Course: WAREHOUSE MANAGEMENT

Lecturer: Dr. Nguyễn Hằng Giang Anh

Email: [email protected]
International University
Vietnam National University HCMC
 Main Content
Chapter 1: Fundamentals of Logistics and Supply Chain Management
(Week 1)

Chapter 2: Introduction to Warehousing (Week 2)

Chapter 3: Inventory, Stock Analysis and Product classifying (Weeks 3-4)

Chapter 4: Warehouse Operations (Weeks 5-6)

Chapter 5: Warehouse Networking (Weeks 7-8)

Chapter 6: Warehouse Management System (Weeks 9)

Chapter 7: Productivity, cost and Service (Week 10)

Chapter 8: Warehouse Health and Safety (Week 11)

International University
Vietnam National University HCMC
2
 Chapter 5
Warehouse Networking

Lecturer: Dr. Nguyễn Hằng Giang Anh

Email: [email protected]
International University
Vietnam National University HCMC
 Main Contents

1. Model Classification

2. Rectilinear-distance facility location models

3. Euclidean-distance facility location models

4. Covering problems

International University
Vietnam National University HCMC
 1.
MODEL
CLASSIFICATION

International University
Vietnam National University HCMC
 Facility location models (1/5)

Facility location problems can be classified in a variety of ways:

1. Based on the number of new facilities: single-facility and multi-
facility location problems.
2. Based on the solution space:
o Continuous location problem: When new facilities can be located any
place within a two- or three-dimensional region.
o Discrete location problem: When new facilities can be located only at a
specific locations.
o Planar location problem: when it is to be located in two-dimensional
space
o 3-D location problem: when it is to be located in three-dimensional
space
International University
Vietnam National University HCMC
 Facility location models (2/5)

3. Based on the size of the facilities:

o by points in two or three-dimensional space.

o by defined areas.

4. Based on the criteria used to determine the location:

o Minisum location problem: locating one or more new facilities
such that the weighted sum of distances among pairs of facilities
is minimized.

o Minimax location problem: locating one or more new facilities

such that the weighted maximum distance travelled from a new
facility to an existing or new facility is minimized.

International University
Vietnam National University HCMC
 Facility location models (3/5)

5. Based on the distance measures used in the model: Suppose

two facilities are located at points represented by (𝑿𝟏 , 𝒀𝟏 ) and
(𝑿𝟐 , 𝒀𝟐 )
o Rectilinear distance: where distances are measured along
paths that are orthogonal to each other.

Rectilinear Distance = Difference in X co-ordinates + Difference in Y co-ordinates

= | 𝑋1 − 𝑋2 | + | 𝑌1 − 𝑌2 |

Example: A material
International Universitytransporter moving along rectilinear aisles in a factory
Vietnam National University HCMC
 Facility location models (4/5)

• Suppose two facilities are located at points represented by (𝑿𝟏 ,

𝒀𝟏 ) and (𝑿𝟐 , 𝒀𝟐 )
o Straight-line distance (Euclidean distance): where distances
are measured along the straight-line path between two points.

Straight-line Distance = (𝑋1 − 𝑋2 )2 +(𝑌1 − 𝑌2 )2

Example:
InternationalAUniversity
straight conveyor segment linking two workstations
Vietnam National University HCMC
 Facility location models (5/5)

• Suppose two facilities are located at points represented by (𝑿𝟏 ,

𝒀𝟏 ) and (𝑿𝟐 , 𝒀𝟐 )
o Chebyshev distance: where distances between two points in
two-dimensional space is the greater of the horizontal and
vertical distance travel.

Chebyshev Distance = max ( 𝑋1 − 𝑋2 , 𝑌1 − 𝑌2 )

Example: Material handling equipment, such as an automated

storage/retrieval (AS/R) machine, that moves vertically while moving
horizontally along an aisle to a reach a storage location.

o Actual distance (flow path distance): where distances are

measured along the actual path between two points.

Example: An automated guided vehicle following a guide path network

International University
Vietnam National University HCMC
 2.
RECTILINEAR-DISTANCE
FACILITY LOCATION
MODELS

International University
Vietnam National University HCMC
Rectilinear-distance facility location problems
(1/2)

• Rectilinear-distance facility location problem represents a

common situation in manufacturing and distribution settings.
It also occurs within many cities, due to the layout of streets.

• We consider:

o Both Single-facility vs. Multi-facility location problems;

o in which the solution space is planar (or two-dimensional);

o Both Minisum vs. Minimax criteria;

o in which the facilities are a mixture of points and areas.

International University
Vietnam National University HCMC
Rectilinear-distance facility location problems
(2/2)

1. Single-facility, Rectilinear Minisum Location Problem

2. Multifacility, Rectilinear Minisum location problem

3. Single-facility, Rectilinear Minimax location problem

4. Multifacility, Rectilinear Minimax location problem

International University
Vietnam National University HCMC
 1. Single-facility, Rectilinear Minisum Location
Problem (1/3)

Notation Meaning
𝑋 = (𝑥, 𝑦) Location of new facility
𝑃𝑖 = (𝑎𝑖 , 𝑏𝑖 ) Location of existing facility 𝑖, 𝑖 = 1, 2, … , 𝑚
“Weight” associated with travel between new facility
𝑤𝑖
and existing facility 𝑖
𝑑(𝑋, 𝑃𝑖 ) Distance between new facility and existing facility 𝑖

• Assumption: The annual cost of travel between the new facility

and existing facility 𝑖 is proportional to the distance between the
points 𝑋 and 𝑃𝑖 at the constant weight 𝑤𝑖 .

