Topic: LP Formulation and Graphical Solutions

For the next few topics, we will be discussing applications of linear programming. Linear
programming was developed by George Dantzig in the 1940’s and remains the most
widely used management science/operations research technique in practice. The term
programming sometimes confuses students, who assume that it means computer
programming. Linear programming actually predates the advent of digital computers. At
that time, programming was a military term that meant planning. Linear programming,
therefore, refers to planning problems that can be described with linear functions, while
nonlinear programming refers to planning problems that involve nonlinear functions.

Characteristics of Linear Programming Problems

A linear program optimizes (either maximizes or minimizes) a single linear objective
function subject to linear constraints (inequalities or equations).

 Decision Variables: Variables whose values are under our control.

 Objective Function: An expression to maximize or minimize.
 Constraints: Restrictions on the values of the decision variables.
 Parameters: Given values, used in the objective function and constraints, to
describe how the system behaves.

Example 1: A very small example of an LP problem is shown below.

Maximize 3𝑥 + 4𝑦 Decision Variables:

subject to 5𝑥 + 8𝑦  40 Parameter Values:

𝑥, 𝑦  0

A feasible solution satisfies all of the constraints. Give examples of feasible solutions.

(𝑥, 𝑦) = (𝑥, 𝑦) =

An infeasible solution violates some constraint(s). Give examples of infeasible solutions.

(𝑥, 𝑦) = (𝑥, 𝑦) =

An optimal solution is a feasible solution that has the best objective value. Can you reason
out (or guess) what the optimal solution is for this problem?

𝑥= 𝑦= Objective Value =

1
Setting Up the LP Model for Beaver Creek Pottery
In every LP modeling situation, we use the following three steps.

1. Define the decision variables.

2. Express the objective as a linear function of the decision variables.
3. Express each constraint as a linear function of the decision variables.

One of the most basic types of LP problems is known as the product mix problem. In these
types of problems, a business wants to determine how many of each product it should
create to maximize its profit (or sometimes, to minimize its cost). Different products sell
for different prices and require different amounts of resources. The company wants to
select the mix of products that best utilize the available resources to create profit. We will
discuss a product mix problem in the following example.

Example 2: The Beaver Creek Pottery Company is a small crafts company that produces
bowls and mugs. Each day, they have 120 lb of clay and 40 hours of labor available. The
profit and resource requirements for each product are summarized below.

Resources Requirements
Labor Clay Profit
Product (hr/unit) (lb/unit) ($/unit)
Bowl 1 4 40
Mug 2 3 50

Formulate a model to determine the number of bowls and mugs that maximize profit.

Step 1: Define Decision Variables.

𝑥1 = # bowls to make each day

𝑥2 = # mugs to make each day

Step 2: Express the Objective (what should be maximized or minimized). Note: We

typically use the letter Z to represent the expression for the objective function.

2
 Resources Requirements
Labor Clay Profit
Product (hr/unit) (lb/unit) ($/unit)
Bowl 1 4 40
Mug 2 3 50

Step 3: Express the Constraints.

Labor Hours:
(40 hr available)

Clay Used:
(120 lb available)

Sign Restrictions:

General Form of a Linear Program

Linear objective
Max (or Min) 𝑍= 𝑐1 𝑥1 + 𝑐2 𝑥2 + ⋯ + 𝑐𝑛 𝑥𝑛 function

subject to 𝑎11 𝑥1 + 𝑎12 𝑥2 + ⋯ + 𝑎1𝑛 𝑥𝑛 ≤ 𝑏1

Linear constraints with all 𝑎21 𝑥1 + 𝑎22 𝑥2 + ⋯ + 𝑎2𝑛 𝑥𝑛 ≥ 𝑏2

variables on LHS, only a
number on RHS. 𝑎31 𝑥1 + 𝑎32 𝑥2 + ⋯ + 𝑎3𝑛 𝑥𝑛 = 𝑏3

Must be ≤, ≥, or = ⋮
(Never <, >, or ≠)
𝑎𝑚1 𝑥1 + 𝑎𝑚2 𝑥2 + ⋯ + 𝑎𝑚𝑛 𝑥𝑛 ≤ 𝑏𝑚

𝑥1 , 𝑥2 , … , 𝑥𝑛 ≥ 0 All variables are

non-negative

3
Assumptions of Linear Programming
Several assumptions are implicit in a linear programming formulation, as detailed below.
After formulating an LP, the modeler should verify that each of these assumptions are
appropriate for the particular problem.

1. Proportionality: Given a variable 𝑥𝑗 , its contribution to the objective is 𝑐𝑗 𝑥𝑗 , and

its contribution to the ith structural constraint is 𝑎𝑖𝑗 𝑥𝑗 . That means, if 𝑥𝑗 doubles,
then so does its contribution to the objective, as well as its contribution to each
constraint.
Implication: LPs can’t model bulk discounts, set-up costs, etc.

2. Additivity: This assumption guarantees that the total cost is the sum of the
individual costs, and that the total contribution to the ith restriction is the sum of
the individual contributions of the individual activities.
Implication: LPs can’t model substitution or interaction effects

3. Divisibility: This assumption ensures that the decision variables can be divided
into fractional levels so that non-integer values for the decision variables are
always permitted.
Implication: In linear programming models, we ignore integrality
restrictions for the number of products produced and allow
fractional answers. (For example, our solution might be to make
1.47 bowls, and we will just report that.) In real-life, variables
often require integer answers. For instance, sometimes we
need to decide whether or not to open a factory, so we will need
binary (0-1, yes/no) decisions. Whenever we require any of the
variables to take on only integer values, we have left the realm
of linear programming and headed into integer programming,
which we will discuss later in the semester.

4. Certainty: The coefficients in the constraints and objective function are all
known with certainty. Any probabilistic element inherent in costs, demands, and
prices are assumed to be approximated by some deterministic value.

Implication: In textbook problems, parameters will always be given with

certainty, so this is not a problem. In real-life, some parameters
might not be constant. Sometimes, people use average values as
inputs, but this can lead to poor decisions in practice. A more
complete approach is to use stochastic programming, which
allows parameters to follow probability distributions. However,
that topic is beyond the scope of this course.

4
Graphical Solution for Beaver Creek (Max Problem)
To solve the LP, we could keep testing individual solutions to see if they are feasible and
select the feasible point with the best objective value. However, there are more efficient
ways to solve LPs. In the lab, we will discuss how to use Excel to solve LPs. For the
remainder of this topic, we will discuss a graphical approach for solving LPs.

As we discussed in Topic #1, we will only use graphical approaches in problems with two
variables. Why?

 Graphs involving three variables are 3D. It is possible to draw these, but it requires
mad drawing skills or special software (more likely).
 It is impossible to graph functions with 4 or more variables.

We will begin by graphing the points that satisfy all of the constraints (known as the
feasible region), and then we will talk about how to find the optimal solution. (NOTE:
Optimal means best. Out of all of the feasible (possible) solutions, we want the best one.)

Since we always have a restriction that our variables are non-

negative, our solutions are always located in the first quadrant
(positive quadrant or upper right-hand region).

The feasible region for our problem is the set of all points that satisfy the
following constraints. We begin by graphing the lines and then select the correct side of
the inequality.

𝑥1 + 2𝑥2 ≤ 40:

Two points on line : (𝑥1 , 𝑥2 ) = (𝑥1 , 𝑥2 ) =

4𝑥1 + 3𝑥2 ≤ 120:

Two points on line : (𝑥1 , 𝑥2 ) = (𝑥1 , 𝑥2 ) =

Find the point where the two lines cross:

𝑥1 + 2𝑥2 = 40
⟹
4𝑥1 + 3𝑥2 = 120

5
Figure 2.2: Coordinates for Graphical Analysis
(Taylor, 2019, p. 39)

Figure 2.6: Graph of Both Model Constraints

(Taylor, 2019, p. 41)
6
All of the points in the blue region are feasible, meaning that they satisfy all of the
constraints for Beaver Creek Pottery. While they are allowed to use any of these possible
solutions, they want to use the solution that has the best objective value (highest profit).

