Vehicle routing at TNT Algorithm Future research Scheduling and routing problems at TNT Some solutions and some future research directions orensen 1 J. Vandenbergh 1 D. Cattrysse 1 M. Sevaux 2 K. S¨ 1 Centre for Industrial Management, KULeuven { kenneth.sorensen, joos.vandenbergh, dirk.cattrysse } @cib.kuleuven.be 2 University of South Brittany marc.sevaux@univ-ubs.fr VIP – Oslo S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation TNT TNT’s express division in 2007 Main business: B2B parcel delivery 4.1 million parcels a week > 200 countries > 2,300 depots, hubs and sortation centres > 26,700 road vehicles 47 aircraft > 75,000 staff Revenue of e 6.01 billion in 2006 Operating income e 580 million S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Vehicle routing at TNT Pick-up in milk-runs (dynamic!) Unloading/sorting/loading in depot/sortation centre FTL and airplane transport to destination depot Unloading/sorting/loading Delivery in milk-runs S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Problem delineation One depot Many (thousands) of known customers S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Problem delineation One depot Many (thousands) of known customers Customers are divided into zones S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Problem delineation One depot Many (thousands) of known customers Customers are divided into zones Zones have a known “centre” S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Problem delineation One depot Many (thousands) of known customers Customers are divided into zones Zones have a known “centre” Routing is done on zone level (e.g. 8 zones per truck) Intra-zone optimization is done by the driver S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Geographic zones Geographic delivery region is divided into zones Historically often based on postal codes Advantages Easy sorting Containers correspond to zones Last-minute changes are possible Easier routing Problem: how to determine zones Robust Easy to sort Easy to route “Right size” S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Two sub-problems Routing problem Minimize transportation cost Minimize distance to baseline solution Route balance Clustering problem: determine zones Later! S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT TNT’s express division Algorithm Routing problem Future research Objective calculation Objective calculation (inter-zone routing only) Transportation cost = total travel time Travel times between clusters Intra-cluster expected travel and drop-off time (estimated a priori) Distance to baseline solution Hamming distance of zone–truck assignments 0 if zone is assigned to the same truck in both solutions 1 if zone is assigned to different truck in both solutions Route balance Sum of squared deviations from average travel time S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT Objective integration Algorithm Operators Future research Structure “Multi-objective” algorithm Normalized weights i ( x ) = f i ( x ) − f min f n i f max − f min i i Additive objective function � α i f n � f ( x ) = i ( x ) with α i = 1 i i S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT Objective integration Algorithm Operators Future research Structure Algorithm operators “Multiple neighbourhood” local search Swap customers Insert customers 2-opt Archive of non-dominated solutions Fixed archive size Minimum Hamming distance to closest solution Perturbation move Break up one or more routes and re-route zones customers S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT Objective integration Algorithm Operators Future research Structure Algorithm overview Algorithm 1 : “Multi-objective” algorithm for routing at TNT Read baseline solution; Set weights α i ; while stopping conditions not satisfied do repeat Local search using neighbourhoods k and weights α i ; k ← ( k + 1) mod k ; until no more improvement ; if solution satisfies archiving conditions then Add solution to archive; Remove dominated solutions from archive; Perturb solution; Update weights α i ; S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT Algorithm Future research Future research Get some results Compare to current way of working Improve the algorithm Add more neighbourhoods Add constructive heuristics for good starting solutions Intelligent neighbourhood search (tabu?) Study diversification strategies . . . Define zones Robust Easy to route Size restrictions . . . S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT Algorithm Future research Future research (cont.) “Real” multi-objective optimization method Implement in commercial vehicle routing software Strategic supply chain evaluation tool Location of new distribution centres Use of different vehicle types . . . S¨ orensen et al. Scheduling and routing problems at TNT
Vehicle routing at TNT Algorithm Future research Scheduling and routing problems at TNT Some solutions and some future research directions orensen 1 J. Vandenbergh 1 D. Cattrysse 1 M. Sevaux 2 K. S¨ 1 Centre for Industrial Management, KULeuven { kenneth.sorensen, joos.vandenbergh, dirk.cattrysse } @cib.kuleuven.be 2 University of South Brittany marc.sevaux@univ-ubs.fr VIP – Oslo S¨ orensen et al. Scheduling and routing problems at TNT
Recommend
More recommend
