IE 311 – Operations Research I

Part 2: Modeling Linear and Integer Programs

2b. Integer Programs

Spring 2021
Recall: Types of Linear Optimization Models

• An optimization problem with linear relations between

continuous variables is called a Linear Program:

• An optimization problem with linear relations between integer

variables is called a (Pure) Integer (Linear) Program:

• An optimization problem with linear relations between

continuous and integer variables is called a Mixed-Integer
Linear Program:

IE 311 2
Is Integer Programming Really Needed?

IE 311 3
Integer Programming Preliminaries

• Integer Programming is needed when

• the continuity assumption on the decision variables is not valid, or
• following type of decisions are to modeled:
• "yes/no",
• logical statements such as "either-or", "if-then", "and", "exclusive-or".
• We will see examples of IP Formulations, and how to implement
and solve such models in Python-Gurobi.
• The solution methods for the IP Models will be covered in IE 312.

IE 311 4
A First Example: Knapsack IP

• Suppose that there is a knapsack of capacity 14 units.

• There are four items, each with a size and a value given as
below:
Item 1 Item 2 Item 3 Item 4
Value 5 7 4 3
Size 8 11 6 4

• Determine the items to pick so as to maximize the value of the

knapsack while respecting its capacity.

IE 311 5
 Knapsack capacity is 14 units.
Item Item Item Item
Knapsack IP: Components (I) 1 2 3 4
Value 5 7 4 3
Size 8 11 6 4
• Parameters (what are given?)
• Number of items (4)
• Value of each item (5, 7, 4, 3)
• Size of each item (8, 11, 6, 4)
• Capacity of the knapsack (14)

• Decision variables (what can be decided/changed?)

• 𝑥𝑖 =1 if item 𝑖 picked, and 0 otherwise (such variables are called
BINARY), 𝑖 = 1, … , 4.

IE 311 6
 Knapsack capacity is 14 units.
Item Item Item Item
Knapsack IP: Components (II) 1 2 3 4
Value 5 7 4 3
Size 8 11 6 4
• Objective function (what is the aim?)
• Maximize the total value of the knapsack

• Constraints (what should be obeyed/respected?)

• Knapsack capacity constraint

• Binary requirements

IE 311 7
 Knapsack IP: Simplest (but Dirty)
Python-Gurobi Implementation s.t.

from gurobipy import GRB,Model

m = Model('Knapsack')

x1 = m.addVar(vtype=GRB.BINARY) # define binary variable x1

x2 = m.addVar(vtype=GRB.BINARY) # define binary variable x2

x3 = m.addVar(vtype=GRB.BINARY) # define binary variable x3

x4 = m.addVar(vtype=GRB.BINARY) # define binary variable x4

m.setObjective(5*x1+7*x2+4*x3+3*x4, GRB.MAXIMIZE) # maximize the value

m.addConstr(8*x1+11*x2+6*x3+4*x4 <= 14) # Constraint on capacity

m.optimize() # Optimize the model

IE 311 8
Capital Budgeting (Winston, Page 478)

• Stockco is considering four investments.

• Investment 1 will yield a net present value (NPV) of $16,000; investment 2, an NPV of $22,000;
investment 3, an NPV of $12,000; and investment 4, an NPV of $8,000.
• Each investment requires a certain cash outflow at the present time: investment 1, $5,000;
investment 2, $7,000; investment 3, $4,000; and investment 4, $3,000.
• Currently, $14,000 is available for investment.
• Formulate an IP whose solution will tell Stockco how to maximize the NPV obtained from
investments 1–4.
• Modify the Stockco formulation to account for each of the following requirements separately:
• Stockco can invest in at most two investments.
• If Stockco invests in investment 2, it must also invest in investment 1.
• If Stockco invests in investment 2, it cannot invest in investment 4.

IE 311 9
Capital Budgeting (Winston, Page 478)

• Let 𝑥𝑖 =1 if investment 𝑖 if picked, and 0 otherwise, 𝑖 =

1, … , 4.
• Then, we have the following base model:

IE 311 10
Capital Budgeting (Winston, Page 478)

• Additional constraints:
• Stockco can invest in at most two investments.

• If Stockco invests in investment 2, it must also invest in

investment 1.

• If Stockco invests in investment 2, it cannot invest in

investment 4.

IE 311 11
 Distribution Problem

• A company is operating 𝑚 supply stations to satisfy the demand of 𝑛

demand centers.
• Each supply station has a capacity of at most 𝑠𝑖 tons of products and
each demand center requires at least 𝑑𝑗 tons of products (you can model
the amount of product as a continuous variable).
• The products are shipped by using trucks. Each truck can carry at most 𝑏
tons of products (you need to model the number of trucks as an integer
variable).
• The cost of sending a truck from supply station 𝑖 to demand center 𝑗 is
𝑐𝑖𝑗, regardless of the amount of product carried by the truck.
• Write a mixed integer linear program to minimize the total cost.
IE 311 12
Distribution Problem: Key Observations

• Decision variables (what can be decided/changed?)

