RI

FTVALLEYUNI
VERSI
TY

BALEROBECAMPUS
Col
l
egeofBusi
nessandEconomi
cs
Depar
tmentofBusi
nessManagement
Four
thy
earf
ir
stsemest
er
Gr
oupAssi
gnment
Cour
se:
-Oper
ati
onsResear
ch
Pr
epar
edby
:gr
oup
Name I
d.No

1.YonaSey
oum …………………….
.. 0086/
13
2.Habt
amuAssef
a………………… 0087/
13
3.Mi
ft
aAbda…………………………….0088/
13
4.Ay
ubAbdel
a…………….
……………. 0016/
13
5.Ashr
ekaHussei
n………………… 0012/
13
6.Tedel
echNugusu…………………… 0065/
13

Submi
tTo:
-Abdur
o.A
Submi
ssi
ondat
e:-09/
02/
2016E.
C

Bal
eRobeEt
hiopi
a
Oper
ati
onResear
chGr
oupAssi
gnment
Par
tone
Model
sandModel
i
ngi
nOper
ati
onsResear
ch
Bothsimpl eandcompl exsy st
emscaneasi l
ybest udiedbyconcentr
ati
ngonsomepor tionor
keyfeat uresinsteadofconcent r
atingoneverydetailofit
.Thisappr
oximati
onorabstract ion,
maintainingonl ytheessent ialelementsofthesystem,whi chmaybeconst ruct
edinv arious
for
msbyest abli
shingrelati
onshipsamongspecifi
edv ari
ablesandpar
ametersofthesystem, i
s
cal
ledamodel .Ingener al
,model satt
empttodescribetheessenceofasituati
onoractivit
yby
abstr
act i
ngf rom real
itysot hatthedecisi
on-makercanst udytherel
ati
onshipamongr elevant
vari
ablesi nisol
ation.

Model sdonot ,andcannot ,repr

esenteveryaspectofr eali
tybecauseoft hei nnumer abl
eand
changingchar acteri
sti
csoft hereal
-l
if
epr oblemst ober epr
esented.Inst
ead,t heyar eli
mited
approximationofr eali
ty.Forexample,tostudythef l
owofmat eri
alsthroughaf actory,ascaled
diagram onpapershowi ngthefactor
yf l
oor,positi
onofequipment ,tool
s,andwor kerscanbe
construct
ed.Itwoul dnotbenecessar ytogi vesuchdetail
sast hecolorofmachi nes, t
heheight
oft hewor kers,ort het emperat
ureoft hebui l ng.Foramodelt
di obeef fecti
ve,itmustbe
r
epr
esent
ati
veoft
hoseaspect
sofr
eal
i
tyt
hatar
ebei
ngi
nvest
igat
edandhaveamaj
ori
mpacton
t
hedeci
si
onsit
uat
ion.
A modeli s constr
ucted to analyze and underst
and the gi
ven system f
orthe purpose of
i
mpr ovi
ngitsperfor
mance.Ther eli
abil
i
tyofthesoluti
onobtainedfrom amodeldependsont he
val
idi
tyofthemodeli nr epr
esentingthesystem underst
udy.Amodel , al
l
owstheopportuni
tyto
examinethebehav i
oralchangesofasy stem wit
houtdist
urbi
ngt heon-goi
ngoper
ati
ons.

1.Cl
assi
fi
cat
ionofORModel
ORmodeli sanabstractrepresent
ati
onofanexi
sti
ngprobl
em si
tuat
ion.Itcanbei
ntheform of
agraphorchar
t,butmostf requent
lyanORmodelconsi
stofasetofmat hemati
cal
rel
ati
onshi
ps.
Thesemathematical
relat
ionshipsaremadeupofnumbersandsymbols.

Ther
ear
emanyway
stocl
assi
fymodel
s:

)Cl
i assi
fi
cat
ionbasedonst
ruct
ure
a)Phy
sicalModels:Thesemodel
spr ov
ideaphy si
calappear
anceoftherealobj
ectunderstudy
ei
therreducedinsizeorscaledup.Physicalmodelsareusefulonl
yi ndesi
gnpr obl
ems
because theyare easyto obser
ve,bui
ld,and descri
be.Since t
hese models can not
mani
pul
atedandar
enotv
eryusef
ulf
orpr
edi
cti
on,pr
obl
emssuchaspor
tfol
i
oanal
ysi
ssel
ect
ion,
 medi
asel
ect
ion,
product
ionschedul
i
ng,
etccannotbeanal
yzedbyphy
sical
model
s.

b)Sy
mbol i
cmodel
s:Thesemodelsusesymbols(l
ett
ers,number
s)andfunct
ionst
orepr
esent
var
iabl
esandt
hei
rrel
ati
onshi
pstodescr
ibet
hepropert
iesoft
hesyst
em.

)Cl
i
i assi
fi
cat
ionbasedonf
unct
ionorpur
pose
Model
sbasedont
hepur
poseoft
hei
rut
il
it
yincl
ude:

a)Descript
ivemodels:Descri
pti
vemodel ssimplydescri
besomeaspectsofasituat
ion,based
onobser v
ati
on,
survey,quest
ionnair
eresul
tsorotheravai
labl
edataofasi
tuat
ionanddonot
recommendany t
hing.Example:Organi
zati
onalchar
t,pl
antlay
outdi
agram,
etc.

b)Pr
edict i
veModels:Thesemodel sindicate―Ifthisoccur
s,thenthatfol
low‖.Theyr el
ate
dependentandi ndependentvar
iablesandpermittryi
ngout,―whatif‖questi
ons.Inother
words,t hesemodelsareusedtopredictout
comesduet oagivensetofalt
ernati
vesforthe
problem.Thesemodel sdonothav eanobjecti
vefunct
ionasapartofthemodeltoev al
uate
decisionalter
nat
ives.

