The Pirelli (Cables) Case Logistics network simulator n Assessment of logistical performance n Detailed analysis of alternative structural changes n scenarios 6 months operation, 10. 000 orders n 36
Green. Trip - GGT Systems Architecture Application. Modelling Application Model Application Server VRP Solver GIS Road data Legacy Systems 37
The VRP Solver - Objects Plans n Locations n Visits n Vehicles n Routes n Dimensions n Constraints n 38
VRP Solver - Algorithms n Construction – Savings – Sweep – Nearest. . . n Improvement, move operators – 2 -opt, Or-opt – Relocate – Exchange – Cross 39
VRP Solver - Search Control n Basic heuristic – Greedy Search (First Improvement) – Steepest Descent (Best Improvement) n Meta-heuristics – Tabu Search – Guided Local Search – Guided Tabu Search 40
Green. Trip - Results VRP Solver -> ILOG Dispatcher n GGT -> Green. Trip AS “Dynamic planner” n “best-until-now” results on OR benchmarks n Industrial Test Cases n Publications n – some 20 scientific papers – reports - “VRP Solving and IIT Survey” 41
Green. Trip Dissemination n n n Kilby, Prosser, Shaw: “Guided Local Search for the VRP”, Proc. MIC 97 De Backer, Furnon: “Metaheuristics in Constraint Programming: Experiments with Tabu Search on the VRP”, Proc. MIC 97 De Backer, Furnon, Kilby, Prosser, Shaw: “Local Search in Constraint Programming: Applications to vehicle routing problems”, CP 97 Scheduling Workshop Hasle: “Green. Trip - the Development of a Generic Toolkit for Vehicle Routing”, NOAS 97 De Backer, Furnon: “Solving vehicle routing problems with Side Constraints Using Constraint Programming”, INFORMS 97 De Backer, Furnon: “Modelling pickup and delivery problems in constraint programming”, INFORMS 98 Bouzoubaa, Hasle, Kloster, Prosser: “The GGT: a Generic Toolkit for VRP Applications and its Modelling Capabilities”, Proc. PACLP 99 42
Green. Trip Papers n De Backer, Furnon, Kilby, Prosser, Shaw: “Solving vehicle routing problems with constraint programming and metaheuristics”, Journal of Heuristics, Special Issue on CP n Kilby, Prosser, Shaw: “A comparison of traditional and constraint-based heuristic methods on vehicle routing problems with side constraints”, Constraints, April 98 n n n De Backer, Furnon: “Local Search in Constraint Programming”, in METAHEURISTICS: Advances and Trends in Local Search Paradigms for Optimization (Voss, Martello, Osman, Roucairol, 1999) Kilby, Prosser, Shaw: “Guided Local Search for the Vehicle Routing problem with Time Windows”, in META-HEURISTICS: Advances and Trends in Local Search Paradigms for Optimization ( Voss, Martello, Osman, Roucairol, 1999) Kilby, Prosser, Shaw: “Dynamic VRPs: A Study of Scenarios” 43 (forthcoming)
Vessel Routing - Ammonia Norsk Hydro Agri n Producer - Consumer Harbours (25) n Fleet (10) n Strong Inventory Constraints n External Trading n Feasible solution n Earlier approach: MIP n Approach taken: Heuristic Sequencing + LP 44 n
HAMMER Problem Producing harbours Quantity Time-window Consuming harbours Fleet of vessels Harbours with stock inventory External orders (laycans) Find the routing plan with the lowest cost so that inventory limits are not exceeded and all external orders included. 45
Combinatorial solution Vessel View: Harbour View: 3 2 6 1 4 H 1: H 2: H 3: 7 5 H 4: H 5: H 6: H 7: Site Route for Vessel 1 Route for Vessel 2 Vessel View: Which harbours, and in which sequence, each vessel will visit. Harbour View: Which vessels, and in which sequence, will call at each harbour. 46
HAMMER - Linear solution Vessel view: max Load min Time Harbour view: Call Stock Time 47
HAMMER - System overview Problem data Initial solver Iterative improver Combinatorial solution Feasibility check Feasible solution Greedy Propagator Update LP solver 48
HAMMER - Working with the system n Initialisation of the problem – Harbours, ships, laycans and planning parameters n Schedule generation – Initial solver - from scratch or existing – Iterative improvement n Analysis and user interaction – plan statistics - slack, unserviced – manually change plan n Lock ship, harbour or time period Flatberg, Haavardtun, Kloster, Løkketangen. (2000): Combining exact and Heuristic methods for solving a Vessel Routing Problem with inventory constraints and time windows. To appear in Ricerca Operativa, special issue on combined constraint programming and OR techniques 49
Research Agenda SAM: VRP n construction heuristics – – – construct and improve restart greedy + limited backtracking IIT by local search and meta-heuristics n exact methods subproblems / limited problems n hybrid methods n dynamic VRP n empirical investigation n 50
Important topics, SAM § § configuration of transportation networks VRPs and TSPs with side constraints in road based and maritime transportation cheapest path problems in large, dynamic network topologies Proposal to Research Council of Norway 51
Research Agenda SAM: Optimisation / CSP over-constrained problems n multi-criterion problems n supply-chain coordination n distributed problem solving n 52
Research Agenda: VRP rich models, large problems n dynamic VRPs n exact methods for limited (sub)problems n over-constrained problems n multi-criteria problems n methodology: problem type - algorithm n cooperating VRP solvers, hybrid methods n decomposition n 53
Issues in Dynamic Fleet Management Talk at ROUTE 2000 - INTERNATIONAL WORKSHOP ON VEHICLE ROUTING SKODSBORG, DENMARK - AUGUST 16 -19, 2000 Geir Hasle Research Director, Department of Optimization SINTEF Applied Mathematics Oslo, Norway Geir. Hasle@math. sintef. no http: //www. oslo. sintef. no/am/ 54