Chapter 6 Network Optimization Problems

Chapter Six
Network Optimization
Problems
Learning Objectives
After completing this chapter, you should be able to
1. Formulate network models for various types of network optimization problems.

2. Describe the characteristics of minimum-cost flow problems, maximum flow prob-
lems, and shortest path problems.
3. Identify some areas of application for these types of problems.
4. Identify several categories of network optimization problems that are special types of
minimum-cost flow problems.
5. Formulate and solve a spreadsheet model for a minimum-cost flow problem, a maxi-
mum flow problem, or a shortest path problem from a description of the problem.
Networks arise in numerous settings and in a variety of guises. Transportation, electrical, and
communication networks pervade our daily lives. Network representations also are widely
used for problems in such diverse areas as production, distribution, project planning, facilities
location, resource management, and financial planning—to name just a few examples. In fact,
a network representation provides such a powerful visual and conceptual aid for portraying
the relationships between the components of systems that it is used in virtually every field of
scientific, social, and economic endeavor.
One of the most exciting developments in management science in recent decades has
been the unusually rapid advance in both the methodology and application of network opti-
mization problems. A number of algorithmic breakthroughs have had a major impact, as
have ideas from computer science concerning data structures and efficient data manipu-
lation. Consequently, algorithms and software now are available and are being used to
solve huge problems on a routine basis that would have been completely intractable a few
decades ago.
This chapter presents the network optimization problems that have been particularly
helpful in dealing with managerial issues. We focus on the nature of these problems and
their applications rather than on the technical details and the algorithms used to solve the
problems.
You already have seen some examples of network optimization problems in Chapter 3. In
particular, transportation problems (described in Section 3.5) have a network representation,
as illustrated in Figure 3.9, and assignment problems (Section 3.6) have a similar network
representation (as described in Chapter 15 on the CD-ROM). Therefore, both transportation
problems and assignment problems are simple types of network optimization problems.
Like transportation problems and assignment problems, many other network optimization
problems (including all the types considered in this chapter) also are special types of linear
programming problems. Consequently, after formulating a spreadsheet model for these prob-
lems, they can be readily solved by Solver.
Section 6.1 discusses an especially important type of network optimization problem called
a minimum-cost flow problem. A typical application involves minimizing the cost of shipping
goods through a distribution network.
194
6.1 Minimum-Cost Flow Problems 195
Section 6.3 presents maximum flow problems, which are concerned with such issues as
how to maximize the flow of goods through a distribution network. Section 6.2 lays the
groundwork by introducing a case study of a maximum flow problem.
Section 6.4 considers shortest path problems. In their simplest form, the objective is to
find the shortest route between two locations.
A supplement to this chapter on the CD-ROM discusses minimum spanning-tree prob-
lems, which are concerned with minimizing the cost of providing connections between all
users of a system. This is the only network optimization problem considered in this book that
is not, in fact, a special type of linear programming problem.
6.1 MINIMUM-COST FLOW PROBLEMS

Before describing the general characteristics of minimum-cost flow problems, let us first look
at a typical example.
An Example: The Distribution Unlimited Co. Problem

The Distribution Unlimited Co. has two factories producing a product that needs to be
shipped to two warehouses. Here are some details.
Factory 1 is producing 80 units.
Factory 2 is producing 70 units.
Warehouse 1 needs 60 units.
Warehouse 2 needs 90 units.
(Each unit corresponds to a full truckload of the product.)
Figure 6.1 shows the distribution network available for shipping this product, where F1 and
F2 are the two factories, W1 and W2 are the two warehouses, and DC is a distribution center.
The arrows show feasible shipping lanes. In particular, there is a rail link from Factory 1 to
Warehouse 1 and another from Factory 2 to Warehouse 2. (Any amounts can be shipped along
these rail links.) In addition, independent truckers are available to ship up to 50 units from
each factory to the distribution center, and then to ship up to 50 units from the distribution
center to each warehouse. (Whatever is shipped to the distribution center must subsequently
be shipped on to the warehouses.) Management’s objective is to determine the shipping plan
The objective is to mini- (how many units to ship along each shipping lane) that will minimize the total shipping cost.
mize the total shipping cost The shipping costs differ considerably among these shipping lanes. The cost per unit shipped
through the distribution
through each lane is shown above the corresponding arrow in the network in Figure 6.2.
network.
To make the network less crowded, the problem usually is presented even more com-
pactly, as shown in Figure 6.3. The number in square brackets next to the location of each
FIGURE 6.1
The distribution network
80 units 60 units
for the Distribution produced F1 W1 needed
Unlimited Co. problem,
where each feasible ship-
ping lane is represented
by an arrow.
DC
70 units 90 units
produced F2 W2 needed
196 Chapter Six Network Optimization Problems
FIGURE 6.2 80 units $700/unit 60 units

The data for the distribu- produced F1 W1 needed
tion network for the Dis-
$3
[5
tribution Unlimited Co.
its nit
00 s m
0
.]
ax
un
/u
un 0/u
problem.
m
ni x . ]
it
0
t
$2
a
0
[5
DC
$4
[5
its nit
.]
00 s m
0
ax
un 0/u
un
/u
m
0
ni x .]
it
$5
t
a
0
[5
70 units $1,000/unit 90 units
produced F2 W2 needed
facility indicates the net number of units (outflow minus inflow) generated there. Thus, the
number of units terminating at each warehouse is shown as a negative number. The number at
the distribution center is 0 since the number of units leaving minus the number of units arriv-
ing must equal 0. The number on top of each arrow shows the unit shipping cost along that
shipping lane. Any number in square brackets underneath an arrow gives the maximum num-
ber of units that can be shipped along that shipping lane. (The absence of a number in square
brackets underneath an arrow implies that there is no limit on the shipping amount there.)
This network provides a complete representation of the problem, including all the necessary
Figure 6.3 illustrates how data, so it constitutes a network model for this minimum-cost flow problem.
a minimum-cost flow Since this is such a tiny problem, you probably can see what the optimal solution must be.
problem can be completely
(Try it.) This solution is shown in Figure 6.4, where the shipping amount along each shipping
depicted by a network.
lane is given in parentheses. (To avoid confusion, we delete the unit shipping costs and ship-
ping capacities in this figure.) Combining these shipping amounts with the unit shipping costs
given in Figures 6.2 and 6.3, the total shipping cost for this solution (when starting by listing
the costs from F1, then from F2, and then from DC) is
Total shipping cost 5 30($700) 1 50($300) 1 30($500) 1 40($1,000)
1 30($200) 1 50($400)
5 $117,000
[80] [–60] [80] [–60]

$700 (30)
F1 W1 F1 W1
$3 0]
(5
[5 00
00
0)
0)
[5
$2
(3
0]
[0] [0]
DC DC
$4 0]
(5
[5 00
0)
00
0)
$5
(3
[5
0]
$1,000 (40)
F2 W2 F2 W2
[70] [–90] [70] [–90]
FIGURE 6.3 FIGURE 6.4

A network model for the Distribution The optimal solution for the Distribution Unlim-
Unlimited Co. problem as a minimum-cost ited Co. problem, where the shipping amounts
flow problem. are shown in parentheses over the arrows.
An Application Vignette
Hewlett-Packard (HP) offers many innovative products to envisioned as flows through a network that can help fill
meet the diverse needs of more than 1 billion customers. some of the projected orders on the right-hand side of the
The breadth of its product offering has helped the com- network. The resulting model is a special type of minimum
pany achieve unparalleled market reach. However, offering cost flow problem (related to the special type discussed in
multiple similar products also can cause serious problems— the next two sections).
including confusing sales representatives and customers— Following its implementation by the beginning of 2005,
that can adversely affect the revenue and costs for any par- this application of a minimum cost flow problem had a dra-
ticular product. Therefore, it is important to find the right matic impact in enabling HP businesses to increase oper-
balance between too much product variety and too little. ational focus on their most critical products. This yielded
With this in mind, HP top management made managing companywide profit improvements of over $500 million
product variety a strategic business priority. HP has been between 2005 and 2008, and then about $180 million
a leader in applying management science to its important annually thereafter. It also yielded a variety of important
business problems for decades, so it was only natural that qualitative benefits for HP.
many of the company’s top management scientists were These dramatic results led to HP winning the prestigious
called on to address this problem as well. first prize in the 2009 Franz Edelman Award for Achieve-
The heart of the methodology that was developed to ment in Operations Research and the Management Sciences.
address this problem involved formulating and apply-
Source: J. Ward and 20 co-authors, “HP Transforms Product Port-
ing a network optimization model. After excluding pro- folio Management with Operations Research,” Interfaces 40,
posed products that do not have a sufficiently high return no. 1 (January–February 2010), pp. 17–32. (A link to this article is
on investment, the remaining proposed products can be provided on our website, www.mhhe.com/hillier5e.)
General Characteristics
This example possesses all the general characteristics of any minimum-cost flow problem.
Before summarizing these characteristics, here is the terminology you will need.
Terminology
1. The model for any minimum-cost flow problem is represented by a network with flow
passing through it.
2. The circles in the network are called nodes.
A supply node has net 3. Each node where the net amount of flow generated (outflow minus inflow) is a fixed posi-
flow going out whereas a tive number is a supply node. (Thus, F1 and F2 are the supply nodes in Figure 6.3.)
demand node has net flow
coming in. 4. Each node where the net amount of flow generated is a fixed negative number is a demand
node. (Consequently, W1 and W2 are the demand nodes in the example.)
5. Any node where the net amount of flow generated is fixed at zero is a transshipment
node. (Thus, DC is the transshipment node in the example.) Having the amount of flow
out of the node equal the amount of flow into the node is referred to as conservation of
flow.
6. The arrows in the network are called arcs.
7. The maximum amount of flow allowed through an arc is referred to as the capacity of
that arc.
Using this terminology, the general characteristics of minimum-cost flow problems (the
model for this type of problem) can be described in terms of the following assumptions.
Assumptions of a Minimum-Cost Flow Problem

1. At least one of the nodes is a supply node.
2. At least one of the other nodes is a demand node.
Since the arrowhead on an 3. All the remaining nodes are transshipment nodes.
arc indicates the direction 4. Flow through an arc is only allowed in the direction indicated by the arrowhead, where the
in which flow is allowed, maximum amount of flow is given by the capacity of that arc. (If flow can occur in both
a pair of arcs pointing in directions, this would be represented by a pair of arcs pointing in opposite directions.)
opposite directions is used
if flow can occur in both 5. The network has enough arcs with sufficient capacity to enable all the flow generated at
directions. the supply nodes to reach all the demand nodes.
197
6. The cost of the flow through each arc is proportional to the amount of that flow, where the
cost per unit flow is known.
7. The objective is to minimize the total cost of sending the available supply through the net-
work to satisfy the given demand. (An alternative objective is to maximize the total profit
The objective is to mini- from doing this.)
mize the total cost of sup-
plying the demand nodes. A solution for this kind of problem needs to specify how much flow is going through each
arc. To be a feasible solution, the amount of flow through each arc cannot exceed the capacity
of that arc and the net amount of flow generated at each node must equal the specified amount
for that node. The following property indicates when the problem will have feasible solutions.
Feasible Solutions Property: Under the above assumptions, a minimum-cost flow problem will
have feasible solutions if and only if the sum of the supplies from its supply nodes equals the sum
of the demands at its demand nodes.
Note that this property holds for the Distribution Unlimited Co. problem, because the sum of
its supplies is 80 1 70 5 150 and the sum of its demands is 60 1 90 5 150.
For many applications of minimum-cost flow problems, management desires a solution
with integer values for all the flow quantities (e.g., integer numbers of full truckloads along
each shipping lane). The model does not include any constraints that require this for feasible
solutions. Fortunately, such constraints are not needed because of the following property.
Integer Solutions Property: As long as all its supplies, demands, and arc capacities have inte-
ger values, any minimum-cost flow problem with feasible solutions is guaranteed to have an
optimal solution with integer values for all its flow quantities.
See in Figure 6.3 that all the assumptions needed for this property to holds are satisfied for
the Distribution Unlimited Co. problem. In particular, all the supplies (80 and 70), demands
(60 and 90), and arc capacities (50) have integer values. Therefore, all the flow quantities in
the optimal solution given in Figure 6.4 (30 three times, 50 two times, and 40) have integer
values. This ensures that only full truckloads will be shipped into and out of the distribution
center. (Remember that each unit corresponds to a full truckload of the product.)
Now let us see how to obtain an optimal solution for the Distribution Unlimited Co. prob-
lem by formulating a spreadsheet model and then applying Solver.
Using Excel to Formulate and Solve Minimum-Cost Flow Problems

Figure 6.5 shows a spreadsheet model that is based directly on the network representation of
the problem in Figure 6.3. The arcs are listed in columns B and C, along with their capacities
(unless unlimited) in column F and their costs per unit flow in column G. The changing cells
Ship (D4:D9) show the flow amounts through these arcs and the objective cell TotalCost
(D11) provides the total cost of this flow by using the equation
Capacity constraints like D11 5 SUMPRODUCT(Ship, UnitCost)
these are needed in any
minimum-cost flow prob- The first set of constraints in the Solver Parameters box, D5:D8 # Capacity (F5:F8), ensures
lem that has any arcs with that the arc capacities are not exceeded.
limited capacity. Similarly, Column I lists the nodes, column J calculates the actual net flow generated at
each node (given the flows in the changing cells), and column L specifies the net amount of
Any minimum-cost flow
problem needs net flow flow that needs to be generated at each node. Thus, the second set of constraints in the Solver
constraints like this for Parameters box is NetFlow (J4:J8) 5 SupplyDemand (L4:L8), requiring that the actual net
every node. amount of flow generated at each node must equal the specified amount.
The equations entered into NetFlow (J4:J8) use the difference of two SUMIF functions to
Excel Tip: SUMIF(A, B, C) calculate the net flow (outflow minus inflow) generated at each node. In each case, the first
adds up each entry in the SUMIF function calculates the flow leaving the node and the second one calculates the flow
range C for which the cor-
entering the node. For example, consider the F1 node (I4). SUMIF(From,I4,Ship) sums each
responding entry in range
A equals B. This function is individual entry in Ship (D4:D9) if that entry is in a row where the entry in From (B4:B9)
especially useful in network is the same as in I4. Since I4 5 F1 and the only rows that have F1 in the From column
problems for calculating the are rows 4 and 5, the sum in the Ship column is only over these same rows, so this sum
net flow generated at a node. is D4 1 D5. Similarly, SUMIF(To,I4,Ship) sums each individual entry in Ship (D4:D9) if
FIGURE 6.5
A spreadsheet model for the Distribution Unlimited Co. minimum-cost flow problem, including the objective cell TotalCost (D11)
and the other output cells NetFlow (J4:J8), as well as the equations entered into these cells and the other specifications needed to set
up the model. The changing cells Ship (D4:D9) show the optimal shipping quantities through the distribution network obtained by
Solver.
A B C D E F G H I J K L
1 Distribution Unlimited Co. Minimum Cost Flow Problem
2
3 From To Ship Capacity Unit Cost Nodes Net Flow Supply/Demand
4 F1 W1 30 $700 F1 80 = 80
5 F1 DC 50 ≤ 50 $300 F2 70 = 70
6 DC W1 30 ≤ 50 $200 DC 0 = 0
7 DC W2 50 ≤ 50 $400 W1 -60 = -60
8 F2 DC 30 ≤ 50 $500 W2 -90 = -90
9 F2 W2 40 $1,000
10
11 Total Cost $117,000
Solver Parameters Range Name Cells J

