View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Ithaca College

Ithaca College
Digital Commons @ IC
Mathematics Honors Theses Mathematics Department

2014

Operations Research
Jihyun Rachel Lee
Ithaca College

Follow this and additional works at: http://digitalcommons.ithaca.edu/math_honors_theses

Part of the Mathematics Commons

Recommended Citation
Lee, Jihyun Rachel, "Operations Research" (2014). Mathematics Honors Theses. Paper 4.

This Thesis is brought to you for free and open access by the Mathematics Department at Digital Commons @ IC. It has been accepted for inclusion in
Mathematics Honors Theses by an authorized administrator of Digital Commons @ IC.
Operations Research

Jihyun Rachel Lee

[email protected]
Honors Thesis
Spring 2014

Acknowledgment: Kazem Mahdavi

[email protected]

0
 Table of Contents

1. Introduction

2. Linear Programming
2.1Standard Maximum program
2.2Duality
2.3Solving LP using excel
2.4Simplex Method

3. Dynamic Programming
3.1 Scientist problem

1
1. Introduction

Since the industrial revolutions, there has been a significant growth in the size
and complexity of organizations. One of the most notable characteristics of this
revolutionary change have been the increase in the division of labor and segmentation
of management responsibilities in these organizations. While the results have been
tremendous, one of the problem was that the divisions and components of
organizations often grew autonomously, losing sight of the overall organizational
objectives and benefits. A related problem was that as the complexity of organization
increased, it became much more difficult to allocate the resources to various activities in
accordance with the most productive, valuable way for the overall organizations. The
need to solve these problem let to emergence of Operation Research (often referred to
as OR).

The word Operation Research came from the military services early in World
War II. As the countries needed to allocate scarce resources to the various military
operations and activities in most effective ways, the British and then the U.S. military
management called a large number of scientists to come up with the strategies. The
scientists were asked to do research on (military) operations. They were instrumental in
winning the Air Battle of Britain as well as the Battle of the North Atlantic. After the war,
OR have been applied to other areas outside the military. By 1950s, a variety of
organizations in business, industry, and government adopted the use of OR.

As implied in the name, operation research is used in the problems regarding

how to conduct and coordinate the operations within an organization. OR has been used
extensively in such diverse areas as manufacturing, transportation, construction,
telecommunications, financial planning, health care, the military, and public services,
etc. The research part of the name came as the OR uses an approach that is similar to
that of scientific fields. Thus, the OR begins by carefully observing and formulating
problems, including gathering data. Then, the scientific/ mathematical model is
constructed to abstracts the essence of the real problem and to hypothesize it. Next,
suitable experiments are conducted to test this hypothesis. Thus, OR involves creative
scientific research. However, as OR is specifically concerned with the practical
management of the organizations, a successful OR provides positive, practical and
understandable conclusions to the decision makers. Additionally, the OR adopts an
organizational point of view – resolve the conflicts of interest among the division of
organizations in a way that is best for the organization as a whole. (See reference [2])

2
Application Vignette 1

Federal Express (FedEx), the  world’s  largest  express  transportation  company,  

delivers more than 6.5 million documents, packages, and other items throughout the
United States and more than220 countries and territories around the world. The logical
challenges in this services are staggering: these millions of daily shipments must be
individually sorted and routed to the correct general location and then delivered to the
exact destination in an extremely short period of time. OR is the technological engine
that make this possible. Since its founding in 1973, OR was used to make decisions
regarding equipment investment, route structure, scheduling, finances, and location of
facilities, etc. As OR was credited with literally saving the company during its early years,
it became the custom to have OR represented at the weekly senior management
meetings and, indeed, several of the senior corporative vice presidents have come up
from the outstanding FedEx OR group. (See reference [2])