International University
Vietnam National University HCMC
 1. Single-facility, Rectilinear Minisum Location
Problem (2/3)
• Objective function: số facility

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑋 = ෍ 𝑤𝑖 𝑑(𝑋, 𝑃𝑖 )
𝑖=1
where
𝑑 𝑋, 𝑃𝑖 = 𝑥 − 𝑎𝑖 + 𝑦 − 𝑏𝑖
• Hence, the problem can be formulated as:
𝑚 𝑚

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑥, 𝑦 = ෍ 𝑤𝑖 𝑥 − 𝑎𝑖 + ෍ 𝑤𝑖 𝑦 − 𝑏𝑖
𝑖=1 𝑖=1

• Since the above function is separable in 𝑥 and 𝑦, the optimum values of

𝒙 and 𝒚 can be obtained independently:
𝑴𝒊𝒏 𝒇 𝒙 = σ𝒎
𝒊=𝟏 𝒘𝒊 𝒙 − 𝒂𝒊

International University
𝑴𝒊𝒏 𝒇 𝒚 = σ𝒎
𝒊=𝟏 𝒘𝒊 𝒚 − 𝒃𝒊
Vietnam National University HCMC
 1. Single-facility, Rectilinear Minisum Location
Problem (3/3)

• 𝑓 𝑥 and 𝑓 𝑦 are piecewise linear functions → Optimum values

of 𝑥 and 𝑦 will be coordinates of existing facilities.
o an optimum value of x will be one of the 𝑎𝑖 values,
o an optimum value of y will be one of the b𝑖 values.
• Median conditions of an optimum solution:
o 𝒙∗ , the optimum 𝒙-coordinate, will be such that no more
than half the total weight is to the left of 𝑥 ∗ , and no more
than half the total weight is to the right of 𝑥 ∗ .
o 𝒚∗ , the optimum 𝒚-coordinate, will be such that no more
than half the total weight is to the left of 𝑦 ∗ , and no more
than half the total weight is to the right of 𝑦 ∗
International University
Vietnam National University HCMC
 Example 1

• Five warehouses currently have the following cooordinate

locations: 𝑃1 = 1, 1 , 𝑃2 = 6, 2 , 𝑃3 = 2, 8 , 𝑃4 = 3, 6 , 𝑃5 = 8, 4 .
• The cost per unit distance traveled between the new warehouse
and each existing warehouse is the same.
• The number of trips per month between the new warehouse and
existing warehouse 1, 2, 3, 4, 5 are 𝟏𝟎, 𝟐𝟎, 𝟐𝟓, 𝟐𝟎, 𝟐𝟓, respectively.
a. Find the optimum location for the new warehouse.
b. If it is impossible to locate the new warehouse in the optimum
location. Instead, there are three feasible locations: 𝑄1 =
3, 5 , 𝑄2 = 4, 5 , and 𝑄3 = 2, 4 . Which is preferred?

International University
Vietnam National University HCMC
 Example 1 – Solution (1/4)

a. Solving for the optimum x-coordinate:

𝒊
Warehouse 𝒊 Coordinate 𝒂𝒊 Weight 𝒘𝒊 ෍ 𝒘𝒉
𝒉=𝟏
1 1 10 10
3 2 25 35 < 50
4 3 20 55 > 50
2 6 20 75
5 8 25 100

- Ordering the x-coordinates of the existing facilities give the sequence

1, 2, 3, 6, and 8, with the corresponding weights 10, 25, 20, 20, and 25.
- The sum of the weights is 100. The partial sum of the ordered
sequence of weights first equal or exceeds one-half the total (50) for
𝑖 = 4 → 𝒙∗ = 𝒂𝟒 = 𝟑.
International University
Vietnam National University HCMC
 Example 1 – Solution (2/4)

Solving for the optimum y-coordinate: Likewise, 𝒚∗ = 𝒃𝟓 = 𝟒.

𝒊
Warehouse 𝒊 Coordinate 𝒃𝒊 Weight 𝒘𝒊 ෍ 𝒘𝒉
𝒉=𝟏
1 1 10 10
2 2 20 30 < 50
5 4 25 55 > 50
4 6 20 75
3 8 25 100

Thus, the optimum location for the new warehouse is 𝑿∗ = 𝒙∗ , 𝒚∗ =

(3, 4)
The total weighted distance resulting from the location 𝑿∗ = (𝟑, 𝟒) is
𝑓(3, 4) = σ5𝑖=1 𝑤𝑖 3 − 𝑎𝑖 + σ5𝑖=1 𝑤𝑖 4 − 𝑏𝑖
= 50+100+125+40+125
International University = 440
Vietnam National University HCMC
 Example 1 – Solution (3/4)

𝑃3

𝑃4

𝑿∗
𝑃5

𝑃1 𝑃2

Using Excel to solve the single-facility, rectilinear location problem

International University
Vietnam National University HCMC
 Example 1 – Solution (4/4)

b. Computing the value of 𝑓(𝑋) for 𝑋 = 𝑄𝑘 , 𝑘 = 1, 2, 3 yeilds:

• 𝑄1 = 3, 5 → 𝑓 3, 5 = 60 + 120 + 100 + 20 + 150 = 450
• 𝑄2 = 4, 5 → 𝑓 4, 5 = 70 + 100 + 125 + 40 + 125 = 460
• 𝑄3 = 2, 4 → 𝑓 2, 4 = 40 + 120 + 100 + 60 + 150 = 470

→ The best site is 𝑸𝟏 = 𝟑, 𝟓

International University
Vietnam National University HCMC
2. Multifacility, Rectilinear Minisum location problem
(1/3)