Set Objective Cell: TotalCost 3 Net Flow
To: Min Capacity F5:F8
By Changing Variable Cells: From B4:B9 4 =SUMIF(From,I4,Ship)-SUMIF(To,I4,Ship)
Ship NetFlow J4:J8 5 =SUMIF(From,I5,Ship)-SUMIF(To,I5,Ship)
Subject to the Constraints: Nodes I4:I8
D5:D8 <= Capacity Ship D4:D9 6 =SUMIF(From,I6,Ship)-SUMIF(To,I6,Ship)
NetFlow = SupplyDemand SupplyDemand L4:L8 7 =SUMIF(From,I7,Ship)-SUMIF(To,I7,Ship)
To C4:C9
Solver Options: 8 =SUMIF(From,I8,Ship)-SUMIF(To,I8,Ship)
TotalCost D11
Make Variables Nonnegative
Solving Method: Simplex LP UnitCost G4:G9
C D
11 Total Cost =SUMPRODUCT(Ship,UnitCost)
that entry is in a row where the entry in To (C4:C9) is the same as in I4. However, F1 never
appears in the To column, so this sum is 0. Therefore, the overall equation for J4 yields
J4 5 D4 1 D5 5 30 1 50 5 80, which is the net flow generated at the F1 node.
While it appears more complicated to use the SUMIF function rather than just entering J4
5 D4 1 D5, J5 5 D8 1 D9, J6 5 D6 1 D7 2 D5 2 D8, and so on, it is actually simpler. The
SUMIF formula only needs to be entered once (in cell J4). It can then be copied down into the
remaining cells in NetFlow (J5:J8). For a problem with many nodes, this is much quicker and
(perhaps more significantly) less prone to error. In a large problem, it is all too easy to miss
an arc when determining which cells in the Ship column to add and subtract to calculate the
net flow for a given node.
The first Solver option specifies that the flow amounts cannot be negative. The second
acknowledges that this is still a linear programming problem.
Running Solver gives the optimal solution shown in Ship (D4:D9). This is the same solu-
tion as displayed in Figure 6.4.
Solving Large Minimum-Cost Flow Problems More Efficiently

Because minimum-cost flow problems are a special type of linear programming problem, and
the simplex method can solve any linear programming problem, it also can solve any mini-
mum-cost flow problem in the standard way. For example, Solver uses the simplex method
to solve this type (or any other type) of linear programming problem. This works fine for
small problems, like the Distribution Unlimited Co. problem, and for considerably larger
ones as well. Therefore, the approach illustrated in Figure 6.5 will serve you well for any
minimum-cost flow problem encountered in this book and for many that you will encounter
subsequently.
However, we should mention that a different approach is sometimes needed in practice to
solve really big problems. Because of the special form of minimum-cost flow problems, it is
possible to greatly streamline the simplex method to solve them far more quickly. In particu-
lar, rather than going through all the algebra of the simplex method, it is possible to execute
the same steps far more quickly by working directly with the network for the problem.
network simplex This streamlined version of the simplex method is called the network simplex method.
method The network simplex method can solve some huge problems that are much too large for the
The network simplex simplex method.
method can solve much
Like the simplex method, the network simplex method not only finds an optimal solu-
larger minimum-cost flow
problems (sometimes with tion but also can be a valuable aid to managers in conducting the kinds of what-if analyses
millions of nodes and described in Chapter 5.
arcs) than can the simplex Many companies now use the network simplex method to solve their minimum-cost flow
method used by Solver. problems. Some of these problems are huge, with many tens of thousands of nodes and arcs.
Occasionally, the number of arcs will even be far larger, perhaps into the millions.
Although Solver does not, other commercial software packages for linear programming
commonly include the network simplex method.
An important advance in recent years has been the development of excellent graphical
interfaces for modeling minimum-cost flow problems. These interfaces make the design of
the model and the interpretation of the output of the network simplex method completely
visual and intuitive with no mathematics involved. This is very helpful for managerial deci-
sion making.
Some Applications
Probably the most important kind of application of minimum-cost flow problems is to the
operation of a distribution network, such as the one depicted in Figures 6.1–6.4 for the Dis-
tribution Unlimited Co. problem. As summarized in the first row of Table 6.1, this kind of
application involves determining a plan for shipping goods from their sources (factories, etc.)
to intermediate storage facilities (as needed) and then on to the customers.
For some applications of minimum-cost flow problems, all the transshipment nodes are
processing facilities rather than intermediate storage facilities. This is the case for solid waste
management, as indicated in Table 6.1. Here, the flow of materials through the network
begins at the sources of the solid waste, then goes to the facilities for processing these waste
materials into a form suitable for landfill, and then sends them on to the various landfill loca-
tions. However, the objective still is to determine the flow plan that minimizes the total cost,
where the cost now is for both shipping and processing.
In other applications, the demand nodes might be processing facilities. For example, in the
third row of Table 6.1, the objective is to find the minimum-cost plan for obtaining supplies
from various possible vendors, storing these goods in warehouses (as needed), and then ship-
ping the supplies to the company’s processing facilities (factories, etc.).
The next kind of application in Table 6.1 (coordinating product mixes at plants) illustrates
that arcs can represent something other than a shipping lane for a physical flow of materials.
TABLE 6.1 Typical Kinds of Applications of Minimum-Cost Flow Problems
Kind of Application Supply Nodes Transshipment Nodes Demand Nodes

Operation of a distribution network Sources of goods Intermediate storage facilities Customers
Solid waste management Sources of solid waste Processing facilities Landfill locations
Operation of a supply network Vendors Intermediate warehouses Processing facilities
Coordinating product mixes at plants Plants Production of a specific product Market for a specific
product
Cash flow management Sources of cash at a Short-term investment options Needs for cash at a
specific time specific time
This application involves a company with several plants (the supply nodes) that can produce
the same products but at different costs. Each arc from a supply node represents the produc-
tion of one of the possible products at that plant, where this arc leads to the transshipment
node that corresponds to this product. Thus, this transshipment node has an arc coming in
from each plant capable of producing this product, and then the arcs leading out of this node
go to the respective customers (the demand nodes) for this product. The objective is to deter-
mine how to divide each plant’s production capacity among the products so as to minimize
the total cost of meeting the demand for the various products.
The last application in Table 6.1 (cash flow management) illustrates that different nodes
can represent some event that occurs at different times. In this case, each supply node repre-
sents a specific time (or time period) when some cash will become available to the company
(through maturing accounts, notes receivable, sales of securities, borrowing, etc.). The supply
at each of these nodes is the amount of cash that will become available then. Similarly, each
demand node represents a specific time (or time period) when the company will need to draw
on its cash reserves. The demand at each such node is the amount of cash that will be needed
then. The objective is to maximize the company’s income from investing the cash between
each time it becomes available and when it will be used. Therefore, each transshipment node
represents the choice of a specific short-term investment option (e.g., purchasing a certificate
of deposit from a bank) over a specific time interval. The resulting network will have a suc-
cession of flows representing a schedule for cash becoming available, being invested, and
then being used after the maturing of the investment.
Special Types of Minimum-Cost Flow Problems

There are five important categories of network problems that turn out to be special types of
minimum-cost flow problems.
One is the transportation problems discussed in Section 3.5. Figure 3.9 shows the net-
work representation of a typical transportation problem. In our current terminology, the
sources and destinations of a transportation problem are the supply nodes and demand nodes,
respectively. Thus, a transportation problem is just a minimum-cost flow problem without
any transshipment nodes and without any capacity constraints on the arcs (all of which go
directly from a supply node to a demand node).
A second category is the assignment problems discussed in Section 3.6. Recall that this
kind of problem involves assigning a group of people (or other operational units) to a group of
tasks where each person is to perform a single task. An assignment problem can be viewed as a
special type of transportation problem whose sources are the assignees and whose destinations
are the tasks. This then makes the assignment problem also a special type of minimum-cost
flow problem with the characteristics described in the preceding paragraph. In addition, each
person is a supply node with a supply of 1 and each task is a demand node with a demand of 1.
transshipment A third special type of minimum-cost flow problem is transshipment problems. This
problem kind of problem is just like a transportation problem except for the additional feature that
A transshipment problem is the shipments from the sources (supply nodes) to the destinations (demand nodes) might
just a minimum-cost flow
also pass through intermediate transfer points (transshipment nodes) such as distribution
problem that has unlimited
capacities for all its arcs. centers. Like a transportation problem, there are no capacity constraints on the arcs. Conse-
quently, any minimum-cost flow problem where each arc can carry any desired amount of
flow is a transshipment problem. For example, if the data in Figure 6.2 were altered so that
any amounts (within the ranges of the supplies and demands) could be shipped into and out
of the distribution center, the Distribution Unlimited Co. would become just a transshipment
problem.1
Because of their close relationship to a general minimum-cost flow problem, we will not
discuss transshipment problems further.
The other two important special types of minimum-cost flow problems are maximum
flow problems and shortest path problems, which will be described in Sections 6.3 and 6.4
after presenting a case study of a maximum flow problem in the next section.
1
Be aware that a minimum-cost flow problem that does have capacity constraints on the arcs is sometimes
referred to as a capacitated transshipment problem. We will not use this terminology.
In case you are wondering why we are bothering to point out that these five kinds of
problems are special types of minimum-cost flow problems, here is one very important rea-
son. It means that the network simplex method can be used to solve large problems of any of
these types that might be difficult or impossible for the simplex method to solve. It is true
that other efficient special-purpose algorithms also are available for each of these kinds of
problems. However, recent implementations of the network simplex method have become so
The network simplex powerful that it now provides an excellent alternative to these other algorithms in most cases.
method can be used to solve This is especially valuable when the available software package includes the network simplex
huge problems of any of
these five special types. method but not another relevant special-purpose algorithm. Furthermore, even after finding
an optimal solution, the network simplex method can continue to be helpful in aiding mana-
gerial what-if sessions along the lines discussed in Chapter 5.
Review 1. Name and describe the three kinds of nodes in a minimum-cost flow problem.
Questions 2. What is meant by the capacity of an arc?
3. What is the usual objective for a minimum-cost flow problem?
4. What property is necessary for a minimum-cost flow problem to have feasible solutions?
5. What is the integer solutions property for minimum-cost flow problems?
6. What is the name of the streamlined version of the simplex method that is designed to solve
minimum-cost flow problems very efficiently?
7. What are a few typical kinds of applications of minimum-cost flow problems?
8. Name five important categories of network optimization problems that turn out to be special
types of minimum-cost flow problems.
6.2 A CASE STUDY: THE BMZ CO. MAXIMUM FLOW PROBLEM

What a day! First being called into his boss’s office and then receiving an urgent telephone
call from the company president himself. Fortunately, he was able to reassure them that he
has the situation under control.
Although his official title is Supply Chain Manager for the BMZ Company, Karl Schmidt
often tells his friends that he really is the company’s crisis manager. One crisis after another.
The supplies needed to keep the production lines going haven’t arrived yet. Or the supplies
have arrived but are unusable because they are the wrong size. Or an urgent shipment to a
key customer has been delayed. This current crisis is typical. One of the company’s most
important distribution centers—the one in Los Angeles—urgently needs an increased flow of
shipments from the company.
Karl was chosen for this key position because he is considered a rising young star. Hav-
ing just received his MBA degree from a top American business school four years ago, he is
the youngest member of upper-level management in the entire company. His business school
training in the latest management science techniques has proven invaluable in improving sup-
ply chain management throughout the company. The crises still occur, but the frequent chaos
of past years has been eliminated.
Karl has a plan for dealing with the current crisis. This will mean calling on management
science once again.
Background
The BMZ Company is a European manufacturer of luxury automobiles. Although its cars
sell well in all the developed countries, its exports to the United States are particularly impor-
tant to the company.
BMZ has a well-deserved reputation for providing excellent service. One key to main-
taining this reputation is having a plentiful supply of automobile replacement parts readily
available to the company’s numerous dealerships and authorized repair shops. These parts are
mainly stored in the company’s distribution centers and then delivered promptly when needed.
One of Karl Schmidt’s top priorities is avoiding shortages at these distribution centers.
6.2 A Case Study: The BMZ Co. Maximum Flow Problem 203
The company has several distribution centers in the United States. However, the clos-
est one to the Los Angeles center is over 1,000 miles away in Seattle. Since BMZ cars are
becoming especially popular in California, it is particularly important to keep the Los Ange-
les center well supplied. Therefore, the fact that supplies there are currently dwindling is a
matter of real concern to BMZ top management—as Karl learned forcefully today.
Most of the automobile replacement parts are produced at the company’s main factory in
Stuttgart, Germany, along with the production of new cars. It is this factory that has been sup-
plying the Los Angeles center with spare parts. Some of these parts are bulky, and very large
numbers of certain parts are needed, so the total volume of the supplies has been relatively
massive—over 300,000 cubic feet of goods arriving monthly. Now a much larger amount will
be needed over the next month to replenish the dwindling inventory.
The Problem
Karl needs to execute a plan quickly for shipping as much as possible from the main factory
to the distribution center in Los Angeles over the next month. He already has recognized that
this is a maximum flow problem—a problem of maximizing the flow of replacement parts
The problem is to maximize from the factory to this distribution center.
the flow of automobile The factory is producing far more than can be shipped to this one distribution center.
replacement parts from the
Therefore, the limiting factor on how much can be shipped is the limited capacity of the com-
factory in Stuttgart,
Germany, to the distribu- pany’s distribution network.
tion center in Los Angeles. This distribution network is depicted in Figure 6.6, where the nodes labeled ST and LA
are the factory in Stuttgart and the distribution center in Los Angeles, respectively. There is
a rail head at the factory, so shipments first go by rail to one of three European ports: Rotter-
dam (node RO), Bordeaux (node BO), and Lisbon (node LI). They then go by ship to ports in
the United States, either New York (node NY) or New Orleans (node NO). Finally, they are
shipped by truck from these ports to the distribution center in Los Angeles.
The organizations operating these railroads, ships, and trucks are independently owned
companies that ship goods for numerous firms. Because of prior commitments to their regular
customers, these companies are unable to drastically increase the allocation of space to any sin-
gle customer on short notice. Therefore, the BMZ Co. is only able to secure a limited amount of
shipping space along each shipping lane over the next month. The amounts available are given
in Figure 6.6, using units of hundreds of cubic meters. (Since each unit of 100 cubic meters is a
little over 3,500 cubic feet, these are large volumes of goods that need to be moved.)
FIGURE 6.6
The BMZ Co. distribu-
tion network from its
main factory in Stuttgart,
Germany, to a distribution
center in Los Angeles.
RO Rotterdam
x.] [50
ts ma un
its
uni ma
[60 x.]
New York [40 units max.]
NY BO ST Stuttgart
[70 units max.]
x.] Bordeaux x.]
a ma
m n its
its .] u
un ax [40
[80 itsm
New Orleans
un LI
LA [50 x.] Lisbon
NO ma
[70 units
0 u nits
Los Angeles max.] [3
FIGURE 6.7
RO
A network model for the
BMZ Co. problem as a
maximum flow problem,
0]
[6
where the number in
square brackets below
[50
each arc is the capacity of
]
NY
that arc.
0]
[4
[8
0]
LA BO ST
[70]
0]
[5
[7
0]
NO
]
[40
[3
0]
LI
Model Formulation
Figure 6.7 shows the network model for this maximum flow problem. Rather than showing
the geographical layout of the distribution network, this network simply lines up the nodes
(representing the cities) in evenly spaced columns. The arcs represent the shipping lanes,
where the capacity of each arc (given in square brackets under the arc) is the amount of ship-
ping space available along that shipping lane. The objective is to determine how much flow
to send through each arc (how many units to ship through each shipping lane) to maximize
the total number of units flowing from the factory in Stuttgart to the distribution center in Los
Angeles.
Figure 6.8 shows the corresponding spreadsheet model for this problem when using
the format introduced in Figure 6.5. The main difference from the model in Figure 6.5 is
the change in the objective. Since we are no longer minimizing the total cost of the flow
through the network, column G in Figure 6.5 can be deleted in Figure 6.8. The objective
cell MaxFlow (D14) in Figure 6.8 now needs to give the total number of units flowing from
Stuttgart to Los Angeles. Thus, the equations at the bottom of the figure include D14 5 I4,
where I4 gives the net flow leaving Stuttgart to go to Los Angeles. As in Figure 6.5, the
equations in Figure 6.8 entered into NetFlow (I4:I10) again use the difference of two
SUMIF functions to calculate the net flow generated at each node. Since the objective is
In contrast to the spread-
sheet model in Figure 6.5, to maximize the flow shown in MaxFlow (D14), the Solver Parameters box specifies that
which minimizes TotalCost this objective cell is to be maximized. After running Solver, the optimal solution shown in
(D11), the spreadsheet the changing cells Ship (D4:D12) is obtained for the amount that BMZ should ship through
model in Figure 6.8 maxi- each shipping lane.
mizes the objective cell
However, Karl is not completely satisfied with this solution. He has an idea for doing even
MaxFlow (D14).
better. This will require formulating and solving another maximum flow problem. (This story
continues in the middle of the next section.)
Review 1. What is the current crisis facing the BMZ Co.?