• We need to know how many tons of products are sent from each supply
station 𝑖 to each demand center 𝑗.
tons of
𝑖 products 𝑗
• We also need to know how many trucks are sent from each supply
station 𝑖 to each demand center 𝑗.
#trucks
𝑖 𝑗
• Note the following relationship:
𝐭𝐨𝐧𝐬 𝐨𝐟 𝐩𝐫𝐨𝐝𝐮𝐜𝐭𝐬
#𝐭𝐫𝐮𝐜𝐤𝐬 =
truck capacity
IE 311 13
 Distribution Problem

IE 311 14
 Machine Allocation Problem

• A machine tool plant owns 𝑚 different machines on which it can process

𝑛 types of jobs.
• A job 𝑗 takes 𝑡𝑖𝑗 time units if processed on machine 𝑖.
• A job cannot be divided between machines, that is, each job must be
processed by exactly one machine.
• Each machine can process more than one job.
• If a machine is used at all, then a setup time of 𝑠𝑖 units is needed.
• Formulate an IP to minimize the sum of total processing and setup
times.

IE 311 15
 Machine Allocation Problem: Key Observations

• Decision variables (what can be decided/changed?)

• We need to know job 𝒋 is processed on which machine 𝒊.

• We also need to know which machine 𝒊 is used at all.

• Note the following relationship:

• If a machine 𝒊 is NOT used at all, then machine 𝒊 CANNOT process
job 𝒋.

IE 311 16
Machine Allocation Problem

IE 311 17
Modeling Conditional Statements

• Let 𝑥, 𝑦, 𝑧 be binary variables. Let us see how to model some

logical relationships:
• 𝑧 = 𝑥 and 𝑦 can be modelled as
𝑥 + 𝑦 − 1 ≤ 𝑧, 𝑧 ≤ 𝑥, 𝑧 ≤ 𝑦.
• 𝑧 = 𝑥 or 𝑦 can be modelled as
𝑥 + 𝑦 ≥ 𝑧, 𝑧 ≥ 𝑥, 𝑧 ≥ 𝑦.
• 𝑧 = 𝑥 xor 𝑦 can be modelled as
𝑥 + 𝑦 = 𝑧 + 2𝑤, 𝑤 ∈ 0,1 .

IE 311 18
Modeling Conditional Statements: Continued

• Let 𝑥, 𝑦, 𝑧 be binary variables. Let us see how to model some

logical relationships:
• y = not 𝑥 can be modelled as
𝑦 = 1 − 𝑥.
•𝑥 → 𝑦 can be modelled as
𝑥 ≤ 𝑦.
• 𝑥 → not 𝑦 can be modelled as
𝑥 ≤ 1 − 𝑦.

IE 311 19
Modeling Conditional Statements: Continued

• Let 𝑓1 𝑥1 , … , 𝑥𝑛 , 𝑓2 𝑥1 , … , 𝑥𝑛 be functions of decision variables

𝑥1 , … , 𝑥𝑛 .
• We will assume that there exist two constants 𝑀1 , 𝑀2 which upper
bound 𝑓1 , 𝑓2 for every value of 𝑥1 , … , 𝑥𝑛 .
• Let 𝑦 be a binary variable. Let us see how to model some logical
relationships:
• 𝑓1 𝑥1 , … , 𝑥𝑛 ≤ 0 or 𝑓2 𝑥1 , … , 𝑥𝑛 ≤ 0 can be modelled as
𝑓1 𝑥1 , … , 𝑥𝑛 ≤ 𝑀1 y, 𝑓2 𝑥1 , … , 𝑥𝑛 ≤ 𝑀2 1 − 𝑦 .
• 𝑓1 𝑥1 , … , 𝑥𝑛 > 0 → 𝑓2 𝑥1 , … , 𝑥𝑛 ≤ 0 can be modelled as
𝑓1 𝑥1 , … , 𝑥𝑛 ≤ 𝑀1 y, 𝑓2 𝑥1 , … , 𝑥𝑛 ≤ 𝑀2 1 − 𝑦 .

IE 311 20
Single-Machine Job Scheduling

• Jobs 1, … , 𝑛 must be processed on a single machine.

• Each job is available for processing after a certain time, called
release time.
• For each job 𝑖, we are given
• its release time 𝑟𝑖 ,
• its processing time 𝑝𝑖 and
• its weight 𝑤𝑖 .
• Formulate as a mixed-integer linear program the problem of
sequencing the jobs without overlap or interruption so that the
sum of the weighted completion times is minimized.