Notation Meaning
n Number of new facilities
m Number of existing facilities
𝑋𝑗 = (𝑥𝑗 , 𝑦𝑗 ) Location of new facility 𝑗, 𝑗 = 1, 2, … , 𝑛
𝑃𝑖 = (𝑎𝑖 , 𝑏𝑖 ) Location of existing facility 𝑖, 𝑖 = 1, 2, … , 𝑚
“Weight” associated with travel between new facility 𝑗
𝑤𝑗𝑖
and existing facility 𝑖
“Weight” associated with travel between new facility 𝑗
𝑣𝑗𝑘
and new facility 𝑘
𝑑(𝑋𝑗 , 𝑃𝑖 ) Distance between new facility 𝑗 and existing facility 𝑖
𝑑(𝑋𝑗 , 𝑋𝑘 ) Distance between new facility 𝑗 and new facility 𝑘
International University
Vietnam National University HCMC
2. Multifacility, Rectilinear Minisum location problem
(2/3)

• Objective: Determine {𝑋𝑗 , 𝑗 = 1, 2 … , 𝑛} that minimizes the weighted sum

of travel between pairs of new facilities, and between each new facility and
the existing ones.
• As before, the objective function is separable in the 𝑥- and 𝑦-coordinates:
𝑴𝒊𝒏 𝒇(𝒙𝟏 , … , 𝒙𝒏 )= σ𝟏≤𝒋<𝒌≤𝒏 𝒗𝒋𝒌 𝒙𝒋 − 𝒙𝒌 + σ𝒏𝒋=𝟏 σ𝒎
𝒊=𝟏 𝒘𝒋𝒊 𝒙𝒋 − 𝒂𝒊 (1)

𝑴𝒊𝒏 𝒇(𝒚𝟏 , … , 𝒚𝒏 )= σ𝟏≤𝒋<𝒌≤𝒏 𝒗𝒋𝒌 𝒚𝒋 − 𝒚𝒌 + σ𝒏𝒋=𝟏 σ𝒎

𝒊=𝟏 𝒘𝒋𝒊 𝒚𝒋 − 𝒃𝒊

• Because of the interactions between pairs of new facilities, obtaining

optimum solutions is not as straightforrward as it was for a single-facility
problem.

International University
Vietnam National University HCMC
2. Multifacility, Rectilinear Minisum location problem
(3/3)

• One approach is to formulate multifacility, rectilinear minisum location

problem as a linear programming problem.
• To do so, we replace the absolute value terms with new decision
variables and add appropriate linear constraints.
• Then, Equation (1) is replaced with the following Equations (2), (3), (4)
and (5).
𝑀𝑖𝑛 σ1≤𝑗≤𝑘≤𝑛 𝑣𝑗𝑘 (𝑝𝑗𝑘 + 𝑞𝑗𝑘 ) + σ𝑛𝑗=1 σ𝑚
𝑖=1 𝑤𝑗𝑖 (𝑟𝑗𝑘 + 𝑠𝑗𝑘 ) (2)
Subject to:
𝑥𝑗 − 𝑥𝑘 + 𝑝𝑗𝑘 − 𝑞𝑗𝑘 = 0 for 1 = 𝑗 < 𝑘 = 𝑛 (3)
𝑥𝑗 − 𝑟𝑗𝑖 + 𝑠𝑗𝑖 = 𝑎𝑖 for 𝑖 = 1, … , 𝑚; 𝑗 = 1, … , 𝑛 (4)
𝑝𝑗𝑘 = 0, 𝑞𝑗𝑘 = 0, 𝑟𝑗𝑖 = 0, 𝑠𝑗𝑖 = 0 for 1 ≤ 𝑗 < 𝑘 ≤ 𝑛; 𝑖 = 1, … , 𝑚 (5)
International University
Vietnam National University HCMC
 Example 2

• Consider a situation where a new electronics product is manufactured

and assembled.
• Four machines currently located in the production area will be used,
along with the two new machines that have been purchased.
• A portion of incoming material moves from new machine 1 to new
machine 2; the remainder moves from new machines to the existing
machines.
• The material flows, expressed in thousands of units, are: 𝑤11 =
10, 𝑤12 = 6, 𝑤13 = 𝑤14 = 𝑤21 = 0, 𝑤22 = 2, 𝑤23 = 16, 𝑤24 = 8, and
𝑣12 = 12.
• The existing machines have the following coordinates: 𝑃1 = (0, 10),
𝑃2 = 15, 0 , 𝑃3 = (25, 30), and 𝑃4 = (40, 15).
• Determine the location of the two new machines?
International University
Vietnam National University HCMC
 Example 2 – Solution (1/3)

The linear programming formulation for x-coordinate is:

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 12 𝑝12 + 𝑞12 + 10 𝑟11 + 𝑠11 + 6 𝑟12 + 𝑠12 + 2 𝑟22 + 𝑠22
+ 16 𝑟23 + 𝑠23 + 8(𝑟24 + 𝑠24 )
Subject to: 𝑥1 − 𝑥2 + 𝑝12 − 𝑞12 = 0
𝑥1 − 𝑟11 + 𝑠11 = 0
𝑥1 − 𝑟12 + 𝑠12 = 15
𝑥2 − 𝑟22 + 𝑠22 = 15
𝑥2 − 𝑟23 + 𝑠23 = 25
𝑥2 − 𝑟24 + 𝑠24 = 40
𝑟11 , 𝑠11 , 𝑟12 , 𝑠12 , 𝑟22 , 𝑠22 , 𝑟23 , 𝑠23 , 𝑟24 , 𝑠24 , 𝑝12 , 𝑞12 ≥ 0
Solve with Solver Excel to get the optimum x-coordinates: 𝑥1 = 15, 𝑥2 =
15, 𝑓 𝑋 = 410
International University
Vietnam National University HCMC
 Example 2 – Solution (2/3)

The linear programming formulation for y-coordinate is:

Min 12 𝑝12 + 𝑞12 + 10 𝑟11 + 𝑠11 + 6 𝑟12 + 𝑠12 + 2(𝑟22 +
𝑠22 ) + 16 𝑟23 + 𝑠23 + 8(𝑟24 + 𝑠24 )
Subject to: 𝑦1 − 𝑦2 − 𝑝12 + 𝑞12 = 0
𝑦1 − 𝑟11 + 𝑠11 = 10
𝑦1 − 𝑟12 + 𝑠12 = 0
𝑦2 − 𝑟22 + 𝑠22 = 0
𝑦2 − 𝑟23 + 𝑠23 = 30
𝑦2 − 𝑟24 + 𝑠24 = 15
𝑟11 , 𝑠11 , 𝑟12 , 𝑠12 , 𝑟22 , 𝑠22 , 𝑟23 , 𝑠23 , 𝑟24 , 𝑠24 , 𝑝12 , 𝑞12 ≥ 0
SolveInternational
with Solver Excel to get the optimum y-coordinates: 𝑦1 = 10,
University
𝑦2 =Vietnam
15, 𝑓National
𝑌 = 390HCMC
University
 Example 2 – Solution (3/3)
• The optimum locations for the two new facilities are 𝑿∗𝟏 = 𝟏𝟓, 𝟏𝟎 , 𝑿∗𝟐 =
(𝟐𝟓, 𝟏𝟓) with the total distance traveled is 410 + 390 = 800

International University
Vietnam National University HCMC
 3. Single-facility, Rectilinear Minimax location
problem (1/2)

• Using the same notation as before, the minimax location problem for a
single new facility is given by:
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑋 = 𝑚𝑎𝑥1≤𝑖≤𝑚 𝑥 − 𝑎𝑖 + 𝑦 − 𝑏𝑖 + 𝑔𝑖 ,
where 𝑔𝑖 is fixed travel associated with existing facility 𝑖.
• To obtain the minimax solution, we let:
𝑐1 = 𝑚𝑖𝑛𝑖𝑚𝑢𝑚 𝑎𝑖 + 𝑏𝑖 − 𝑔𝑖 (6)
𝑐2 = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚(𝑎𝑖 + 𝑏𝑖 + 𝑔𝑖 ) (7)
𝑐3 = 𝑚𝑖𝑛𝑖𝑚𝑢𝑚(−𝑎𝑖 + 𝑏𝑖 − 𝑔𝑖 ) (8)
𝑐4 = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚(−𝑎𝑖 + 𝑏𝑖 + 𝑔𝑖 ) (9)
𝑐5 = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚(𝑐2 − 𝑐1 , 𝑐4 − 𝑐3 ) (10)

International University
Vietnam National University HCMC
 3. Single-facility, Rectilinear Minimax location
problem (2/2)

• As shown in Francis et al. (1992), optimum solutions to the minimax

location problem are all points on the line segment connecting the
point:
𝒙∗𝟏 , 𝒚∗𝟏 = 𝟎. 𝟓 𝒄𝟏 − 𝒄𝟑 , 𝒄𝟏 + 𝒄𝟑 + 𝒄𝟓

and the point

𝒙∗𝟐 , 𝒚∗𝟐 = 𝟎. 𝟓 𝒄𝟐 − 𝒄𝟒 , 𝒄𝟐 + 𝒄𝟒 − 𝒄𝟓

• The maximum distance equal (𝑐5Τ2).

International University
Vietnam National University HCMC
 Example 3
Worksta Coordinates Workst Coordinates
tion (i) 𝒂𝒊 𝒃𝒊 ation (i) 𝒂𝒊 𝒃𝒊
• A portable restroom and canteen
1 10 10 16 135 190
facility is located on a site. It is
2 10 45 17 150 130
desired that the maximum
distance any workers have to 3 15 145 18 155 25

travel to reach the facility is 4 15 75 19 165 25

minimized. 5 20 110 20 170 200

6 35 185 21 175 180
• Travel is best approximated with
7 40 150 22 180 40
rectilinear distances.
8 45 15 23 180 65
• The coordinates for the 30
9 50 90 24 180 170
workstations on the site are given
10 65 140 25 185 10
in the next table.
11 70 200 26 185 165
• Fixed travel for each workstation 12 85 60 27 190 35
is assumed to be negligible.
13 100 45 28 190 100
14 110 200 29 200 65
International University
Vietnam National University HCMC 15 115 70 30 200 195
 Example 3 – Solution (1/3)
Coordinates Workst Coordinates
Workst
𝒂 𝒊 + 𝒃𝒊 −𝒂𝒊 + 𝒃𝒊 ation 𝒂 𝒊 + 𝒃𝒊 −𝒂𝒊 + 𝒃𝒊
ation (i) 𝒂𝒊 𝒃𝒊 𝒂𝒊 𝒃𝒊
(i)
1 10 10 20 0 16 135 190 325 55
2 10 45 55 35 17 150 130 280 -20
3 15 145 160 130 18 155 25 180 -130
4 15 75 90 60 19 165 25 190 -140
5 20 110 130 90 20 170 200 370 30
6 35 185 220 150 21 175 180 355 5
7 40 150 190 110 22 180 40 220 -140
8 45 15 60 -30 23 180 65 245 -115
9 50 90 140 40 24 180 170 350 -10
10 65 140 205 75 25 185 10 195 -175
11 70 200 270 130 26 185 165 350 -20
12 85 60 145 -25 27 190 35 225 -155
13 100 45 145 -55 28 190 100 290 -90
14 110 200
International University
310 90 29 200 65 265 -135
Vietnam115
15 National70
University
185HCMC -45 30 200 195 395 -5
 Example 3 – Solution (2/3)

𝑐1 = 𝑚𝑖𝑛𝑖𝑚𝑢𝑚 𝑎𝑖 + 𝑏𝑖 − 𝑔𝑖 = 20
𝑐2 = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚 𝑎𝑖 + 𝑏𝑖 + 𝑔𝑖 = 395
𝑐3 = 𝑚𝑖𝑛𝑖𝑚𝑢𝑚(−𝑎𝑖 + 𝑏𝑖 − 𝑔𝑖 ) = -175
𝑐4 = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚(−𝑎𝑖 + 𝑏𝑖 − 𝑔𝑖 )= 150
𝑐5 = 𝑚𝑎𝑥𝑖𝑚𝑢𝑚(𝑐2 − 𝑐1 , 𝑐4 − 𝑐3 )= 375
→ Optimum solutions are all points on the line with coordinates:
𝑥1∗ , 𝑦1∗ = 97.5, 110
𝑥2∗ , 𝑦2∗ = 122.5, 85
The maximum distance is 187.5