Questions 2. When formulating this problem in network terms, what is flowing through BMZ’s distribution
network? From where to where?
3. What is the objective of the resulting maximum flow problem?
6.3 Maximum Flow Problems 205
FIGURE 6.8
A spreadsheet model for the BMZ Co. maximum flow problem, including the equations entered into the objective cell MaxFlow
(D14) and the other output cells NetFlow (I4:I10), as well as the other specifications needed to set up the model. The changing cells
Ship (D4:D12) show the optimal shipping quantities through the distribution network obtained by Solver.
A B C D E F G H I J K
1 BMZ Co. Maximum Flow Problem
2
3 From To Ship Capacity Nodes Net Flow Supply/Demand
4 Stuttgart Rotterdam 50 ≤ 50 Stuttgart 150
5 Stuttgart Bordeaux 70 ≤ 70 Rotterdam 0 = 0
6 Stuttgart Lisbon 30 ≤ 40 Bordeaux 0 = 0
7 Rotterdam New York 50 ≤ 60 Lisbon 0 = 0
8 Bordeaux New York 30 ≤ 40 New York 0 = 0
9 Bordeaux New Orleans 40 ≤ 50 New Orleans 0 = 0
10 Lisbon New Orleans 30 ≤ 30 Los Angeles -150
11 New York Los Angeles 80 ≤ 80
12 New Orleans Los Angeles 70 ≤ 70
13
14 Maximum Flow 150
Solver Parameters Range Name Cells I

Set Objective Cell: MaxFlow 3 Net Flow
To: Max Capacity F4:F12
By Changing Variable Cells: From B4:B12 4 =SUMIF(From,H4,Ship)-SUMIF(To,H4,Ship)
Ship MaxFlow D14 5 =SUMIF(From,H5,Ship)-SUMIF(To,H5,Ship)
Subject to the Constraints: NetFlow I4:I10
I5:I9 = SupplyDemand 6 =SUMIF(From,H6,Ship)-SUMIF(To,H6,Ship)
Nodes H4:H10
Ship <= Capacity
Ship D4:D12 7 =SUMIF(From,H7,Ship)-SUMIF(To,H7,Ship)
Solver Options: SupplyDemand K5:K9 8 =SUMIF(From,H8,Ship)-SUMIF(To,H8,Ship)
Make Variables Nonnegative To C4:C12
Solving Method: Simplex LP 9 =SUMIF(From,H9,Ship)-SUMIF(To,H9,Ship)
10 =SUMIF(From,H10,Ship)-SUMIF(To,H10,Ship)
C D
14 Maximum Flow =I4
6.3 MAXIMUM FLOW PROBLEMS

Like a minimum-cost flow problem, a maximum flow problem is concerned with flow through
a network. However, the objective now is different. Rather than minimizing the cost of the
flow, the objective now is to find a flow plan that maximizes the amount flowing through the
network. This is how Karl Schmidt was able to find a flow plan that maximizes the number of
units of automobile replacement parts flowing through BMZ’s distribution network from its
factory in Stuttgart to the distribution center in Los Angeles.
Except for the difference in objective (maximize flow versus minimize cost), the character-
istics of the maximum flow problem are quite similar to those for the minimum-cost flow
problem. However, there are some minor differences, as we will discuss after summarizing
the assumptions.
Assumptions of a Maximum Flow Problem
1. All flow through the network originates at one node, called the source, and terminates at
one other node, called the sink. (The source and sink in the BMZ problem are the factory
and the distribution center, respectively.)
The network for transport of natural gas on the Norwegian import terminals). However, in addition to the usual sup-
Continental Shelf, with approximately 5,000 miles of subsea ply-and-demand constraints, the model also includes con-
pipelines, is the world’s largest offshore pipeline network. straints involving pressure-flow relationships, maximum
Gassco is a company entirely owned by the Norwegian delivery pressures, and technical pressure bounds on pipe-
state, which operates this network. Another company that lines. Therefore, this model is a generalization of the model
is largely state owned, StatoilHydro, is the main Norwegian for the maximum flow problem described in this section.
supplier of natural gas to markets throughout Europe and This key application of management science, along
elsewhere. with a few others, has had a dramatic impact on the effi-
Gassco and StatoilHydro together use management ciency of the operation of this offshore pipeline network.
science techniques to optimize both the configuration of The resulting accumulated savings were estimated to be
the network and the routing of the natural gas. The main approximately $2 billion in the period 1995–2008.
model used for this routing is a multicommodity network-
flow model in which the different hydrocarbons and con-
taminants in natural gas constitute the commodities. The Source: F. Rømo, A. Tomasgard, L. Hellemo, M. Fodstad, B. H.
Eidesen, and B. Pedersen, “Optimizing the Norwegian Natural Gas
objective function for the model is to maximize the total Production and Transport,” Interfaces 39, no. 1 (January–February
flow of the natural gas from the supply points (the off- 2009), pp. 46–56. (A link to this article is provided on our website,
shore drilling platforms) to the demand points (typically www.mhhe.com/hillier5e.)
2. All the remaining nodes are transshipment nodes. (These are nodes RO, BO, LI, NY, and
NO in the BMZ problem.)
3. Flow through an arc is only allowed in the direction indicated by the arrowhead, where the
maximum amount of flow is given by the capacity of that arc. At the source, all arcs point
away from the node. At the sink, all arcs point into the node.
The objective is to find a 4. The objective is to maximize the total amount of flow from the source to the sink. This
flow plan that maximizes amount is measured in either of two equivalent ways, namely, either the amount leaving
the flow from the source to the source or the amount entering the sink. (Cells D14 and I4 in Figure 6.8 use the amount
the sink.
leaving the source.)
The source and sink of a maximum flow problem are analogous to the supply nodes and
demand nodes of a minimum-cost flow problem. These are the only nodes in both problems
that do not have conservation of flow (flow out equals flow in). Like the supply nodes, the
source generates flow. Like the demand nodes, the sink absorbs flow.
However, there are two differences between these nodes in a minimum-cost flow problem
and the corresponding nodes in a maximum flow problem.
One difference is that, whereas supply nodes have fixed supplies and demand nodes have
fixed demands, the source and sink do not. The reason is that the objective is to maximize the
Although a maximum flow flow leaving the source and entering the sink rather than fixing this amount.
problem has only a single The second difference is that, whereas the number of supply nodes and the number of
source and a single sink,
demand nodes in a minimum-cost flow problem may be more than one, there can be only one
variants with multiple
sources and sinks also can source and only one sink in a maximum flow problem. However, variants of maximum flow
be solved, as illustrated in problems that have multiple sources and sinks can still be solved by Solver, as you now will
the next subsection. see illustrated by the BMZ case study introduced in the preceding section.
Continuing the Case Study with Multiple Supply Points

and Multiple Demand Points
Here is Karl Schmidt’s idea for how to improve upon the flow plan obtained at the end of Sec-
tion 6.2 (as given in column D of Figure 6.8).
The company has a second, smaller factory in Berlin, north of its Stuttgart factory, for pro-
ducing automobile parts. Although this factory normally is used to help supply distribution
centers in northern Europe, Canada, and the northern United States (including one in Seattle),
it also is able to ship to the distribution center in Los Angeles. Furthermore, the distribution
center in Seattle has the capability of supplying parts to the customers of the distribution cen-
ter in Los Angeles when shortages occur at the latter center.
206
6.3 Maximum Flow Problems 207
FIGURE 6.9
A network model for the HA
expanded BMZ Co. prob-
]
lem as a variant of a maxi- [40
BN
[6
mum flow problem, where
0]
the number in square
]
]
[20
[30
brackets below each arc is
SE RO BE
the capacity of that arc. [20]
[40
] ]
[60
0]
NY
[1
[5
0]
] [40
[80 ]
LA BO ST
[70]
]
[70 [50
]
NO 0]
[4
[30
]
LI
In this light, Karl now has developed a better plan for addressing the current inventory
shortages in Los Angeles. Rather than simply maximizing shipments from the Stuttgart fac-
tory to Los Angeles, he has decided to maximize the total shipments from both factories to the
distribution centers in both Los Angeles and Seattle.
Figure 6.9 shows the network model representing the expanded distribution network that
encompasses both factories and both distribution centers. In addition to the nodes shown in
Figures 6.6 and 6.7, node BE is the second, smaller factory in Berlin; nodes HA and BN are
additional ports used by this factory in Hamburg and Boston, respectively; and node SE is the
distribution center in Seattle. As before, the arcs represent the shipping lanes, where the num-
ber in square brackets below each arc is the capacity of that arc, that is, the maximum number
of units that can be shipped through that shipping lane over the next month.
The corresponding spreadsheet model is displayed in Figure 6.10. The format is the same
as in Figure 6.8. However, the objective cell MaxFlow (D21) now gives the total flow from
Stuttgart and Berlin, so D21 5 I4 1 I5 (as shown by the equation for this objective cell given
at the bottom of the figure).
The changing cells Ship (D4:D19) in this figure show the optimal solution obtained for
the number of units to ship through each shipping lane over the next month. Comparing this
solution with the one in Figure 6.8 shows the impact of Karl Schmidt’s decision to expand the
distribution network to include the second factory and the distribution center in Seattle. As
indicated in column I of the two figures, the number of units going to Los Angeles directly
has been increased from 150 to 160, in addition to the 60 units going to Seattle as a backup
for the inventory shortage in Los Angeles. This plan solved the crisis in Los Angeles and won
Karl commendations from top management.
Some Applications
The applications of maximum flow problems and their variants are somewhat similar to those for
minimum-cost flow problems described in the preceding section when management’s objective
is to maximize flow rather than to minimize cost. Here are some typical kinds of applications.
1. Maximize the flow through a distribution network, as for the BMZ Co. problem.
2. Maximize the flow through a company’s supply network from its vendors to its processing
facilities.
3. Maximize the flow of oil through a system of pipelines.
4. Maximize the flow of water through a system of aqueducts.
5. Maximize the flow of vehicles through a transportation network.
FIGURE 6.10
A spreadsheet model for the expanded BMZ Co. problem as a variant of a maximum flow problem with sources in both Stuttgart and
Berlin and sinks in both Los Angeles and Seattle. Using the objective cell MaxFlow (D21) to maximize the total flow from the two
sources to the two sinks, Solver yields the optimal shipping plan shown in the changing cells Ship (D4:D19).
1 BMZ Co. Expanded Maximum Flow Problem
2
3 From To Ship Capacity Nodes Net Flow Supply/Demand
4 Stuttgart Rotterdam 40 ≤ 50 Stuttgart 140
5 Stuttgart Bordeaux 70 ≤ 70 Berlin 80
6 Stuttgart Lisbon 30 ≤ 40 Hamburg 0 = 0
7 Berlin Rotterdam 20 ≤ 20 Rotterdam 0 = 0
8 Berlin Hamburg 60 ≤ 60 Bordeaux 0 = 0
9 Rotterdam New York 60 ≤ 60 Lisbon 0 = 0
10 Bordeaux New York 30 ≤ 40 Boston 0 = 0
11 Bordeaux New Orleans 40 ≤ 50 New York 0 = 0
12 Lisbon New Orleans 30 ≤ 30 New Orleans 0 = 0
13 Hamburg New York 30 ≤ 30 Los Angeles -160
14 Hamburg Boston 30 ≤ 40 Seattle -60
15 New Orleans Los Angeles 70 ≤ 70
16 New York Los Angeles 80 ≤ 80
17 New York Seattle 40 ≤ 40
18 Boston Los Angeles 10 ≤ 10
19 Boston Seattle 20 ≤ 20
20
21 Maximum Flow 220
Solver Parameters Range Name Cells I

Set Objective Cell: MaxFlow 3 Net Flow
To: Max Capacity F4:F19
By Changing Variable Cells: From B4:B19 4 =SUMIF(From,H4,Ship)-SUMIF(To,H4,Ship)
Ship MaxFlow D21 5 =SUMIF(From,H5,Ship)-SUMIF(To,H5,Ship)
Subject to the Constraints: NetFlow I4:I14
Nodes H4:H14 6 =SUMIF(From,H6,Ship)-SUMIF(To,H6,Ship)
I6:I12 = SupplyDemand
Ship <= Capacity Ship D4:D19 7 =SUMIF(From,H7,Ship)-SUMIF(To,H7,Ship)
SupplyDemand K6:K12
Solver Options: 8 =SUMIF(From,H8,Ship)-SUMIF(To,H8,Ship)
To C4:C19
Make Variables Nonnegative 9 =SUMIF(From,H9,Ship)-SUMIF(To,H9,Ship)
Solving Method: Simplex LP
C D
21 Maximum Flow =I4+I5
Solving Very Large Problems

The expanded BMZ network in Figure 6.9 has 11 nodes and 16 arcs. However, the networks
for most real applications are considerably larger, and occasionally vastly larger. As the num-
ber of nodes and arcs grows into the hundreds or thousands, the formulation and solution
approach illustrated in Figures 6.8 and 6.10 quickly becomes impractical.
Fortunately, management scientists have other techniques available for formulating and
solving huge problems with many tens of thousands of nodes and arcs. One technique is to
6.4 Shortest Path Problems 209
reformulate a variant of a maximum flow problem so that an extremely efficient special-

purpose algorithm for maximum flow problems still can be applied. Another is to reformulate
the problem to fit the format for a minimum-cost flow problem so that the network simplex
method can be applied. These special algorithms are available in some software packages, but
not in Solver. Thus, if you should ever encounter a maximum flow problem or a variant that
is beyond the scope of Solver (which won’t happen in this book), rest assured that it probably
can be formulated and solved in another way.
Review 1. How does the objective of a maximum flow problem differ from that for a minimum-cost flow
Questions problem?
2. What are the source and the sink for a maximum flow problem? For each, in what direction do
all their arcs point?
3. What are the two equivalent ways in which the total amount of flow from the source to the
sink can be measured?
4. The source and sink of a maximum flow problem are different from the supply nodes and
demand nodes of a minimum-cost flow problem in what two ways?
5. What are a few typical kinds of applications of maximum flow problems?
6.4 SHORTEST PATH PROBLEMS