It can be shown that, if an optimal solution exists, then one (or more) of the feasible
region’s corners (or extreme points) must be optimal. Thus, we can calculate the
objective value at each extreme point and verify that the optimal solution is the corner with
the highest profit.

Profit at Corner Points:

x1 = 0 bowls
x2 = 0 mugs
Z = $0

Figure 2.12: Solutions at All Corner Points

(Taylor, 2019, p. 46)

Optimal Solution: #bowls = 𝑥1 = #mugs = 𝑥2 =

Optimal Profit = 40𝑥1 + 50𝑥2 =

Example 3: To ensure that we understand the optimization process, let’s solve a different
version of the same problem. Suppose that all of the constraints stay the same, but that the
unit profit for bowls increases to $70 and for mugs decreases to $20.

New Objective Function: Maximize Profit = 𝑍 =

The optimal solution for this problem.

Optimal Solution: #bowls = 𝑥1 = #mugs = 𝑥2 =

Optimal Profit = 70𝑥1 + 20𝑥2 =

7
Standard Form of LPs

There are different formats in which LPs can be presented. In all cases, we need to ensure
that the objective function and constraints are all linear. We should also ensure that all
variables are on the left-hand-side, and only a constant is on the right-hand-side.

When we put LPs in standard form, we need to have all of the constraints as equations
(“=” form, not inequalities). That means we need a way to convert inequalities to
equations. We can accomplish this by adding new variables to the problem. Their sole
purpose is to measure the amount of slack (or surplus) in a constraint.

To convert an inequality in ≤ form to an equation, we add a slack variable, 𝑠𝑖 ≥ 0. Then 𝑠𝑖

represents the amount by which the left-hand-side expression is below the given right-
hand-side value.

Example 4: Convert 2𝑥1 + 𝑥2 ≤ 6 into an equation, using slack variable 𝑠1 .

We can demonstrate how the value of the slack variable 𝑠1 indicates whether or not the
original inequality was satisfied.

Extra capacity
Value of 2𝑥1 + Satisfies inequality
𝑥1 𝑥2 (𝑠1 = 6 − 2𝑥1 − Interpretation
𝑥2 2𝑥1 + 𝑥2 ≤ 6?
𝑥2 )?
1 2 2(1) + 2 = 4 Yes/No 6−4=2

2 2 2(2) + 2 = 6 Yes/No 6−6=0

3 2 2(3) + 2 = 8 Yes/No 6 − 8 = −2

If we have an inequality in ≥ form, we can convert it to an equation by subtracting a surplus

variable, 𝑠𝑖 ≥ 0. Then 𝑠𝑖 represents the amount by which the left-hand side expression
exceeds the given right-hand side value.

8
Example 5: Convert 3𝑥1 + 2𝑥2 ≥ 6 into an equation, using surplus variable 𝑠2 .

We can demonstrate how the value of the excess variable 𝑠2 indicates whether or not the
original inequality was satisfied.

Satisfies inequality Amount over limit

𝑥1 𝑥2 Value of 3𝑥1 + 2𝑥2 Interpretation
3𝑥1 + 2𝑥2 ≥ 6? (𝑠2 = 3𝑥1 + 2𝑥2 − 6)?
2 2 3(2) + 2(2) = 10 Yes/No 10 − 6 = 4

2 0 3(2) + 2(0) = 6 Yes/No 6−6=0

1 1 3(1) + 2(1) = 5 Yes/No 5 − 6 = −1

Conclusions:
 If the slack/surplus variable is not negative, it indicates that the solution
satisfies the corresponding inequality.
o If the slack/surplus variable is strictly positive, it indicates that the
solution is feasible and that the constraint is non-binding (the solution
is not right up against the limit).
o If the slack/surplus variable equals zero, it indicates that the solution
is feasible and that the constraint is binding (lies on the border line).
 If the slack/surplus variables is negative, it indicates that the solution
violates the corresponding inequality.