Application Vignette 2

Prostate cancer is the most common form of cancer diagnosed in men. Like
many other forms of cancer, radiation therapy is a common method to treat this cancer,
where the goal is to have enough radiation dosage in the tumor region to kill the
malignant cells while minimizing the radiation exposure to the healthy structure near
the tumor. This radiation therapy also uses OR: the research process (optimization
which will be explained soon) is done in a matter of minutes by an automated
computerized planning system that can be operated readily by medical personnel when
beginning  the  procedure  of  inserting  the  seeds  into  the  patient’s  prostate.  OR  thus  has  
had a great impact on both health care costs and quality of life for treated patients
because of tis much greater effectiveness and the substantial reduction in side effects. It
is estimated that annual cost saving would be about $500 million when all U.S. clinics
adopt this procedure, as it will eliminate the need for pretreatment planning meeting
and post-operation CT scan, as well as providing a more efficient surgical procedure and
reducing the need to treat subsequent side effects. This approach is also anticipated to
be applied to treatment of breast, cervix, esophagus, biliary tract, pancreas, head and
neck, and eyes. (See reference [2])

3
 2. Linear Programming
The development of linear programming (LP) has been ranked among the most important
scientific advances of the mid-20th century. Linear programming, the very basic but powerful
tool of OR, involves the general problem of allocating limited resources among competing
activities in a best possible way; Linear programming helps to select the level of certain
activities that compete for scarce resources to make the best outcome possible for the
organization as a whole.

As its name implies, LP deals with mathematical model that uses only linear functions. The
word programming indicates planning. Thus, LP is planning of activities to obtain an optimal
result – result that reaches the specified goal best among all feasible alternatives.

In more technical term, linear programming may be defined as maximizing or minimizing a

linear function subject to linear constraints.

2.1 Standard Maximum Problem

We are given an m-vector, b= (𝑏 ,…,

𝑏 ) , an n-vector, c = (𝑐 ,…,
𝑐 ) , and an m x n matrix,

𝑎 𝑎 ⋯ 𝑎
𝑎 𝑎 ⋯ 𝑎
A= ⋮ ⋮ ⋱ ⋮
𝑎 𝑎 ⋯ 𝑎

of real numbers.
Find an n-vector, x =(𝑥 ,…, 𝑥 ) , to maximize

𝑐 𝑥 = 𝑐 𝑥 + ⋯+ 𝑐 𝑥

Subject to the constraints (𝐴𝑥 ≤ 𝑏)

𝑎 𝑥 + 𝑎 𝑥 + ⋯ + 𝑎 𝑥   ≤ 𝑏
𝑎 𝑥 + 𝑎 𝑥 + ⋯+ 𝑎 𝑥 ≤ 𝑏
⋮
𝑎 𝑥 + 𝑎 𝑥 + ⋯+ 𝑎 𝑥 ≤ 𝑏
and  (𝑥 ≥ 0)

𝑥 ≥ 0, 𝑥 ≥ 0, … 𝑥 ≥ 0

The Standard Minimum Problem:

Find an m-vector, y =(𝑦 ,…, 𝑦 )   , to minimize

4
 𝑦 𝑏 = 𝑦 𝑏 + ⋯+ 𝑦 𝑏
Subject to the constrains (𝑦 𝐴 ≤ 𝑐 )

𝑦 𝑎 + 𝑦 𝑎 + ⋯+ 𝑦 𝑎   ≤ 𝑐
𝑦 𝑎 + 𝑦 𝑎 + ⋯+ 𝑦 𝑎 ≤𝑐
⋮
𝑦 𝑎 + 𝑦 𝑎 + ⋯+ 𝑦 𝑎 ≤𝑐
and  (𝑦 ≥ 0)

𝑦 ≥ 0, 𝑦 ≥ 0, … 𝑦 ≥ 0

Terminology

Objective function: the function to be maximized or minimized.