The most common applications of shortest path problems are for what the name suggests—
finding the shortest path between two points. Here is an example.
An Example: The Littletown Fire Department Problem

Littletown is a small town in a rural area. Its fire department serves a relatively large geo-
graphical area that includes many farming communities. Since there are numerous roads
The objective is to find throughout the area, many possible routes may be available for traveling to any given farming
the shortest route from the community from the fire station. Since time is of the essence in reaching a fire, the fire chief
fire station to the farming wishes to determine in advance the shortest path from the fire station to each of the farming
community.
communities.
Figure 6.11 shows the road system connecting the fire station to one of the farming com-
munities, including the mileage along each road. Can you find which route from the fire sta-
tion to the farming community minimizes the total number of miles?
Model Formulation for the Littletown Problem

Figure 6.12 gives the network representation of this problem, which ignores the geographical
layout and the curves in the roads. This network model is the usual way of representing a
shortest path problem. The junctions now are nodes of the network, where the fire station and
farming community are two additional nodes labeled as O (for origin) and T (for destinat-
ion), respectively. Since travel (flow) can go in either direction between the nodes, the lines
FIGURE 6.11 8
The road system between
the Littletown Fire Sta-
tion and a certain farming 6
community, where A,
B, . . . , H are junctions
A F 4
and the number next to
each road shows its dis- 6
3 1 4 D 3
tance in miles. 6
Fire 6 Farming
Station B 3 G Community
5 5
E 2
2
4
4 C 7 H 7
Incorporated in 1881, Canadian Pacific Railway (CPR) was flow of blocks of railcars as flow through a network where
North America’s first transcontinental railway. CPR trans- each node corresponds to both a location and a point in
ports rail freight over a 14,000-mile network extending time. This representation then enabled the application of
from Montreal to Vancouver and throughout the U.S. network optimization techniques. For example, numerous
Northwest and Midwest. Alliances with other carriers shortest path problems are solved each day as part of the
extend CPR’s market reach into the major business centers overall approach.
of Mexico as well. This application of management science is saving CPR
Every day CPR receives approximately 7,000 new ship- roughly US$100 million per year. Labor productivity, loco-
ments from its customers going to destinations across North motive productivity, fuel consumption, and railcar veloc-
America and for export. It must route and move these ship- ity have improved very substantially. In addition, CPR now
ments in railcars over the network of track, where a railcar provides its customers with reliable delivery times and has
may be switched a number of times from one locomotive received many awards for its improvement in service. This
engine to another before reaching its destination. CPR application of network optimization techniques also led to
must coordinate the shipments with its operational plans CPR winning the prestigious First Prize in the 2003 inter-
for 1,600 locomotives, 65,000 railcars, over 5,000 train crew national competition for the Franz Edelman Award for
members, and 250 train yards. Achievement in Operations Research and the Management
CPR management turned to a management science con- Sciences.
sulting firm, MultiModal Applied Systems, to work with CPR
employees in developing a management science approach Source: P. Ireland, R. Case, J. Fallis, C. Van Dyke, J. Kuehn, and
M. Meketon, “The Canadian Pacific Railway Transforms Opera-
to this problem. A variety of management science tech- tions by Using Models to Develop Its Operating Plans,” Interfaces
niques were used to create a new operating strategy. How- 34, no. 1 (January–February 2004), pp. 5–14. (A link to this article is
ever, the foundation of the approach was to represent the provided on our website, www.mhhe.com/hillier5e.)
links connecting the nodes now are referred to as links2 instead of arcs. A link between a pair of
In a shortest path problem, nodes allows travel in either direction, whereas an arc allows travel in only the direction indi-
travel goes from the origin
cated by an arrowhead, so the lines in Figure 6.12 need to be links instead of arcs. (Notice that
to the destination through
a series of links (such as the links do not have an arrowhead at either end.)
roads) that connect pairs Have you found the shortest path from the origin to the destination yet? (Try it now before
of nodes (junctions) in the reading further.) It is
network.
OSASBSESFST
with a total distance of 19 miles.
This problem (like any shortest path problem) can be thought of as a special kind of min-
imum-cost flow problem (Section 6.1) where the miles traveled now are interpreted to be the
cost of flow through the network. A trip from the fire station to the farming community is inter-
preted to be a flow of 1 on the chosen path through the network, so minimizing the cost of this
flow is equivalent to minimizing the number of miles traveled. The fire station is considered to
be the one supply node, with a supply of 1 to represent the start of this trip. The farming com-
munity is the one demand node, with a demand of 1 to represent the completion of this trip. All
the other nodes in Figure 6.12 are transshipment nodes, so the net flow generated at each is 0.
Figure 6.13 shows the spreadsheet model that results from this interpretation. The format
is basically the same as for the minimum-cost flow problem formulated in Figure 6.5, except
FIGURE 6.12
A 6 F
The network representa- 8
tion of Figure 6.11 as a 3 D 4
1 4 3
shortest path problem. 6
6 6
(Origin) O B 3 G T (Destination)
5 5
4 2 2 7
E
7 4
C H
2
Another name sometimes used is undirected arc, but we will not use this terminology.
210
FIGURE 6.13
A spreadsheet model for the Littletown Fire Department shortest path problem, including the equations entered into the objective cell
TotalDistance (D29) and the other output cells SupplyDemand (K4:K13). The values of 1 in the changing cells OnRoute (D4:D27)
reveal the optimal solution obtained by Solver for the shortest path (19 miles) from the fire station to the farming community.
1 Littletown Fire Department Shortest Path Problem
2
3 From To On Route Distance Nodes Net Flow Supply/Demand
4 Fire St. A 1 3 Fire St. 1 = 1
5 Fire St. B 0 6 A 0 = 0
6 Fire St. C 0 4 B 0 = 0
7 A B 1 1 C 0 = 0
8 A D 0 6 D 0 = 0
9 B A 0 1 E 0 = 0
10 B C 0 2 F 0 = 0
11 B D 0 4 G 0 = 0
12 B E 1 5 H 0 = 0
13 C B 0 2 Farm Comm. 21 = 21
14 C E 0 7
15 D E 0 3
16 D F 0 8
17 E D 0 3
18 E F 1 6
19 E G 0 5
20 E H 0 4
21 F G 0 3
22 F Farm Comm. 1 4
23 G F 0 3
24 G H 0 2
25 G Farm Comm. 0 6
26 H G 0 2
27 H Farm Comm. 0 7
28
29 Total Distance 19
Solver Parameters I
Set Objective Cell: TotalDistance 3 Net Flow
To: Min
By Changing Variable Cells: 4 =SUMIF(From,H4,OnRoute)-SUMIF(To,H4,OnRoute)
OnRoute 5 =SUMIF(From,H5,OnRoute)-SUMIF(To,H5,OnRoute)
Subject to the Constraints:
NetFlow = SupplyDemand 6 =SUMIF(From,H6,OnRoute)-SUMIF(To,H6,OnRoute)
7 =SUMIF(From,H7,OnRoute)-SUMIF(To,H7,OnRoute)
Solver Options:
Make Variables Nonnegative 8 =SUMIF(From,H8,OnRoute)-SUMIF(To,H8,OnRoute)
Solving Method: Simplex LP 9 =SUMIF(From,H9,OnRoute)-SUMIF(To,H9,OnRoute)
Range Name Cells
Distance F4:F27
From B4:B27 11 =SUMIF(From,H11,OnRoute)-SUMIF(To,H11,OnRoute)
NetFlow I4:I13 12 =SUMIF(From,H12,OnRoute)-SUMIF(To,H12,OnRoute)
Nodes H4:H13
OnRoute D4:D27 13 =SUMIF(From,H13,OnRoute)-SUMIF(To,H13,OnRoute)
SupplyDemand K4:K13
To C4:C27 C D
TotalDistance D29 29 Total Distance =SUMPRODUCT(OnRoute,Distance)
now there are no arc capacity constraints and the unit cost column is replaced by a column of
distances in miles. The flow quantities given by the changing cells OnRoute (D4:D27) are 1 for
each arc that is on the chosen path from the fire station to the farming community and 0 other-
wise. The objective cell TotalDistance (D29) gives the total distance of this path in miles. (See
the equation for this cell at the bottom of the figure.) Columns B and C together list all the verti-
cal links in Figure 6.12 twice, once as a downward arc and once as an upward arc, since either
direction might be on the chosen path. The other links are only listed as left-to-right arcs, since
This spreadsheet model is this is the only direction of interest for choosing a shortest path from the origin to the destination.
like one for a minimum- Column K shows the net flow that needs to be generated at each of the nodes. Using the
cost flow problem with no
equations at the bottom of the figure, each column I cell then calculates the actual net flow at
arc capacity constraints
except that distances that node by adding the flow out and subtracting the flow in. The corresponding constraints,
replace unit costs and Nodes (H4:H13) 5 SupplyDemand (K4:K13), are specified in the Solver Parameters box.
travel on a chosen path is The solution shown in OnRoute (D4:D27) is the optimal solution obtained after running
interpreted as a flow of 1 Solver. It is exactly the same as the shortest path given earlier.
through this path.
Just as for minimum-cost flow problems and maximum flow problems, special algorithms
are available for solving large shortest path problems very efficiently, but these algorithms
are not included in Solver. Using a spreadsheet formulation and Solver is fine for problems of
the size of the Littletown problem and somewhat larger, but you should be aware that vastly
larger problems can still be solved by other means.
Except for more complicated variations beyond the scope of this book, all shortest path problems
share the characteristics illustrated by the Littletown problem. Here are the basic assumptions.
Assumptions of a Shortest Path Problem
1. You need to choose a path through the network that starts at a certain node, called the
origin, and ends at another certain node, called the destination.
2. The lines connecting certain pairs of nodes commonly are links (which allow travel in
either direction), although arcs (which only permit travel in one direction) also are allowed.
3. Associated with each link (or arc) is a nonnegative number called its length. (Be aware
that the drawing of each link in the network typically makes no effort to show its true
The objective is to find the
length other than giving the correct number next to the link.)
shortest path from the ori- 4. The objective is to find the shortest path (the path with the minimum total length) from the
gin to the destination. origin to the destination.
Some Applications
Not all applications of shortest path problems involve minimizing the distance traveled from
the origin to the destination. In fact, they might not even involve travel at all. The links (or arcs)
might instead represent activities of some other kind, so choosing a path through the network
corresponds to selecting the best sequence of activities. The numbers giving the “lengths” of
the links might then be, for example, the costs of the activities, in which case the objective
would be to determine which sequence of activities minimizes the total cost.
Here are three categories of applications.
1. Minimize the total distance traveled, as in the Littletown example.
2. Minimize the total cost of a sequence of activities, as in the example that follows in the
subsection below.
3. Minimize the total time of a sequence of activities, as in the example involving the Quick
Company at the end of this section.
An Example of Minimizing Total Cost

Sarah has just graduated from high school. As a graduation present, her parents have given
her a car fund of $21,000 to help purchase and maintain a certain three-year-old used car for
college. Since operating and maintenance costs go up rapidly as the car ages, Sarah’s parents
tell her that she will be welcome to trade in her car on another three-year-old car one or more
times during the next three summers if she determines that this would minimize her total net
cost. They also inform her that they will give her a new car in four years as a college gradu-
ation present, so she should definitely plan to trade in her car then. (These are pretty nice
parents!)
Table 6.2 gives the relevant data for each time Sarah purchases a three-year-old car. For
example, if she trades in her car after two years, the next car will be in ownership year 1 dur-
ing her junior year, and so forth.
Sarah needs a schedule for When should Sarah trade in her car (if at all) during the next three summers to minimize her
trading in her car that will total net cost of purchasing, operating, and maintaining the car(s) over her four years of college?
minimize her total net cost.
Figure 6.14 shows the network formulation of this problem as a shortest path problem.
Nodes 1, 2, 3, and 4 are the end of Sarah’s first, second, third, and fourth years of college,
respectively. Node 0 is now, before starting college. Each arc from one node to a second node
corresponds to the activity of purchasing a car at the time indicated by the first of these two
nodes and then trading it in at the time indicated by the second node. Sarah begins by purchas-
ing a car now, and she ends by trading in a car at the end of year 4, so node 0 is the origin and
node 4 is the destination.
The number of arcs on the path chosen from the origin to the destination indicates how
many times Sarah will purchase and trade in a car. For example, consider the path
0 1 3 4
This corresponds to purchasing a car now, then trading it in at the end of year 1 to purchase a
second car, then trading in the second car at the end of year 3 to purchase a third car, and then
trading in this third car at the end of year 4.
Since Sarah wants to minimize her total net cost from now (node 0) to the end of year 4
(node 4), each arc length needs to measure the net cost of that arc’s cycle of purchasing, main-
taining, and trading in a car. Therefore,
Arc length 5 Purchase price 1 Operating and maintenance costs 2 Trade-in value
For example, consider the arc from node 1 to node 3. This arc corresponds to purchasing a car
at the end of year 1, operating and maintaining it during ownership years 1 and 2, and then
trading it in at the end of ownership year 2. Consequently,
Length of arc from 1 to 3 5 12,000 1 2,000 1 3,000 2 6,500
5 10,500 (in dollars)
TABLE 6.2 Operating and Maintenance Costs Trade-in Value at End

Sarah’s Data Each Time for Ownership Year of Ownership Year
She Purchases a Three-
Year-Old Car Purchase
Price 1 2 3 4 1 2 3 4
$12,000 $2,000 $3,000 $4,500 $6,500 $8,500 $6,500 $4,500 $3,000
FIGURE 6.14 25,000

Formulation of the
problem of when Sarah 17,000 10,500
should trade in her car as
10,500
a shortest path problem.
The node labels measure (Origin) 0 1 2 3 4 (Destination)
the number of years from 5,500 5,500 5,500 5,500
now. Each arc represents 10,500
purchasing a car and then
trading it in later.
17,000
The sum of the arc lengths The arc lengths calculated in this way are shown next to the arcs in Figure 6.14. Adding up
on any path through this the lengths of the arcs on any path from node 0 to node 4 then gives the total net cost for that
network gives the total net
cost of the corresponding
particular plan for trading in cars over the next four years. Therefore, finding the shortest path
plan for trading in cars. from the origin to the destination identifies the plan that will minimize Sarah’s total net cost.
Figure 6.15 shows the corresponding spreadsheet model, formulated in just the same way
as for Figure 6.13 except that distances are now costs. Thus, the objective cell TotalCost (D23)
The objective cell now
is TotalCost instead of
now gives the total cost that is to be minimized. The changing cells OnRoute (D12:D21) in
TotalDistance. the figure display the optimal solution obtained after running Solver. Since values of 1 indi-
cate the path being followed, the shortest path turns out to be
0 2 4
Trade in the first car at the end of year 2.

