Overview of Lecture 10

▪ Integer Programming
▪ Integer Programming Formulation
• Why not using the LP solution?
▪ Popular Integer Programming Examples

Lecture 10 1
Introduction
▪ Linear Programming Assumption
• Fractional and real number (non-integer) solutions
are allowed.
▪ Rounded-off solution of LP solution?
• If the unit cost is small, it may be okay.
• However, if the unit cost is large, it may not be
proper.
• 𝑥1 = 8,000.4 nailes 8,000 nails (may be fine).
• 𝑥1 = 7.4 jet airliners The difference may be
several million dollars.
▪ An optimal Integer Solution is Required.

Lecture 10 2
Integer Proramming Model Types
▪ Total Integer Model
▪ 0-1 Integer Model
▪ Mixed Integer Model

Lecture 10 3
Total Integer Model (1)
▪ Machine Shop
• Two types of machines: press and lathe.
Machine Space Price (thousands) Profit
Press 15 8 100
Lathe 30 4 150
Total capacity 200 40 (budget)

• Two decision variables.

𝑥1 : number of press machines (integer).
𝑥2 : number of lathe machines (integer).
• Objective: Maximize profit

Lecture 10 4
Total Integer Model (2)
▪ Formulation

Maximize 𝑍 = 100𝑥1 + 150𝑥2 Budget constraint

Subject to 8𝑥1 + 4𝑥2 ≤ 40
15𝑥1 + 30𝑥2 ≤ 200 Space constraint
𝑥1 , 𝑥2 ≥ 0 and integer

Lecture 10 5
0-1 Integer Model (1)
▪ Community Council
• 4 facilities: Swimming pool (SP), Tennis center (TC),
Athletic field (AF), Gymnasium (GY).
Facility Cost (thousands) Land Usage (pp/day)
SP 35 4 300
TC 10 2 90
AF 25 7 400
GY 90 3 150
Total capacity 120 (budget) 12

• Four decision variables

𝑥1 : construction of SP, 𝑥2 : construction of TC,
𝑥3 : construction of AF, 𝑥4 : construction of GY.

Lecture 10 6
0-1 Integer Model (2)
• 𝑥𝑖 : 0 or 1.
• If 𝑥𝑖 =0, Facility i is not constructed.
• If 𝑥𝑖 =1, Facility i is constructed.
▪ Mutually Exclusive Constraint
• SP and TC cannot be built together due to land
constraint.
• If 𝑥1 =1 (SP), then 𝑥2 =0 (TC).
• If 𝑥2 =1 (TC), then 𝑥1 =0 (SP).
• Possible cases: (𝑥1 , 𝑥2) = 0, 0 , 1,0 , 0,1 .
• Impossible cases: (𝑥1 , 𝑥2)=(1,1).
• 𝑥1 + 𝑥2 ≤ 1.

Lecture 10 7
0-1 Integer Model (3)
▪ Objective
• Maximize total daily usage.
▪ Formulation
Maximize 𝑍 = 300𝑥1 + 90𝑥2 + 400𝑥3 + 150𝑥4 Budget constraint
Subject to 35𝑥1 + 10𝑥2 + 25𝑥3 + 90𝑥4 ≤ 120
4𝑥1 + 2𝑥2 + 7𝑥3 + 3𝑥4 ≤ 12 Land constraint
𝑥1 + 𝑥2 ≤ 1 Mutually exclusive
𝑥1 , 𝑥2 , 𝑥3 , 𝑥4 = 0 or 1 constraint

Lecture 10 8
0-1 Integer Model (4)
▪ Mutually Exclusice Constraint
• 𝑥1 + 𝑥2 ≤ 1.
▪ More Multiple Choice Constraint
• Exactly 2 of the four facilities must be built.
𝑥1 + 𝑥2 + 𝑥3 + 𝑥4 = 2.
• No more than 2 facilities must be built.
𝑥1 + 𝑥2 + 𝑥3 + 𝑥4 ≤ 2.
▪ Conditional Constraint
• If SP is not built, NO TC.
• If 𝑥1 = 0, then 𝑥2 = 0.
• Possible cases: (𝑥1 , 𝑥2 ) = 0, 0 , 1,0 , 1,1 .
• Impossible cases: (𝑥1 , 𝑥2 )=(0,1).
• 𝑥2 ≤ 𝑥1 .
▪ Corequisite Constraint
• If SP (or TC) is built, TC (or SP) must be built.
• 𝑥1 = 𝑥2 .