Feasible: a vector, x for the standard maximum problem or y for the standard minimum
problem if it satisfies the corresponding constraints.
Constraint set: the set of feasible vector
A linear programming problem is said to be feasible if the constraint set is not empty;
otherwise it is said to be infeasible.
A feasible maximum/ minimum problem is said to be unbounded if the objective
function can assume arbitrarily large positive/ negative values at feasible vectors;
otherwise, it is said to be bounded. There are three possibilities for a linear
programming problem. It may be bounded feasible, it may be unbounded feasible, and
it may be infeasible.
The value of a bounded feasible maximum/ minimum problem is the maximum/
minimum value of the objective function as the variable range over the constraint set.
A feasible vector at which the objective function achieves the value is called optimal.
(See reference [1])

Example 1

Find numbers 𝑥 𝑎𝑛𝑑  𝑥 that maximize the sum 𝑥 +   𝑥 subject to the constraints 𝑥 ≥ 0, 𝑥 ≥
0, and

𝑥 + 2𝑥   ≤ 4
       4𝑥 + 2𝑥   ≤ 12
−𝑥 + 𝑥   ≤ 1

This standard maximum problem has two unknowns and five constraints. We have the non-
negativity constraints, 𝑥 ≥ 0  𝑎𝑛𝑑  𝑥 ≥ 0. The objective function here is 𝑥 +   𝑥 .

5
One way to approach this is to draw graphs.

Drawing graphs helps us see that the maximum occurs along an entire edge or a corner point of
the bounded regions.

Without drawing graphs, we can also solve this problem. The function 𝑥 +   𝑥 is constant on
line with slope -1, and we can move this line 𝑥 +   𝑥 =1 further from the origin. Thus, we find a
point where the line of slope -1 is farthest from the origin and still lies within the constraints.
This is where  𝑥 + 2𝑥 = 4 and        4𝑥 + 2𝑥 = 12 intersect:  𝑥 =  𝑎𝑛𝑑    𝑥 = .   Thus the
solution to the objective function is   (See reference [1]).

2.2 Duality
Every linear program has a dual linear program.

Definition. The dual of the standard maximum problem,

Maximize 𝑐 𝑥
Subject to the constraints 𝐴𝑥 ≤ 𝑏 and 𝑥 ≥ 0

is defined to be the standard minimum problem

Minimize 𝑦 𝑏
Subject to the constraints 𝑦 𝐴 ≤ 𝑐 and 𝑦 ≥ 0

6
Let’s  recall  the  numerical  example  from  above:  Find  𝑥  𝑎𝑛𝑑  𝑥 to maximize 𝑥 +   𝑥 subject to
the constraints 𝑥 ≥ 0, 𝑥 ≥ 0, 𝑎𝑛𝑑

𝑥 + 2𝑥   ≤ 4
4𝑥 + 2𝑥 ≤ 12
−𝑥 + 𝑥   ≤ 1
The dual of this standard maximum problem is the standard minimum problem:

Find 𝑦 , 𝑦 , 𝑎𝑛𝑑  𝑦  to minimize 4𝑦 + 12𝑦 + 𝑦  subject to the constraints: 𝑦 ≥ 0, 𝑦 ≥

0, 𝑦 ≥ 0 , and

𝑦 + 2𝑦 − 𝑦 ≥ 1
2𝑦 + 2𝑦 + 𝑦 ≥ 1

The general standard maximum problem and the dual standard minimum problem may be
simultaneously exhibited in the display:

Thus the example could be expressed as

Example 2

Find the dual to the following standard minimum problem. Find 𝑦 , 𝑦 , 𝑎𝑛𝑑  𝑦 to minimize𝑦 +
2𝑦 + 𝑦 , subject to the constraints 𝑦 ≥ 0 for all 𝑖, and

𝑦 − 2𝑦 + 𝑦 ≥ 2
−𝑦 + 𝑦 + 𝑦 ≥ 4
2𝑦 + 𝑦 ≥ 6

7
 𝑦 +𝑦 +𝑦 ≥2

We can easily set up the table:

𝑥 𝑥 𝑥 𝑥
𝑦 1 -1 2 1 1
𝑦 -2 1 0 1 2
𝑦 1 1 1 1 1
2 4 6 2

Thus, the dual for the problem is to maximize

2𝑥 4𝑥 + 6𝑥 + 2𝑥

subject to the constraints 𝑥 ≥ 0 for all 𝑖, and

𝑥 − 𝑥 + 2𝑥 + 𝑥 ≤ 1
−2𝑥 + 𝑥 + 𝑥 ≤ 2
𝑥 +𝑥 +𝑥 +𝑥 ≤1

By solving the dual of the problem, we can make sure that we have the correct answer.