Trade in the second car at the end of year 4.
The length of this path is 10,500 1 10,500 5 21,000, so Sarah’s total net cost is $21,000, as
given by the objective cell. Recall that this is exactly the amount in Sarah’s car fund provided
by her parents. (These are really nice parents!)
An Example of Minimizing Total Time

The Quick Company has learned that a competitor is planning to come out with a new kind
of product with great sales potential. Quick has been working on a similar product that had
been scheduled to come to market in 20 months. However, research is nearly complete and
Quick’s management now wishes to rush the product out to meet the competition.
There are four nonoverlapping phases left to be accomplished, including the remaining
research (the first phase) that currently is being conducted at a normal pace. However, each
phase can instead be conducted at a priority or crash level to expedite completion. These are
the only levels that will be considered for the last three phases, whereas both the normal level
and these two levels will be considered for the first phase. The times required at these levels
are shown in Table 6.3.
Management now has allocated $30 million for these four phases. The cost of each phase
The objective is to mini- at the levels under consideration is shown in Table 6.4.
mize the total time for the Management wishes to determine at which level to conduct each of the four phases to
project.
minimize the total time until the product can be marketed, subject to the budget restriction of
$30 million.
Figure 6.16 shows the network formulation of this problem as a shortest path problem.
Each node indicates the situation at that point in time. Except for the destination, a node is
identified by two numbers:
1. The number of phases completed.
2. The number of millions of dollars left for the remaining phases.
The origin is now, when 0 phases have been completed and the entire budget of $30 million
is left. Each arc represents the choice of a particular level of effort (identified in parentheses
below the arc) for that phase. [There are no crash arcs emanating from the (2, 12) and
The sum of the arc lengths (3, 3) nodes because this level of effort would require exceeding the budget of $30 million
on any path through this for the four phases.] The time (in months) required to perform the phase with this level of
network gives the total effort then is the length of the arc (shown above the arc). Time is chosen as the measure of
time of the corresponding
plan for preparing the new arc length because the objective is to minimize the total time for all four phases. Summing the
product. arc lengths for any particular path through the network gives the total time for the plan cor-
responding to that path. Therefore, the shortest path through the network identifies the plan
that minimizes total time.
FIGURE 6.15 A B C D E F G H I J
A spreadsheet model 1 Sarah's Car Purchasing Problem
that formulates Sarah’s
2
problem as a shortest path Trade-in
3
problem where the objec- Operating & Value at End Purchase
tive is to minimize the 4 Maint. Cost of Year Price
total cost instead of the 5 Year 1 $2,000 $8,500 $12,000
total distance. The bottom 6 Year 2 $3,000 $6,500
of the figure shows the 7 Year 3 $4,500 $4,500
equations entered in the 8 Year 4 $6,500 $3,000
objective cell TotalCost 9
(D23) and the other out- 10
put cells Cost (E12:E21)
11 From To On Route Cost Nodes Net Flow Supply/Demand
and NetFlow (H12:H16).
12 Year 0 Year 1 0 $5,500 Year 0 1 = 1
After applying Solver,
the values of 1 in the 13 Year 0 Year 2 1 $10,500 Year 1 0 = 0
changing cells OnRoute 14 Year 0 Year 3 0 $17,000 Year 2 0 = 0
(D12:D21) identify the 15 Year 0 Year 4 0 $25,000 Year 3 0 = 0
shortest (least expen- 16 Year 1 Year 2 0 $5,500 Year 4 -1 = -1
sive) path for scheduling 17 Year 1 Year 3 0 $10,500
trade-ins. 18 Year 1 Year 4 0 $17,000
19 Year 2 Year 3 0 $5,500
20 Year 2 Year 4 1 $10,500
21 Year 3 Year 4 0 $5,500
22
23 Total Cost $21,000
Range Name Cells E

11 Cost
Cost E12:E21
From B12:B21 12 =PurchasePrice+OpMaint1-TradeIn1
NetFlow H12:H16 13 =PurchasePrice+OpMaint1+OpMaint2-TradeIn2
Nodes G12:G16
OnRoute D12:D21 14 =PurchasePrice+OpMaint1+OpMaint2+OpMaint3-TradeIn3
OpMaint1 C5 15 =PurchasePrice+OpMaint1+OpMaint2+OpMaint3+OpMaint4-TradeIn4
OpMaint2 C6
16 =PurchasePrice+OpMaint1-TradeIn1
OpMaint3 C7
OpMaint4 C8 17 =PurchasePrice+OpMaint1+OpMaint2-TradeIn2
PurchasePrice E5 18 =PurchasePrice+OpMaint1+OpMaint2+OpMaint3-TradeIn3
SupplyDemand J12:J16
19 =PurchasePrice+OpMaint1-TradeIn1
To C12:C21
TotalCost D23 20 =PurchasePrice+OpMaint1+OpMaint2-TradeIn2
TradeIn1 D5 21 =PurchasePrice+OpMaint1-TradeIn1
TradeIn2 D6
TradeIn3 D7 H
TradeIn4 D8
11 Net Flow
Solver Parameters 12 =SUMIF(From,G12,OnRoute)-SUMIF(To,G12,OnRoute)
Set Objective Cell: TotalCost 13 =SUMIF(From,G13,OnRoute)-SUMIF(To,G13,OnRoute)
To: Min 14 =SUMIF(From,G14,OnRoute)-SUMIF(To,G14,OnRoute)
By Changing Variable Cells:
OnRoute 15 =SUMIF(From,G15,OnRoute)-SUMIF(To,G15,OnRoute)
Subject to the Constraints: 16 =SUMIF(From,G16,OnRoute)-SUMIF(To,G16,OnRoute)
NetFlow = SupplyDemand
C D
Solver Options: 23 Total Cost =SUMPRODUCT(OnRoute,Cost)
Make Variables Nonnegative
Solving Method: Simplex LP
All four phases have been completed as soon as any one of the four nodes with a first label of
4 has been reached. So why doesn’t the network just end with these four nodes rather than hav-
ing an arc coming out of each one? The reason is that a shortest path problem is required to have
only a single destination. Consequently, a dummy destination is added at the right-hand side.
TABLE 6.3
Remaining Design of Initiate Production
Time Required for
Level Research Development Manufacturing System and Distribution
the Phases of Prepar-
ing Quick Co.’s New Normal 5 months — — —
Product Priority 4 months 3 months 5 months 2 months
Crash 2 months 2 months 3 months 1 month
TABLE 6.4
Remaining Design of Initiate Production
Cost for the Phases of
Level Research Development Manufacturing System and Distribution
Preparing Quick Co.’s
New Product Normal $3 million — — —
Priority 6 million $6 million $ 9 million $3 million
Crash 9 million 9 million 12 million 6 million
FIGURE 6.16 5 2
3 2, 21 3, 12 4, 9
Formulation of the Quick r i t y) (Priority) (Priority)
o
Co. problem as a shortest (Pri (C 3 (C 1 0
1, 27 ra ra
path problem. Except for 2 sh sh
( ) )
5 l) Cra 5 2
the dummy destination, ma sh) 2, 18 3, 9 4, 6
the arc labels indicate, (Origin) N or 3 (Priority) (Priority) 0 (Destination)
(
ty)
first, the number of phases 4 iori (C 3 (C 1
0, 30 1, 24 (Pr ra ra T
completed and, second, (Priority) 2 sh sh
(Cra ) ) 0
the amount of money left (C 2 s h ) 5 2
ra 3 2, 15 3, 6 4, 3
(in millions of dollars) sh (Priority) (Priority)
) ity)
for the remaining phases. rior (C 3 (C 1 0
1, 21 (P ra ra
2 sh sh
Each arc length gives the (Cra ) )
time (in months) to per- sh) 5 2
2, 12 3, 3 4, 0
form that phase. (Priority) (Priority)
When real travel through a network can end at more than one node, an arc with length 0 is
inserted from each of these nodes to a dummy destination so that the network will have just a
single destination.
Since each of the arcs into the dummy destination has length 0, this addition to the network
The objective cell now does not affect the total length of a path from the origin to its ending point.
is TotalTime instead of Figure 6.17 displays the spreadsheet model for this problem. Once again, the format is the
TotalDistance.
same as in Figures 6.13 and 6.15, except now the quantity of concern in column F and the
objective cell TotalTime (D32) is time rather than distance or cost. Since Solver has already
been run, the changing cells OnRoute (D4:D30) indicate which arcs lie on the path that mini-
mizes the total time. Thus, the shortest path is
0, 30 1, 21 2, 15 3, 3 4, 0 T
with a total length of 2 1 3 1 3 1 2 1 0 5 10 months, as given by TotalTime (D32). The
resulting plan for the four phases is shown in Table 6.5. Although this plan does consume the
TABLE 6.5
Phase Level Time Cost
The Optimal Solution
Remaining research Crash 2 months $ 9 million
Obtained by Solver for
Development Priority 3 months 6 million
Quick Co.’s Shortest
Design of manufacturing system Crash 3 months 12 million
Path Problem Initiate production and distribution Priority 2 months 3 million
Total 10 months $30 million
entire budget of $30 million, it reduces the time until the product can be brought to market
from the originally planned 20 months down to just 10 months.
Given this information, Quick’s management now must decide whether this plan provides
the best trade-off between time and cost. What would be the effect on total time of spending
a few million more dollars? What would be the effect of reducing the spending somewhat
instead? It is easy to provide management with this information as well by quickly solving
some shortest path problems that correspond to budgets different from $30 million. The ulti-
mate decision regarding which plan provides the best time–cost trade-off then is a judgment
decision that only management can make.
FIGURE 6.17
A spreadsheet model that formulates the Quick Co. problem as a shortest path problem where the objective is to minimize the total
time instead of the total distance, so the objective cell is TotalTime (D32). The other output cells are NetFlow (I4:I20). The values of
1 in the changing cells OnRoute (D4:D30) reveal the shortest (quickest) path obtained by Solver.
1 Quick Co. Product Development Scheduling Problem
2
3 From To On Route Time Nodes Net Flow Supply/Demand
4 (0, 30) (1, 27) 0 5 (0, 30) 1 = 1
5 (0, 30) (1, 24) 0 4 (1, 27) 0 = 0
6 (0, 30) (1, 21) 1 2 (1, 24) 0 = 0
7 (1, 27) (2, 21) 0 3 (1, 21) 0 = 0
8 (1, 27) (2, 18) 0 2 (2, 21) 0 = 0
9 (1, 24) (2, 18) 0 3 (2, 18) 0 = 0
10 (1, 24) (2, 15) 0 2 (2, 15) 0 = 0
11 (1, 21) (2, 15) 1 3 (2, 12) 0 = 0
12 (1, 21) (2, 12) 0 2 (3, 12) 0 = 0
13 (2, 21) (3, 12) 0 5 (3, 9) 0 = 0
14 (2, 21) (3, 9) 0 3 (3, 6) 0 = 0
15 (2, 18) (3, 9) 0 5 (3, 3) 0 = 0
16 (2, 18) (3, 6) 0 3 (4, 9) 0 = 0
17 (2, 15) (3, 6) 0 5 (4, 6) 0 = 0
18 (2, 15) (3, 3) 1 3 (4, 3) 0 = 0
19 (2, 12) (3, 3) 0 5 (4, 0) 0 = 0
20 (3, 12) (4, 9) 0 2 (T) -1 = -1
21 (3, 12) (4, 6) 0 1
22 (3, 9) (4, 6) 0 2
23 (3, 9) (4, 3) 0 1
24 (3, 6) (4, 3) 0 2
25 (3, 6) (4, 0) 0 1
26 (3, 3) (4, 0) 1 2
27 (4, 9) (T) 0 0
28 (4, 6) (T) 0 0
29 (4, 3) (T) 0 0
30 (4, 0) (T) 1 0
31
32 Total Time 10
(continued)
FIGURE 6.17 I
Range Name Cells
(continued)
3 Net Flow
From B4:B30
NetFlow I4:I20 4 =SUMIF(From,H4,OnRoute)-SUMIF(To,H4,OnRoute)
Nodes H4:H20 5 =SUMIF(From,H5,OnRoute)-SUMIF(To,H5,OnRoute)
OnRoute D4:D30
SupplyDemand K4:K20 6 =SUMIF(From,H6,OnRoute)-SUMIF(To,H6,OnRoute)
Time F4:F30 7 =SUMIF(From,H7, OnRoute)-SUMIF(To,H7,OnRoute)
To C4:C30
TotalTime D32
Solver Parameters 10 =SUMIF(From,H10,OnRoute)-SUMIF(To,H10,OnRoute)
Set Objective Cell: TotalTime 11 =SUMIF(From,H11,OnRoute)-SUMIF(To,H11,OnRoute)
To: Min
By Changing Variable Cells: 12 =SUMIF(From,H12,OnRoute)-SUMIF(To,H12,OnRoute)
OnRoute 13 =SUMIF(From,H13,OnRoute)-SUMIF(To,H13,OnRoute)
Subject to the Constraints:
NetFlow = SupplyDemand 14 =SUMIF(From,H14,OnRoute)-SUMIF(To,H14,OnRoute)
Solver Options:
Make Variables Nonnegative 16 =SUMIF(From,H16,OnRoute)-SUMIF(To,H16,OnRoute)
Solving Method: Simplex LP 17 =SUMIF(From,H17,OnRoute)-SUMIF(To,H17,OnRoute)
C D
32 Total Time =SUMPRODUCT(OnRoute,Time)
Review 1. What are the origin and the destination in the Littletown Fire Department example?
Questions 2. What is the distinction between an arc and a link?
3. What are the supply node and the demand node when a shortest path problem is interpreted
as a minimum-cost flow problem? With what supply and demand?
4. What are three measures of the length of a link (or arc) that lead to three categories of applica-
tions of shortest path problems?
5. What is the objective for Sarah’s shortest path problem?
6. When does a dummy destination need to be added to the formulation of a shortest path problem?
7. What kind of trade-off does the management of the Quick Co. need to consider in making its
final decision about how to expedite its new product to market?
6.5 Summary Networks of some type arise in a wide variety of contexts. Network representations are very useful
for portraying the relationships and connections between the components of systems. Each component
is represented by a point in the network called a node, and then the connections between components
(nodes) are represented by lines called arcs (for one-way travel) or links (for two-way travel).
Frequently, a flow of some type must be sent through a network, so a decision needs to be made
about the best way to do this. The kinds of network optimization models introduced in this chapter pro-
vide a powerful tool for making such decisions.
The model for minimum-cost flow problems plays a central role among these network optimization
models, both because it is so broadly applicable and because it can be readily solved. Solver solves
spreadsheet formulations of reasonable size, and the network simplex method can be used to solve
larger problems, including huge problems with tens of thousands of nodes and arcs. A minimum-cost
flow problem typically is concerned with optimizing the flow of goods through a network from their
points of origin (the supply nodes) to where they are needed (the demand nodes). The objective is to
minimize the total cost of sending the available supply through the network to satisfy the given demand.
One typical application (among several) is to optimize the operation of a distribution network.
Special types of minimum-cost flow problems include transportation problems and assignment prob-
lems (discussed in Chapter 3) as well as two prominent types introduced in this chapter: maximum flow
problems and shortest path problems.
Chapter 6 Solved Problems 219
Given the limited capacities of the arcs in the network, the objective of a maximum flow problem is
to maximize the total amount of flow from a particular point of origin (the source) to a particular termi-
nal point (the sink). For example, this might involve maximizing the flow of goods through a company’s
supply network from its vendors to its processing facilities.
A shortest path problem also has a beginning point (the origin) and an ending point (the destination),
but now the objective is to find a path from the origin to the destination that has the minimum total
length. For some applications, length refers to distance, so the objective is to minimize the total distance
traveled. However, some applications instead involve minimizing either the total cost or the total time
of a sequence of activities.
Glossary arc A channel through which flow may occur nodes, shown as a line between the nodes.
from one node to another, shown as an arrow (Section 6.4), 210
between the nodes pointing in the direction in network simplex method A streamlined
which flow is allowed. (Section 6.1), 197 version of the simplex method for solving
capacity of an arc The maximum amount of minimum-cost flow problems very efficiently.
flow allowed through the arc. (Section 6.1), 197 (Section 6.1), 200
conservation of flow Having the amount of flow node A junction point of a network, shown as a
out of a node equal the amount of flow into that labeled circle. (Section 6.1), 197
node. (Section 6.1), 197 origin The node at which travel through the
demand node A node where the net amount of network is assumed to start for a shortest path
flow generated (outflow minus inflow) is a fixed problem. (Section 6.4), 212
negative number, so that flow is absorbed there. sink The node for a maximum flow problem at
(Section 6.1), 197 which all flow through the network terminates.
destination The node at which travel through (Section 6.3), 205
the network is assumed to end for a shortest path source The node for a maximum flow problem
problem. (Section 6.4), 212 at which all flow through the network originates.
dummy destination A fictitious destination (Section 6.3), 205
introduced into the formulation of a shortest supply node A node where the net amount of
path problem with multiple possible termination flow generated (outflow minus inflow) is a fixed
points to satisfy the requirement that there be just positive number. (Section 6.1), 197
a single destination. (Section 6.4), 216 transshipment node A node where the amount
length of a link or arc The number (typically a of flow out equals the amount of flow in.
distance, a cost, or a time) associated with includ- (Section 6.1), 197
ing the link or arc in the selected path for a short- transshipment problem A special type of
est path problem. (Section 6.4), 212 minimum-cost flow problem where there are
link A channel through which flow may no capacity constraints on the arcs.
occur in either direction between a pair of (Section 6.1), 201
Learning Aids for This Chapter in Your MS Courseware

