lecture4网络问题

NETWORK PROBLEMS 1.224J/ESD.204J TRANSPORTATION OPERATIONS, PLANNING AND CONTROL: CARRIER SYSTEMS Professor Cynthia Barnhart Professor Nigel H.M. Wilson Fall 2003 1/2/2004 Barnhart - 1.224J 2 Announcements –C. Barnhart open office hours from 1:00-2:30 on Wednesday this week –Reference: Network Flows: Theory, Algorithms, and Applications (Ahuja, Magnanti, Orlin) 1/2/2004 Barnhart - 1.224J 3 Outline • Network Introduction • Properties of Network Problems • Minimum Cost Flow Problem • Shortest Path • Maximum Flow Problem • Assignment Problem • Transportation Problem • Circulation Problem • Multicommodity Flow problem • Matching Problem 1/2/2004 Barnhart - 1.224J 4 Networks • Very common in transportation • Can be physical networks: transportation networks such as railroads and highways) • Network flows sometimes also arise in surprising ways (problems that on the surface might not appear to involve any networks at all). • Sometimes the nodes and arcs have a temporal dimension that models activities that take place over time. Many scheduling applications have this flavor (crew scheduling; location and layout theory; warehousing and distribution; production planning and control) 1/2/2004 Barnhart - 1.224J 5 Graphs & Networks • Network problems are defined on graphs – Undirected and directed graphs G=(N, A) • N=set of nodes • A=set of feasible links/ arcs – Trees (connected graph; no cycles) – Bipartite graphs: two sets of nodes with arcs that join only nodes between the 2 sets • Additional numerical information such as: – b(i) representing supply and demand at each node i; – uij representing the capacity of each arc ij – lij representing the lower bound on flow for each arc ij – cij representing the cost of each arc ij. 5 1 3 4 2 Undirected Graph 5 1 3 4 2 Directed Graph b(4) c12, u12 b(2)b(1) c21, u21 c25, u25 c54, u54 c45, u45c35, u35 c12, u12 b(3) b(5) 5 1 3 4 2 Network 1/2/2004 Barnhart - 1.224J 6 Formulating network flow problems • Concise formulation • Node arc incidence matrix: • Columns = arcs • Rows = nodes • Outgoing arc ⇒ +1 • Incoming arc ⇒ -1 • No arc incident ⇒ 0 • Sum of rows of A = 0 (1) Balance constraints: flow out minus flow in must equal the supply/demand at the node (2) Flow lower bound constraints (usually lower bound is not stated and equal to 0) (3) Capacity constraints (4) Integrality constraints { } { } )4....(..............................),(, )3....(..............................),(, )2.....(..............................),(, )1..(..........),( . ),(:),(: ),( AjiZX AjiUX AjiLX NiibXX ts XCMinimize ij ijij ijij Aijj ij Ajij ij Aji ijij ∈∀∈ ∈∀≤ ∈∀≥ ∈∀=− + ∈∈ ∈ ∑∑ ∑ 1/2/2004 Barnhart - 1.224J 7 Properties of Network Problems • Solving the LP relaxation of network problems with integer problem data, yields an integer solution • Network problems are special cases of LPs and any algorithm for a LP can be directly applied. • Network flow problems have a special structure which results in substantial simplification of general methods (Network Simplex) • Many algorithms used to solve network problems (discussed in 1.225) • Multicommodity problems are more complicated 1/2/2004 Barnhart - 1.224J 8 Minimum Cost Flow Problem • Objective: determine the least cost movement of a commodity through a network in order to satisfy demands at certain nodes from available supplies at other nodes. • Applications: – distribution of a product from manufacturing plants to warehouses, or from warehouses to retailers; – the flow of raw material and intermediate goods through the various machining stations in a production line; – the routing of vehicles through an urban street network; • G=(N,A): directed network defined by a set N of nodes and a set A of m directed arcs. – Cij: cost per unit flow on arc (i,j) A. – Uij: capacity; maximum amount that can flow on arc (i,j) – Lij: lower bound; minimum amount that must flow on the arc – B(i): supply or demand at node i N • If b(i)>0 => node i is a supply node • If b(i)<0 => node i is a demand node • If b(i)=0 => node i is a transshipment node – Xij: decision variables; represent the quantity of flow on arc (i,j)∈A ∈ ∈ 1/2/2004 Barnhart - 1.224J 9 Example: Minimum Cost Flow Warehouse distribution problem Company A currently serves its 4 customers from 3 warehouses. It costs $cij to transport a unit from warehouse i to the customer j. Transportation from the plant P to the warehouses is free. Transportation of the products from the warehouse to the customers is done by truck. Company A cannot send more than 100 units of product from each warehouse to each customer. Finally, there is a demand for Dj units of the product in region j. Company A would like to determine how many units of product they should store at each warehouse and how many units of product to send from each warehouse to each customer, in order to minimize costs. W1 W2 W3 A B C D P Da=100 Db=50 Dc=80 Dd=170 S=400 Cij A B C D W1 5 2 6 7 W2 4 4 3 1 W3 3 8 5 3 1/2/2004 Barnhart - 1.224J 10 Approach • Decision variables? – Xij= flow on each arc • Objective Function: • Conservation of Flow Constraints: • Arc capacity constraints: There is a limit such that a given warehouse-customer route (i,j) can be used by at most uij units. ∑ ∈Aji ijij xCMIN ),( * { }{ }∑ ∑∈ ∈ ∈∀=−Ajij Aijj ijiij Nibxx),(: ),(: , 0≥ijx Ajiux ijij ∈∀≤ ),(, 1/2/2004 Barnhart - 1.224J 11 Warehouse Distribution-Formulation 0,,,,,,,,,,,, 100,,,,,,,,, 170 80 50 100 0 0 0 400 .. 358313447625( 3333,2222,1111321 3333,2222,1111 321 321 321 321 33333 22222 11111 321 333322221111 ≥ ≤ =++ =++ =++ =++ =−−−− =−−−− =−−−− −=−−− +++++++++++ DCBADCBADCBAPPP DCBADCBADCBA DDD CCC BBB AAA DCBAP DCBAP DCBAP PPP DCBADCBADCBA XXXXXXXXXXXXXXX XXXXXXXXXXXX XXX XXX XXX XXX XXXXX XXXXX XXXXX XXX ts XXXXXXXXXXXXMin Cij 0 0 0 5 2 6 7 4 4 3 1 3 8 5 3 P-W1 P-W2 P-W3 W1-A W1-B W1-C W1-D W2-A W2-B W2-C W2-D W3-A W3-B W3-C W3-D P -1 -1 -1 -400 W1 1 -1 -1 -1 -1 0 W2 1 -1 -1 -1 -1 0 W3 1 -1 -1 -1 -1 0 A 1 1 1 100 B 1 1 1 50 C 1 1 1 80 D 1 1 1 170 N o d e s Node- Arc Matrix Di 1/2/2004 Barnhart - 1.224J 12 In OPL Studio… 1/2/2004 Barnhart - 1.224J 13 Transportation Problem • Objective: Transport the goods from the suppliers to the consumers at minimum cost given that: – there are m suppliers and n consumers (m can be different from n). – The ith supplier can provide si units of a certain good and the jth consumer has a demand for dj units. – We assume that total supply equals total demand. • Applications: distribution of goods from warehouses to customers. • G=(N1 U N2, A): directed network defined by a set N1+N2 of nodes and a set A of m*n directed arcs. – N1: supply nodes; N2: demand nodes; |N1|=m; |N2|=n; – (i,j)∈A such that i∈N1 and j∈N2 – Cij: unit transportation cost of transporting goods from supplier i to consumer j, per unit flow on arc (i,j) ∈A. – Uij: capacity ; maximum amount that can flow on arc (i,j) – Lij: lower bound; minimum amount that must flow on the arc – B(i): • b(i)= si for all i ∈N1 • b(i)= -dj for all j ∈N2 – Xij: decision variables; represent the quantity of goods flowing from i to j 1/2/2004 Barnhart - 1.224J 14 Assignment Problem • Objective: Pair, at minimum possible cost, each object in set N1 with exactly one object in set N2. • Special case of the transportation problem where number of suppliers equals number of customers and each supplier has unit supply, each consumer has unit demand. • Applications: assigning people to projects, truckloads to truckers, jobs to machines, tenants to apartments, swimmers to events in a swimming set, school graduates to internships. • G=(N1 U