2.3 Simplex Method

Example 3

Maximize 𝑝 =   50𝑥   +   80𝑥

Subject to 𝑥 ≥ 0, 𝑥 ≥ 0   and

𝑥 + 2𝑥 ≤ 32
3𝑥 + 4𝑥 ≤ 84

We can re-write these equations - using the slack variables - as:

𝑥 + 2𝑥 + 𝑠 = 32
3𝑥 + 4𝑥 +  𝑟 = 84
𝑝 − 50𝑥   −   80𝑥 = 0

Here, the non-basic variables are : 𝑥 , 𝑥 , 𝑎𝑛𝑑  𝑝 and the basic variables are: 𝑠  𝑎𝑛𝑑  𝑟

8
Then, we create the simplex tableau.

𝑥 𝑥 𝑠 𝑟 𝑝
𝑠 1 2 1 0 0 32
𝑟 3 4 0 1 0 84

𝑝 -50 -80 0 0 1 0

Here, we pick 𝑥 as the pivot column, which is farthest from the origin and thus has the highest
impact. Then among the two non-basic variable rows, we pick 𝑠 to be the pivot row because;

When 𝑥 = 0,
𝑠 = 32 − 2𝑥 ;  𝑥 = 16
𝑟 = 84 − 4𝑥 ;  𝑥 = 21

We need to pick a smaller number to remove the risk of crossing over the feasible reason.
Picking a bigger number has a chance of getting s and r as negative values when doing simplex
method.

𝑥 𝑥 𝑠 𝑟 𝑝
𝑠 1 2 1 0 0 32
𝑟 3 4 0 1 0 84
𝑝 -50 -80 0 0 1 0

Now, we can replace the basic variable, s, with non-basic variable, 𝑥 . Then, we need to have
the value of 𝑥 = 1 while the value of every other variable in pivot column to be 0. Thus, we
first divide the pivot row by 2.

𝑥 𝑥 𝑠 𝑟 𝑝
𝑥 1/2 1 1/2 0 0 16
𝑟 3 4 0 1 0 84
𝑝 -50 -80 0 0 1 0

9
Then, we do the operation:

( - row) − 4 * (pivot row) and

(𝑝 - row) + 80*(pivot row)

𝑥 𝑥 𝑠 𝑟 𝑝
𝑥 1/2 1 1/2 0 0 16
𝑟 1 0 -2 1 0 20
𝑝 -10 0 40 0 1 1280

Then, we do the same process again: select the remaining 𝑥 to be pivot column, and r to be
the pivot column. We also replacing basic variable, 𝑟, with non- basic variable, 𝑥 .  

𝑥 𝑥 𝑠 𝑟 𝑝
𝑥 1/2 1 1/2 0 0 16
𝑥 1 0 -2 1 0 20
𝑝 -10 0 40 0 1 1280

We then reduce value of 𝑥 to be 1 and make the remaining basic variables in pivot column to
be 0 by doing operations:

2*(𝑥 -row) – (pivot row) and

(𝑝-row) + (pivot row)

𝑥 𝑥 𝑠 𝑟 𝑝
𝑥 0 1 3/2 -1/2 0 6
𝑥 1 0 -2 1 0 20
𝑝 0 0 60 10 1 1480

We have completed the simplex method. Thus, this simplex tableau tells us that:

𝑥 = 20
𝑥 =6

10
and the solution to the objective function to be 1480. And this is correct because our objective
function was  𝑝 = 50𝑥   +   80𝑥 ; 50 (20) + 80(6) = 1480.

2.4 Excel Solver to solve the standard maximum/minimum problem

One of the many linear programming program, I looked into the Excel solver – which was most
accessible to me.

Example 4

The WYNDOR GLASS CO. produces high-quality glass products, including windows and
glass doors. It has three plants: aluminum frames and hardware are made in Plant 1, wood
frames are made in Plant 2, and Plant 3 produces the glass and assembles the products. Due to
declining  earnings,  top  management  has  decided  to  revamp  the  company’s product line: while
some unprofitable products are being discontinued, two new products having large sales
potential are being launched.