International University
Vietnam National University HCMC
 Example 3 – Solution (3/3)

International University
Vietnam National University HCMC
4. Multifacility, Rectilinear Minimax location problem

• An extension of the minimax location problem for multifacility is given by:

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 (𝑚𝑎𝑥 1≤𝑖≤𝑚 ൫𝑤𝑗𝑖 𝑥𝑗 − 𝑎𝑖 + 𝑦𝑗 − 𝑏𝑖 + 𝑔𝑗𝑖 , 𝑣𝑗𝑘 ൫ห𝑥𝑗 −

1≤𝑗<𝑘≤𝑛
𝑥𝑘 ห + 𝑦𝑗 − 𝑦𝑘 ൯ + ℎ𝑗𝑘 ൯,

where 𝑔𝑗𝑖 and ℎ𝑗𝑘 are fixed “costs” associated with new-and-existing
facility pairs and pairs of new facilities, respectively.

• This problem can be converted to a linear programming problem. Instead,

we can use SOLVER to solve a small-size example

International University
Vietnam National University HCMC
 Example 4

Recall Example 1, which consists of five warehouses with the following

coordinate locations: 𝑃1 = 1, 1 , 𝑃2 = 6, 2 , 𝑃3 = 2, 8 , 𝑃4 = 3, 6 , 𝑃5 =
8, 4 .
Suppose two new warehouses are added with the goods flow divided
between the two new warehouses.
Suppose the flow data are as follows: 𝑤11 = 10, 𝑤12 = 20, 𝑤13 =
20, 𝑤14 = 𝑤15 = 𝑤21 = 𝑤22 = 0, 𝑤23 = 5, 𝑤24 = 20, 𝑤25 = 25, 𝑣12 = 10.

→ Solve with SOLVER, we obtain a solution: 𝑥1∗ , 𝑦1∗ = 2.07, 3.07 and
𝑥2∗ , 𝑦2∗ = 7.02, 6.49 with the objective function 𝑓 𝑋 ∗ = 100.00

International University
Vietnam National University HCMC
 3.
EUCLIDEAN-DISTANCE
FACILITY LOCATION
MODELS

International University
Vietnam National University HCMC
Euclidean-distance facility location problems (1/2)

• Although straight-line travel is rare, solving location

problems by assuming Euclidean travel can be useful
- In many cases, the actual distance traveled will be
between Rectilinear and Euclidean distance. In another
words, Rectilinear and Euclidean distance provide
upper and lower bounds on the distance actually
traveled.
- It is assumed that the “real” optimum location in the
neighborhood of the location obtained using either or both
distance measures.

International University
Vietnam National University HCMC
Euclidean-distance facility location problems (2/2)

1. Single-facility, Squared Euclidean minisum location

problem – Gravity Problem

2. Multifacility, Squared Euclidean minisum location problem

3. Single-facility, Euclidean Minisum Location Problem

4. Multifacility, Euclidean Minisum location problem

5. Single-facility, Euclidean Minimax location problem

6. Multifacility, Euclidean Minimax location problem

International University
Vietnam National University HCMC
 1. Single-facility, Squared Euclidean minisum
location problem (1/2)

• Assumption: The “cost of travel” between a single new

facility and multiple existing facilities is proportional to the
square of the Euclidean distance between the facilities.
• The gravity problem can be formulated as follows:
𝑚

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑥, 𝑦 = ෍ 𝑤𝑖 𝑥 − 𝑎𝑖 2 + 𝑦 − 𝑏𝑖 2

𝑖=1

How can we solve this problem?

International University
Vietnam National University HCMC
 1. Single-facility, Squared Euclidean minisum
location problem (2/2)

• Take the partial derivatives of 𝑓(𝑥, 𝑦) with respect to x and y,

setting them equal to 0, and solving for 𝒙∗ and 𝒚∗ yields the
following unique solution:
𝒎 𝒎

𝒙∗ = ෍ 𝒘𝒊 𝒂𝒊 / ෍ 𝒘𝒊
𝒊=𝟏 𝒊=𝟏
𝒎 𝒎

𝒚∗ = ෍ 𝒘𝒊 𝒃𝒊 / ෍ 𝒘𝒊
𝒊=𝟏 𝒊=𝟏

• The optimum solution to the gravity problem is the center

of gravity or centroid location.
International University
Vietnam National University HCMC
 Example 5

• Consider a maintenance department, in which a new

general-purpose machine tool is to be installed, and
where five existing machines have the following
coordinate locations: 𝑃1 = 1, 1 , 𝑃2 = 6, 2 , 𝑃3 = 2, 8 , 𝑃4 =
3, 6 , and 𝑃5 = 8, 4 .
• The cost of material movement is assumed to be
proportional to the square of the straight-line distance
between the new machine tool and the existing machine
tools. The proportionality constants are 10, 20, 25, 20, and
25, respectively.
• Define the location for the new machine tool.