Chapter 6 Excel Files: Sarah Example
Distribution Unlimited Example Quick Example
BMZ Example Supplement to Chapter 6 on the CD-ROM:
Expanded BMZ Example Minimum Spanning-Tree Problems
Littletown Fire Department Example
Solved Problems (See the CD-ROM or Website for the Solutions)

6.S1. Distribution at Heart Beats factories and shipped to two warehouses. The product is then
Heart Beats is a manufacturer of medical equipment. The com- shipped on demand to four third-party wholesalers. All ship-
pany’s primary product is a device used to monitor the heart ping is done by truck. The product distribution network is
during medical procedures. This device is produced in two shown below. The annual production capacity at factories 1 and
2 is 400 and 250, respectively. The annual demand at wholesal- 6.S2. Assessing the Capacity of a Pipeline
ers 1, 2, 3, and 4 is 200, 100, 150, and 200, respectively. The Network
cost of shipping one unit in each shipping lane is shown on the Exxo 76 is an oil company that operates the pipeline network
arcs. Because of limited truck capacity, at most 250 units can shown below, where each pipeline is labeled with its maximum
be shipped from Factory 1 to Warehouse 1 each year. Formu- flow rate in million cubic feet (MMcf) per day. A new oil well
late and solve a network optimization model in a spreadsheet to has been constructed near A. They would like to transport oil
determine how to distribute the product at the lowest possible from the well near A to their refinery at G. Formulate and solve
annual cost. a network optimization model to determine the maximum flow
rate from A to G.
15 D
WS [−200] 14
2 B 6
20 12
4 G
$60 A 2 F
20 8
16
15
C
$40 $35 23
[400] WH WS [−100]
F1 E
[250] 1 2
$35 $55 6.S3. Driving to the Mile-High City

Sarah and Jennifer have just graduated from college at the Uni-
versity of Washington in Seattle and want to go on a road trip.
$25 WH $50 WS They have always wanted to see the mile-high city of Denver.
[250] F2 [−150]
2 3 Their road atlas shows the driving time (in hours) between vari-
ous city pairs, as shown below. Formulate and solve a network
optimization model to find the quickest route from Seattle to
$65
Denver.
Seattle
WS Butte Billings
[−200]
4 3 10
4
9 7
Portland 7
7 12 7
Boise
6 14
Cheyenne
7
Salt Lake 1
City 5
Grand 4 Denver
Junction
Problems
We have inserted the symbol E* to the left of each problem Destination
(or its parts) where Excel should be used (unless your instruc-
tor gives you contrary instructions). An asterisk on the problem 1 2 3 Supply
number indicates that at least a partial answer is given in the Source
back of the book. 1 6 7 4 40
2 5 8 6 60
6.1. Read the referenced article that fully describes the man-
Demand 30 40 30
agement science study summarized in the application vignette
presented in Section 6.1. Briefly describe how the model for a
special type of minimum-cost flow problem was applied in this a. Formulate a network model for this problem as a
study. Then list the various financial and nonfinancial benefits minimum-cost flow problem by drawing a network
that resulted from this study. similar to Figure 6.3.
6.2.* Consider the transportation problem having the follow- E* b. Formulate and solve a spreadsheet model for this
ing data. problem.
Chapter 6 Problems 221
6.3. The Makonsel Company is a fully integrated company based on the mileage to the warehouse. These formulas and the
that both produces goods and sells them at its retail outlets. mileage data are shown below.
After production, the goods are stored in the company’s two
warehouses until needed by the retail outlets. Trucks are used to Charge
transport the goods from the two plants to the warehouses, and Vendor per Shipment Warehouse 1 Warehouse 2
then from the warehouses to the three retail outlets. 1 $300 1 40¢/mile 1,600 miles 400 miles
Using units of full truckloads, the first table below shows 2 $200 1 50¢/mile 500 miles 600 miles
each plant’s monthly output, its shipping cost per truckload sent 3 $500 1 20¢/mile 2,000 miles 1,000 miles
to each warehouse, and the maximum amount that it can ship per
month to each warehouse. Whenever one of the company’s two factories needs a ship-
For each retail outlet (RO), the second table below shows ment of speakers to assemble into the boomboxes, the company
its monthly demand, its shipping cost per truckload from each hires a trucker to bring the shipment in from one of the ware-
warehouse, and the maximum amount that can be shipped per houses. The cost per shipment is given next, along with the num-
month from each warehouse. ber of shipments needed per month at each factory.
Management now wants to determine a distribution plan
(number of truckloads shipped per month from each plant to Unit Shipping Cost
each warehouse and from each warehouse to each retail outlet) Factory 1 Factory 2
that will minimize the total shipping cost.
Warehouse 1 $200 $700
a. Draw a network that depicts the company’s distri- Warehouse 2 400 500
bution network. Identify the supply nodes, trans- Monthly demand 10 6
shipment nodes, and demand nodes in this network.
b. Formulate a network model for this problem as a Each vendor is able to supply as many as 10 shipments per
minimum-cost flow problem by inserting all the month. However, because of shipping limitations, each vendor
necessary data into the network drawn in part a. is only able to send a maximum of six shipments per month to
(Use the format depicted in Figure 6.3 to display each warehouse. Similarly, each warehouse is only able to send
these data.) a maximum of six shipments per month to each factory.
E* c. Formulate and solve a spreadsheet model for this Management now wants to develop a plan for each month
problem. regarding how many shipments (if any) to order from each ven-
6.4. The Audiofile Company produces boomboxes. How- dor, how many of those shipments should go to each warehouse,
ever, management has decided to subcontract out the production and then how many shipments each warehouse should send to
of the speakers needed for the boomboxes. Three vendors are each factory. The objective is to minimize the sum of the pur-
available to supply the speakers. Their price for each shipment chase costs (including the shipping charge) and the shipping
of 1,000 speakers is shown below. costs from the warehouses to the factories.
a. Draw a network that depicts the company’s supply
Vendor Price network. Identify the supply nodes, transshipment
1 $22,500 nodes, and demand nodes in this network.
2 22,700 b. This problem is only a variant of a minimum-cost
3 22,300 flow problem because the supply from each vendor
is a maximum of 10 rather than a fixed amount of
Each shipment would go to one of the company’s two ware- 10. However, it can be converted to a full-fledged
houses. In addition to the price for each shipment, each vendor minimum-cost flow problem by adding a dummy
would charge a shipping cost for which it has its own formula demand node that receives (at zero cost) all the
Unit Shipping Cost Shipping Capacity

To
From Warehouse 1 Warehouse 2 Warehouse 1 Warehouse 2 Output
Plant 1 $425 $560 125 150 200
Plant 2 510 600 175 200 300
Unit Shipping Cost Shipping Capacity

To
From RO1 RO2 RO3 RO1 RO2 RO3
Warehouse 1 $470 $505 $490 100 150 100
Warehouse 2 390 410 440 125 150 75
Demand 150 200 150 150 200 150
unused supply capacity at the vendors. Formu- 6.6. Reconsider Problem 6.5. Suppose now that, for admin-
late a network model for this minimum-cost flow istrative convenience, management has decided that all 130
problem by inserting all the necessary data into units per month needed at the distribution center in Los Angeles
the network drawn in part a supplemented by this must come from the Stuttgart factory (node ST) and all 50 units
dummy demand node. (Use the format depicted in per month needed at the distribution center in Seattle must come
Figure 6.3 to display these data.) from the Berlin factory (node BE). For each of these distribution
E* c. Formulate and solve a spreadsheet model for the centers, Karl Schmidt wants to determine the shipping plan that
company’s problem. will minimize the total shipping cost.
6.5.* Consider Figure 6.9 (in Section 6.3), which depicts the a. For the distribution center in Los Angeles, for-
BMZ Co. distribution network from its factories in Stuttgart and mulate a network model for this problem as a
Berlin to the distribution centers in both Los Angeles and Seat- minimum-cost flow problem by inserting all the
tle. This figure also gives in brackets the maximum amount that necessary data into the distribution network shown
can be shipped through each shipping lane. in Figure 6.6. (Use the format depicted in Figure 6.3
In the weeks following the crisis described in Section 6.2, to display these data.)
the distribution center in Los Angeles has successfully replen- E* b. Formulate and solve a spreadsheet model for the
ished its inventory. Therefore, Karl Schmidt (the supply chain problem formulated in part a.
manager for the BMZ Co.) has concluded that it will be suf- c. For the distribution center in Seattle, draw its distri-
ficient hereafter to ship 130 units per month to Los Angeles and bution network emanating from the Berlin factory
50 units per month to Seattle. (One unit is a hundred cubic at node BE.
meters of automobile replacement parts.) The Stuttgart factory d. Repeat part a for the distribution center in Seattle
(node ST in the figure) will allocate 130 units per month and the by using the network drawn in part c.
Berlin factory (node BE) will allocate 50 units per month out of
E* e. Formulate and solve a spreadsheet model for the
their total production to cover these shipments. However, rather
problem formulated in part d.
than resuming the past practice of supplying the Los Angeles
distribution center from only the Stuttgart factory and supplying f. Add the total shipping costs obtained in parts b and
the Seattle distribution center from only the Berlin factory, Karl e. Compare this sum with the total shipping cost
has decided to allow either factory to supply either distribution obtained in part c of Problem 6.5 (as given in the
center. He feels that this additional flexibility is likely to reduce back of the book).
the total shipping cost. 6.7. Consider the maximum flow problem formulated
The following table gives the shipping cost per unit through in Figures 6.7 and 6.8 for the BMZ case study. Redraw
each of these shipping lanes. Figure 6.7 and insert the optimal shipping quantities (cells
Unit Shipping Cost to Node

To
From LI BO RO HA NO NY BN LA SE
Node
ST $3,200 $2,500 $2,900 — — — — — —
BE — — $2,400 $2,000 — — — — —
LI — — — — $6,100 — — — —
BO — — — — $6,800 $5,400 — — —
RO — — — — — $5,900 — — —
HA — — — — — $6,300 $5,700 — —
NO — — — — — — — $3,100 —
NY — — — — — — — $4,200 $4,000
BN — — — — — — — $3,400 $3,000
Karl wants to determine the shipping plan that will minimize D4:D12 in Figure 6.8) in parentheses above the respective
the total shipping cost. arcs. Examine the capacities of these arcs. Explain why these
a. Formulate a network model for this problem as a arc capacities ensure that the shipping quantities in parenthe-
minimum-cost flow problem by inserting all the ses must be an optimal solution because the maximum flow
necessary data into the distribution network shown cannot exceed 150.
in Figure 6.9. (Use the format depicted in Figure 6.3 6.8. Read the referenced article that fully describes the man-
to display these data.) agement science study summarized in the application vignette
E* b. Formulate and solve a spreadsheet model for this presented in Section 6.3. Briefly describe how a generalization
problem. of the model for the maximum flow problem was applied in this
c. What is the total shipping cost for this optimal study. Then list the various financial and nonfinancial benefits
solution? that resulted from this study.
Chapter 6 Problems 223
E*6.9. Formulate and solve a spreadsheet model for the maxi- industries now has sharply curtailed Texago’s capacity to ship
mum flow problem shown at the top of the next column, where oil from the four oil fields to the four refineries and to ship petro-
node A is the source, node F is the sink, and the arc capacities leum products from the refineries to the distribution centers.
are the numbers in square brackets shown next to the arcs. Using units of thousands of barrels of crude oil (and its equiva-
lent in refined products), the following tables show the maxi-
B D mum number of units that can be shipped per day from each oil
[7]
field to each refinery and from each refinery to each distribution
[6] center.
[9] [2]
A [3] F
Refinery
[4] New St.
[7] [9]
Oil Field Orleans Charleston Seattle Louis
C E
[6] Texas 11 7 2 8
California 5 4 8 7
6.10. The diagram depicts a system of aqueducts that origi- Alaska 7 3 12 6
nate at three rivers (nodes R1, R2, and R3) and terminate at a Middle East 8 9 4 15
major city (node T), where the other nodes are junction points in
the system.
Distribution Center
A
Kansas San
D Refinery Pittsburgh Atlanta City Francisco
R1
New Orleans 5 9 6 4
Charleston 8 7 9 5
T Seattle 4 6 7 8
B St. Louis 12 11 9 7
E
R2
The Texago management now wants to determine a plan for
how many units to ship from each oil field to each refinery and
from each refinery to each distribution center that will maximize
the total number of units reaching the distribution centers.
R3 C
F a. Draw a rough map that shows the location of Tex-
ago’s oil fields, refineries, and distribution cen-
Using units of thousands of acre feet, the following tables ters. Add arrows to show the flow of crude oil and
show the maximum amount of water that can be pumped through then petroleum products through this distribution
each aqueduct per day. network.
b. Redraw this distribution network by lining up all the
To nodes representing oil fields in one column, all the
From A B C nodes representing refineries in a second column,
and all the nodes representing distribution centers in
R1 75 65 —
R2 40 50 60 a third column. Then add arcs to show the possible
R3 — 80 70 flow.
c. Use the distribution network from part b to formu-
late a network model for Texago’s problem as a
To To variant of a maximum flow problem.
From D E F From T E* d. Formulate and solve a spreadsheet model for this
A 60 45 — D 120 problem.
B 70 55 45 E 190 6.12. Read the referenced article that fully describes the man-
C — 70 90 F 130 agement science study summarized in the application vignette
presented in Section 6.4. Briefly describe how network opti-
The city water manager wants to determine a flow plan that will mization models (including for shortest path problems) were
maximize the flow of water to the city. applied in this study. Then list the various financial and nonfi-
a. Formulate this problem as a maximum flow prob- nancial benefits that resulted from this study.
lem by identifying a source, a sink, and the trans- E*6.13. Reconsider the Littletown Fire Department problem
shipment nodes, and then drawing the complete presented in Section 6.4 and depicted in Figure 6.11. Due to
network that shows the capacity of each arc. maintenance work on the one-mile road between nodes A and
E* b. Formulate and solve a spreadsheet model for this B, a detour currently must be taken that extends the trip between
problem. these nodes to four miles.
6.11. The Texago Corporation has four oil fields, four refin- Formulate and solve a spreadsheet model for this revised
eries, and four distribution centers in the locations identified problem to find the new shortest path from the fire station to the
in the next tables. A major strike involving the transportation farming community.
6.14. You need to take a trip by car to another town that you j
have never visited before. Therefore, you are studying a map to
determine the shortest route to your destination. Depending on 1 2 3
which route you choose, there are five other towns (call them A, i
B, C, D, E) through which you might pass on the way. The map 0 $8,000 $18,000 $31,000
shows the mileage along each road that directly connects two 1 10,000 21,000
towns without any intervening towns. These numbers are sum- 2 12,000
marized in the following table, where a dash indicates that there
is no road directly connecting these two towns without going Management wishes to determine at what times (if any) the trac-
through any other towns. tor should be replaced to minimize the total cost for the tractor(s)
over three years.
Miles between Adjacent Towns a. Formulate a network model for this problem as a
Town A B C D E Destination shortest path problem.
E* b. Formulate and solve a spreadsheet model for this
Origin 40 60 50 — — —
A 10 — 70 — —
problem.
B 20 55 40 — 6.16. One of Speedy Airlines’s flights is about to take off
C — 50 — from Seattle for a nonstop flight to London. There is some flex-
D 10 60 ibility in choosing the precise route to be taken, depending upon
E 80 weather conditions. The following network depicts the possible
routes under consideration, where SE and LN are Seattle and
a. Formulate a network model for this problem as London, respectively, and the other nodes represent various
a shortest path problem by drawing a network intermediate locations.
where nodes represent towns, links represent
roads, and numbers indicate the length of each 3.5
A D
link in miles.
E* b. Formulate and solve a spreadsheet model for this 3.4
3.4
problem.
4.6
c. Use part b to identify your shortest route. 3.6