Forexample,S=a+bA+cIisamodelt hatdescri
beshowthesal
es(S)ofaproductchangesi
n
adver
ti
singexpendi
tur
es(
A)anddisposalpersonali
ncome(I
).Her
e,a,b,andcareparameter
s
whosevaluesmustbeest
imat
ed.

c)Normat i
ve(Optimizati
on)models:Thesemodel sprovi
det he―best ‖or―Opt i
mal ‖sol
utionto
probl
emssubj ectto cert
ainlimitati
onsont heuseofr esources.Thesemodel spr ov
ide
recommendedcour sesofact i
on.Forexampl e,inmat hemat i
calpr ogrammi ng,model sare
formulatedforoptimizi
ngthegivenobj ect
ivefunct
ion,subjecttor estr
icti
onsonr esourcesin
thecontextofthepr obl
em underconsiderati
onandnon- negativi
tyofv ari
ables.Thesemodel s
arealsocaledpr
l escri
pti
vemodel s,becausetheyprescri
bewhatt hedeci si
onmakeroughtt o
do.

i
i
i)Cl
assi
fi
cat
ionBasedonTi
meRef
erence
a) Stati
cModels:St
ati
cmodel
srepresentasystem atsomespecif
iedt
imeanddonotaccount
forchangesoverti
me.Forexample,aninventorymodelcanbedev el
opedandsol
vedt o
determi
neaneconomicorderquantit
yfort henextper i
odassumi
ngt hatt
hedemandi n
planni
ngperi
odwouldr
emainthesameast hatfortoday
.

b)Dy namicmodels:I
ndy namicmodel
s,ti
mei sconsi
deredasoneoft hevari
ablesandall
ows
theimpactofchangesduetochangeinti
me.Thus,sequencesofi
nter
rel
ateddecisi
onsover
aper i
odoft imearemadet oselecttheopti
malcour seofacti
ont ooptimizethegiv
en
object
ive.Dy
namicprogrammingi
sanexampleofady namicmodel.

i
v)Cl
assi
fi
cat
ionbasedonDegr
eeofcer
tai
nty
a) Determi
nist
icModel s:I fal
lthe parameters,constant s and f uncti
onalrelat
ionshi
ps are
assumedt obeknownwi t
hcertai
ntywhent hedecisioni smade,t hemodeli ssai dtobe
determi
nist
ic.Thus,insuchacase,t heout comeassoci atedwi t
hapar t
icul
arcourseof
acti
onisknown.Thati s,f
oraspecificsetofi nputvalues,t hereisauni quely
,determined
outputwhichr epr
esentsthesoluti
onoft hemodelunderconsi derat
ionofcer tai
nty
.The
resul
ts oft he model
s assume si
ngl
ev alue.Linearpr ogr
ammi ng models are exampl
es of
det
ermi
nist
icmodel
s.
b)Probabi l
isti
c( Stochast i
c)model s:Model si n whi
ch atleastone par ameterordeci sion
vari
abl ei sar andom v ariabl
ear ecalledprobabil
i
sti
c( orst ochasti
c).Sinceatl eastone
decisionv ari
ablei sr andom,adependentv ari
abl
ewhichi st hef unctionofi ndependent
vari
abl e(s)willalsober andom.Thi smeansconsequencesorpayof fduet ocert
ainchanges
inthei ndependentv ari
abl ecannotbepr oducedwithcertainty.Howev er,iti
spossi bl
et o
predictapat ternofv aluesofbot ht hevari
ablesbytheirprobabili
tydistri
buti
on.Insurance
againstr i
skoff i
re,acci dents,si
ckness,etc.areexampleswher ethepat ternofev entsis
studiedi nthef orm ofapr obabil
it
ydistr
ibuti
on.

)Cl
v assi
fi
cat
ionBasedonMet
hodofsol
uti
onorQuant
if
icat
ion
a)Heur
ist
icModel
:Thesemodel
sempl
oysomeset
sofrul
eswhich,t
houghper
hapsnotopt
imal
,
dofaci
li
tat
esol
uti
onsofpr
obl
emswhenappl
i
edinaconsi
stentmanner
.

b)Anal
yticalModel
s:Thesemodelshav easpecificmathematicalst
ruct
ureandthuscanbe
solvedbyknownanal yti
calormat hemat
icaltechni
ques.Anyopt i
mizat
ionmodel(whi
ch
requir
esmaximizat
ionormini
mizat
ionofanobjecti
vefunct
ion)isananal
yti
cal
model.

Si
c) mulati
on Model s:Thesemodel shavea mat hematicalst
ructur
ebutar enotsol ved by
applyi
ngmat hemat i
calt
echniquestogetasolut
ion.I
nstead,asimulat
ionmodelisessent
iall
y
acomput er-
assistedexperimentat
iononamat hemati
calstruct
ureofar eal
-l
if
eproblem in
ordertodescri
beandev aluatei
tsbehavi
orundercert
ainassumptionsoveraperi
odoftime.

Simulati
onmodelsar emoreflexibl
ethanmat hemati
calonesandt heref
ore,canbeusedt o
repr
esentacompl exsystem whichotherwisecannotberepresent
edmat hemat
ical
l
y.These
model sdonotpr
ovidegener
alsoluti
onl
ikethoseofmathemati
calModels.
2.Adv
ant
ageofModel
s
Model
singeneralareusedasanai
dforanal
yzi
ngcompl
expr
obl
ems.Howev
er,amodelcan
al
soser
veotherpurposesas:

i
) Amodelpr ovi
deseconomyi nrepr
esentati
onofthereali
ti
esofthesy stem.Thatis,models
helpdeci si
onmaker stovi
suali
zeasy stem sothathe/shecanunder standt hesystem ‘
s
structureoroperat
ioninabett
erany.Forexampl
e,iteasi
ertorepresentaf act
oryl
ayouton
papert han to const
ructi
t.Itischeapert otr
youtmodi fi
cati
onsofsuch sy stemsby
rearrangementonpaper.

