You are on page 1of 23

Chapter 9 Integer programming I .

lntegerprogramming
-

④ Mathematical IMP )
programming
Max fest )

sit .
I t.SE/Rn

Constraint set / feasible region .

b- Ies .
I is a
-

I Jeo ES f- CIO ) f- CI > Je ES


if ,
sit .
> t ,

then Teo is an
-

① Lp is a
special case of MP .

Lis fins is linear . i.e .


fix > = ITE

Cii > S = { I l AE -

- I .
I > 5 Y

② LP S " "
: E 2x EIR

Mixed E
: at least one .
but not all
components of are

required to be
integers .

In this chapter ,
we 'll
only consider IP and MIP which can

be reduced to LP by dropping the integer requirement on I .


( ILP ) Max Z = I IMILP ) Max 2- = E'sit d-tu
St .
AE -
-
I set . CA . .
Ar )
( Ky ) -

-
I
£7,5 ,

tht £30 , iht


=

I 35 ,

The LP obtained by dropping the integer requirement


on I is called corresponding LP or LP relaxation .
TH0n

capitalbudgethgproblem.t
( I)

firm has an investment budget of $14 c thousand dollars !

Now there 4
, are investment projects :

Pi Pz Ps Pa

thdoisarmsd
'
cash
required 5 it 7 3
,

13 15
expected return 12 11 .

A project can be selected or not .


One can not select a

fraction of a
project .

Question :
How to place money to maximize the total expected
return ?

I
let selected
Kj = 1 .
if project j is .

+ 0 is selected
.

if project j not .

binary decision variable .

Max
#

Sit .
7

=
let 's extend the fundamental model
by considering more constraints :

4) At most 3 be selected
projects can .

⑦ At most one
of the first 3
projects can be selected .

(3) If P, is selected ,
then P2 should also be selected .

(4) If P, is selected ,
then Ps can not be selected .
Is Either Pz or Ps should be selected ,
but not both .

(6) If Ps is not selected ,


then Pa is not selected .
LI)Restictihgavai5abletotakeoh#feeralaues .

LA ) suppose k can take values from f 4 8 13T


only
on . , .

Exercise
N can
only take on values from l O , 4
.
8 . 13 } .

= .
(B) restricting to take on discontinuous values .

Z can take either 0 Or between lol u .

Exercise
K can take either K= a to or Lex su .
(TI) luorelogicalconstraihts

( At constraint feasibility .
:
fcshebisactiueoo.no
Assume the , is bounded .
:

I M ,
sit .
HIS E bt M VI .

f if
let y = O ,
ten > Eb is known to be active .

1 . otherwise .

if y o

\
-

-
.
-

if y I
-
-
.
-

fins E

\ y=o or
-

1 .
(B) Alternative constraint .

Consider a situation : H E 2 or a > 6 .


Generally speaking , consider an alternative constraint :

f Bes Eb ①
{
, ,

facts Ebz ②


if either ① or is
satisfied ,
but not
necessarily both ,

we can .

/
W, = O ,
if f. Fn > Eb ,
is active

1 .
otherwise .

=/
>

Wa 0 ,
if false > Ebz is active

1 .
otherwise ,

then ,
the alternative constraint can be modeled as :

f- ( Ze bit

I
, ) E 141W ,

f- z ( Ji , E bat 14h02

÷: ?
'

. .. . .

if exactly one constraint is satisfied ,


then .
③ should be

W, t wz = I .
Wz = I -
W,

I
, cus I bit Mw ,

f- z Cn > E bz t M C l -
Wi )

w. on .
(C) Conditional constraint .
:
ficxssb.imphiesfzcsesebz.LT
if 22 , -1322 > 2 then 522 -
Tks E3

This satisfied
implication is not .

only when

2K , -132k 72 d 522 -
THS > 3 .

A B .

By De Morgan 's law ,

ATB =
At UB .

original constraint is equivalent as

2Mt 3. Nz E 2 .
and / or 52 , -7ns E 3 .

( at least one
of them is active ( satisfied ) ) .

.
?
(D) k -

fold alternative :

at least be of set
of constraints should be
suppose a

satisfied .

fick ) E bi 5=1,2 .
-
i '

, m .

>
Assume .
: ficus E bi t Mi t i

{ / satisfied
let wi = o ,
if ith constraint should be active .

1 . otherwise

then .
?
(IV) Nonhhearfuuctions

(A.) fixed cost problem .

In 3 different machines need to be rented


a
production problem ,

to produce 3 products ( P. . Pz ,
Ps ) .

P, Pz Ps limitation ,

machine rent
( per week ) $200 $150 $100

unit labor requirement .


3 hrs 2 hrs Cohrs 150 hrs per week

materials 4 units 3 units 4 units 160 units perweek .

unit cost $6 $4 $8

$12 $8 $15
selling price .

Question : Formulate it as an IP so as to Max the total


weekly

profit .

let
Kj be number
of Pj produced in each week .

j't .
2.3 .

Obj fun :
Max weekly profit
=
total revenue - total cost .

total revenue I
#

total cost =
production cost t machine rent .


P, Pz Ps
machine rent
( per week ) $200 $150 $100
( tj )

f Pj
let O
if
yj = ,
is not produced .