d. If each number in the table represented your cost
(in dollars) for driving your car from one town to 4.7 3.2 3.6
SE B E LN
the next, would the answer in part c now give your
minimum-cost route? 3.3
4.
e. If each number in the table represented your time

2
3.8
(in minutes) for driving your car from one town to
3.5
the next, would the answer in part c now give your
minimum-time route? 3.4
C F
6.15.* At a small but growing airport, the local airline com-
pany is purchasing a new tractor for a tractor-trailer train to bring
luggage to and from the airplanes. A new mechanized luggage The winds along each arc greatly affect the flying time
system will be installed in three years, so the tractor will not be (and so the fuel consumption). Based on current meteorologi-
needed after that. However, because it will receive heavy use, cal reports, the flying times (in hours) for this particular flight
so that the running and maintenance costs will increase rapidly are shown next to the arcs. Because the fuel consumed is so
as it ages, it may still be more economical to replace the trac- expensive, the management of Speedy Airlines has established a
tor after one or two years. The next table gives the total net policy of choosing the route that minimizes the total flight time.
discounted cost associated with purchasing a tractor (purchase a. What plays the role of distances in interpreting this
price minus trade-in allowance, plus running and maintenance problem to be a shortest path problem?
costs) at the end of year i and trading it in at the end of year j E* b. Formulate and solve a spreadsheet model for this
(where year 0 is now). problem.
Case 6-1
Aiding Allies
Commander Votachev steps into the cold October night His two years of work training revolutionaries east of the Ural
and deeply inhales the smoke from his cigarette, savoring Mountains has proven successful; his troops now occupy
its warmth. He surveys the destruction surrounding him— seven strategically important cities in the Russian Federa-
shattered windows, burning buildings, torn roads—and smiles. tion: Kazan, Perm, Yekaterinburg, Ufa, Samara, Saratov, and
Case 6-1 Aiding Allies 225
Orenburg. His siege is not yet over, however. He looks to the launch a counter attack against Votachev to recapture the cities
west. Given the political and economic confusion in the Rus- he currently occupies. (The map is shown at the end of the case.)
sian Federation at this time, he knows that his troops will be The president also explains that all Starlifters and transports
able to conquer Saint Petersburg and Moscow shortly. Com- leave Boston or Jacksonville. All transports that have traveled
mander Votachev will then be able to rule with the wisdom across the Atlantic must dock at one of the NATO ports to
and control exhibited by his communist predecessors Lenin unload. Palletized load system trucks brought over in the trans-
and Stalin. ports will then carry all troops and materials unloaded from the
Across the Pacific Ocean, a meeting of the top security and ships at the NATO ports to the three strategic Russian cities not
foreign policy advisors of the United States is in progress yet seized by Votachev. All Starlifters that have traveled across
at the White House. The president has recently been briefed the Atlantic must land at one of the NATO airfields for refuel-
about the communist revolution masterminded by Commander ing. The planes will then carry all troops and cargo from the
Votachev and is determining a plan of action. The president NATO airfields to the three Russian cities.
reflects upon a similar October long ago in 1917, and he fears
the possibility of a new age of radical Communist rule accom- a. Draw a network showing the different routes troops and
panied by chaos, bloodshed, escalating tensions, and possibly supplies may take to reach the Russian Federation from the
nuclear war. He therefore decides that the United States needs United States.
to respond and to respond quickly. Moscow has requested assis- b. Moscow and Washington do not know when Commander
tance from the United States military, and the president plans to Votachev will launch his next attack. Leaders from the two
send troops and supplies immediately. countries therefore have agreed that troops should reach each
The president turns to General Lankletter and asks him of the three strategic Russian cities as quickly as possible.
to describe the preparations being taken in the United States The president has determined that the situation is so dire
to send the necessary troops and supplies to the Russian that cost is no object—as many Starlifters, transports, and
Federation. trucks as are necessary will be used to transfer troops and
General Lankletter informs the president that along with cargo from the United States to Saint Petersburg, Moscow,
troops, weapons, ammunition, fuel, and supplies, aircraft, and Rostov. Therefore, no limitations exist on the number of
ships, and vehicles are being assembled at two port cities with troops and amount of cargo that can be transferred between
airfields: Boston and Jacksonville. The aircraft and ships will any cities.
transfer all troops and cargo across the Atlantic Ocean to the The president has been given the information in the next
Eurasian continent. The general hands the president a list of the table about the length of the available routes between cities.
types of aircraft, ships, and vehicles being assembled along with Given the distance and the speed of the transportation
a description of each type. The list is shown next. used between each pair of cities, how can the president most
Transportation Type Name Capacity Speed

Aircraft C-141 Starlifter 150 tons 400 miles per hour
Ship Transport 240 tons 35 miles per hour
Vehicle Palletized Load 16,000 kilograms 60 miles per hour
System Truck
All aircraft, ships, and vehicles are able to carry both troops quickly move troops from the United States to each of the
and cargo. Once an aircraft or ship arrives in Europe, it stays three strategic Russian cities? Highlight the path(s) on the
there to support the armed forces. network. How long will it take troops and supplies to reach
The president then turns to Tabitha Neal, who has been nego- Saint Petersburg? Moscow? Rostov?
tiating with the NATO countries for the last several hours to use c. The president encounters only one problem with his first
their ports and airfields as stops to refuel and resupply before plan: He has to sell the military deployment to Congress.
heading to the Russian Federation. She informs the president Under the War Powers Act, the president is required to con-
that the following ports and airfields in the NATO countries will sult with Congress before introducing troops into hostilities
be made available to the U.S. military. or situations where hostilities will occur. If Congress does
Ports Airfields not give authorization to the president for such use of troops,
the president must withdraw troops after 60 days. Congress
Napoli London
also has the power to decrease the 60-day time period by
Hamburg Berlin
Rotterdam Istanbul passing a concurrent resolution.
The president knows that Congress will not authorize sig-
The president stands and walks to the map of the world pro- nificant spending for another country’s war, especially when
jected on a large screen in the middle of the room. He maps voters have paid so much attention to decreasing the national
the progress of troops and cargo from the United States to three debt. He therefore decides that he needs to find a way to get
strategic cities in the Russian Federation that have not yet been the needed troops and supplies to Saint Petersburg, Moscow,
seized by Commander Votachev. The three cities are Saint and Rostov at the minimum cost.
Petersburg, Moscow, and Rostov. He explains that the troops Each Russian city has contacted Washington to commu-
and cargo will be used both to defend the Russian cities and to nicate the number of troops and supplies the city needs at
From To (Kilometers) From To Cost