i
i
)The pr obl
em can be viewed i
nits ent
ir
ety
,wi t
h allthe components bei
ng considered
si
multaneousl
y.
i
i
i)Modelsserveasai dst
ot r
ansmiti
deasandv i
sual
izat
ionamongpeopl eintheorganizati
on.
Forexample,processchartcanhel
pt hemanagementt ocommuni cateaboutbetterwor k
methodstowor ker
s.

i
v)Amodelall
owsust oanalyzeandexperi
menti
nacompl exsi
tuat
iontoadegreethatwould
beimpossi
blei
ntheactualsyst
em andit
senvi
ronment.Forexample,
theexper
imentalf
ir
ing
ofsat
ell
i
temaybecostl
yandr equi
reyear
sofpr
eparat
ion.

v
)Model
ssi
mpl
i
fyhei
nvest
igat
ionconsi
der
abl
yandpr
ovi
deapower
ful
andf
lexi
ble

3.Met
hodol
ogyofOper
ati
onsResear
ch
Foreffecti
veuseofORtechniques,iti
sessent
ialtofol
l
ow somest epsthatarehelpf
ulfor
deci
sion-maker
stomakebet
tersoluti
on.Thef
lowdiagr
am repr
esent
ingthemethodol
ogyofOR
i
sshownas:

Sy
stem ModelBui
l
ding
Abst
ract
ion Choosinga Logi
c
part
icul
ar -Est
abli
shrel
ationshi
psamong
Real aspectof vari
abl
esandpar ameter
soft
he
Wor l
d r
eali
tywhich syst
em.
probl
em - Def i
neobjecti
vestobeachi
eved

Sol
vet
heModel
Appl
ysui
tabl
eORtechni
quetogetsol
uti
on
i
nter
msofdecisi
onvari
abl
es.

Modi
fyt
heModel Test
ingt
heModelandi
tssol
uti
on

- Putval
uesofdecisionvari
ablesi
n
themodelunderconsider
ation
- Seewhethersol
utioninval
idornot
Not

I
mpl
ement
ati
onandCont
rol

- I
nterpretsol
uti
onvalues
- Puttheknowledge(result) Accept
abl
gai
nedf ormthesolutionto e
workt hr
oughorganizati
onal
pol
icies

St
ep1.Obser
vat
ionanddef
ini
ngapr
obl
em
The f ir
stst ep in OR pr ocess i st he i denti
fi
cat i
on ofa pr oblem t hatexi stsi s a sy st
em
(organizat i
on).Thesy stem mustbecont inuouslyandcl oselyobservedsot hatproblemscanbe
i
dent ifi
edassoonast heyoccurorant icipated.Problemsar enotal wayst heresultsofcr isi
s;but
i
nst eadf requentlyinv olveanant icipatoryorpl anningsit
uation.Oncei thasbeendet ermined.
thatapr oblem exist s,thepr oblem mustbecl earlyandconci sel
ydef ined.Becausei mpr operl
y
definingapr oblem caneasi l
yr esultinnosol uti
onorani nappropr i
atesol ut
ion.Si ncet he
existenceofapr obl em impl i
est hatt heobj ecti
vesoft hefir
m ar enotbei ngmeti nsomeway ,
the
goal s(objectivesoft heor ganization)mustal sobecl earl
ydef i
ned.

St
ep2.For
mul
ati
ngamodel
Modelf ormul
ati
oninvol
vesananal y
si softhesystem underst
udy,det
ermini
ngobj
ect
iveofthe
decision-
maker,andalt
ernati
vecour seofact i
on,etc.,soast ounderst
andanddescri
be,in
preciseter
ms,theprobl
em thatanorganizat
ionfaces.
Themaj
orst
epswhi
chhav
etobet
akeni
ntoconsi
der
ati
onf
orf
ormul
ati
ngt
hemodel
are:

 Probl
em Component s.Thef i
r stcomponentoft hepr oblem tobedef i
nedi sthedeci si
on
makerwhoi snotsatisfi
edwi tht heexist
ingst ateofaffairs.Thei nteracti
onwi ththedeci
sion
makerwill
helptheORspeci alistinknowi nghis/herobject i
ves.Thati s,eit
herhe/ shehas
al
readyobtai
nedsomesol utionoft hepr oblem andwant storetaini t
, orhewant stoimprove
i
ttoahigherdegree.Ifthedeci sionmakerhasconf l
icti
ngmul ti
pleobj ecti
ves,he/shemaybe
advi
sedtoranktheobj ecti
vesi nt heorderofpr efer
ence; overlappingobj ecti
vesmaybe
el
iminat
ed.

 Deci
sionenv
ironment

Iti
sdesi r
abletoknow aboutther esour
cessuchasmanager s,employ
ee’sequipment
’s,etc.
whicharer equir
edt ocarr
youtt hepol i
ciesoft heorganizat
ionconsi
deri
ngt hesoci
aland
ecologi
calenvi
ronmentinwhichtheorganizati
onfunct
ions.Knowledgeofsuchfactor
swillhel
p
i
nmodi fyi
ngtheinit
ial
setofdeci
sion-maker‘sobj
ecti
ves.

 Al
ter
nat
ivecour
sesofAct
ion

Thepr obl
em ar i
sesonlywhent herearesever
alcour
sesofact i
onav ai
labl
ef orasol
uti
on.An
exhausti
velistofcourseofacti
oncanbepr epar
edinprocessofgoingthroughtheabovest
eps
offormulati
ngt heproblem.Coursesofact
ionwhichar
enotf easi
blewithrespectt
oobject
ives
andresourcesmayber uledout
.

 Measur
eofef
fect
iveness

Acer t
ainmeasureofeff
ecti
venessorperformancei
sr equiredinordert
oeval
uat
et hemer i
tof
theseveralcour
sesofacti
on.Theper f
ormanceoreffectivenesscanbemeasuredindiffer
ent
unit
ssuchasbi r
r(netpr
ofi
ts)
,percent
age( shar
eofmar ketdesir
ed),t
imedi
mension(ser
viceor
waiti
ngtime).

