Integer Linear programming problem

(P) Max./ Min. c1x1  c2 x2 cn xn

subject to

a11x1  a12 x2  ...  a1n xn  ,  ,= b1

a21x1  a22 x2  ...  a2n xn  ,  ,= b2

a31x1  a32 x2  ...  a3n xn  ,  ,= b3 (1)

am1x1  am2 x2  ...  amn xn  ,  ,= bm

x1, x2 ,.., xn  0 and integers

Two methods:

Cutting plane method

Branch and bound method.

Cutting plane method:

Simplex method

Gomory’s constraint

Dual simplex method

P1: Solve the following integer LP problem

Minimize Z= 2x1+3x2

subject
 x1+x2 ≥6 ; 7x1+x2 ≥ 14 ;

x1, x2 ≥ 0 and integers.

Soln: Consider the following LP problem

Minimize Z= 2x1+3x2

subject

x1+x2 ≥6 ; 7x1+x2 ≥ 14; x1, x2 ≥ 0.

Now, consider the following maximize type LP problem for the given minimize type LP
problem:

Maximize P = -2x1-3x2

subject

x1+x2 ≥6 ; 7x1+x2 ≥ 14 ; x1, x2 ≥ 0.

This implies that

Maximize P = -2x1-3x2

subject

-x1-x2 ≤ -6 ; -7x1-x2 ≤ -14;

x1, x2 ≥ 0.

Now, the standard form:

Maximize P = -2x1-3x2

subject

-x1-x2 +x3 = -6 ;

-7x1-x2 +x4 = -14;

x1, x2 , x3, x4 ≥ 0

where x3 and x4 are slack variables.

x3 = -6 and x4= -14.

Initial table:

C -2 -3 0 0
CBV BV x1 x2 x3 x4 Soln Ratio
0 x3 -1 -1 1 0 -6 6
0 x4 -7 -1 0 1 -14 2
Zj-Cj 2 3 0 0 0
Ratio -2/7 -3 - -

Min.{ -14,-6} = -14. Max.{ -2/7, -3}= -2/7

Entering variable-x1 and leaving variable-x4

First iteration table:

C -2 -3 0 0
CBV BV x1 x2 x3 x4 Soln Ratio
0 x3 0 -6/7 1 -1/7 -4 28
-2 x1 1 1/7 0 -1/7 2
Zj-Cj 0 19/7 0 2/7
Ratio -19/6 -2

Entering variable-x4 and leaving variable-x3

Second iteration table

C -2 -3 0 0
CBV BV x1 x2 x3 x4 Soln Ratio
0 x4 0 -6 -7 1 28
-2 x1 1 1 -1 0 6
Zj-Cj 0 1 2 0 -12
Ratio

Solution to the LP problem:

x1=6 , x2=0 and Min.Z= - Max.P= 12.

integer LP problem is x1=6 , x2=0 and Min.Z= 12.

P2: Using cutting plane method, solve the following integer L P problem

Maximize Z= 5x1+8x2

subject to

x1+2x2≤ 8 ; 4x1+x2≤ 10, x1 , x2 ≥ 0 and integers.

Soln:

The LP problem corresponding to the given Integer LP problem.

Maximize Z= 5x1+8x2

subject to

x1+2x2≤ 8 ; 4x1+x2≤ 10, x1 , x2 ≥ 0 .

The Std. form of the LP problem corresponding to the given Integer LP problem.

Maximize Z= 5x1+8x2+0x3+0x4

subject to

x1+2x2 +x3 = 8 ;

4x1+x2+ x4 = 10,

x1 , x2, x3, x4 ≥ 0 .

where x3 and x4 are slack variables.

Now, the initial basic feasible solution:

x3= 8 and x4 =10 .

Initial table

C 5 8 0 0
CBV BV x1 x2 x3 x4 Soln Ratio
0 x3 1 2 1 0 8 8/2=4
0 x4 4 1 0 1 10 10/1=10
Zj-Cj -5 -8 0 0 0
Entering variable – x2 and Leaving variable –x3.