Product 1: A glass door with aluminum framing

Product 2: Wood-framed window

Product 1 requires some of the production capacity in Plant 1 and 3, while Product 2 needs only
Plants 2 and 3. The marketing division has concluded that the company could sell as much of
either product as could be produced by these plants. However, because both products would
be competing for the same production capacity in Plant 3, it is not clear which mix of the two
products would be most profitable.

The OR team develops this problem. Determine what production rates should be for the
two  products  to  maximize  the  company’s  total  profit,  subject to the restriction of limited
production capacities of the three plants. (Here, the production refers to the number of
batches produced per week) (See reference [2])

The data for this problem is given:

11
Using the datag given, we can express this as a standard maximum linear programming.

Let

𝑥 = number of batches of product 1 produced per week

𝑥 = number of batches of product 2 produced per week
Z = total profit per week (in thousand dollars) from producing these two products

Thus, in this problem, 𝑥  and 𝑥 are the decision variable and the objective function is

Z = 3𝑥 + 5𝑥 .

We can also have equations for constraints. The Plant 1 can only available for 4 hours per week
and product 1 needs one hour of capacity of the Plant 1; the Plant 2 has 12 available hours of
capacity when the product 2 needs two hours per product; and while the Plant 3 has 18 hours
of capacity, product 1 and 2 needs respectively three and two hours each. Thus, the constraints
are:

𝑥 ≤4
2𝑥 ≤ 12
3𝑥 + 2𝑥 ≤ 18
And
𝑥 ≥ 0, 𝑥 ≥ 0

We can formulate this model on a spreadsheet using excel (See reference [3])

Where
E5 = C5*E9 + D5*E9
E6 = C6*E9 + D6*E9 =>
E7 = C7*E9 + D7*E9

12
And E8 = C8*C9 + D8*D9 (the objective function). Thus, this spreadsheet is constructed to
maximize the value of E8.

EXCEL SOLVER (See reference [3])

Excel has a powerful tool called Solver that uses the simplex method to find an optimal
solution. Before going into solver, we have to make sure we have all the required components:

Decision variable (C9 and D9), the objective function and its value (E8), and each functional
constraints (row 5, 6, and 7).

After you press Solver in the Data tab, you will get

And for the constraints: you click add

13
And keep adding all the constraints

Then, at last, you select the linear model with non-negative variable.

14
Then the Solver will provide the solution to the objective function

Thus, the solution suggest that to maximize the profit, the Wyndor Glass CO. needs to produce
2 batches of Product 1 and 6 batches of Product 2 in a week, which will lead to the maximum
profit of $36,000 (See reference [3]).

Application Vignette 3

Welch’s,  Inc.,  is  the  world’s  largest  processor  of  Concord  and  Niagara  grapes,  with  annual  sales  
above $550 million per year. Every September, growers deliver grapes to processing plants that
then press the raw grapes into juice. Time must pass before the grape juice is ready for
conversion into finished jams, jellies, juices and concentrates. Deciding how to allocate the
grape crop is a complex task given changing demand and uncertain crop quality and quantity.
Typical decision include what recipes to use for major product groups, the transfer of grape
juice between plants, and the mode of transportation for these transfers.

In the beginning, the OR team developed a linear programming model with 8,000 decision
variables that focused on the component level of detail. Then, they made the model more
useful by aggregating demand by product group rather than by components. Reduced to 324
decision variables and 361 functional constraints, the model then was incorporated into a
spreadsheet.

The company had revised the spreadsheet every month since 1994 to provide senior
management with information on the optimal logistics plan generated by the Solver. This saved
approximately $150,000 in the first year alone. (See reference [2])

15
 3. Dynamic Programming
Application Vignette 4

Six days after Saddam Hussein ordered his Iraqi military forces to invade Kuwait on August 2,
1990, the United States started deploying many of its own military units and cargo to the region
for intervention. The military operation called Operation Desert Storm was launched in January
17, 1991 as 35 nations (led by the U.S.) formed a coalition in attempt to expel the Iraqi troops
from Kuwait. This led to a decisive victory for the coalition forces, which liberated Kuwait and
penetrated Iraq.