 Col
l
ect
ingDat
aandConst
ruct
ingaMat
hemat
ical
Model

Aftertheprobl
em i
sclearl
ydefinedandunderstood,t
henextstepi
stocol
lectrequi
reddat
aand
then f ormulate a mat hemat ical model . Modelconst ruction consi sts of
hypot hesizi
ng
rel
ationshi
psbetweenvari
ablessubjecttoandnotsubjectt
ocontr
olbydecisi
on-maker
.Cert
ain
basiccomponentsrequi
redinev er
ydecisi
onproblem modelar
e:

o Control
lable(deci
sion)Vari
ables-Thesearethei
ssuesorf actor
sintheproblem whose
val
uesar etobedet ermined(inthef
orm ofnumeri
calvalues)bysolvi
ngthemodel .The
possibl
ev aluesassignedtothesevar
iabl
esarecal
leddecisionalt
ernati
ves(str
ategiesor
coursesofact i
ons).Example,inLPPthenumberofunitsproducedisadecisionv ari
abl
e.

o Uncontr
oll
ablevar
iabl
e.Thesearethefact
orswhosenumer i
calval
uedependsupont he
ext
ernalenvi
ronmentprevai
l
ingaroundtheorgani
zat
ion.Theval
uesofthesevari
abl
esare
notunderthecont
rolofthedeci
sion-
makerandarealsotermedasstat
eofnat ur
e.

 Obj
ect
ivef
unct
ion.

Itisar epr
esent
ati
onof( i
)thecri
ter
iont
hatexpr
essesthedeci
sion-maker‘
smannerof
eval
uat
ing thedesi
rabi
l
ityofalt
ernat
iveval
uesofdecisi
on v
ari
ables,and (
ii
)how t
hat
 cr
it
eri
onistobeoptimized( minimizedormaxi
mized.A st
atedobj
ect
ivehel
pst
ofocus
at
tent
iononwhatt
heproblem actual
l
yi s.

 Const
rai
ntsorLi
mit
ati
ons

Thesearetherestr
icti
onsonthev al
uesofthedeci
sionv
ariabl
es.Theserest
ri
cti
onscanari
se
duetolimit
edresourcessuchasspace,money,manpower,materi
al,et
c.Theconst
rai
ntsmay
beinthefor
m ofequationsori
nequal
it
ies.

 Funct
ional
rel
ati
onshi
ps

Inadecisi
onpr
obl
em,thedecisi
onvariabl
esintheobj
ecti
vefuncti
onandintheconst
rai
ntsare
connect
edbyaspeci
fi
cf unct
ionalr
elati
onshi
p.Ageneraldeci
sionprobl
em modelmighttake
thefor
m:

Opt
imi
ze(
MaxorMi
n)Z=f
(x)Subj
ect

t
otheconst
rai
nts:

gi
(x)(
≤,=≥)bi
;i=1,2….
m andx≥0Wher
e,x=a

v
ect
orofdeci
si
onv
ari
abl
es(
x1,
x2,
x3,
xn)

f
(x)=Cr
it
eri
onorobj
ect
ivef
unct
iont
obeopt
imi
zedgi
(x)

=t
hei
thconst
rai
nt

bi
=fi
xedamountoft
hei
thr
esour
ce

Amodeli sr efer
redt
oasal i
nearmodelifallf
unct
ionalr
elat
ionshi
psamongdeci sionv
ari
abl
es
X1,X2,Xn i
nf (
x)andg(x)ar
eofalinearfor
m.Butifoneormor eofther
elat
ionshipsar
enon–
l
inear,t
hemodel i
ssai
dtobeanon-li
nearmodel.

 Par
ameters-Thesear econstant
sint hef
unct
ionalr
elat
ionshi
ps.Par
amet
erscanbe
det
ermi
nisti
corprobabi
li
sti
cinnatur
e.

St
ep3.Sol
vi
ngt
heMat
hemat
ical
Model
Thisinvol
vesobt
aini
ng thenumer i
calval
uesofdeci si
onvari
abl
es.Obtai
ning t
hesev al
ues
dependsonthespeci
fi
cform ortype,ofmathemati
calmodel
s.Sol
vingt
hemodelr equi
resthe
useofvari
ousmathemati
calt
oolsandnumericalpr
ocedur
es.I
ngener
al

Ther
ear
etwocat
egor
iesofmet
hodsusedf
orsol
vi
nganORmodel
.

 Opti
mizati
onmodel .Thesemodel syiel
dthebestv alueforthedecisi
onvar
iablesbot hfor
unconstr
ained and const r
ained pr
oblems.I n const r
ained probl
ems,t hese v alues
simul
taneouslysati
sfyalloftheconstrai
ntsandpr ovideanoptimaloraccept abl
ev alue
fort
heobj ect
ivefuncti
onormeasur eofeffect
iveness.Thesoluti
onsoobtainedi scalled
theopti
mal sol
uti
ont otheProblem.
  Heuri
sti
c Model.These met hods yi
eldv al
ues ofthe var
iabl
es thatsat
isf
y al
lthe
const
rai
nts,butnotnecessar
il
ypr ov
ideoptimalsol
uti
on.However,t
heseval
uesprovi
de
anacceptabl
evalueoftheobj
ectiv
efuncti
on.

Heuri
sti
c Methods ar
e somet
imes descr
ibed as ―r
ules oft
humb‖whi
ch wor
k.These
methodsareusedwhenobt ai
ningopti
malsoluti
oniseitherv
eryt
imeconsumi
ngorthe
modeliscomplex.
St
ep4.Val
i
dat
ing(
Test
ing)t
hesol
uti
on
Aftersolvi
ngthemat hemati
calmodel,itisimport
anttorevi
ewthesol
uti
oncareful
l
ytosee
thatvaluesmakesenseandt hatt
her esult
ingdeci
sionscanbeimpl
emented.Someofthe
reasonsforval
idat
ingthesol
utionar
e:

(
i)Themat hemat
icalmodel
maynothav
eenumer
atedal
lthel
i
mit
ati
onsoft
hepr
obl
em
underconsi
der
ation.

