Network Models Part I

MSCI 603
Introduction
— Transportation, Assignment, and Transshipment models are
special types of LP problems
— They also belong to a class of problems called “Minimum Cost
Network Flow” problems

Assignment
Transportation Shortest Path

Transshipment Problems Max Flow

Minimum Cost Network Flow Problems

MSCI 603 [email protected]

2
Example 1: Transportation Problem
— A power company has three plants that can supply four customer
zones with electricity. The supply and demands, as well as the
cost of sending 1 million kwh of electricity from each plant to
each customer, are shown in the table below.
Customers
Plants 1 2 3 4 Supply
1 $8 $6 $10 $9 35
2 $9 $12 $13 $7 50
3 $14 $9 $16 $5 40
Demand 45 20 30 30

— Formulate an LP that will minimize the cost of supplying these

customers while meeting demand.
MSCI 603 [email protected]
3
Example 1: Transportation Problem

MSCI 603 [email protected]

4
General Transportation Formulation
— For an unbalanced problem:
m n
min z = åå cij xij
i =1 j =1
n
s.t. åx
j =1
ij £ si "i = 1,2,  , m

åx
i =1
ij ³ d j "j = 1,2,  , n

xij ³ 0 "i, j

— For a balanced problem, what changes do we make?

MSCI 603 [email protected]
5
Unbalanced Problems
— If supply > demand:
— Add a dummy demand node with demand = ås -åd
i
i
j
j
— Cost on that arc = 0

— If demand > supply:

— Add a dummy supply node with supply = åd -ås
j
j
i
i
— Cost on that arc = ?

MSCI 603 [email protected]

6
Important Concept
— If all the supplies and demands for a transportation problem
are integers, then the transportation problem will have an
optimal solution in which all the variables are integers.

MSCI 603 [email protected]

7
Overview of Transportation Simplex
— A variation of the simplex method that solves transportation
problems efficiently

— As with “regular” simplex, we must first find a basic

feasible solution using one of the following methods:
— Northwest corner method
— Minimum cost method

— We then improve the solution:

— Find the entering variable by pricing out non-basic variables
— Update the simplex tableau and repeat until an optimal solution
is obtained
MSCI 603 [email protected]
8
Example 2: Using Transportation Simplex
— Formulate as a transportation problem and solve it using the
transportation simplex method.

cij Customer 1 Customer 2 Customer 3 Customer 4 Supply

Plant 1 8 6 10 9 35
Plant 2 9 12 13 7 50
Plant 3 14 9 16 5 40
Demand 45 20 30 30

MSCI 603 [email protected]

9
Example 2: Find an initial bfs
— Northwest Corner Method
1. In the upper left (northwest) corner, set x11 = min{s1, d1}
— If x11 = s1:
— Cross out the first row
— Set s1 = 0 and change d1 to d1 – s1
— If x11 = d1 :
— Cross out the first column
— Set d1 = 0 and change s1 to s1 – d1
2. Repeat step one with the next most northwest corner that is
not in a crossed-out row or column
3. You will eventually come to a point where there is only one
cell that can be assigned a value equal to its row supply or
column demand, then cross out both that row and column.
MSCI 603 [email protected]
10
Example 2: Find an initial bfs

MSCI 603 [email protected]

11
Minimum-Cost Method for Finding a BFS

• The northwest corner method does not utilize shipping costs.

• It can yield an initial bfs that has a very high shipping cost.

Minimum-Cost Method
1. Find the variable with the smallest shipping cost (call it 𝑥!" ).

2. Then assign 𝑥!" its largest possible value, min{𝑠! , 𝑑" }.

3. As in the northwest corner method, cross out row i or column j and reduce the supply
or demand of the noncrossed-out row or column by the value of 𝑥!" .

4. Then choose from the cells that do not lie in a crossed-out row or column the cell
with the minimum shipping cost and repeat the procedure.

5. Continue until there is only one cell that can be chosen.

MSCI 603 [email protected]
12
Minimum-Cost Method for Finding a BFS

MSCI 603 [email protected]

13
Minimum-Cost Method for Finding a BFS

MSCI 603 [email protected]

14
Improve the solution
A. Pricing out NBVs (determine the entering variable)
— Let ui and vj be the dual variables for supply constraints i and
demand constraints j, where i = 1, 2, …, m and j = 1, 2, …, n
— Arbitrarily set u1 = 0
— We know that in any simplex tableau, basic variables have
reduced costs = 0, so cij = 0
— We also know have the relationship cij = ui + v j - cij , based on
information from dual variables (see the textbook for
reasoning)
— Finally, for a min problem, a bfs is optimal if all cij £ 0
— If this is not the case, then we enter into the basis the most positive cij

MSCI 603 [email protected]

15
Improve the solution

MSCI 603 [email protected]

16
What is a loop?
— An ordered sequence of at least 4 different cells where
— Any two consecutive cells lie in either the same row or
column;
— No three consecutive cells lie in the same row or column; and
— The last cell in the sequence has a row or column in common
with the first cell in the sequence.