First iteration table

C 5 8 0 0
CBV BV x1 x2 x3 x4 Soln Ratio
8 x2 1/2 1 1/2 0 4 8
0 x4 7/2 0 -1/2 1 6 12/7
Zj-Cj -1 0 4 0 32

Entering variable – x1 and Leaving variable –x4

Second iteration table

C 5 8 0 0
CBV BV x1 x2 x3 x4 Soln Ratio
8 x2 0 1 4/7 -1/7 22/7
5 x1 1 0 -1/7 2/7 12/7
Zj-Cj 0 0 27/7 2/7 236/7

Since Zj-Cj ≥ 0, for all j, the current table is optimal. We stop the process.

Optimal solution soln:

x1=12/7 , x2=22/7 and Max. Z= 236/7.

Since x1 and x2 are not integer, it is not optimal solution to the given problem.

Now, x1 = 1+5/7

x2 = 3+1/7

Since fractional part of x1 is high, select x1.

Now, we have

x1+0x2+(-1/7)x3+(2/7)x4=12/7

x1+0x2+(-1+6/7)x3+(2/7)x4=1+5/7

Gomory’s constraint:

(6/7)x3+(2/7)x4 ≥ 5/7
That is, -(6/7)x3-(2/7)x4 ≤ -5/7

This implies that

-(6/7)x3-(2/7)x4 +x5 = -5/7

where x5 is a slack variable.

Adding the Gomory’s constraint to the optimal table

Initial table (DSM) for integer soln:

C 5 8 0 0 0
CBV BV x1 x2 x3 x4 x5 Soln
8 x2 0 1 4/7 -1/7 0 22/7
5 x1 1 0 -1/7 2/7 0 12/7
0 x5 0 0 -6/7 -2/7 1 -5/7 x
Zj-Cj 0 0 27/7 2/7 0 236/7
Ratio - - -27/6 -1 -

Entering variable-x4 and Leaving variable x5

First iteration table for integer soln:

C 5 8 0 0 0
CBV BV x1 x2 x3 x4 x5 Soln
8 x2 0 1 1 0 -1/2 7/2
5 x1 1 0 -1 0 1 1
0 x4 0 0 3 1 -7/2 5/2
Zj-Cj 0 0 3 0 66/2

Since x2 is not integer, it is not optimal solution to the given problem.

Now, x2 = 3+1/2

Now, we have

0x1+x2+x3+0x4+(-1/2)x5=7/2

0x1+x2+x3+ (-1+1/2)x5 =3+1/2

Gomory’s constraint:

(1/2)x5 ≥ 1/2

That is, -(1/2)x5 ≤ -1/2

-(1/2)x5 +x6 = -1/2

where x6 is a slack variable.

Adding the Gomory’s constraint to the optimal table :

Second Iteration table (DSM) for integer soln:

C 5 8 0 0 0 0
CBV BV x1 x2 x3 x4 x5 x6 Soln
8 x2 0 1 1 0 -1/2 0 7/2
5 x1 1 0 -1 0 1 0 1
0 x4 0 0 3 1 -7/2 0 5/2
0 x6 0 0 0 0 -1/2 1 -1/2 X
Zj-Cj 0 0 3 0 1 0 66/2
Ratio - - - - -2 -

Entering variable-x5 and Leaving variables-x6

Third iteration table:

C 5 8 0 0 0 0
CBV BV x1 x2 x3 x4 x5 x6 Soln
8 x2 0 1 1 0 0 -1 4
5 x1 1 0 -1 0 0 2 0
0 x4 0 0 3 1 0 -7 6
0 x5 0 0 0 0 1 -2 1
Zj-Cj 0 0 3 0 1 0 32

Since the values of x1 and x2 are integers, this is the optimal solution . We stop the computation.

The optimal solution of the given problem:

x1=0, x2 = 4 and Max.Z=32