Operations Research

Special Cases of Linear Programming

Amjad Ali
References

1. “Introduction to Operations Research” by Frederick S. Hillier

2. “Operations Research: An Introduction“ by Hamdy A. Taha

3. Video Lectures by Prof G. Srinivasan, IIT Madras

Linear Programming

Special Cases of Linear Programming

Degeneracy Alternative Unbounded Infeasible

Optima Solution Solution
Linear Programming

Special Cases of Linear Programming

1. Degeneracy
2. Alternative Optima
3. Unbounded Solution
4. Infeasible Solution
ALTERNATIVE OPTIMA - Special Case of Linear Programming
EXAMPLE: Solve the following LP Model

Max Z = 2 x1 + 4 x2
Subject to
x1 + 2 x2 ≤ 5
x1 + x2 ≤ 4

x1, x2 ≥ 0
ALTERNATIVE OPTIMA - Special Case of Linear Programming
STANDARD FORM
STANDARD FORM

Max Z = 2 x1 + 4 x2 Z - 2 x1 - 4 x2 = 0
Subject to Subject to

x1 + 2 x2 ≤ 5 x1 + 2 x2 + s1 =5
x1 + x2 ≤ 4 x1 + x2 + s2 = 4
x1, x2 ≥ 0 x1, x2, s1, s2 ≥ 0
ALTERNATIVE OPTIMA - Special Case of Linear Programming
Negative Coefficients

Basic x1 x2 s1 s2 Sol
Z -2 -4 0 0 0
s1 1 2 1 0 5
s2 1 1 0 1 4 Z - 2 x 1 - 4 x2 = 0
x1 + 2 x2 + s1 =5
x1 + x2 + s2 = 4
x1, x2, s2, s2 ≥ 0
ALTERNATIVE OPTIMA - Special Case of Linear Programming
Pivot Column

Basic x1 x2 s1 s2 Sol Ratio

Z -2 -4 0 0 0
s1 1 2 1 0 5 5/2 = 2.5 min ratio
s2 1 1 0 1 4 4/1 = 4
ALTERNATIVE OPTIMA - Special Case of Linear Programming
Pivot Column

Basic x1 x2 s1 s2 Sol Ratio

Z -2 -4 0 0 0
s1 1 2 1 0 5 5/2 = 2.5 min ratio
s2 1 1 0 1 4 4/1 = 4 Alternative
Z 0 0 2 0 10 OBSERVATION Optima
x2 0.5 1 0.5 0 2.5 Optimal Coefficients of non-basic variable in Z Row
s2 0.5 0 -0.5 1 1.5 Solution should be non-zero

But here x1 coefficient is 0

Infinite Solutions
x1 can enter to find another optimal solution
ALTERNATIVE OPTIMA - Special Case of Linear Programming

Basic x1 x2 s1 s2 Sol Ratio Max Z = 2 x1 + 4 x2

Z -2 -4 0 0 0
s1 5 If we again enter non-basic s2 and
1 2 1 0 perform computation. What will
s2 1 1 0 1 4 happen?
Z 0 0 2 0 10
x2 0.5 1 0.5 0 2.5 2.5/0.5 = 5 x1 = 0
s2 0.5 0 -0.5 1 1.5 1.5/0.5 = 3 x2 = 2.5 Z = 10
Z 0 0 2 0 10
x1 = 3
x2 0 1 1 -1 1 2nd Optimal x2 = 1 Z = 10
x1 1 0 -1 2 3 Solution
ALTERNATIVE OPTIMA - Special Case of Linear Programming