(
ii
)Cer
tai
naspectoft
hepr
obl
em mayhav
ebeenov
erl
ooked,
omi
tt
edorsi
mpl
i
fied,

(
ii
i)Thedat
amayhav
ebeeni
ncor
rectest
imat
edorr
ecor
ded,
per
hapswhenent
eredi
ntot
he
comput
er.

St
ep5.I
mpl
ement
ingt
hesol
uti
on
Thedecisi
on-makerhasnotonlytoidenti
fygood decisi
on al
ter
nat
ivesbutalso t
o sel
ect
al
ter
nat
ivesthatarecapabl
eofbei
ngimplement
ed.Iti
si mpor
tanttoensur
ethatanysolut
ion
i
mplementedisconti
nual
lyr
evi
ewedandupdat
edintheli
ghtofachangi
ngenvir
onment.

St
ep6.Modi
fyi
ngt
heModel
Foramat hemat i
calmodelt obeusef ul
, t
hedegr eetowhi chi tactual l
yrepresentst hesystem or
problem bei ng modeled mustbe est ablished.I fdur i
ng v ali
dation,t he solut
ion cannotbe
i
mpl ement ed,oneneedst o(a)identif
yconst r
aintthatwer eomi t
teddur i
ngtheor iginalpr
oblem
formulat
ionor( b)findifsomeoft heoriginalconst r
aintswer eincorrectandneedt obemodi fi
ed.
Inallsuchcases,onemustr eturntot hemodelf ormulationst epandcar efullymaket he
appropri
atemodi fi
cati
onst or epresentmor eaccur atelythegi venpr oblem.Amodelmustbe
appli
cablef orar easonabletimeper i
odandshoul dbeupdat edf rom t i
met otime,t aki
ngint o
consider
at i
ont hepast,present,andfutureaspect softhepr oblem.

St
ep7.Est
abl
i
shi
ngcont
rol
overt
hesol
uti
on
Thedy namicenv i
ronmentandchangeswithintheenv i
ronmentcanhav esi gnif
icantimpli
cat
ionsr
egar
ding
thecontinui
ngv ali
dit
yofmodelsandt hei
rsoluti
ons.Thus,acont r
olpr ocedurehast obeestabl
i
shedfor
detect
ingsigni
fi
cantchangesi
ndecisi
onv ari
abl
esoft heproblem sot hatsuitabl
eadj ust
mentscanbemade
i
nt hesoluti
onwithouthav
ingt
obuil
damodel ever
ytimeasi gnif
icantchangeoccur s.

4.Basi
cOper
ati
onsResear
chModel
s
Ther
eisnouniquesetofprobl
emswhi
chcanbesol
vedbyusi
ngORmodel
sort
echni
ques.SomeOR
modelsort
echni
quesincl
ude:
Al
l
ocat
ionModel
s

All
ocationmodel sar eusedt oal l
ocateresour cest oact i
vateinsuchawayt hatsomemeasur e
ofeffectiv
eness( obj
ectivef unction)isopt i
mi zed.Mat hemat i
calprogrammingisthebroadt erm
ofORt echni quesusedt osol veallocat
ionpr obl ems.
Ifthemeasur eofef fectivenesssuchaspr of i
t ,cost,etc.,i
sr epr
esentedasal i
nearfunctionof
severalvariablesandi flimitationsofr esour ces( const r
aints)canbeexpr essedasasy stem of
l
ineari nequaliti
esorequal it
ies,t heallocationpr oblem iscl assif
iedasl i
nearprogrammi ng
probl
ems.
But,i
ft heobj ecti
v eunctionofanyoral lconst rai
ntscannotbeexpr essedasasy st
em ofl inear
equalit
iesori nequalit
ies,t heal locat
ionpr obl em iscl assif
iedasanon- li
nearprogrammi ng
probl
em.

Whent hesoluti
onv al
uesordecisi
onv ar
iabl esfort hepr oblem arerestri
ctedtobeingi nt
eger
val
ues,t heprobl
em isclassi
fiedasan i ntegerpr ogrammi ng.Thepr oblem havingmul ti
ple,
confl
ict
ingandi ncommensurableobject
iv ef unct
ion( goal s)subjecttol inearconstr
aint
si s
cal
ledgoalprogrammi ng.I
fdecisi
onv ar
iablesi nthel inearpr ogr
ammi ngpr oblem dependon
chance,suchaproblem i
scall
edi edast
scal
l ochasticpr ogrammi ngproblem.

Ifresour
cessuchaswor kers,machi nesorsal esmencanbeassi gnedt operf
orm acer tai
n
numberofactivit
iessuchasj obsort err
it
ori
esonone-t oonebasissoast omini
mizetotalti
me,
costordi st
ance invol
ved in performing a given act
ivi
ty,such probl
emsar e cl
assi
fied as
assignmentproblems.Conver sel
y,iftheresourcescanbeusedf ormor et hanoneacti
vity
,the
all
ocati
onproblem iscl
assif
iedasat r
ansportat
ionprobl
em.

I
nvent
oryModel
Inv
entoryModelsdealwit
htheproblem ofdet
erminati
onofhowmucht oorderatapointi
ntime
andwhent oplaceanorder.Themai nobjecti
veist ominimizethesum oft hreeconf
li
cti
ng
i
nventorycost
s:TheCostofholdingorcarryi
ngextrainv
entory
,thecostofshortageordel
ayin
deli
ver
yofitemswheni ti
sneeded,acostoforderi
ngorset-
up.