International University
Vietnam National University HCMC
 Example 5 – Solution
Objective function:
𝑓 𝑥, 𝑦
= 10 𝑥 − 1 2 + 𝑦 − 1 2 + 20 𝑥 − 6 2
+ 𝑦−2 2

+ 25 𝑥 − 2 2 + 𝑦 − 8 2 + 20 𝑥 − 3 2+ 𝑦−6 2

+ 25[ 𝑥 − 8 2 + 𝑦 − 4 2 ]
Take the partial derivatives of 𝑓(𝑥, 𝑦) with respect to x and y:
𝜕𝑓(𝑥,𝑦)
= 20 𝑥 − 1 + 40 𝑥 − 6 + 50 𝑥 − 2 + 40 𝑥 − 3 +
𝜕𝑥
σ5𝒊=𝟏 𝒘𝒊 𝒂𝒊 880
50 𝑥 − 8 = 0 → 𝑥∗ = σ5𝒊=𝟏 𝒘𝒊
= = 4.4
200
𝜕𝑓(𝑥,𝑦)
= 20 𝑥 − 1 + 40 𝑥 − 2 + 50 𝑥 − 8 + 40 𝑥 − 6 +
𝜕𝑦
σ5𝒊=𝟏 𝒘𝒊 𝒃𝒊 940
50 𝑥 − 4 = 0 → 𝑦∗ = σ5𝒊=𝟏 𝒘𝒊
= = 4.7
200
If the optimum location is not feasible, the closest feasible location to
International University
the optimum
Vietnam National University HCMC location should be chosen
2. Multifacility, Squared Euclidean minisum location
problem (1/2)
• The multifacility extension of the gravity problem can be
formulated as follows:
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑥1 , 𝑦1 , 𝑥2 , 𝑦2 , … , 𝑥𝑛 , 𝑦𝑛
= ෍ 𝑣𝑗𝑘 (𝑥𝑗 − 𝑥𝑘 )2 +(𝑦𝑗 − 𝑦𝑘 )2
1≤𝑗<𝑘≤𝑛
𝑛 𝑚

+ ෍ ෍ 𝑤𝑗𝑖 (𝑥𝑗 − 𝑎𝑖 )2 +(𝑦𝑗 − 𝑏𝑖 )2

𝑗=1 𝑖=1

• As with the single-facility gravity problem, partial derivatives of

the above equation with respect to each decision variable are
computed and set equal to 0.
• Two sets of linear equations result, one involving the x-
coordinates of the new facilities and the other involving the y-
coordinates of the new facilities.
International University
Vietnam National University HCMC
 Example 6
Example 8
• A new electronics product was to be manufactured and
assembled. Two new machines are to be located among
four machines currently located in the production area.
• The material flows, expressed in thousands of units, are as
follows: 𝑤11 = 10, 𝑤12 = 6 , 𝑤13 = 𝑤14 = 𝑤21 = 0, 𝑤22 = 2 ,
𝑤23 = 16, 𝑤24 = 8 , and 𝑣12 = 12 . The existing machines
have the following coordinates: 𝑃1 = 0,10 , 𝑃2 =
15,0 , 𝑃3 = 25,30 , and 𝑃4 = (40,15).
• If material handling cost is proportional to the weighted sum
of the squares of the straight-line distances traveled, where
should the two new machines be located?

International University
Vietnam National University HCMC
 Example 6 – Solution (1/2)

Objective function for x-coordinate:

𝑓 𝑥1 , 𝑥2 = 10𝑥12 + 6(𝑥1 − 15)2 +2(𝑥2 − 15)2 +16

(𝑥2 − 25)2 +8(𝑥2 − 40)2 +12 (𝑥1 − 𝑥2 )2

Take the partial derivatives of 𝑓 𝑥1 , 𝑥2 with respect to 𝑥1 and 𝑥2 :

𝜕𝑓 𝑥1 , 𝑥2
= 20𝑥1 + 12 𝑥1 − 15 + 24(𝑥1 −𝑥2 ) = 0 ⇔ 7𝑥1 − 3𝑥2 = 22.5
𝜕𝑥1

𝜕𝑓 𝑥1 ,𝑥2
= 4 𝑥2 − 15 + 32 𝑥2 − 25 + 16 𝑥2 − 40 + 24(𝑥2 −𝑥1 ) = 0
𝜕𝑥2

⇔19 𝑥2 - 6 𝑥1 = 375

Solving the system of two linear equations give 𝑥1∗ = 13.5 and 𝑥2∗ = 24.
International University
Vietnam National University HCMC
 Example 6 – Solution (2/2)

Objective function for y-coordinate:

𝑓 𝑦1 , 𝑦2 = 10(𝑦1 − 10)2 +6𝑦12 + 2𝑦22 + 16(𝑦2 − 30)2 +8(𝑦2 − 15)2 +12

(𝑦1 − 𝑦2 )2

Take the partial derivatives of𝑓 𝑦1 , 𝑦2 with respect to 𝑦1 and 𝑦2 :

𝜕𝑓 𝑦1 , 𝑦2
= 20(𝑦1 −10) + 12𝑦1 + 24(𝑦1 −𝑦2 ) = 0 ⇔ 7𝑦1 − 3𝑦2 = 25
𝜕𝑦1

𝜕𝑓 𝑦1 ,𝑦2
= 4𝑦2 + 32 𝑦2 − 30 + 16 𝑦2 − 15 + 24(𝑦2 −𝑦1 ) = 0 ⇔ 19 𝑦2 -6
𝜕𝑦2

𝑦1 = 300

Solving the system of two linear equations give 𝑦1∗ = 11.96 and 𝑦2∗ = 19.57.
Let’s solve with SOLVER
International University
Vietnam National University HCMC
 3. Single-facility, Euclidean minisum location
problem

• The formulation is:

𝑚

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑥, 𝑦 = ෍ 𝑤𝑖 𝑥 − 𝑎𝑖 2 + 𝑦 − 𝑏𝑖 2 1/2

𝑖=1

• Rather than pursue more exact methods, we will use SOLVER

tool to solve the single-facility, Euclidean minisum location
problem.

International University
Vietnam National University HCMC
 Example 7

Recall Example 5 with the consideration of Euclidean travel in

obtaining the minisum location for the new machine

Solving with SOLVER yields the location for the new machine:
(3.70, 5.45) with the objective function value of 342.85.

