1. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Arc hub location problems as network design problems with routing Elena Fernández- Dpt EIO-UPC Ivan Contreras- CIRRELT- Montréal Seminario de Geometría Tórica Jarandilla 12-15 de noviembre 2010

2. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre  Which set of facilities to open ? Location  How to satisfy the customers demands from open facilities ?  From which facility does the customer receive service ? Allocation  How is service provided ? Routing  Are facilities somehow connected ? Routing  Which are the possible (or preferable) connections between Network design customers or between customers and facilities ? Decisions in discrete location problems on networks

3. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre where customers obtain service from where flows between pairs of customers are consolidated and rerouted connect customers and facilities  Connect customers and facilities  Connect facilities between them What are facilities used for? Routing Which are the possible connections ? Network design

4. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre If customers move to facilities to recieve service ⋮ the routing of each customer is trivial Customers receive service from/at facilities

5. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre If several customers are visited in the same route ⋮ the design of the routes may become difficult Customers receive service from facilities

6. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre There exists communication between each pair of customers. Flows are consolidated and re-routed at facilities (which must be connected) Facilities used to reroute flows between pairs of customers HUB LOCATION

7. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Connection between facilities by means of a tree HUB LOCATION Facilities used to reroute flows between pairs of customers

8. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre G=(V, E) f i : set-up cost for facilities i  V d ij : per unit routing cost from i to ji, j  V W ij : flow between i and j i, j  V MINIMUM TOTAL COST Set-up costs + Flow Routing costs HUB LOCATION TO FIND Network design Hubs are used to consolidate and reroute flow between customers A set of facilities (hubs) to open Subset of edges to connect hubs among them Subset of edges to connect customers to their allocated hubs Location Assignment i j

9. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre i j m k HUB LOCATION: Typical asumptions  Transfer between hubs  Collection  Distribution Discount factors to routing costs Full interconnetion of hubs Paths: i-k-m-j Triangle inequality     Hub location problems are NP-hard

10. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Hub arcs i i j j j Campbell JF, Ernst AT, Krishnamoorthy M (2005a) Hub arc location problems: Part i-introduction and results. Manag Sci 51(10):1540–1555 Campbell JF, Ernst AT, Krishnamoorthy M (2005b) Hub arc location problems: Part ii-formulations and optimal algorithms. Manag Sci 51(10):1556–1571

11. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Index  Hub arc location problems  Formulation based on properties of supermodular functions  Comparison of formulations  Some computational results

12. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Hub arc location problems G=(V, E) complete undirected graph n=|V|, m=|E|=n(n-1)/2 K={(i, j)  V  V: there is demand between i y j}; k  K commodity i i j q : Maximum (exact) number of hub arcs p : Maximum (exact) number of hub nodes  Commodities demand is routed via hub arcs  If an arc hub is set-up then hub nodes are also established at both endnodes

13. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre d ij : unit routing cost from i to ji, j  V g e : set-up cost for hub arc ee  E c u : set-up cost for hub vertex uu  V F ek : routing cost for commodity k  K via hub arc e=(u,v) k  K, e  E i i j gege u v i j d ij To find:  Hub arcs to set-up  Assignment of commodities to hub arcs Such that the overall cost is minimized Hubs set-up cost (both arcs and nodes) + Commodities routing costs F ek = W ij (  d iu +  d uv +  d vj ) cucu Hub arc location problems

14. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Hub arc location problems i i j General model: If we allow G to have loops, then we can locate both hub arcs and independent hub nodes.

15. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre  q <  p(p-1)/2  unfeasible (if exactly p hub nodes must be open)  q = p(p-1)/2 y g e =0,  e p -hub (nodes) location problem  q ≥ min{ m, p(p-1)/2 } the constraint on the number of hub arcs is redundant  If c u =0  u, and p ≥ 2q, problem of locating only hub arcs. Hub arc location problems

16. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation I Variables: |E| + |V| + |E||K|

17. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation I  Extension of UFLP  Many variables ( x ek 4-index variables)  (|K|+2)(1+|E|) constraints

18. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation II (based on propertirs of supermodular functions) Hub arc minimization problem Minimization of supermodular function Maximization of submodular functions Nemhauser, Wolsey, Maximizing submodular set functions: formulations and analysis of algorithms, in P. Hansen, ed., Studies on Graphs and Discrete Programming, N-H (1981)

19. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Supermodular functions Let E be a finite set, and f : P ( E )  ℝ Definition: f is supermodular if f(S  T) + f(S  T) ≥ f(S) + f(T)  S, T  E Characterization: f supermodular  f(S  {e}) - f(S)  f(S  {e’, e}) - f(S  {e’}) Characterization: f supermodular and non-increasing  f(T) ≥ f(S) +  e  T\S [ f(S  {e}) - f(S)]  S, T  E The maximization of supermodular functions is “easy” The minimization of supermodular functions is “difficult”

20. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre For S ⊆ E h k (S) = Min e’  S F e’k assignment cost associated with k  K Proposition: h k is supermodular and non-increasing, for all k  K Supermodular functions Optimization problem: Find T* such that h k (T*) =Min S ⊆ E h k (S) Find Min  k = h k (T*)  k ≥ h k (S) +  e  T*\S  e k (S) for all S  E Find (z e ) e  E, z e  {0,1} s.t. Min  k  k ≥ h k (S) +  e  S  e k (S) z e for all S  E Find (z e ) e  E, z e  {0,1} s.t. Min  k  k ≥ (Min e’  S F e’k )+  e  S (F ek -Min e’  S F e’k ) - z e for all S  E Corollary: h k (T) ≥ h k (S) +  e  T\S  e k (S) for all k  K, S, T  E whereh k (S  {e})- h k (S )= (F ek - Min e’  S F e’k ) - (a) - =min {a, 0}

21. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation II Min  k  k ≥ (Min e’  S F e’k ) +  e  S (F ek - Min e’  S F e’k ) - z e for all S  E Remark: Even if there is an exponential number of constraints (subsets S) there is a small number of possible values of Min e’  S F e’k. The candidate values are F e h k for e h  E. Find (z e ) e  E, z e  {0,1} s.t. Min  k  k ≥ F e h k +  e  S (F ek - F e h k ) - z e for all e h  E

22. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation II (based on propertirs of supermodular functions) f(S) = g(S)+c(V(S))+  k  K h k (S) supermodular g(S) =  e  E g e supermodular ĉ(S) = c(V(S))=  u  V(S) c u supermodular h k (S) =Min e  S F ek supermodular and non-increasing S⊆ E,S⊆ E, Hub arc minimization problem Minimization of supermodular function

23. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation II  |K|+|E|+|V| variables (variables with1-2 indices)  (|K|+2)(1+|E|) constraints “Saving” in allocation cost for using additional hub arc e

24. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation I vs Formulation II  |K|+|E|+|V|  (|K|+2)(1+|E|)  |E||K|+|E|+|V|  (|K|+2)(1+|E|) Variables Constraints Theorem: (LP bounds ) v LPF1 =v LPF2 FIFII

25. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation I vs Formulation II: APset InstanceNatural FormulationSupermodular Formulation |V|  LP bound LP % gap Final % gap Time (sec)nodesLP bound LP % gap Final % gap Time (sec)nodes 100.21677085.210.004.801191677085.210.0020.00101 100.51731043.840.002.00571731043.840.009.5053 100.81761582.150.001.00211761582.150.005.8029 200.21885010.800.0040.00131885010.800.001136.0035 200.51940950.330.009.9051940950.330.0090.609 200.81947370.00 3.1001947370.00 17.900 250.21917170.520.00303.40211917170.520.3510800.0032 250.51961650.660.0088.30111961650.660.00811.0013 250.81973870.310.0037.1071973870.310.00258.605 400.21964492.682.3010800.0028memory 400.5200711.450.00 828.700memory 400.8200711.450.00 407.600memory 500.2---10800.00-memory 500.5200436.80.380.2710800.0053memory 500.8201074.020.060.003156.305memory p=3, q=9, Xpress, CPU limit:3 hours g e = ( c u + c v ) coeff, e=(u,v); coeff = 0.15

26. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation I

27. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Formulation II Separate Given ¿  k  K, h t.q. ? Brute force: |K||E| (O(|V 4 |)

28. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre  x   0 Separation problem Given a polyhedron P and a point x*, to identify if x*  P. If it does not, to find a valid inequality for P,  x   0 such that  x*>  0. x*  x   0

29. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Separation of constraints Given to know if there exists k  K, h s.t. Proposition: For k given, the maximum of S hk, is attained for h=r k Concave, piecewise linear; with break values F ek (k fixed) First index such that the slope is no longer positive

30. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Preliminary Results nodesalpha LP %gap Final % gap Times (sec) Nodes LP % gap Final % gap Times (sec) nodes 100.25.2104.81195.2101.0343 100.53.8402573.8400.8833 100.82.1501212.1500.5521 200.20.80040130.8006.6533 200.50.3309.950.3302.469 200.80.0003.100.0001.080 250.20.520303.4210.52011.057 250.50.66088.3110.6609.11711 250.80.31037.170.3104.375 400.22.682.310800282.680923.4124 400.50.000828.700.00062.360 400.80.000407.600.00029.370 500.2--10800-1.2409438.8265 500.50.380.2710800530.380810.5959 500.80.0603156.350.060145.235

31. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre Preliminary Results nodesalphaLP % gap Final % gapTimes(sec)nodes 600.2memory2.1006887.43183 600.5memory1.560183059 600.8memory0.960931.2235 750.2memory1.721.261348374 750.5memory1.17012216105 750.8memory0.9805922.3955 900.2memory0.80030479.695 900.5memory8.418.272129358 900.8memory0.6905822.4533 1000.2memory- -10800 - 1000.5memory0.63027033.755 1000.8memory9.178.813820028

32. Location problems on networks with routing ▪ E Fernández ▪ TGS 2010 ▪ Jarandilla 12-15 Noviembre  Arc hub location problems (involve routing decisions)  General Problem  Two alternative formulacions  Minimization of supermodular function  Efficient solution of separation problem  Promising preliminary results Summary