Lecture 10 9
Mixed Integer Model (1)

▪ Nacy Smith
• Three investments: condominium (C), land (L),
municipal bond (M).
Investment Cost (thousands) Return (thousands) Capacity
C 50 9 4
L 12 1.5 15
M 8 1 20
Total capacity 250 (budget)

• Three decision variables

𝑥1 : Number of C (Integer)
𝑥2 : acres of L
𝑥3: Number of M (Integer)

Lecture 10 10
Mixed Integer Model (1)
▪ Objective
• Maximize total return.
▪ Formulation
Maximize 𝑍 = 9𝑥1 + 1.5𝑥2 + 𝑥3 Budget constraint
Subject to 50𝑥1 + 12𝑥2 + 8𝑥3 ≤ 250
Condo. constraint
𝑥1 ≤4
𝑥2 ≤ 15 Land. constraint
𝑥3 ≤ 20 Bond constraint
𝑥1 , 𝑥3 ≥ 0 and integer
𝑥2 ≥ 0

Lecture 10 11
Graphical Solution (1)
Maximize 𝑍 = 100𝑥1 + 150𝑥2
Subject to 8𝑥1 + 4𝑥2 ≤ 40
15𝑥1 + 30𝑥2 ≤ 200
𝑥1 , 𝑥2 ≥ 0 and integer

▪ Linear Programming Solution

• 𝑥1 , 𝑥2 = 2.22, 5.56 .
▪ Rounded Solution
• 𝑥1 , 𝑥2 = 2, 6 .
• This violates the second constraint.
• Infeasible soltuion.
▪ Rounded-Down Solution
• 𝑥1 , 𝑥2 = 2, 5 .
• Feasible solution with 𝑍 = 950.
• Optimal integer solution?

Lecture 10 12
Graphical Solution (2)
𝑥2
Maximize 𝑍 = 100𝑥1 + 150𝑥2
10 Subject to 8𝑥1 + 4𝑥2 ≤ 40
A
B 15𝑥1 + 30𝑥2 ≤ 200
5 C 𝑥1 , 𝑥2 ≥ 0 and integer

5 10 𝑥1

▪ Linear Programming Solution

• Point B: 𝑥1 , 𝑥2 = 2.22, 5.56 with Z = 1056.
▪ Rounded-Down Solution
• Point C: 𝑥1 , 𝑥2 = 2, 5 with Z=950.
▪ Point A: 𝑥1 , 𝑥2 = 1, 6 with Z=1,000.
▪ Rounded-Down solution is not an optimal integer solution.

Lecture 10 13
Popular 0-1 Integer Progamming Model
▪ Capital Budgeting Example

▪ Fixed Charge and Facility Location Example

▪ Set Covering Example

Lecture 10 14
Capital Budgeting Example (1)

▪ University Bookstore at Tech

• 5 projects: 1. Web site , 2. Warehouse, 3. Clothing
department, 4. Computer department, 5. ATMs.
Project Year 1 Cost Year 2 Cost Year 3 Cost Return
1 55 40 25 120
2 45 35 20 85
3 60 25 -- 105
4 50 35 30 140
5 30 30 -- 70
Total capacity 150 110 60
• 5 Decision Variables
𝑥1 , 𝑥2, 𝑥3 , 𝑥4, 𝑥5 = 0 or 1

Lecture 10 15
Capital Budgeting Example (2)
▪ Mutually Exclusive Constraint
• Project 3 and Project 4 cannot be selected together.
• 𝑥3 + 𝑥4 ≤ 1.
▪ Objective
• Maximize return.
▪ Formulation
Maximize 𝑍 = 120𝑥1 + 85𝑥2 + 105𝑥3 + 140𝑥4 + 70𝑥5 Year 1
Subject to 55𝑥1 + 45𝑥2 + 60𝑥3 + 50𝑥4 + 30𝑥5 ≤ 150 Year 2
40𝑥1 + 35𝑥2 + 25𝑥3 + 35𝑥4 + 30𝑥5 ≤110
Year 3
25𝑥1 + 20𝑥2 + 0𝑥3 + 30𝑥4 + 0𝑥5 ≤ 60
𝑥3 + 𝑥4 ≤1 Mutually
Exclusive
𝑥1, 𝑥2 , 𝑥3 , 𝑥4 , 𝑥5 = 0 𝑜𝑟 1
Constraint

Lecture 10 16
Fixed Charge and Facility Location Example (1)

▪ 6 Farms and 3 plants

• Shipping potatoes from farmers to a plants.
▪ Farm Information ▪ Plant Information
Farm Fixed Cost/year Harvest (tons) Plant Demand
1 405 11.2 A 12
2 390 10.5 B 10
3 450 12.8 C 14
4 368 9.3
5 520 10.8
6 465 9.6

Denote 𝑓𝑖 ,
i=1,...,6

Lecture 10 17
Fixed Charge and Facility Location Example (2)
▪ Shipping Cost/ton
Farm Plant A Plant B Plant C
1 18 15 12
2 13 10 17
Denote
𝑐𝑖𝑗 ,
3 16 14 18 i=1,...,6
4 19 15 16 j=A,B,C
5 17 19 12
6 14 16 12

▪ Decision Variables
• 𝑦𝑖 = 0 or 1, 𝑖 = 1, … , 6. If 𝑦𝑖 =1, Farm i is chosen.
• 𝑥𝑖𝑗 : amount of potatoes shipped from farm i to
plant j. 𝑖 = 1, … , 6 and 𝑗 = 𝐴, 𝐵, 𝐶.