MSCI 603 [email protected]

17
What is a loop?

Theorem:

MSCI 603 [email protected]

18
Improve the solution
B. Pivoting
— Find the loop involving the entering variable
— Counting only the cells in the loop, label those found above
“even” (“odd”) if they are an even (odd) number of cells
away from the entering variable.
— Find the odd cell whose variable value is the smallest. Call
this value θ and perform the pivot:
— Decrease the value of each odd cell by θ
— Increase the value of each even cell by θ

MSCI 603 [email protected]

19
Improve the solution
— If θ is zero, the entering variable will equal 0, and an odd
variable that has a current value of 0 will leave the basis. In
this case, a degenerate bfs existed before and will result after
the pivot.

— If more than one odd cell in the loop equals θ , you may
arbitrarily choose one of these odd cells to leave the basis;
again, a degenerate bfs will result.

— In NWC method if supply and demand are equal, then

randomly choose one of them (either eliminate row or
column). This results in a degenerate solution.

MSCI 603 [email protected]

20
Example 2: Improve the solution

MSCI 603 [email protected]

21
Summary:
Transportation Simplex Method
1. If the problem is unbalanced, balance it.
2. Find an initial bfs
3. Set u1 = 0 and determine ui and vj such that ui + vj = cij for all
basic variables
4. If ui + vj – cij ≤ 0 for all non-basic variables, the current solution
is optimal. If not, enter the variable with the most positive
reduced cost into the basis using the pivoting procedure. This
yields a new bfs.
5. Using the new bfs, return to Steps 3 and 4.

For a maximization problem, replace Step 4 with Step 4' below:

If ui + vj – cij ≥ 0 for all NBV, the current solution is
optimal. If not, enter the variable with the most negative
reduced cost into the basis using the pivoting procedure.
MSCI 603 [email protected]
22
Example 3: The Assignment Problem
— Three architects are asked to inspect three major projects.
Each Architect must be assigned to complete one project.
The time it takes each one of them to inspect each project is
given below:
Time Project 1 Project 2 Project 3
Adam 17 4 10
Betty 15 5 8
Chris 18 7 11

— Architects need to be assigned so that the total inspection

time is minimized.
MSCI 603 [email protected]
23
General Assignment Formulation
m n
min z = åå cij xij
i =1 j =1
n
s.t. åx
j =1
ij = 1 "i = 1,2,  , m

åx
i =1
ij = 1 "j = 1,2,  , n

xij = 0 or 1 "i, j

— An assignment problem is a balanced transportation

problem in which all supplies and demands are equal to 1.

MSCI 603 [email protected]

24
General Assignment Formulation
— All the supplies and demands for the assignment problem
are integers. Hence, all variables in the optimal solution
must be integers.

— We can ignore the restrictions that 𝑥!" ∈ {0,1} and solve the
assignment problem as a balanced transportation problem.

— The transportation simplex is inefficient since the problem is

highly degenerate.

MSCI 603 [email protected]

25
General Assignment Formulation

MSCI 603 [email protected]

26
Solving Assignment Problems: Hungarian Method
1. Find the minimum element in each row of the m x m cost matrix.
Create a new matrix by subtracting from each cost the minimum cost
in its row. In this new matrix, find the minimum element in each
column. Create a new matrix by subtracting from each cost the
minimum cost in its column. This is now the reduced cost matrix

2. Draw the minimum number of lines (horizontal, vertical, or both)

needed to cover all the zeros in the reduced cost matrix. If m lines are
required, an optimal solution is available among the covered zeros in
the matrix; stop. If fewer than m lines are needed, go to the next step.

3. Find the smallest nonzero element, k, in the reduced cost matrix that
is uncovered by the lines drawn in step 2. Subtract k from each
uncovered element in the matrix and add k to each element that is
covered by two lines. Return to step 2.

MSCI 603 [email protected]

27
Requirements for Hungarian Method
— The problem must be balanced
— If not, add dummy points with zero cost

— The objective function must be a minimization

— If not, multiply the profit matrix through by -1 and solve as a
minimization problem

MSCI 603 [email protected]

28
Example 4:
— Five employees are available to perform four jobs. The time
it takes each person to perform each job is given in the table
below. Determine the assignment of employees to jobs that
minimizes the total time required to perform the four jobs.

Time Job 1 Job 2 Job 3 Job 4

1 22 18 30 18
2 18 - 27 22
3 26 20 28 28
4 16 22 - 14
5 21 - 25 28

MSCI 603 [email protected]

29
Example 4:

MSCI 603 [email protected]

30
Example 4:

MSCI 603 [email protected]

31
Intuitive Justification of the Hungarian Method
— If a constant is added to each cost in a row (or column) of a
balanced transportation problem, then the optimal solution
to the problem is unchanged.
— Steps 1 and 3 create a new cost matrix having the same
optimal solution as the original problem. They also ensure
that all costs remain nonnegative.
— Any feasible assignment in which all the 𝑥!" ’s that equal 1
have zero costs must be optimal.
— When step 2 indicates that m lines are required to cover all
the zeros in the cost matrix, an optimal solution to the
original problem has been found.
MSCI 603 [email protected]
32