1 .
otherwise .

rent
of machine
j
f O
if bj O
=
-

-
,

tj , if yj =L .

weekly machine rent =


-

obj fun : ( 1224+822+1523 ) -


close , -142kt 8ns )

( 150 Yz
200g 10093 )
-

,
t t

= 6k , + GHz +723 -

200g ,
-

15092 -
10093 .

P , Ps Ps limitation
unit labor requirement 3 hrs 2 hrs Ghrs 150 hrs week
per
.

materials Quarts 3 units aunts 160 units perweek .

IP model :

Max 6k , + GHz -1723 -

200g ,
-

15092 -
10093 .

St . 3K , +222+623 E- 150 ( labor )

4K , t 322 +42cg E 160 ( material )

Ni ,
Nz . 23 30 .
Tht .

Yi O or 1 i =L 2 3
J J
.
.
, ,

Question : lls it a correct model ?


We should have that
if y =o then so =o
-
-
.
.

j j

i.e if y; = I
Kj
30

|
. ,

if yj =o, Kj =o

to bounded from
model this ,
assume
Kj is above .
i.e .
OE
Kj
E
Mj .

Then ,

if y; =L
Kj
30

{ Eng Mj
E
.
o .

if yj =o, Kj =o

IP model :

Max 6k , + GHz -1723 -

200g ,
-

15092 -
10093 .

St . 3k , +222+623 E- 150 ( labor )

4K , t 322 +42cg E 160 ( material )

O E
Nj E
Mj Yj 5- i. 2.3
Ni ,
Nz . 23 30 .
Tht .

Yg . o or 1 ,
g- =L . 2 ,
3 .

"

µ
In general , fixed cost

f
T
ten >
O
=
ft da ,
no

O K = 0
,

( OER EU )
(d > O )

f
let y = l
,
if a>o

O
, if x=o

then .

=/
tea > Cida
if y I
-

-
,

0 ,
if y=o .

fix, Cy talk

to:: : :
=
LB ) piecewise linear curve .

Define Si Ii -_ i. 2.3 ) as the

(1) amount of k lies in the ith

interval .

2=8 , -182 83

\
t

OE S, E

o es ,

O E 83 E-

y= -

Question Is the formulation ?


:

complete

We should also have that ,

Ss > 0 81=4
if then ①

/
,

if 83>0 . then 82=6 ③

for ① if 8220 S =4

/ if
,
,
,

82=0 .
OES .
a- 4

let y ,
=

f l ,
if Sz >O

O . otherwise .

for ② if Is so 82=6

{
, ,

83=0 0£82 Eb
if .

=/ if
let ya I ,
8370

0 , otherwise .
82=6
if

{
.yz=l ,
y, =L , .

it 92--0.9 0£82 EG

;
'

if yz=0 o s. o
-
- .
. ,

As for 83 ,

8370 83
if E5

/
Of
,
or
ya > I .

if 83=0 ,
or
yz=o ,
83=0

linear model is :

2=8 -182+83 9=58 -1821-383

{
,
,
,

4g ,
E S , Ed byz Esa Eby ,
O ES 's £592 .

, ,

Y ,
Fo or 1, ya = o or 1 .

verify :

Y ,
=
ya =D ;

Y , =L , 92=0 .

Y, =D , yz= I

y , =
42=1
In general ,
N linear segments .
( piecewise linear function )

Lj :
length of jth linear Segment .

for jth linear


segment : (
j=i ,
- . .

. N - i ) .

=/ if exceeds the jth interval


Gj I ,
U .

O . otherwise .

Then L, S, Li

{
1St E E
y
:
, ,

2nd - N ith
-
:
Lj Yj E
Sj E
Lj Yj -
i

Nth : O E Sw E LN YN -
i .
(2) 2K ,
OER E3

9=1 9

-5in
-
K ,

,
4EkE7

geneal

if
( if
W, = I OEK E3 se , K OEM E3

(
=
, ,

o , otherwise o , otherwise

=/ if if
I GEN ET se GENET

{
wz , sea =
,

o . otherwise 0 .
otherwise

=/
if 8EkE9 Us if 8Eks9
{
↳ = I ,
se ,

o, otherwise o , otherwise .

ze= #

Question : How to model y ?

When EN E3 Wi= I y 2K ,

{
O or ,
=

4 EN ET or wz=i y =9 Nz
-

8Ek±9 or Ws =L .
9=-5+23

y= #
Question : How to model K . .
Na .
Us ?

for ki :

if he , = I OER , E3

{
,

it w , =o se , =o -

for as :

if ET
|
wz =/ ,
HE ka

if
-

wz=o . see =o

for x3 .

if Ws 8EK3E9
{
=L .

if wz =o , 23=0

Also ,
exactly one
of these 3 Wi 's is 1 .

in general ,

C 9Wz C

{
N= Kitkat Kz , y -_ 22 , t -
Kz > t -

5h03 -12cg )

803 9h03
I Tfw :c
on ' ' 3W ' ' " w- ' as ⇐

bit wz t WS = , . , .
Exercise

(
0 ,
a = o

ton =

1 ,
LEK EZ

2 ,
z ex EQ

You might also like