International University
Vietnam National University HCMC
4. Multifacility, Euclidean minisum location problem

• The formulation is:

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑥1 , 𝑦1 , 𝑥2 , 𝑦2 , … , 𝑥𝑛 , 𝑦𝑛
= ෍ 𝑣𝑗𝑘 [(𝑥𝑗 − 𝑥𝑘 )2 +(𝑦𝑗 − 𝑦𝑘 )2 ]1/2
1≤𝑗<𝑘≤𝑛
𝑛 𝑚
1/2
+ ෍ ෍ 𝑤𝑗𝑖 (𝑥𝑗 − 𝑎𝑖 )2 +(𝑦𝑗 − 𝑏𝑖 )2

𝑗=1 𝑖=1

• We will use SOLVER tool to solve the multifacility, Euclidean

minisum location problem.

International University
Vietnam National University HCMC
 Example 8

Two new special purpose machine tools are to be added to a

maintenance department that contained five existing machines
having the following coordinate locations: 𝑃1 = 1,1 , 𝑃2 =
6,2 , 𝑃3 = 2,8 , 𝑃4 = 3,6 , 𝑃5 = 8,4 .
The flow data for the location problem are as follows: 𝑤11 =
10, 𝑤12 = 𝑤13 = 20, 𝑤14 = 𝑤15 = 𝑤21 = 𝑤22 = 0, 𝑤23 = 5, 𝑤24 =
20, 𝑤25 = 25, 𝑣12 = 5.

Solving with SOLVER yields the location for the 2 new tools:
𝑥1∗ , 𝑦1∗ = (3.94, 4.25) and 𝑥2∗ , 𝑦2∗ = (4.48, 5.25) with the
objective function value of 340.04.

International University
Vietnam National University HCMC
 5. Single-facility, Euclidean minimax location
problem

• When using a minimax criterion in locating a single new facility,

one is seeking the center of a circle with the smallest diameter
that “covers” the m existing facilities. The formulation is:
2 2 1/2
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓 𝑋 = 𝑚𝑎𝑥1≤𝑖≤𝑚 𝑥 − 𝑎𝑖 + 𝑦 − 𝑏𝑖 + 𝑔𝑖

• As with the rectilinear formulation, fixed “costs” are included for

each existing facility.

• We will use SOLVER tool to solve the single-facility, Euclidean

minimax location problem.

International University
Vietnam National University HCMC
 Example 9

• A conveyor belt is to be located in a fulfillment DC, so that

pickers can reach the belt quickly in order to reduce traveling
distance. As a result of the need to minimize the maximum travel
distance, the location problem is formulated as a single-facility,
Euclidean minimax location problem.
• Locations of the major picking areas have the following
coordinates: 𝑃1 = 10, 10 , 𝑃2 = 80, 20 , 𝑃3 = 25, 85 , 𝑃4 =
30, 60 , 𝑃5 = 85, 45 , 𝑃6 = 60, 15 , 𝑃7 = 80, 70 , 𝑃9 =
30, 45 , 𝑃10 = 45, 25 , with no fixed “costs” incorporated in the
solution

Solving with SOLVER yields the location for the conveyor belt:
𝒙, 𝒚 = (𝟒𝟐. 𝟗𝟏, 𝟒𝟐. 𝟒𝟒) with the objective function value of 46.21.

International University
Vietnam National University HCMC
6. Multifacility, Euclidean minimax location problem

• The multifacility extension has the formulation:

1/2
𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 (𝑚𝑎𝑥 1≤𝑖≤𝑚 ൬𝑤𝑗𝑖 (𝑥𝑗 − 𝑎𝑖 )2 +(𝑦𝑗 − 𝑏𝑖 )2 +
1≤𝑗<𝑘≤𝑛
2 2 1
𝑔𝑗𝑖 , 𝑣𝑗𝑘 [ 𝑥𝑗 − 𝑥𝑘 + 𝑦𝑗 − 𝑦𝑘 ]2 + ℎ𝑗𝑘 ൰)

where 𝒈𝒋𝒊 and 𝒉𝒋𝒌 are fixed “costs” associated with new-and-
existing facility pairs and pairs of new facilities, respectively.
• Rather than pursue more exact methods, we will use SOLVER
tool to solve the multifacility, Euclidean minimax location
problem

International University
Vietnam National University HCMC
 Example 10

Recalling Example 8, with Euclidean distances, where should the

two new machines be located in order to minimize the maximum
weighted distance traveled?

Solving with SOLVER yields the locations for the two new
machines 𝑥1∗ , 𝑦1∗ = (4.00, 5.00), and 𝑥2∗ , 𝑦2∗ = 6.23, 4.45 with
the objective function value of 72.11.

International University
Vietnam National University HCMC
 4.
COVERING PROBLEMS

International University
Vietnam National University HCMC
 Covering problems

• Covering problems are discrete-space location problems and

consist of two types:
- Total cover problem or Set covering problem: determines the
minimum number of new facilities required to cover a set of
existing facilities;
- Partial cover problem or Maximum cover problem: :
determines the locations of a specified number of new
facilities to maximize the number of existing facilities covered
by a specified number of new facilities
• By cover, we mean that the distance between the new facility and
an existing facility is less than some specified maximum value.
International University
Vietnam National University HCMC
 Total cover problem (1/2)

Notation Meaning
𝑐𝑗 Cost of assigning a new facility to site 𝑗, 𝑗 = 1, 2, … , 𝑛
Cover coefficient
𝑎𝑖𝑗 1 𝑖𝑓 𝑐𝑢𝑠𝑡𝑜𝑚𝑒𝑟 𝑜𝑟 𝑒𝑥𝑖𝑠𝑡𝑖𝑛𝑔 𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦 𝑖 𝑖𝑠 𝑐𝑜𝑣𝑒𝑟𝑒𝑑 𝑏𝑦 𝑠𝑖𝑡𝑒 𝑗
=ቊ
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