Lecture 10 18
Fixed Charge and Facility Location Example (3)

▪ Objective
• Minimize fixed cost +shipping cost.
• Fixed cost =σ6𝑖=1 𝑓𝑖 𝑦𝑖
• Shipping cost=σ6𝑖=1 σ𝑗=𝐴,𝐵,𝐶 𝑐𝑖𝑗 𝑥𝑖𝑗
▪ Farm 1 Constraint
• Total amount shipped from Farm 1 if Farm 1 is
chosen (𝑦1 = 1)
• 𝑥1𝐴 + 𝑥1𝐵 + 𝑥1𝐶 ≤11.2 tons.
• If Farm 1 is not chosen (𝑦1 = 0)
• 𝑥1𝐴 = 𝑥1𝐵 = 𝑥1𝐶 = 0.
• 𝑥1𝐴 + 𝑥1𝐵 + 𝑥1𝐶 ≤ 11.2𝑦1.
• 𝑥1𝐴 + 𝑥1𝐵 + 𝑥1𝐶 − 11.2𝑦1 ≤ 0.

Lecture 10 19
Fixed Charge and Facility Location Example (4)

▪ Plant A Constraint
• Total amount shipped to Plant A:
𝑥1𝐴 + 𝑥2𝐴 + 𝑥3𝐴 + 𝑥4𝐴 + 𝑥5𝐴 + 𝑥6𝐴 =12
▪ Model Formulation
Minimize σ6𝑖=1 𝑓𝑖 𝑦𝑖 +σ6𝑖=1 σ𝑗=𝐴,𝐵,𝐶 𝑐𝑖𝑗 𝑥𝑖𝑗
Subject to σ𝑗=𝐴,𝐵,𝐶 𝑥1𝑗 − 11.2𝑦1 ≤ 0
σ𝑗=𝐴,𝐵,𝐶 𝑥2𝑗 − 10.5𝑦2 ≤ 0
σ𝑗=𝐴,𝐵,𝐶 𝑥3𝑗 − 12.8𝑦3 ≤ 0
Farm
σ𝑗=𝐴,𝐵,𝐶 𝑥4𝑗 − 9.3𝑦4 ≤ 0 Constraint
σ𝑗=𝐴,𝐵,𝐶 𝑥5𝑗 − 10.8𝑦5 ≤ 0
σ𝑗=𝐴,𝐵,𝐶 𝑥6𝑗 − 9.6𝑦6 ≤ 0
Plant
σ6𝑖=1 𝑥𝑖𝐴 = 12, σ6𝑖=1 𝑥𝑖𝐵 = 10, σ6𝑖=1 𝑥𝑖𝐶 = 14
Constraint
𝑥𝑖𝑗 ≥ 0
𝑦𝑖 = 0 𝑜𝑟 1

Lecture 10 20
Set Covering Problem (1)
▪ American Parcel Service
• Select minimum set of package distribution hubs
from 12 cities such that there is a hub within 300
miles of every city.

Lecture 10 21
Set Covering Problem (2)
Cities Cities within 300 miles
1. Atlanta Atlanta, Charlotte, Nashville
2. Boston Boston, New York
3. Charlotte Atlanta, Charlotte, Richmond
4. Cincinnati Cincinnati, Detroit, Indianapolis, Nashville, Pittsburgh
5. Detroit Cincinnati, Detroit, Indianapolis, Milwaukee, Pittsburgh
6. Indianapolis Cincinnati, Detroit, Indianapolis, Milwaukee, Nashville, St. Lo
uis
7. Milwaukee Detroit, Indianapolis, Milwaukee
8. Nashville Atlanta, Cincinnati, Indianapolis, Nashville, St. Louis
9. New York Boston, New York, Richmond
10. Pittsburgh Cincinnati, Detroit, Pittsburgh, Richmond
11. Richmond Charlotte, New York, Pittsburgh, Richmond
12. St. Louis Indianapolis, Nashville, St. Louis

Lecture 10 22
Set Covering Problem (3)
▪ American Parcel Service
• Select minimum set of package distribution hubs

• Decision Variables
𝑥𝑖 , 𝑖 = 1 𝑡𝑜 12.
𝑥𝑖 =0, if city i is not selected as a hub.
𝑥𝑖 =1, if selected.

Lecture 10 23
Set Covering Problem (4)
▪ Objective Function
• Minimize the number of hubs.
• Minimize σ12
𝑖=1 𝑥𝑖 .

▪ Set Covering Constraint

• Each city be within 300 miles of a hub.
• Atlanta: 𝑥1 + 𝑥3 + 𝑥8 ≥ 1 (at least one must equal 1).
• Do for all the cities.

Lecture 10 24