Compet
it
ive(
GameTheor
y)Model
Thesemodel
sar eusedtocharact
eri
zethebehav
ioroft
woormor
eopponent
s(cal
l
edpl
ayer
s)
whocompetefortheachi
evementofconfl
i
cti
nggoal
s.
Networ
kModels
Thesemodel sareappli
edtot hemanagement( pl
anni ng,cont
rol
li
ngandscheduli
ng)oflarge-
scaleproject
s.PERT/ CPM Techniqueshelpinident if
y i
ngpotenti
altr
oubl
espotsinapr oject
throughtheidenti
fi
cati
onofthecri
tical
path.Theset echniquesi
mpr ov
eproj
ectcoor
dinat
ionand
enabletheef f
ici
entuseofr esour
ces.Networkmet hodsar ealsousedtodeter
minetime-cost
tradeoff
,resourceall
ocat
ionandupdatingofacti
v i
tytime.

Deci
sionanal
ysi
sModel
Thesemodel sdealwit
htheselecti
onofanopt imalcour
sefact
iongi
venthepossi
blepayoff
s
andtheirassociat
edpr
obabi
l
iti
esofoccurrence.Thesemodelsarebr
oadl
yappl
iedtoprobl
ems
i
nvolvi
ngdecisionmaki
ngunderriskanduncert
aint
y.
 Par
ttwo
Speci
alI
ssuesonTr
anspor
tat
ionpr
obl
em
Ther
ear
eanumberofspeci al
issuesi nrelat
iontothetransport
ati
onmodel.Theyar
e:
1.Determi ni
ngifthereareal t
ernat
eopt i
mal sol
uti
ons.
2.Defi
ni ngandhandl i
ngdegener acy(toofewoccupiedcell
stopermit
evaluati
onofasol ution).
3. Avoi di
ngunaccept ableorprohi
bitedrouteassi
gnments.
4.Dealingwithproblemsi nwhi chsupplyanddemandar enotequal
.
5.Solvingmaximizationpr obl
ems

Answer
:
1.Al
ter
nat
eOpt
imal
Sol
uti
ons
Somet i
mes,transpor
tat
ion pr
obl
ems have mul
tipl
e opti
malsol uti
ons.In such
i
nstances,i
tcanbeusefulforamanagert
obeawar eofal
ternat
esoluti
ons,because
thi
sgivesthemanageranopt i
onofbri
ngi
ngnon-
quanti
tat
iveconsi
derati
onsintothe
deci
sion.

Inthecaseoft het r
ansportati
onpr oblem,theexistenceofanal ter
natesol uti
oni s
si
gnaledbyanempt ycell‘sevaluati
onequalt ozero.Infact,youmayhav enot edthat
cel
lB-1hadanev al
uationequalt ozeroi nthefi
nalsoluti
onoft heHarleypr obl
em.We
canfindoutwhatt hatalt
ernatesol uti
oni sbyreall
ocati
ngt hemaxi mum numberof
uni
tspossibl
earoundthest epping-stonepathforthatcell
.Afterthecellevaluati
onof,
andreal
locat
ingunit
st ocellB-1,wef i
ndt hefol
l
owingt abletobeanal ternateoptimal
sol
uti
on.

Al
ter
nat
eopt
imal
sol
uti
onf
ort
heHar
leypr
obl
em

To:
Proj
ec Pr oj
ec Pr oj
ec Supply
Fr
om: t#1 t#2 t#3
4 2 8
Far
mA 100 100

5 1 9
Far
mB 50 150 200

7 6 3
Far
mC 200 200

Demand 50 150 300 500

2.Degener
acy
Asol
uti
onisdegener
atei
fthenumberofoccupi
edcellsi
slessthanthenumberof
r
owsplust
henumberofcolumnsminusone.i
.e.
,ther
earetoofewoccupi
edcel
lsto
enableal lt
heempt ycel l
stobeev aluat
ed.Inthecaseoft hestepping-stonemet hod,
thi
smeanst hattherewi l
lbeatl eastoneempt ycel lforwhichanev aluationpat h
cannotbeconst r
ucted.Fort heMODImet hod,i
tmeanst hatitwillbei mpossi bleto
determinealloftherowandcol umni ndexnumber s.Obvi
ously,somemodi fi
cat i
onhas
tobemadet odetermineifsuchadegener atesol
ut i
onisoptimal.Themodi ficationi s
totreatsomeoft heempt ycell
sasoccupi edcell
s.Thisisaccompl ishedbypl acinga
del
ta()i noneoft heempt ycell
s.Thedel t
arepresentsanext remelysmal lquant i
ty
(e.
g.,0.001uni t
);itissosmal lthatsupplyanddemandf orther ow andcol umn
i
nvolvedwi llbeunaf fectedevenwi thoutmodifyingotherquant i
tiesi nt her ow or
col
umn, andsosmal lthattot
alcostwill
notchange.

Thepur poseoft hedel taistoenabl eeval

uati
onoft heremainingempt ycells.The
choiceofl ocat ionfort hedeltacanbesomewhatt ri
cky:someempt ycellsmaybe
unsuitableiftheydonotenabl eev al
uati
onsoft heremainingempt ycells.Mor eover
,
thedel t
acannotbepl acedinacel lwhi
chlatert
urnsoutt obeinanegat i
veposi ti
onof
acel lpathinv olvedinr eal
locati
onbecausedel t
awi l
lbet he―smal l
estquant ityina
negativeposition‖andshi f
ti
ngt hatminutequanti
tyaroundt hecellpathwi l
lleavethe
soluti
onv ir
tuallyunchanged.Consequent l
y,a
certai
namountoft rialander r
ormaybenecessar ybeforeasatisfactor
yl ocati
oncan
beident i
fi
edf ordelta.