Recall our Beaver Creek Pottery example.

Maximize 𝑍= 40𝑥1 + 50𝑥2
subject to 𝑥1 + 2𝑥2 ≤ 40 (Labor)
4𝑥1 + 3𝑥2 ≤ 120 (Clay)
𝑥1 , 𝑥2 ≥ 0

To put this into standard form, we need to add two slack variables.

𝑠1 = unused labor (hours)

𝑠2 = unused clay (lb)

Then the problem can be re-written in standard form as follows.

9
 Maximize 𝑍= 40𝑥1 + 50𝑥2

subject to

2.7 Finding the Values of All Variables at Extreme Points

We can use the graph and the model
formulation to determine the values of all
variables (𝑥1 , 𝑥2 , 𝑠1 , 𝑠2 ) at each corner or 𝑥1 = 0 4𝑥1 + 3𝑥2 = 120
extreme point.

We first use the graph to see which two lines

cross at each corner. Then we solve for the
values of 𝑥1 and 𝑥2 . On the graph to the right,
we have labeled the equation for each edge of 𝑥1 + 2𝑥2 = 40
the feasible region.

𝑥2 = 0

With values for 𝑥1 and 𝑥2 , we can plug in to find the values of 𝑠1 and 𝑠2 . Recall that we had

𝑥1 + 2𝑥2 + 𝑠1 = 40, so that means 𝑠1 = 40 − 𝑥1 − 2𝑥2

4𝑥1 + 3𝑥2 + 𝑠2 = 120, so that means 𝑠2 = 120 − 4𝑥1 − 3𝑥2

At extreme point O
Two lines crossing:

𝑥1 = 𝑥2 =

𝑠1 = 40 − 𝑥1 − 2𝑥2 = 40 − 0 − 2(0) = 40 (40 hours of unused labor)

𝑠2 = 120 − 4𝑥1 − 3𝑥2 = 120 − 4(0) − 3(0) = 120 (120 lb of unused clay)

10
At extreme point A
Two lines crossing:

𝑥1 = 𝑥2 =

𝑠1 = 40 − 𝑥1 − 2𝑥2 = 40 − 0 − 2(20) = 0 (Used all 40 hours of labor)

𝑠2 = 120 − 4𝑥1 − 3𝑥2 = 120 − 4(0) − 3(20) = 60 (60 lb of unused clay)

At extreme point B
Two lines crossing: 4𝑥1 + 3𝑥2 = 120 and 𝑥1 + 2𝑥2 = 40

4𝑥1 + 3𝑥2 = 120

4𝑥1 + 8𝑥2 = 160
5𝑥2 = 40 ⇒ 𝑥2 = 8
⇒ 𝑥1 = 40 − 2𝑥2 = 40 − 2(8) = 24

𝑥1 = 24 𝑥2 = 8

𝑠1 = 40 − 𝑥1 − 2𝑥2 = 40 − 24 − 2(16) = 0 (Used all 40 hours of labor)

𝑠2 = 120 − 4𝑥1 − 3𝑥2 = 120 − 4(24) − 3(8) = 0 (Used all 120 lb of clay)

At extreme point C
Two lines crossing: 4𝑥1 + 3𝑥2 = 120 and 𝑥2 = 0

⇒ 4𝑥1 = 120 ⇒ 𝑥1 = 30

𝑥1 = 30 𝑥2 = 0

𝑠1 = 40 − 𝑥1 − 2𝑥2 = 40 − 30 − 2(0) = 10 (10 hours of unused labor)

𝑠2 = 120 − 4𝑥1 − 3𝑥2 = 120 − 4(30) − 3(0) = 0 (Used all 120 lb of clay)

This information is summarized in the following figure.