The logical challenge was quickly transporting the needed troops and cargo to the war zone.
A typical airlift mission carrying troops and cargo from the U.S. to the Persian Gulf required a
three-day round-trip, visited seven or more different airfields, burned almost one million
pounds of fuel, and cost $280,000. Operation Desert Storm was the largest airlift in history, in
which the Military Airlift Command (MAC) commanded more than100 such missions daily.

Operation research – specifically, Dynamic Programming - was applied to meet this

challenge in developing the decision support system to schedule and route each airlift mission.
The stages in the dynamic programming formulation correspond to the airfields in the network
of flight legs relevant to the mission. For a given airfield, the states are characterized by the
departure time from the airfield and the remaining available duty for the current crew. The
objective function to be minimized is a weighted sum of several measures of performance: the
lateness of deliveries, the flushing time of the mission, the ground time, and the number of
crew changes. The constraints include a lower bound on the load carried by the mission and
upper bounds on the availability of crew and ground-support resources at airfields. This
application of dynamic programming, thus, is recognized for having a dramatic impact on the
ability to deliver the necessary cargo and personnel to the Persian gulf quickly to support the
Operation Desert Storm. (See reference [2])

Characteristics of Dynamic Programming

Dynamic programming is useful method for making a sequence of interrelated decision;

systematic procedure for determining the optimal combination of decisions. Dynamic
programming does not have a standard mathematical model like the linear programming; it is a
general type of problem solving where model has to be developed to fit each situation.

While dynamic programming process is said to be created to fit its own situation, there
are some basic features and characteristics:

16
1) The problem can be divided into stages, with the policy decision required at each stage. It
requires making a sequence of interrelated decisions, where each decision corresponds to one
stage of the problem.

2) Each stage has number of states associated with the beginning of that stage – the states are
the various possible conditions in which the system might be at that stage of the problem

3) The effect of the policy decision at each stage is to transform the current state to a state
associated with the beginning of the next stage.

4) The solution procedure is designed to find an optimal policy for the overall problem.

5) An optimal policy for the remaining stages is independent of the policy decisions adopted in
previous stages. Thus, the optimal decision depends on the current state and not on how you
got there – it is called the principle of optimality for dynamic programming.

6) The solution procedure begins by finding the optimal policy for the last stage.

7) A recursive relationship that identifies the optimal policy for stage n, given the optimal policy
for stage n+1, is available.

N = number of stages

n = label for current  stage  (n=  1,2,…,  N)

𝑠 = current state for stage n

𝑥 = decision variable for stage n

𝑥 ∗ = optimal value of 𝑥  (given 𝑠 )

𝑓 (𝑠 , 𝑥 ) =  contribution  of  stages  n,  n+1,  …N  to  objective  function  if  system  starts  in  
state 𝑠 at stage n, immediate decision is 𝑥 , and optimal decisions are made thereafter.

𝑓 ∗ (𝑠 ) =   𝑓 (𝑠 , 𝑥 ∗ )

The recursive relationship will always be of the form

𝑓 ∗ (𝑠 ) = max  { 𝑓 (𝑠 , 𝑥 ∗ )} or 𝑓 ∗ (𝑠 ) = min  { 𝑓 (𝑠 , 𝑥 ∗ )}

8) When using the recursive relationship, the solution procedure starts at the end and moves
backward stage by stage until it finds the optimal policy starting at the initial stage – which
yields an optimal solution for the entire problem (See reference [2]).

17
Example 5

A government space project is conducting research on a certain engineering problem that must
be solved before people can fly safely to Mars. Three research teams are currently trying three
different approaches for solving this problem. The estimate has been made that, under present
circumstances, the probability that the respective teams – call them 1,2 and 3 – will not
succeed is 0.40, 0.60, and 0.80, respectively. Thus, the current probability that all three teams
will fail is (0.4)(0.6)(0.8) = 0.192. Because the objective is to minimize the probability of failure,
two more top scientist have been assigned to the project.