Basic x1 x2 s1 s2 Sol
Z -2 -4 0 0 0 The Simplex Method determines only
s1 1 2 1 0 5
two optimal solutions
s2 1 1 0 1 4
0 0 2 0 instead of
Z 10
x2 0.5 1 0.5 0 2.5 infinite solutions.
s2 0.5 0 -0.5 1 1.5 As this is a two-variable problem,
Z 0 0 2 0 10 You can solve it Graphically. Now we
see, how can we detect alternative
x2 0 1 1 -1 1 optima in Graphical Method and
x1 1 0 -1 2 3 compare it with Algebraic Method.
ALTERNATIVE OPTIMA - Special Case of Linear Programming
Max Z = 2 x1 + 4 x2
Basic x1 x2 s1 s2 Sol Subject to
x1 + 2 x2 ≤ 5
Z -2 -4 0 0 0
x1 + x2 ≤ 4
s1 1 2 1 0 5 x1, x 2 ≥ 0
Any point on the line segment
s2 1 1 0 1 4 BC represents an Alternative
Z 0 0 2 0 10 optimum, with the same
objective value Z=10.
x2 0.5 1 0.5 0 2.5
s2 0.5 0 -0.5 1 1.5
B
Z 0 0 2 0 10
x2 0 1 1 -1 1
C
x1 1 0 -1 2 3 A D
ALTERNATIVE OPTIMA - Special Case of Linear Programming
We can easily find all the points on the line
Basic x1 x2 s1 s2 Sol segment BC from the graph.
But how can we determine all the points when
Z -2 -4 0 0 0 we solve the problem using Algebraic version
s1 1 2 1 0 5 of Simplex Method?
s2 1 1 0 1 4
Z 0 0 2 0 10
x2 0.5 1 0.5 0 2.5
s2 0.5 0 -0.5 1 1.5
B
Z 0 0 2 0 10
x2 0 1 1 -1 1
C
x1 1 0 -1 2 3 A D
ALTERNATIVE OPTIMA - Special Case of Linear Programming
Using Simplex Method algebraically,
Basic x1 x2 s1 s2 Sol we found two solutions.
Z -2 -4 0 0 0 One is at point B
and
s1 1 2 1 0 5
the other is at point C
s2 1 1 0 1 4
Z 0 0 2 0 10
x2 0.5 1 0.5 0 2.5
s2 0.5 0 -0.5 1 1.5
B
Z 0 0 2 0 10
x2 0 1 1 -1 1
C
x1 1 0 -1 2 3 A D
ALTERNATIVE OPTIMA - Special Case of Linear Programming

The Simplex Method determines

B
only the two corner points ( B and C in this case)
C
To determine all the points on the line segment BC, we use A D

weighted average (non-negative) of the points B and C.

ALTERNATIVE OPTIMA - Special Case of Linear Programming
Weighted Average Method
All points inside BC = fraction ( B ) + fraction ( C )
= α ( B ) + (1- α) ( C )
= α ( 0, 2.5 ) + (1- α) ( 3, 1 )
= ( 0 α , 2.5 α ) + (3 (1- α) , 1 (1- α) )
= ( 0 , 2.5 α ) + ( 3 - 3 α , 1- α )
= ( 0 + 3 - 3 α , 2.5 α + 1- α )
= (3-3α , 1.5 α + 1 )
= ( (3 - 3 α) , (1.5 α + 1) )
ALTERNATIVE OPTIMA - Special Case of Linear Programming
Weighted Average Method All points = ( (3 - 3 α) , (1.5 α + 1) )
Taking α = 0
Avg vs Weighted Avg. = ( 3 , 1 ) - Point C in the graph
Taking α = 1
= ( 0 , 2.5 ) - Point B in the graph

Taking α = 1/2
= ( 1.5 , 1.75 ) - Mid Point

You can calculate all optimal solutions by using all values of

α ( where 0 ≤ α ≤ 1)
ALTERNATIVE OPTIMA - Special Case of Linear Programming
In practice,
Alternative optima are useful because they allow to choose from many solutions without
affecting the objective value.

In the above example, the solution at B shows that activity 2 (or product 2 in case of Product-Mix
problem) only is at a positive level, whereas at C both activities (product 1 and 2 in case of Product-
Mix) are positive. It may be advantageous from the standpoint of sales competition to produce two
products rather than one. In this case, the solution at C is recommended.

NOTE:
You also have the option to choose from infinite optimal solutions whichever you feel the best for your
organization at that moment.