Thetechniquecanbedemonst rat
edf orthedegener
atealt
ernat
esolut
ionoftheHar
ley
probl
em.Supposet hataft
ersomeexper i
mentati
on,cel
lA-1hasbeenselect
edforthe
l
ocation ofdelta.The r esul
ting index numbers gener
ated usi
ng MODIand t he
i
mpr ovementpot enti
alforempt ycel l
sbasedondel taincellA-1areshowni nt
he
fol
lowingtabl
e.Thisconfir
mst hatthesoluti
onisopti
mal

Har
leyAl
ter
nat
eSol
uti
onModi
fi
edf
orDegener
acy
+4 0 +8
To:
Proj
ec Pr ojec Proj
ec Supply
Fr
om: t#1 t#2 t#3
4 2 8
0 Far
mA Δ +2 100 100

5 1 9
+1 Far
mB 50 150 0 200

7 6 3
-
5 Far
mC 200 200
+8 +11

Demand 50 150 300 500

3.Unaccept
abl
eRout
es
I
nsomecases,anor i
gin-
dest
inat
ioncombi nati
onmaybeunaccept abl
e.Thi
smaybe
duet oweatherfactor
s,equi
pmentbr eakdowns,laborpr
oblems,orski
llr
equi
rement
s
thatei
therprohi
bit
,ormakeundesirabl
e,certai
ncombinat
ions(r
outes)
.

Supposet
hati
ntheHar
leypr
obl
em r
out
eA-
3wassuddenl
yunav
ail
abl
ebecauseof
recentflooding.Inor dertopreventt hatroutefrom appeari
ngi nthefi
nalsol
ution(asit
origi
nall
ydi d),t
hemanagercoul dassi gnauni tcosttothatcellthatwaslargeenough
tomaket hatr outeuneconomi caland,hence,pr ohibi
titsoccur r
ence.Oner ul
eof
thumbwoul dbet oassignacostt hatis10t i
mest helargestcostinthetabl
e(orav ery
big+M) .Then,t hi
sr evi
sedpr oblem coul dbesol vedusingei t
heroft hemethodswe
hav ediscussedear l
ier.Notet hatt hepr ohi
bitedroutemayappeari nanon- opt
imal
soluti
on, butitwillbeelimi
natedbyt heti
met heoptimalsolut
ionisreached.

4.Unequal
Suppl
yandDemand
Upt othispoint,exampl eshavei nvolvedcasesi nwhi chsupplyanddemandwer eequal.
Asy oumi ghtguess,t herearesituationsi nwhichthet woar enotequal .Whensucha
sit
uationisencount er ed,i
tisnecessar ytomodi fytheor i
ginalpr obl
em sot hatsupply
anddemandar eequal .Thi
sisaccompl ishedbyaddi ngei t
heradummycol umnora
dummyr ow;adummyr ow isaddedi fsupplyisl esst handemandandadummy
columni saddedi fdemandi slesst hansuppl y.Thedummyi sassi gneduni tcost
sof
zeroforeachcel l
,andi tisgi
venasuppl y(ifarow)orademand( ifacolumn)equalt o
thediffer
encebet weensuppl yanddemand.Quant i
ti
esindummyr out
esi ntheopti
mal
soluti
onar enotshi pped.Rat her,theyser vetoindicat
ewhi chsuppl i
erwi llhol
dt he
excesssuppl y,andhow much,orwhi chdest i
nati
onwi llnotr eceiveit
st ot
aldemand,
andhowmuchi twi l
l beshort.

Let‘ s consider s an example.Suppose t hatFarm Ci nt he Harleyproblem has

experiencedanequi pmentbr eakdown,andi twil
lbeabl et osuppl yonl y120cubi c
yardsoft opsoi lforaper i
odofti
me.Ther ef
ore,t
otalsupplywi llbe80uni tslessthan
totaldemand.Thi swillrequi
readdingadummyor igi
nwi thasuppl yof80uni ts.The
fi
nalsol uti
oni sshowni nthefol
lowingtable.Weinterpretthesol ut
ionindicati
ngthat
Project#3wi llbeshor tby80uni tsperweekunt i
ltheequi pmenti srepair
ed.Not e,
though,t hatthi sanalysi
shasconsi der
edonl ytr
anspor t
ationcost s,andt hatother
factors,suchar eshortagecostsorschedul esofthepr oj
ects,maydi ctatesomeot her
courseofact ion.

Ift
heintuit
iveapproachi susedt oobtai
nt heiniti
alf
easiblesol
utionwhenadummyi s
i
nvolv
ed, makeassi gnmentstot hedummyl ast.Hence,beginbyassigningunit
stothe
cel
lwiththelowestnonzer ocost,thenthenextl owestnonzerocost,andsoon.Forthe
Harl
eypr obl
em thi
swoul dmeant hatunitswoul dbeassi gnedf
irsttocellB-
2because
i
tscostof$1i sthel owestnonzerocellcost.

Sol
uti
onusi
ngt
heDummyOr
igi
n

To:
Proj
ec Proj
ec Proj
ec Suppl
y
Fr
om: t#1 t#2 t#3
4 2 8
Far
mA 50 50 100

5 1 9
Far
mB 150 50 200

7 6 3
Far
mC 120 120
 0 0 0
Dummy 80 80

Demand 50 150 300 500

5.Max
imi
zat
ion
Somet r
anspor t
ationt y peproblemsconcer nprofitsorr evenuesr atherthancost s.In
suchcases,t heobj ect ivei
st omaxi mi zeratherthant omi nimize.Suchpr obl
emscan
behandl edbyaddi ngoneaddi ti
onalst epatt hestar t
:Identi
fythecel lwiththelargest
profi
tandsubt ractal ltheothercell prof i
tsfr
om thatv alue.Thenr eplacethecellprofi
ts
witht her esulti
ngv alues.Thesev al uesr ef
lecttheoppor tunit
ycost sthatwoul dbe
i
ncur redbyusi ngr outeswi thuni tpr of i
tsthatarel esst hant hel argestunitprofit
.
Replacet heor i
ginaluni tprofi
tswi t
ht heseoppor tunitycostsol utions.Thiswi l
lbe
i
dent icaltomaxi mi zingt het ot
alpr of i
t.Forexampl e,supposei nt heHar leyproblem,
thecel lvalueshadbeenuni tprofi
tsi nst eadofunitcost s.CellB-
3hadt helargestdoll
ar
value:$9.Hence,eachcel l‘sdollaramountwoul dbesubt ractedf r
om 9.Forcel lA-1,
ther esulti
ngoppor tunitycostwoul dhav ebeen9- 4=5andsoon.Cel lB-3woul dhav e
anoppor tunit
ycostof0maki ngitthemostdesi rabler oute.