IE 311 21
 Single-Machine Job Scheduling: Key Observations
• Decision variables (what can be decided/changed?)
• We need to know the completion time of each job 𝒊.
• We also need to know whether job 𝒊 is processed before job 𝒋 or not.
• If job 𝒊 is processed before job 𝒋, then we must have
job 𝑖 job 𝑗

𝐜𝐨𝐦𝐩𝐥. 𝐭𝐢𝐦𝐞 proc. time 𝐜𝐨𝐦𝐩𝐥. 𝐭𝐢𝐦𝐞

𝐨𝐟 𝐣𝐨𝐛 𝒊 + of job 𝑗
≤ 𝐨𝐟 𝐣𝐨𝐛 𝒋
• If job 𝒋 is processed before job 𝒊, then we must have
job 𝑗 job 𝑖
𝐜𝐨𝐦𝐩𝐥. 𝐭𝐢𝐦𝐞 proc. time 𝐜𝐨𝐦𝐩𝐥. 𝐭𝐢𝐦𝐞
IE 311 𝐨𝐟 𝐣𝐨𝐛 𝒋
+ of job 𝑖 ≤ 𝐨𝐟 𝐣𝐨𝐛 𝒊
22
Single-Machine Job Scheduling

IE 311 23
 Set Covering (Winston, Page 486)

• There are six cities in Kilroy County.

• The county wants to determine the minimum number of fire stations needed to ensure
that at least one fire station is within 15 minutes (driving time) of each city.
• The times (in min) required to drive between the cities in Kilroy County are shown below:
City 1 City 2 City 3 City 4 City 5 City 6
City 1 0 10 20 30 30 20
City 2 10 0 25 35 20 10
City 3 20 25 0 15 30 20
City 4 30 35 15 0 15 25
City 5 30 20 30 15 0 14
City 6 20 10 20 25 14 0
• Formulate an IP that will tell Kilroy how many fire stations should be built and where they
should be located.
IE 311 24
Set Covering (Winston, Page 486)

• Example: By building a fire station in City 1, we cover cities 1 and 2.

• Parameters:

• Decision Variables:

• Model:

IE 311 25
 Fixed Charge (Winston, Page 480)

• Gandhi Cloth Co. is capable of manufacturing three types of clothing: shirts, shorts, and pants.
• The manufacture of each type of clothing requires Gandhi to have the appropriate machinery
available.
• The machinery needed to manufacture each type of clothing must be rented at the following
rates: shirt machinery, $200/week; shorts machinery, $150/week; pants machinery, $100/week.
• The amounts of cloth and labor required in the manufacturing together with the variable unit
cost and selling price for each type of clothing are as follows:
Clothing Labor(hrs) Cloth (yards) Price($) Cost($)
Shirt 3 4 12 6
Shorts 2 3 8 4
Pants 6 4 15 8
• Each week, 150 hours of labor and 160 sq yd of cloth are available.
• Formulate an IP to maximize Gandhi’s profit (assume that production amounts are integer).

IE 311 26
Fixed Charge (Winston, Page 480)

• Decision variables:

• Objective:
• Labor Availability:
• Cloth Availability:
• Big-M Constraint:
• Variable Domains:

• NOTE: We need to choose Mj in such a way that 𝑥𝑗 ≤ 𝑀𝑗 is redundant when 𝑦𝑗 = 1.

For instance, 𝑀1 = 𝑀2 = 𝑀3 = 1000 work (best values: 𝑀1 = 40, 𝑀2 = 53, 𝑀3 = 25).
 Fixed Charge-General Model
• Let us define the following parameters:

• Model:

IE 311 28
IP with Piecewise Linear Functions (Winston, Page 492)

• Euing Gas produces two types of gasoline (gas 1 and gas 2) from two types
of oil (oil 1 and oil 2).
• Each gallon of gas 1 must contain at least 50 percent oil 1, and each gallon
of gas 2 must contain at least 60 percent oil 1.
• Each gallon of gas 1 can be sold for 12¢, and each gallon of gas 2 can be
sold for 14¢.
• Currently, 500 gallons of oil 1 and 1,000 gallons of oil 2 are available.
• As many as 1,500 more gallons of oil 1 can be purchased at the following
prices: first 500 gallons, 25¢ per gallon; next 500 gallons, 20¢ per gallon;
next 500 gallons, 15¢ per gallon.
• Formulate an IP that will maximize Euing’s profits.

IE 311 29
IP with Piecewise Linear Functions

• Decision variables:

• Constraints:

• Objective function:
• We need to understand the details of the purchasing cost.