Boston Berlin 7,250 km Boston Berlin $50,000 per Starlifter
Boston Hamburg 8,250 Boston Hamburg $30,000 per transport
Boston Istanbul 8,300 Boston Istanbul $55,000 per Starlifter
Boston London 6,200 Boston London $45,000 per Starlifter
Boston Rotterdam 6,900 Boston Rotterdam $30,000 per transport
Boston Napoli 7,950 Boston Napoli $32,000 per transport
Jacksonville Berlin 9,200 Jacksonville Berlin $57,000 per Starlifter
Jacksonville Hamburg 9,800 Jacksonville Hamburg $48,000 per transport
Jacksonville Istanbul 10,100 Jacksonville Istanbul $61,000 per Starlifter
Jacksonville London 7,900 Jacksonville London $49,000 per Starlifter
Jacksonville Rotterdam 8,900 Jacksonville Rotterdam $44,000 per transport
Jacksonville Napoli 9,400 Jacksonville Napoli $56,000 per transport
Berlin Saint Petersburg 1,280 Berlin Saint Petersburg $24,000 per Starlifter
Hamburg Saint Petersburg 1,880 Hamburg Saint Petersburg $3,000 per truck
Istanbul Saint Petersburg 2,040 Istanbul Saint Petersburg $28,000 per Starlifter
London Saint Petersburg 1,980 London Saint Petersburg $22,000 per Starlifter
Rotterdam Saint Petersburg 2,200 Rotterdam Saint Petersburg $3,000 per truck
Napoli Saint Petersburg 2,970 Napoli Saint Petersburg $5,000 per truck
Berlin Moscow 1,600 Berlin Moscow $22,000 per Starlifter
Hamburg Moscow 2,120 Hamburg Moscow $4,000 per truck
Istanbul Moscow 1,700 Istanbul Moscow $25,000 per Starlifter
London Moscow 2,300 London Moscow $19,000 per Starlifter
Rotterdam Moscow 2,450 Rotterdam Moscow $5,000 per truck
Napoli Moscow 2,890 Napoli Moscow $5,000 per truck
Berlin Rostov 1,730 Berlin Rostov $23,000 per Starlifter
Hamburg Rostov 2,470 Hamburg Rostov $7,000 per truck
Istanbul Rostov 990 Istanbul Rostov $2,000 per Starlifter
London Rostov 2,860 London Rostov $4,000 per Starlifter
Rotterdam Rostov 2,760 Rotterdam Rostov $8,000 per truck
Napoli Rostov 2,800 Napoli Rostov $9,000 per truck
a minimum for reinforcement. After analyzing the requests, used between the United States and the Russian Federation
General Lankletter has converted the requests from numbers on the network.
of troops, gallons of gasoline, and so on, to tons of cargo for d. Once the president releases the number of planes, ships, and
easier planning. The requirements are listed below. trucks that will travel between the United States and the Russian
Federation, Tabitha Neal contacts each of the American cities
City Requirements and NATO countries to indicate the number of planes to expect
Saint Petersburg 320,000 tons at the airfields, the number of ships to expect at the docks,
Moscow 440,000 tons and the number of trucks to expect traveling across the roads.
Rostov 240,000 tons Unfortunately, Tabitha learns that several additional restrictions
exist that cannot be immediately eliminated. Because of air-
Both in Boston and Jacksonville, there are 500,000 tons field congestion and unalterable flight schedules, only a limited
of the necessary cargo available. When the United States number of planes may be sent between any two cities. These
decides to send a plane, ship, or truck between two cities, plane limitations are given below.
several costs occur: fuel costs, labor costs, maintenance
costs, and appropriate port or airfield taxes and tariffs. These
costs are listed next. Maximum Number of
The president faces a number of restrictions when trying From To Airplanes
to satisfy the requirements. Early winter weather in northern Boston Berlin 300
Russia has brought a deep freeze with much snow. Therefore, Boston Istanbul 500
General Lankletter is opposed to sending truck convoys in the Boston London 500
Jacksonville Berlin 500
area. He convinces the president to supply Saint Petersburg
Jacksonville Istanbul 700
only through the air. Moreover, the truck routes into Rostov
Jacksonville London 600
are quite limited, so that from each port, at most 2,500 trucks Berlin Saint Petersburg 500
can be sent to Rostov. The Ukrainian government is very sen- Istanbul Saint Petersburg 0
sitive about American airplanes flying through its air space. London Saint Petersburg 1,000
It restricts the U.S. military to at most 200 flights from Berlin Berlin Moscow 300
to Rostov and to at most 200 flights from London to Rostov. Istanbul Moscow 100
(The U.S. military does not want to fly around the Ukraine London Moscow 200
and is thus restricted by the Ukrainian limitations.) Berlin Rostov 0
How does the president satisfy each Russian city’s mili- Istanbul Rostov 900
London Rostov 100
tary requirements at minimum cost? Highlight the path to be
Case 6-2 Money in Motion 227
In addition, because some countries fear that citizens will Tabitha learns that all shipping lanes have no capacity limits
become alarmed if too many military trucks travel the pub- due to the American control of the Atlantic Ocean.
lic highways, they object to a large number of trucks traveling The president realizes that due to all the restrictions, he
through their countries. These objections mean that a limited will not be able to satisfy all the reinforcement requirements
number of trucks are able to travel between certain ports and of the three Russian cities. He decides to disregard the cost
Russian cities. These limitations are listed below. issue and instead to maximize the total amount of cargo he
can get to the Russian cities. How does the president maxi-
From To Maximum Number of Trucks
mize the total amount of cargo that reaches the Russian Fed-
Rotterdam Moscow 600 eration? Highlight the path(s) used between the United States
Rotterdam Rostov 750 and the Russian Federation on the network.
Hamburg Moscow 700
Hamburg Rostov 500
Napoli Moscow 1,500
Napoli Rostov 1,400
St.
Boston Perm
Petersburg
Kazan Yekaterinburg
Moscow
Samara Ufa
London Hamburg Berlin
Orenburg
Jacksonville Saratov
Rotterdam
Rostov
Napoli Istanbul
Case 6-2
Money in Motion
Jake Nguyen runs a nervous hand through his once finely Associates held in the Japanese market. Because the Japanese
combed hair. He loosens his once perfectly knotted silk tie. And market had performed better than expected over the past year,
he rubs his sweaty hands across his once immaculately pressed Jake had increased investments in Japan from $2.5 million to
trousers. Today has certainly not been a good day. $15 million only one month ago. At that time, one dollar was
Over the past few months, Jake had heard whispers circu- worth 80 yen.
lating from Wall Street—whispers from the lips of investment No longer. Jake realizes that today’s devaluation of the yen
bankers and stockbrokers famous for their outspokenness. They means that one dollar is worth 125 yen. He will be able to liqui-
had whispered about a coming Japanese economic collapse— date these investments without any loss in yen, but now the dol-
whispered because they had believed that publicly vocalizing lar loss when converting back into U.S. currency would be huge.
their fears would hasten the collapse. He takes a deep breath, closes his eyes, and mentally prepares
And, today, their very fears have come true. Jake and his himself for serious damage control.
colleagues gather around a small television dedicated exclu- Jake’s meditation is interrupted by a booming voice calling
sively to the Bloomberg channel. Jake stares in disbelief as he for him from a large, corner office. Grant Hill, the president of
listens to the horrors taking place in the Japanese market. And Grant Hill Associates, yells, “Nguyen, get the hell in here!”
the Japanese market is taking the financial markets in all other Jake jumps and looks reluctantly toward the corner office
East Asian countries with it on its tailspin. He goes numb. As hiding the furious Grant Hill. He smooths his hair, tightens his
manager of Asian foreign investment for Grant Hill Associates, tie, and walks briskly into the office.
a small West Coast investment boutique specializing in currency Grant Hill meets Jake’s eyes upon his entrance and continues
trading, Jake bears personal responsibility for any negative yelling, “I don’t want one word out of you, Nguyen! No excuses;
impacts of the collapse. And Grant Hill Associates will experi- just fix this debacle! Get all of our money out of Japan! My gut
ence negative impacts. tells me this is only the beginning! Get the money into safe U.S.
Jake had not heeded the whispered warnings of a Japanese bonds! NOW! And don’t forget to get our cash positions out of
collapse. Instead, he had greatly increased the stake Grant Hill Indonesia and Malaysia ASAP with it!”
Jake has enough common sense to say nothing. He nods his always can find on-the-minute exchange rates for most curren-
head, turns on his heels, and practically runs out of the office. cies in the world (see Table 1).
Safely back at his desk, Jake begins formulating a plan to The table states that, for example, 1 Japanese yen equals
move the investments out of Japan, Indonesia, and Malaysia. His 0.008 U.S. dollars. By making a few phone calls, he discovers
experiences investing in foreign markets have taught him that the transaction costs his company must pay for large currency
when playing with millions of dollars, how he gets money out of transactions during these critical times (see Table 2).
a foreign market is almost as important as when he gets money Jake notes that exchanging one currency for another one
out of the market. The banking partners of Grant Hill Associates results in the same transaction cost as a reverse conversion.
charge different transaction fees for converting one currency into Finally, Jake finds out the maximum amounts of domestic cur-
another one and wiring large sums of money around the globe. rencies his company is allowed to convert into other currencies
And now, to make matters worse, the governments in East in Japan, Indonesia, and Malaysia (see Table 3).
Asia have imposed very tight limits on the amount of money an
individual or a company can exchange from the domestic cur- a. Formulate Jake’s problem as a minimum-cost flow problem,
rency into a particular foreign currency and withdraw it from and draw the network for his problem. Identify the supply
the country. The goal of this dramatic measure is to reduce the and demand nodes for the network.
outflow of foreign investments out of those countries to prevent b. Which currency transactions must Jake perform to con-
a complete collapse of the economies in the region. Because of vert the investments from yens, rupiahs, and ringgits
Grant Hill Associates’ cash holdings of 10.5 billion Indonesian into U.S. dollars to ensure that Grant Hill Associates has
rupiahs and 28 million Malaysian ringgits, along with the hold- the maximum dollar amount after all transactions have
ings in yen, it is not clear how these holdings should be con- occurred? How much money does Jake have to invest in
verted back into dollars. U.S. bonds?
Jake wants to find the most cost-effective method to con- c. The World Trade Organization forbids transaction limits
vert these holdings into dollars. On his company’s website, he because they promote protectionism. If no transaction limits
TABLE 1
Currency Exchange Rates
To
From Yen Rupiah Ringgit U.S. Dollar Canadian Dollar Euro Pound Peso
Japanese yen 1 50 0.04 0.008 0.01 0.0064 0.0048 0.0768
Indonesian rupiah 1 0.0008 0.00016 0.0002 0.000128 0.000096 0.001536
Malaysian ringgit 1 0.2 0.25 0.16 0.12 1.92
U.S. dollar 1 1.25 0.8 0.6 9.6
Canadian dollar 1 0.64 0.48 7.68
European euro 1 0.75 12
English pound 1 16
Mexican peso 1
TABLE 2
Transaction Cost (Percent)
To
Yen — 0.5 0.5 0.4 0.4 0.4 0.25 0.5
Rupiah — 0.7 0.5 0.3 0.3 0.75 0.75
Ringgit — 0.7 0.7 0.4 0.45 0.5
U.S. dollar — 0.05 0.1 0.1 0.1
Canadian dollar — 0.2 0.1 0.1
Euro — 0.05 0.5
Pound — 0.5
Peso —
TABLE 3
Transaction Limits in Equivalent of 1,000 Dollars
To
Yen — 5,000 5,000 2,000 2,000 2,000 2,000 4,000
Rupiah 5,000 — 2,000 200 200 1,000 500 200
Ringgit 3,000 4,500 — 1,500 1,500 2,500 1,000 1,000
Case 6-3 Airline Scheduling 229
exist, what method should Jake use to convert the Asian limits, what currency transactions should Jake perform to
holdings from the respective currencies into dollars? convert the Asian holdings from the respective currencies
d. In response to the World Trade Organization’s mandate into dollars?
forbidding transaction limits, the Indonesian government e. Jake realizes that his analysis is incomplete because he has
introduces a new tax to protect its currency that leads to a not included all aspects that might influence his planned
500 percent increase in transaction costs for transactions of currency exchanges. Describe other factors that Jake should
rupiahs. Given these new transaction costs but no transaction examine before he makes his final decision.
Case 6-3
Airline Scheduling
Richard Cook is very concerned. Until recently, he has always (including some new ones) for the coming year that could feasi-
had the golden touch, having successfully launched two start- bly be flown by the four airplanes.
up companies that made him a very wealthy man. However, A little over a decade ago, Richard had been an honor gradu-
the timing could not have been worse for his latest start-up— ate of a leading MBA program. He had enjoyed the management
a regional airline called Northwest Commuter that operates on science course he took then and he has decided to apply spread-
the west coast of the United States. All had been well at the sheet modeling to analyze his problem.
beginning. Four airplanes had been leased and the company The leasing cost for each airplane is $30,000 per day. At the
had become fairly well established as a no-frills airline provid- end of the day, an airplane might remain in the city where it
ing low-cost commuter flights between the west coast cities of landed on its last flight. Another option is to fly empty overnight
Flight Expected
Number From To Depart Arrive Revenue ($000)
1257 Seattle San Francisco 8:00 AM 10:00 AM 37
2576 Seattle Portland 9:30 AM 10:30 AM 20
8312 Seattle San Francisco 9:30 AM 11:30 AM 25
1109 Seattle San Francisco 12:00 PM 2:00 PM 27
2498 Seattle Portland 3:00 PM 4:00 PM 18
8423 Seattle Portland 6:30 PM 7:30 PM 27
7922 Portland Seattle 9:00 AM 10:00 AM 20
5623 Portland San Francisco 9:30 AM 11:00 AM 23
2448 Portland San Francisco 11:00 AM 12:30 PM 19
1842 Portland Seattle 12:00 PM 1:00 PM 21
4361 Portland San Francisco 4:00 PM 5:30 PM 29
1288 San Francisco Seattle 8:00 AM 10:00 AM 32
3335 San Francisco Portland 8:30 AM 10:00 AM 26
9348 San Francisco Seattle 10:30 AM 12:30 PM 24
7400 San Francisco Seattle 12:00 PM 2:00 PM 27
7328 San Francisco Portland 12:00 PM 1:30 PM 24
6386 San Francisco Portland 4:00 PM 5:30 PM 28
6923 San Francisco Seattle 5:00 PM 7:00 PM 32
Seattle, Portland, and San Francisco. Achieving fast turnaround to another city to be ready to start a flight from there the next
times between flights had given Northwest Commuter an impor- morning. The cost of this latter option is $5,000.
tant competitive advantage. Then the cost of jet fuel began spi- The table above shows the 22 possible flights that are being
raling upward and the company began going heavily into the red considered for the coming year. The last column gives the esti-
(like so many other airlines at the time). Although some of the mated net revenue (in thousands of dollars) for each flight, given
flights were still profitable, others were losing a lot of money. the average number of passengers anticipated for that flight.
Fortunately, jet fuel costs now are starting to come down, but
it has become clear to Richard that he needs to find new ways a. To simplify the analysis, assume for now that there is vir-
for Northwest Commuter to become a more efficient airline. tually no turnaround time between flights so the next flight
In particular, he wants to start by dropping unprofitable flights can begin as soon as the current flight ends. (If an immediate
and then identifying the most profitable combination of flights next flight is not available, the airplane would wait until the
next scheduled flight from that city.) Develop a network that ground for unloading and loading passengers between the
displays some of the feasible routings of the flights. (Hint: arrival of a flight and the departure of the next flight by the
Include separate nodes for each half hour between 8:00 am same airplane. (Most airlines use a considerably longer turn-
and 7:30 pm in each city.) Then develop and apply the corre- around time.) Does this change the number of flights that can
sponding spreadsheet model that finds the feasible combina- be flown?
tion of flights that maximizes the total profit. d. Richard now is considering having each of the four airplanes
b. Richard is considering leasing additional airplanes to achieve carry freight instead of flying empty if it flies overnight to
economies of scale. The leasing cost of each one again would another city. Instead of a cost of $5,000, this would result in
be $30,000 per day. Perform what-if analysis to determine net revenue of $5,000. Adapt the spreadsheet model used in
whether it would be worthwhile to have 5, 6, or 7 airplanes part c to find the feasible combination of flights that maxi-
instead of 4. mizes the total profit. Does this change the number of air-
c. Now repeat part a under the more realistic assumption that planes that fly overnight to another city?
there is a minimum turnaround time of 30 minutes on the
Case 6-4
Broadcasting the Olympic Games
The management of the WBC television network has been each link in the network is shown in the diagram below (in
celebrating for days. What a coup! After several unsuccessful GB/s). WBC can divide the transmission and route it through
attempts in recent decades, they finally have hit the big jackpot. multiple paths of the network from A to G, so long as the total
They have won the bidding war to gain the rights to broadcast bandwidth required on each link does not exceed the capacity
the next Summer Olympic Games! of that link.
The price was enormous. However, the advertising income
also will be huge. Even if the network loses some money in the a. By utilizing the entire computer network, what is the maxi-
process, the gain in prestige should make it all worthwhile. After mum bandwidth available (in GB/s) for transmission from
all, the entire world follows these games closely every four years. the general site of the Olympic Games (node A) to the home
Now the entire world receiving the feed of the broadcast from the studios (node G)? Set up and solve a linear programming
WBC network will learn what a preeminent network it is. spreadsheet model.
7 F
B
13 4 9
5
A E
6
9 G
3
6 10
12
D
C 8
However, reality also is setting in for WBC management. b. WBC would like to expand the capacity of the network so it
Telecasting the entire Olympic Games will be an enormously can handle the peak requirement of 35 GB/s from the Olym-
complex task. Many different sporting events will be occurring pics site (A) to the home studios (G). WBC can increase the
simultaneously in far-flung venues. An unprecedented amount capacity of each link of the computer network by install-
of live television and live-on-the-Internet coverage of the vari- ing additional fiber optic cables. The table on the next page
ous sporting events needs to be planned. shows the existing capacity of each network segment (in
Due to the high amount of bandwidth that will be required GB/s), the maximum additional capacity that can be added
to transmit the coverage of the games back to its home studios, (in GB/s), and the cost to increase the capacity (in millions of
WBC needs to upgrade its computer network. It operates a pri- dollars per unit GB/s added). Make a copy of the spreadsheet
vate computer network as shown in the network diagram in the model used to solve part a and make any revisions necessary
right-hand column. The games will be held near node A. WBC’s to solve this new problem.
home studios are located at node G. At peak times, coverage
of the games will require 35 GB/s (GB per second) to be sent Note: This case will be continued in the next chapter (Case 7-4),
through the network from node A to node G. The capacity of so we suggest that you save your spreadsheet model from part b.
Additional Cases 231
Network Segment Cost per GB/s

Existing Maximum Additional of Additional Capacity
From To Capacity (GB/s) Capacity (GB/s) ($million)
A B 13 6 2.8
A C 6 4 2.5
A D 10 3 2.8
B D 9 4 2.5
B E 5 5 3.1
B F 7 3 1.6
C D 8 5 3.9
D E 3 2 2.8
D G 12 5 1.6
E F 4 2 4.6
E G 6 4 2.9
F G 9 5 1.8
Additional Cases
Additional cases for this chapter also are available at the
University of Western Ontario Ivey School of Business website,
cases.ivey.uwo.ca/cases, in the segment of the CaseMate area
designated for this book.

Chapter 6 Network Optimization Problems

Uploaded by

Copyright:

Available Formats

Chapter 6 Network Optimization Problems

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 6 Network Optimization Problems

Uploaded by

Copyright:

Available Formats

Chapter Six

1. Formulate network models for various types of network optimization problems.

6.1 MINIMUM-COST FLOW PROBLEMS

An Example: The Distribution Unlimited Co. Problem

FIGURE 6.2 80 units $700/unit 60 units

[80] [–60] [80] [–60]

[70] [–90] [70] [–90]

FIGURE 6.3 FIGURE 6.4

Assumptions of a Minimum-Cost Flow Problem

Using Excel to Formulate and Solve Minimum-Cost Flow Problems

Solver Parameters Range Name Cells J

Solving Large Minimum-Cost Flow Problems More Efficiently

TABLE 6.1 Typical Kinds of Applications of Minimum-Cost Flow Problems

Kind of Application Supply Nodes Transshipment Nodes Demand Nodes

Special Types of Minimum-Cost Flow Problems

6.2 A CASE STUDY: THE BMZ CO. MAXIMUM FLOW PROBLEM

Review 1. What is the current crisis facing the BMZ Co.?

Solver Parameters Range Name Cells I

6.3 MAXIMUM FLOW PROBLEMS

Continuing the Case Study with Multiple Supply Points

Solver Parameters Range Name Cells I

Solving Very Large Problems

reformulate a variant of a maximum flow problem so that an extremely efficient special-

6.4 SHORTEST PATH PROBLEMS

An Example: The Littletown Fire Department Problem

Model Formulation for the Littletown Problem

An Example of Minimizing Total Cost

TABLE 6.2 Operating and Maintenance Costs Trade-in Value at End

FIGURE 6.14 25,000

Trade in the first car at the end of year 2.

An Example of Minimizing Total Time

Range Name Cells E

Learning Aids for This Chapter in Your MS Courseware

Solved Problems (See the CD-ROM or Website for the Solutions)

$35 $55 6.S3. Driving to the Mile-High City

Unit Shipping Cost Shipping Capacity

Unit Shipping Cost Shipping Capacity

Unit Shipping Cost to Node

c. Use part b to identify your shortest route. 3.6

e. If each number in the table represented your time

Transportation Type Name Capacity Speed

From To (Kilometers) From To Cost

Network Segment Cost per GB/s

You might also like