11
 x1 = 0
x2 = 0
s1 = 40
s2 = 120

Figure 2.14: Solutions at Points A, B, and C with Slack

(Taylor, 2019, p. 49)

Graphical Solution for Minimization Problems

So far, we have considered a problem to maximize profit. It is also possible to solve
problems where the objective function should be minimized.

The process of graphing the constraints is exactly the same, and we can still draw the arrow
to help up find increasing objective lines.

The only difference is that since we are minimizing, we want to find the feasible point with
the lowest objective function value. This means that we will want to push the lines in the
opposite direction of the arrow.

Example 6: Suppose we had a problem with the same constraints as the Beaver Creek
Pottery problem; however, suppose that the objective was to minimize 40𝑥1 + 50𝑥2
instead of maximizing it.

Minimize 𝑍= 40𝑥1 + 50𝑥2

subject to 𝑥1 + 2𝑥2 ≤ 40
4𝑥1 + 3𝑥2 ≤ 120
𝑥1 , 𝑥2 ≥ 0

Find the solution that minimizes this objective function.

12
 Optimal Solution: 𝑥1 = 𝑥2 =

Optimal Objective Value = 40𝑥1 + 50𝑥2 =

Graphical Solution for Fertilizer Problem (Min Problem)

It is more common to have minimization objective functions when the constraints are in ≥
form. To illustrate that, consider a more typical minimization problem as stated below.

Example 7: A farmer wants to mix fertilizer for his spring crops. He intends to create a
custom mix by combining two brands: Super-Gro and Crop-Quick. The cost and chemical
composition of each brand are summarized in the table below.
Chemical Contribution
Nitrogen Phosphate Cost
Brand (lb/bag) (lb/bag) ($/bag)
Super-Gro 2 4 6

Crop-Quick 4 3 3

The final blend must contain at least 16 pounds of nitrogen and at least 24 pounds of
phosphate. Our goal is to determine how many bags to use of each brand in order to
minimize total cost while satisfying the requirements.

13
Step 1: Define Decision Variables.
𝑥1 = # bags of Super-Gro to use

𝑥2 = # bags of Crop-Quick to use

Step 2: Express the Objective (what should be maximized or minimized).

Step 3: Express the Constraints.

Nitrogen:
(at least 16 lb)

Phosphate:
(at least 24 lb)

Sign Restrictions:

Shade the feasible region below.

Figure 2.16: Constraint Lines for Fertilizer Model

(Taylor, 2019, p. 53)

14
 4𝑥1 + 3𝑥2 = 24

2𝑥1 + 4𝑥2 = 16

Figure 2.17: Feasible Solution Values

(Taylor, 2019, p. 53)

Optimal Solution: Bags of SG = 𝑥1 = Bags of CQ = 𝑥2 =

Optimal Cost = 6𝑥1 + 3𝑥2 =

We can put the fertilizer problem into standard form by using surplus variables.
Minimize 𝑍= 6𝑥1 + 3𝑥2
subject to 2𝑥1 + 4𝑥2 ≥ 16 (Nitrogen)
4𝑥1 + 3𝑥2 ≥ 24 (Phosphate)
𝑥1 , 𝑥2 ≥ 0

To put this into standard form, we need to subtract two surplus variables.