1 𝑖𝑓 𝑎 𝑛𝑒𝑤 𝑓𝑎𝑐𝑖𝑙𝑖𝑡𝑦 𝑖𝑠 𝑙𝑜𝑐𝑎𝑡𝑒𝑑 𝑎𝑡 𝑠𝑖𝑡𝑒 𝑗

𝑥𝑗 =ቊ
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

• Application: product delivery, bus routes, network defense, truck

dispatching, airline crew scheduling, assembly line balancing, and
warehouse location.
International University
Vietnam National University HCMC
 Total cover problem (2/2)

• The objective is to cover all customers and to do so at

minimum cost. The total cover problem can be formulated as
follows:
𝑛

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑧 = ෍ 𝑐𝑗 𝑥𝑗
𝑗=1

subject to σ𝑛𝑗=1 𝑎𝑖𝑗 𝑥𝑗 ≥ 1 , 𝑖 = 1, … , 𝑚 (11)

𝑥𝑗 = 0, 1 , 𝑗 = 1, … , 𝑛
• Constraint (11) ensures that each of the 𝑚existing facilities is
covered by at least one occupied site.
• A variety of solution procedures have been developed for
covering problems, ranging from heuristics to cutting-plane
methods. For our purposes, we continue to use SOLVER
International University
Vietnam National University HCMC
 Example 11 (1/2)

• A retail firm is considering a major change in its distribution system.

Specifically, in the past, it shipped merchandise from a few DCs to its
stores on a weekly basis.
• Consumer surveys indicate lost sales because of out-of-stock situations
occurring. The competition seems to be able to maintain its inventory in
its store without incurring excessive inventory costs.
• Among the options being considered is locating DCs such that all retail
stores can be replenished within 24 hours of a request for
merchandise. To evaluate the feasibility of such an approach, a region
of the country with numerous stores is to serve as a test region.
• The following table shows a matrix of cover coefficients for 8 stores
and 10 potential sites for distribution centers.
1 𝑖𝑓 𝑠𝑡𝑜𝑟𝑒 𝑖 𝑐𝑎𝑛 𝑏𝑒 𝑟𝑒𝑝𝑙𝑒𝑛𝑖𝑠ℎ𝑒𝑑 within 24ℎ 𝑓𝑟𝑜𝑚 𝑠𝑖𝑡𝑒 𝑗
𝑎𝑖𝑗 = ቊ
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

International University
Vietnam National University HCMC
 Example 11 (2/2)

Distribution center site

Store 𝒊
1 2 3 4 5 6 7 8 9 10
1 1 1 0 0 0 0 0 1 1 1
2 1 1 1 0 0 0 0 0 0 0
3 1 1 0 1 1 0 0 1 1 0
4 0 0 1 0 1 1 1 0 1 1
5 0 0 1 1 0 1 1 1 1 0
6 0 0 0 1 1 1 1 0 0 1
7 0 1 0 1 0 1 0 1 1 0
8 1 1 1 0 1 0 0 1 1 1

International University
Vietnam National University HCMC
 Example 11 – Solution

Solving with SOLVER yields the following results:

International University
Vietnam National University HCMC
 Partial cover problem

• Locating a specified number of new facilities so that the number

of existing facilities that are covered is maximized.
• The partial cover problem can be formulated as follows:
𝑚

𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑧 = ෍ 𝑚𝑎𝑥𝑗 𝑎𝑖𝑗 𝑥𝑗

𝑖=1

subject to σ𝑛𝑗=1 𝑥𝑗 ≤ 𝑘 (12)

𝑥𝑗 = 0, 1 , 𝑗 = 1, … , 𝑛
• Constraint (12) ensures that at most, 𝑘 new facilities are available
for assignment to sites.
• We continue to use SOLVER to solve the problem.
International University
Vietnam National University HCMC
 Example 12 (1/2)

• A telecommunications company must decide where to locate cell

phone towers. The budget allows no more than three to be installed
in a particular city.
• The city has been divided into 10 zones and eight potential sites have
been identified for the cell phone towers. A survey has been conducted
to estimate how many cell phone customers will need to be served
within each zone.
• It is desired to locate the cell phone towers so the largest number of
customers are covered. The 𝑎𝑖𝑗 cover coefficients reflect the maximum
number of potential cell phone customers in zone 𝑖.
• Since the number of customers is independent of where the cell phone
towers are located, the cover coefficients are identical for all values of 𝑗
that provide coverage for a given zone.
• The following table shows a matrix of cover coefficients for the 10
zones and eight potential sites.
International University
Vietnam National University HCMC
 Example 12 (2/2)

Cell Phone Tower Site

Zone 𝒊
1 2 3 4 5 6 7 8
1 50 50 0 0 0 0 0 0
2 25 0 0 0 0 25 0 0
3 0 0 15 0 0 15 0 15
4 0 0 25 0 0 25 25 25
5 0 0 50 0 0 0 50 50
6 0 0 0 35 0 0 35 0
7 0 0 0 50 0 0 0 0
8 0 0 0 35 35 0 0 0
9 0 45 0 0 45 0 0 0
10 30
International University 30 0 0 30 0 0 0
Vietnam National University HCMC
 Example 12 – Solution

Solving with SOLVER yields the following results:

International University
Vietnam National University HCMC
Reference
1. Tompkins, James A., White, J. A., Bozer, Y. A.,
Tanchoco, J. M. A. Facilities Planning, 4th ed
(Chapter 10). Wiley 2010.
2. Francis, R.L., McGinnis, L.F. and White, J.A., Facility
layout and Location: An Analytical Approach, 2nd ed.,
Prientice-Hall, Englewood Cliffs, NJ, 1992.

International University
Vietnam National University HCMC