The table is added to give more data about this problem – estimated probability that the
respective teams will fail when 0,1, or 2 additional scientists are added to that team.

In this case, stage n (n = 1,2,3) is research team n, and the state 𝑠 is the number of new
scientist still available for allocation to the remaining teams. The decision variables, 𝑥 are the
number of additional scientists sent to team n. We let 𝑝 (𝑥 ) be the probability of failure of
team i if it is assigned 𝑥 additional scientists. Thus, the objective function for this problem is to

Minimize ∏ 𝑝 (𝑥 ) =  𝑝 (𝑥 )  𝑝 (𝑥 )  𝑝 (𝑥 )

Subject to

∑ 𝑥 =2

And 𝑥 are nonnegative integers

Consequently, 𝑓 (𝑠 , 𝑥 ) for this problem is

𝑓 (𝑠 , 𝑥 ) =   𝑝 (𝑥 ) ∙ min 𝑝 (𝑥 ),

Where the minimum is taken over 𝑥 ,…   𝑥 such that

18
 ∑ 𝑥 =𝑠

And 𝑥 are nonnegative integers,

For n = 1,2,3. Thus,

𝑓 ∗  (𝑠 ) = 𝑚𝑖𝑛 , ,,…, 𝑓 (𝑠 , 𝑥 ),

Where

𝑓 (𝑠 , 𝑥 ) = 𝑝 (𝑥 ) ∙ 𝑓 ∗  (𝑠 −   𝑥 )

(With 𝑓 ∗ defined to be 1). The diagram summarizes these relationship

Thus, the recursive relationship relating the 𝑓 ∗ , 𝑓 ∗ , 𝑎𝑛𝑑  𝑓 ∗ functions in this case is

𝑓 ∗  (𝑠 ) = 𝑚𝑖𝑛 , ,,…, {𝑝 (𝑥 ) ∙ 𝑓 ∗  (𝑠 −   𝑥 )} for n = 1,2,

And when n=3,

𝑓 ∗ =   𝑚𝑖𝑛 , ,,…, 𝑝 (𝑥 ).

The resulting dynamic programming calculations are as follows:

n= 3,

𝑠 𝑓 ∗  (𝑠 ) 𝑥∗
0 .8 0
1 .5 1
2 .3 2

19
n=2,

𝑓 ∗  (𝑠 , 𝑥 ) = 𝑝 (𝑥 ) ∙ 𝑓 ∗  (𝑠 −   𝑥 ) 𝑓 ∗  (𝑠 ) 𝑥∗

𝑥
𝑠 0 1 2

0 .48 - - .48 0

1 .3 .32 - .30 0

2 .18 .2 .16 .16 2

𝑓 ∗  (𝑠 , 𝑥 ) = 𝑝 (𝑥 ) ∙ 𝑓 ∗  (𝑠 −   𝑥 ) 𝑓 ∗  (𝑠 ) 𝑥∗

𝑥
𝑠 0 1 2

2 .064 .06 .072 .060 1

Thus, the optimal solution must have 𝑥 ∗ =1, which makes 𝑠 = 2-1 =1, so that 𝑥 ∗ = 0, which
makes 𝑠 = 1 – 0 = 1, so that 𝑥 ∗ = 1. Therefore, teams 1 and 3 should each receive one
additional scientist. The new probability that all three teams will fail would then be 0.060 (See
reference [2] and [3]).

20
Reference

[1] Ferguson,  Thomas.  “LINEAR  PROGRAMMING:  A  Concise  Introduction.”  UCLA MATH. N.p.,

n.d. Web. 5 May 2014. <http://www.math.ucla.edu/~tom/LP.pdf>.

[2] Lieberman, Gerald J., and Frederick S. Hillier. Introduction to Operations Research. New York

City: McGraw-Hill Science/Engineering/Math, 2009. Print.

[3] Wang,  Jin.  “Operation  Research  II.”  College of Management, NCTU. N.p., n.d. Web. 5 Mar.

2009. <http://ocw.nctu.edu.tw/upload/classbfs1210015708111683.pdf>.

21