N2,A): directed network defined by a set N1+N2 of nodes and a set A of m directed arcs. – |N1|=|N2|=m; – (i,j) ∈ A such that i∈N1 and j∈N2 – Cij: cost per unit flow on arc (i,j)∈A. – Uij:=1 for all (i,j)∈A – Lij: lower bound; minimum amount that must flow on the arc – B(i): supply or demand at node i∈N • B(i) =1 for all i ∈ N1 • B(i) =-1 for all i ∈ N2 – Xij: decision variables; represent the quantity flow on arc (i,j) ∈ A 1/2/2004 Barnhart - 1.224J 15 Assignment Problem Trucking company TC needs to pick up 4 loads of products at different locations. Currently, 4 truck drivers are available to pick up those shipments. The cost of having driver i pick-up shipment j is illustrated in the above table. Formulate the problem of assigning each driver to a load, in order to minimize costs T1 T2 T3 T4 Load 1 Load 2 Load 3 Load 4 Cij 1 2 3 4 1 6 4 5 n/a 2 n/a 3 6 n/a 3 5 n/a 4 3 4 7 5 5 5 1/2/2004 Barnhart - 1.224J 16 Assignment problem- Formulation Cij 6 4 5 3 6 5 4 3 7 5 5 5 1-1 1-2 1-3 2-2 2-3 3-1 3-3 3-4 4-1 4-2 4-3 4-4 T1 -1 -1 -1 -1 T2 -1 -1 -1 T3 -1 -1 -1 -1 T4 -1 -1 -1 -1 -1 L1 1 1 1 1 L2 1 1 1 1 L3 1 1 1 1 1 L4 1 1 1 Node- Arc Matrix bi N o d e s 0,,,,,,,,, 1,,,,,,,,, 1 1 1 1 1 1 1 1 .. )555734563546( 44434241,34333123,22131211 44434241,34333123,22131211 4434 43332313 422212 413111 44434241 343331 2322 131211 444342413433312322131211 ≥ ≤ =+ =+++ =++ =++ −=−−−− −=−−− −=−− −=−−− +++++++++++ XXXXXXXXXXXX XXXXXXXXXXXX XX XXXX XXX XXX XXXX XXX XX XXX ts XXXXXXXXXXXXMin 1/2/2004 Barnhart - 1.224J 17 In OPL Studio 1/2/2004 Barnhart - 1.224J 18 Shortest Path •Objective: Find the path of minimum cost (or length) from a specified source node s to another specified sink t, assuming that each arc (i,j) ∈ A has an associated cost (or length) cij. • Applications: • project scheduling; cash flow management; message routing in communication systems; traffic flow through congested city. • G=(N,A): directed network defined by a set N of nodes and a set A of m directed arcs. –Cij: cost per unit flow on arc (i,j) ∈ A. –Uij: capacity ; maximum amount that can flow on arc (i,j) –Lij:=0 for all (i,j) –b(i): supply or demand at node i ∈ N •b(s)=1 ; b(t)=-1; • b(i)=0 for all other nodes –Xij: decision variables; represent the quantity flow on arc (i,j) ∈ A • If want to determine shortest path from source to all every other node in the system, then: b(s)=(n-1); b(i)=-1 for all other nodes. 1/2/2004 Barnhart - 1.224J 19 Example: Product Distribution A producer wishes to distribute a sample of its product to product testers. In order to minimize the cost of this distribution, it has decided to use its existing supply network. The goal then is to minimize the distance traveled by each sample. The network G consists of a set of nodes, N, between which the distribution is non-stop. Let |N|=n; we want to deliver a sample to each of the n-1 cities in our distribution network. We denote the set of these connections, or arcs, as A, and the travel time along such a connection, starting at some point i and connecting to a point j, by Tij. If we number the plant as node 0, provide a formulation of the problem given that we must deliver only one sample to each destination. 1/2/2004 Barnhart - 1.224J 20 Example: Network Representation Network Representation n=4 Equivalent to shortest path problem 0 1 3 2 S=n-1=3 D=1 D=1D=1 1/2/2004 Barnhart - 1.224J 21 Product Distribution: Formulation • Decision variables? – Xij= flow on each arc • Objective Function: • Constraints: (conservation of flow and flow non-negativity) ∑ ∈Aji ijij xTMIN ),( * { }{ }∑ ∑∈ ∈ ∈∀=−Ajij Aijj ijiij Nibxx),(: ),(: , 0≥ijx 1/2/2004 Barnhart - 1.224J 22 Product Distribution--Application COSTS Tij 1 2 3 0 16 24 55 1 n/a 32 30 2 15 n/a 17 3 13 11 n/a TO j F R O M i N={0,1,2,3} 0,,,,,, 1111010100 1101101010 