IE 311 30
IP with Piecewise Linear Functions: Purchasing Cost

• As many as 1,500 more gallons of oil 1 can be purchased at the following prices:
• First 500 gallons, 25¢ per gallon;
• Next 500 gallons, 20¢ per gallon;
• Next 500 gallons, 15¢ per gallon.

• The purchasing cost of oil 1 as a function of 𝑧 is given by the following piecewise

linear function:

IE 311 31
IP with Piecewise Linear Functions: Purchasing Cost

• Observe that the purchasing cost is a nonconvex function.

• Minimizing a nonconvex piecewise linear function is
“difficult”.
IE 311 32
 IP with Piecewise Linear Functions: Purchasing Cost

• Minimizing a nonconvex piecewise linear function is “difficult”:

• However, if we knew in which piece 𝑧 lies in, then the resulting problem
would be “easy”.
• We create three copies of 𝑧 variable as 𝑧1, 𝑧2, 𝑧3 corresponding to each
piece and control them via three binary variables 𝑤1, 𝑤2, 𝑤3.

IE 311 33
IP with Piecewise Linear Functions: Complete Model

• Linearized Objective Function:

• Piecewise Linearization Constraints:

• Oil 1 requirement in Gas 1 and Gas 2:

• Availability of Oil 1 and Oil 2:

• Nonnegativity:

IE 311 34
Dinner Seating

• You are asked to determine the seating arrangement of 8 diners,

4 girls (A, B, C, D) and 4 boys (E, F, G, H) on a circular table.
• No boy must be seated next to a boy.
• A and E should not be seated next to each other.
• C and H should be seated next to each other.
• If A and F are seated next to each other, then D and F must be seated
next to each other.
• Formulate an integer linear program to determine if a feasible seating
arrangement exists.

IE 311 35
Dinner Seating

• Let 𝑖 = 𝐴, … , 𝐻 denote the girls and boys.

• Let 𝑗 = 1, … , 8 denote the seats.
• Decision Variables:

• This is a FEASIBILITY problem, therefore, there is no objective.

• Assignment Constraints:

IE 311 36
 Dinner Seating
• No boy must be seated next to a boy.

• A and E should not be seated next to each other.

• C and H should be seated next to each other.

• If A and F are seated next to each other, then D and F must be seated
next to each other.

IE 311 37
Dorian Auto: “Either-Or” (Winston, Page 488)

• Dorian Auto is considering manufacturing three types of autos: compact, midsize,

and large.
• The resources required for, and the profits yielded by, each type of car is shown
below: Resource Compact Midsize Large
Steel required 1.5 tons 3 tons 5 tons
Labor required 30 hours 25 hours 40 hours
Profit yielded $2,000 $3,000 $4,000

• Currently, 6,000 tons of steel and 60,000 hours of labor are available.
• For production of a type of car to be economically feasible, at least 1,000 cars of
that type must be produced.
• Formulate an IP to maximize Dorian’s profit.

IE 311 38
Dorian Auto: Either-Or (Winston, Page 488)

• Decision variables:

• Objective:
• Labor Availability:
• Steel Availability:
• Big-M Constraint:

• Variable Domains:
• NOTE: We need to choose Mj in such a way that 𝑥𝑗 ≤ 𝑀𝑗 is redundant when 𝑦𝑗 = 1,
e.g., 𝑀1 = 𝑀2 = 𝑀3 = 100000 work (best values: 𝑀1 = 𝑀2 = 2000, 𝑀3 = 1200).
IE 311 39
Box Problem: “And”

• A company sells seven types of boxes, ranging in volume from 17 to 33 cubic feet. The
demand and size of each box type are given as follows:
Box 1 Box 2 Box 3 Box 4 Box 5 Box 6 Box 7
Size 33 30 26 24 19 18 17
Demand 400 300 500 700 200 400 200
• The cost (in dollars) of producing each box is equal to the box's volume. Moreover, a fixed
cost of $1000 is incurred to produce any of a box type.
• All demand should be met. If the company desires, a demand for a box may be satisfied by
a box of larger size.
• If both box type 5 and 7 are manufactured, then the total fixed cost is $1500 instead of
$2000.
• The company must manufacture at least 5 types of boxes.
• Formulate a mixed-integer linear program to minimize the cost of meeting the demand of
the boxes.

IE 311 40
Box Problem: “And”

• Parameters:

• Decision Variables:

• Objective Function:

IE 311 41
 Box Problem: “And”

• Relation between 𝑥 and 𝑧 variables:

• Demand constraint:

• Fixed charge constraint (what is a reasonable value for 𝑀?):

• Relation between y and w variables:

• At least 5 box types to be produced:

• Variable Domains:

IE 311 42