𝑠1 = amount over the nitrogen limit

𝑠2 = amount over the phospate limit

15
Then the problem can be re-written in standard form as follows.

Minimize 𝑍= 6𝑥1 + 3𝑥2

subject to

As before, we could use algebra to

determine the value of each
variable (𝑥1 , 𝑥2 , 𝑠1 , 𝑠2 ) at every
corner. These values are
summarized in the graph to the
left. To check that you
understand the process, try to
find these values using algebra.
Figure 2.19: Graph of the Fertilizer Example
(Taylor, 2019, p. 56)

Multiple (or Alternative) Optimal Solutions

As we stated earlier, when an optimal solution exists, we are guaranteed that at least one of
the extreme points (corners) is optimal. So far, we have only seen examples where exactly
one corner is optimal. We now examine an example where multiple extreme points are
tied for the best objective value. This situation can happen in either maximization
problems or minimization problems, and it can only happen if the slope of the objective
function is equal to the slope of one of the constraints.

Example 8: Suppose that the objective (profit) function for the Beaver Creek Pottery
example changes from maximizing 𝑍 = 40𝑥1 + 50𝑥2 to maximizing 𝑍 = 40𝑥1 + 30𝑥2 . This
means that the problem would be given as follows.

Maximize 𝑍 = 40𝑥1 + 30𝑥2

subject to 𝑥1 + 2𝑥2 ≤ 40 (Labor)
4𝑥1 + 3𝑥2 ≤ 120 (Clay)
𝑥1 , 𝑥2 ≥ 0

Solve this problem graphically. Remember that we are maximizing 𝑍 = 40𝑥1 + 30𝑥2 ..

16
Figure 2.6: Graph of Both Model Constraints
(Taylor, 2019, p. 41)

The graph below shows that both points B and C are tied for the best objective value. The
same is true for any point on the line segment that connects them. That is, the entire line
segment between points B and C is optimal with a profit of $1200.

Figure 2.20: Graph with Multiple Optimal Solutions

(Taylor, 2019, p. 57)

Whenever multiple options exist, a good consultant always tells the client about all options.
In our case, we would tell Beaver Creek Pottery that they can maximize their profit by
selecting any combination of bowls and mugs that lies on the segment between B and C.

Recall that line segment AB corresponds to the line 𝑥1 + 2𝑥2 = 40. Give an objective
function that would result in alternative optimal solutions along AB.

Slope of AB =

17
 Objective Function with the same slope:

Infeasible Problems
Several times, we have mentioned that something will be true whenever an optimal
solution exists. There are two cases in which no optimal solution exists. One of these
situations occurs when there are no points that simultaneously satisfy all of the constraints.

Example 9: Consider the following LP in two variables.

Maximize 𝑍 = 5𝑥1 + 3𝑥2

subject to 4𝑥1 + 2𝑥2 ≤ 8
𝑥1 ≥ 4
𝑥2 ≥ 6
𝑥1 , 𝑥2 ≥ 0

We begin by graphing each constraint, and we then look for the points that satisfy all of
them simultaneously.
To help us, the line 4𝑥1 + 2𝑥2 = 8 is shown below. Draw the regions where 4𝑥1 + 2𝑥2 ≤ 8,
and where 𝑥1 ≥ 4 and 𝑥2 ≥ 6.

4𝑥1 + 2𝑥2 = 8

What do you conclude?

18
In real-life, infeasible problems are a major problem. Sometimes, they are the result of a
typo and the actual problem is feasible. If not, we need to examine what is causing the
infeasibilities and adjust either our model or the underlying conflicts in the system.

Unbounded Problems
The other case in which no optimal solution exists is when the objective value can increase
infinitely (in a maximization problem) or decrease infinitely (in a minimization problem).
There are feasible solutions to these problems, but no solution is optimal (best).

Example 10: Consider the following LP in two variables.

Maximize 𝑍 = 2𝑥1 + 𝑥2
subject to 𝑥1 ≥ 2
𝑥2 ≤ 3
𝑥1 , 𝑥2 ≥ 0

We can graph the feasible region and objective lines in the following figure.
Recall that the objective is to maximize 2𝑥1 + 𝑥2 .

What do you conclude?

 For this problem, the objective increases as you move right, and the feasible region
extends forever to the right.
 As a result, the objective value can increase infinitely, so we say that the objective is
unbounded.