1010111001 3000000111 .. )111317153032552416( 32312321131203,02,01 323123211312030201 323123211312030201 323123211312030201 323123211312030201 323123211312030201 ≥ =−−++++++ =++−−++++ =++++−−++ −=++++++−−− ++++++++ xxxxxxxxx xxxxxxxxx xxxxxxxxx xxxxxxxxx xxxxxxxxx ts xxxxxxxxxMIN COSTS Tij 16 24 55 32 30 15 17 13 11 Node/Arcs 0-1 0-2 0-3 1-2 1-3 2-1 2-3 3-1 3-2 0 -1 -1 -1 0 0 0 0 0 0 -3 1 1 0 0 -1 -1 1 0 1 0 1 2 0 1 0 1 0 -1 -1 0 1 1 3 0 0 1 0 1 0 1 -1 -1 1 Node-Arc Matrix bi 1/2/2004 Barnhart - 1.224J 23 In OPL… 1/2/2004 Barnhart - 1.224J 24 Maximum Flow Problem • Objective: Find a feasible flow that sends the maximum amount of flow from a specified source node s to another specified sink t. • Maximum flow problem incurs no cost but is restricted by arc/ node capacities. • Applications: – determining the maximum steady state flow of petroleum products in a pipeline network, cars in a road network, messages in a telecom network; electricity in an electrical network. • G=(N,A): directed network defined by a set N of nodes and a set A of m directed arcs. – Cij=0 for all (i,j) in A; introduce arc from t to s such that Cts=-1 – Uij; Uts= unlimited – Lij: lower bound – b(i)=0 for all i ∈N – Xij: decision variables; represent the quantity of flow on arc (i,j) ∈A • Solution: maximize the flow on arc (t,s), but any flow on arc (t,s) must travel from node s to node t through the arcs in A [because each b(i)=0]. Thus, the solution to the minimum cost flow problem will maximize the flow from s to t. 1/2/2004 Barnhart - 1.224J 25 Circulation Problem • Objective: Find a feasible flow that honors the lower and upper bounds Lij and Uij imposed on arc flows Xij. Since we never introduce any exogenous flow into the network or extract any flow from it, all the flow circulates around the network. We wish to find the circulation that has the minimum cost. • Minimum cost flow problem with only transshipment nodes. • Applications: design of routing schedule for a commercial airline where bound Lij=1 on arc (i,j) if airline needs to provide service between cities i and j , and so must dispatch an airplane on this arc (actually the nodes will represent a combination of both a physical location and a time of day). • G=(N,A): directed network defined by a set N of nodes and a set A of m directed arcs. – Cij: cost per unit flow on arc (i,j)∈A. – Uij: capacity ; maximum amount that can flow on arc (i,j) – Lij: lower bound; minimum amount that must flow on the arc – B(i) = 0 for all i∈N – Xij: decision variables; represent the quantity flow on arc (i,j)∈A 1/2/2004 Barnhart - 1.224J 26 Multicommodity Flow Problem • Minimum cost flow problem => deals with single commodity over a network. • Multicommodity flow problems => several commodities use the same underlying network. – Objective: Allocate the capacity of each arc to the individual commodities in a way that minimizes overall costs. – Commodities may either be differentiated by their physical characteristics or simply by their origin destination pairs. – Different commodities have different origins and destinations, and commodities have separate balance constraints at each node. – Commodities share the arc capacities • Applications: transportation of passengers from different origins to different destinations within a city; routing of non-homogeneous tankers; worldwide shipment of different varieties of grains from countries that produce grains to countries that consume it; the transmission of messages in a communication network between different origin-destination pairs. ¾ MCF problems are network flow problems with side constraints (integrality property doesn’t hold) 1/2/2004 Barnhart - 1.224J 27 Multicommodity Flow A company produces 2 types of products A and B at 3 plants (P1, P2, P3). It then ships these products to 3 market zones (M1, M2, M3). For k=1,2; i=1,2 and j=1,..,3 the following data is given: - The costs of shipping one unit of product k from plant i to zone j - The maximum number of units that can be shipped from each plant to each market zone - The demand for product k at market zone j - The distribution channel P1-M1 and P2-M3 cannot carry more than 10 and 35 units respectively Formulate the problem of minimizing transportation costs as an LP P1 P2 M1 M2 M3 A B DA1 DB1 DA2 DB2 DA3 DB3 Cij M1 M2 M3 Supply P1 6 4 5 150 P2 5 3 6 130 P1 10 3 9 140 P2 8 4 6 170 Demand A 100 70 110 Demand B 150 30 130 P r o d u c t A P r o d u c t B 1/2/2004 Barnhart - 1.224J 28 Multi-commodity Flow },{},3,2,1{},2,1{,0 10 35 }3,2,1{},,{, }2,1{},,{, )( , 32, 11, ,, ,, ,, BAkMMMjPPiX X X MMMjBAkDX PPiBAkSX XCMin ijk k MPk k MPk jk i ijk ik j ijk k i ijk j ijk ∈∀∈∀∈∀≥ ≤ ≤ =∀=∀= =∀=∀= ∑ ∑ ∑ ∑ ∑∑∑ 0, 10 35 130 110 30 70 150 100 170 130 140 150 .. )6489 31063 5546( ,, 32,32, 11,11, 33,32,31, 33,32,31, 23,22,21, 23,22,21, 13,12,11, 13,12,11, 32,22,12, 32,22,12, 31,21,11, 31,21,11, 32,22,12,31, 21,11,32,22, 12,31,21,11, ≥ ≤+ ≤+ =++ =++ =++ =++ =++ =++ =++ =++ =++ =++ ++++ ++++ +++ ijBijA MPBMPA MPBMPA MPBMPBMPB MPAMPAMPA MPBMPBMPB MPAMPAMPA MPBMPBMPB MPAMPAMPA MPBMPBMPB MPAMPAMPA MPBMPBMPB MPAMPAMPA MPBMPBMPBMPB MPBMPBMPAMPA MPAMPAMPAMPA XX XX XX XXX XXX XXX XXX XXX XXX XXX XXX XXX XXX ts XXXX XXXX XXXXMin 1/2/2004 Barnhart - 1.224J 29 Multi-commodity Cij 6 4 5 5 3 6 10 3 9 8 4 6 1-1 1-2 1-3 2-1 2-2 2-3 1-1 1-2 1-3 2-1 2-2 2-3 P1 1 1 1 = 150 P2 1 1 1 = 130 M1 -1 -1 = -100 M2 -1 -1 = -70 M3 -1 -1 = -110 P1 1 1 1 = 140 P2 1 1 1 = 170 M1 -1 -1 = -150 M2 -1 -1 = -30 M3 -1 -1 = -130 Cap P1-M1 1 1 <= 35 Cap P2-M3 1 1 <= 10 P r o d u c t A P r o d u c t B A & B Node- Arc Matrix bi Product A Product B 1/2/2004 Barnhart - 1.224J 30 In OPL… 1/2/2004 Barnhart - 1.224J 31 Matching problem • Objective: The matching seeks a matching that optimizes some criteria. • A matching in a graph G=(N,A) is a set of arcs with the property that every node is incident to at most one arc in the set; thus a matching induces a pairing of some of the nodes in the graph using the arcs in A. In a matching each node is matched with at most one other node, and some nodes might not be matched with any other node. • Bipartite matching problems: matching problems on bipartite graphs (graphs with two distinct sets of nodes): assignment and transportation problem Set Partitioning 1/2/2004 Barnhart - 1.224J 33 Example A company needs to hire extra drivers to run a special shuttle service on Saturday. The shuttle service will last from 6 AM to 6 PM, with shifts of 3 hours minimum. The transit company has found 3 potential drivers. Drivers want to be compensated as follows: Shift Type Hours Cost ($/hr) Total Cost 1 6-12 20 120 2 12-18 19 114 3 6-9 22 66 4 9-12 17 51 5 12-15 18 54 6 15-18 19 57 7 6-12 19 114 8 12-18 18 108 9 6-9 19 57 10 9-12 22 66 11 12-18 18 108 12 6-12 20 60 13 9-15 21 126 14 15-18 19 57 Driver A Driver B Driver C What should the company do in order to minimize its costs ? 1/2/2004 Barnhart - 1.224J 34 Model Cost 120 114 66 51 54 57 114 108 57 66 108 60 126 57 1 2 3 4 5 6 7 8 9 10 11 12 13 14 6-9 1 1 1 1 1 = 1 9-12 1 1 1 1 1 1 = 1 12-15 1 1 1 1 1 = 1 15-18 1 1 1 1 1 = 1 S h i f t s A B C Node- Arc Matrix bi Let S be the set of four 3-hour shifts (6-9; 9-12; 12-15; 15-18). Let P be the set of 14 combinations (a combination is defined as the combination of driver and rate charged). Let Xj=1 if package j is chosen, 0 otherwise Let δij=1 if shift i is covered by package j, 0 otherwise. The only constraint is that every shift should be worked by somebody. { } PjX SiX ts XCMin j j jij j jj ∈∀∈ ∈∀=∑ ∑ = = ,1,0 ,1 .. 14 1 14 1 δ 1/2/2004 Barnhart - 1.224J 35 In OPL…