Notes:
 Be very skeptical of unbounded solutions in real-life. It is highly unlikely that you
will find a solution with infinite profit! It is much more likely that you omitted a
constraint that should have been included.
 You cannot conclude that the problem is unbounded just because the feasible
region extends to infinity. It depends on how the objective function behaves.

19
  In the fertilizer problem, you already saw an example of a problem with an
unbounded feasible region that still had an optimal solution.

You can even have an optimal solution for max problems with unbounded regions. For
example, what if we were maximizing −𝑥1 + 𝑥2 for the previous feasible region?

Summary of the Solution Approach

We have now seen everything that can happen when solving a linear programming model.
For any linear programming problem, only the following options exist.

 The problem has optimal solution(s).

o If exactly one corner of the feasible region has the best objective value, we
say that the problem has a unique optimal solution.
o If multiple corners of the feasible region are tied for the best objective value,
then everything between them is also optimal. We say that the problem has a
multiple optimal solutions or alternative optimal solutions.
 No optimal solution exists.
o No points simultaneously satisfy all of the constraints, so there is not even a
feasible point to the overall problem. We say that the problem is infeasible.
o Feasible solutions exist, but the objective value can be pushed all the way to
infinity for max problems (or negative infinity for min problems). As a result,
no solution is best (or optimal). We say that the problem is unbounded.
 There are only three possibilities for the number of optimal solutions.
o 0 optimal solutions (problem is infeasible or unbounded)
o 1 optimal solution (unique extreme point is optimal)
o Infinite number of optimal solutions (entire edge is optimal)
o There are never exactly 2 optimal solutions. If you find two optimal corners,
then everything between them is also optimal, so there are really an infinite
number of optimal solutions.

20
Things to remember
 If no points can satisfy all the constraints, then the problem is infeasible.
 If the feasible region is bounded (doesn’t extend forever in any direction), then the
objective can never be unbounded. Either a single corner is optimal, or multiple
corners could be tied for the best.
 If the feasible is unbounded (extends infinitely in some direction), then you need to
see if the objective improves in that direction. If it does, then the objective is
unbounded (can improve infinitely). If not, then the optimal solution occurs at some
corner(s) of the feasible region.

A flow chart for the graphical solution process is shown on the next page.

21
 Are there any points that
The problem
simultaneously satisfy all the No is infeasible.
constraints?

Yes

Clarification: Clarification:
 An unbounded region stretches to  For a MAX problem,
infinity in some direction. Is the feasible region Calculate the objective value at each the best corner has
No the largest objective
 A bounded region can be contained in unbounded? corner, and find the best corner.
a circle of finite radius. value.
 For a MIN problem,
the best corner has
the smallest objective
value.
No
Yes

Clarification:
 Look at the direction that the objective
gradient points. The gradient is the
arrow from the origin to the point The best
corresponding to the objective Is there more than one corner is the
Does the objective improve in a best corner?
No unique optimal
coefficients. It points toward
increasing objective values. direction of unboundedness? solution.
 In a MAX problem, the objective can go
to if the gradient points in a
direction that extends forever in the
feasible region.
 For a MIN problem, the objective can Yes
Yes
go to - if an arrow opposite of the
gradient points in a direction that
extends forever in the feasible region. Alternative optimal Note: If there is more
The problem is solutions exist at the best than one best corner, you
unbounded. corners and all points should see that the
between them. objective function has the
same slope as one of the
constraints.

22
Practice Exercises
Chapter 2 (pages 64-73), Problems: 1, 2, 3, 4, 5, 6, 7, 8, 16, 17, 18, 22, 23, 25, 26, 40, 46, 52,
58, 59, 60

Reference
Taylor, B. W. (2019). 2: Linear programming: Modeling formulation and graphical solution.
In Introduction to management science (13th ed., pp. 33-75). Pearson.

23