Theremainderofthest epsfordevelopinganini
ti
alfeasiblesol
uti
on,ev al
uati
onof
emptycel
ls,andreal
locati
onareidenticalt
othoseusedforcostmi ni
mi zati
on.When
theopt
imaldist
ri
buti
onplanhasbeeni denti
fi
ed,
usetheoriginal
cel
lvalues(i.
e.,
prof
it
s)
tocomputethetot
alprofi
tfort
hatplan.

Speci
alSi
tuat
ionsonAssi
gnmentpr
obl
em
Whenwesol
veassi
gnmentpr
obl
ems,
ther
ear
ecaseswhi
char
etr
eat
eddi
ff
erent
lyf
rom
t
heusual
way.

Thenumberofr
owsdoesnotequal
thenumberofcol
umns
TheHungar i
anMet hodofsol
v i
nganassi gnmentprobl
em requi
rest hatt
henumber
ofcolumnsshoul dbeequaltothenumberofr ows.Theyareequal,theprobl
em i
s
balancedpr obl
em,andwhennot ,itiscall
edanunbalancedproblem.Thus,where
thereare5wor ker
sand4machi nes,orwhenthereare4wor kersand6machi nes,
forinstance,wehav eunbalancedsi t
uati
onsi nwhichone-to-onemat chisnot
possibl
e.Incaset hemachinesar einexcess,t
heexcessmachi ne(s)wouldremai
n
i
dleandsoi sthecasewhenmenar einexcess-thenumberofexcesspeopl ewoul
d
getanassi gnment.

I
nsuchsi t
uati
ons,dummycolumn( s)
/row(s)
,whicheveri
ssmall
erinnumber,are
i
nsertedwithzerosasthecostelements.Forexample,whent
hegiv
encostmat r
ix
i
soft hedimension4*5,adummyr owwoul dbeincl
uded.I
neachcol
umninrespect
ofthi
sr ow,a
‗zero‘wouldbeplaced.Aft
erthi
soperationofintr
oduci
ngdummycolumns/rows,
theproblem i
ssolvedi
ntheusualmanner.

Ex
ampl
e:
 Acompanyhas4machinestodo3j obs.Eachjobcanbeassi
gnedtooneandonl
y
onemachi
ne.Thecostofeachjoboneachmachi nei
sgivenbel
ow.Deter
minet
he
j
obassi
gnmentswhichwil
lmini
mizethetotal
cost.
Machi
ne
W X Y Z
A 18 24 28 32
Job B 8 13 17 18
C 10 15 19 22

Thepr
obl
em i
nvol
vesmaxi
mizat
ionr
athert
hanmi
nimi
zat
ion
I
nsomesi tuati
ons, theassi gnmentpr obl em maycal lf
ormaxi mizati
onofpr ofi
t,r
evenue,
etc.ast heobj ective.Forexampl e,wemaybef acedwi ththeproblem ofassi gnment
ofsal esmeni ndi fferentregionsi nwhicht heycandi splaydif
ferentqual i
ti
esin
maki ngsal es( r
eflectedi namount sofsal esexecut edbyt hem).Obv iously,
assignmentwoul dbemadei nsuchawayt hatthetotalexpectedr ev
enuei s
maxi mi zed.
Fordeal i
ngwi thamaxi mi zati
onpr obl em,wef i
rstchangei tintoanequi val
ent
mi ni
mi zat i
onpr oblem.Thi si sachievedbysubt racti
oneachoft heel ementsoft he
givenpay -offmat rixf r
om aconst ant(v alue)k.Thus, wemaysi mpl yputanegat i
ve
signbef oreeachoft hepay -
offv al
ues( whichi sequi val
entt osubt ract
ingeach
valuef rom zer o).Usual l
y,t helargestv alueofal lvaluesint hegi venmat r
ixis
l
ocat edandt heneachoneoft hev aluesi ssubtractf r
om it(thelar gestvalueis
takensoast oav oidt heappear anceofnegat i
vesi gns).Thenthepr oblem issolved
thesamewayasami nimizationproblem i s.

Ex
ampl
e:
Acompanypl anstoassign5salesmento5distri
ctsinwhichitoperates.Esti
mates
ofsalesrevenuei nthousandsofbir
rsforeachsal esmanindi f
ferentdist
ri
ctsare
giv
eni nthefollowingtabl
e.Inyouropini
on,whatshouldbet heplacementoft he
salesmeniftheobjecti
veistomaximizetheexpectedsalesrevenue?

Dist
ri
ct
Salesman D1 D2 D3 D4 D5
S1 40 46 48 36 48
S2 48 32 36 29 44
S3 49 35 41 38 45
S4 30 46 49 44 44
S5 37 41 48 43 47
Soluti
on:Si ncei
tisamaximizati
onprobl
em,wewouldfi
rstsubtr
acteachoft he
entri
esint hetabl
efr
om thelar
gestone,whi
chequal
s49here.Theresul
tantdata
aregiveninTablebel
ow.

Oppor
tuni
tyLossMat
ri
x
Di
str
ict
Sal
esman D1 D2 D3 D4 D5
S1 9 3 1 13 1
S2 1 17 13 20 5
 S3 0 14 8 11 4
S4 19 3 0 5 5
S5 12 8 1 6 2

Now,
wewi l
lproceedasusual.
St
ep1:Subt
ractvalueineachrowf
rom ev
eryv
aluei
nther
ow.Ther
esul
ti
ngv
alues
ar
egi
venintablebelow.

ReducedcostTabl
e1
Di
str
ict
Sal
esman D1 D2 D3 D4 D5
S1 8 2 0 12 0
S2 0 16 12 19 4
S3 0 14 8 11 4
S4 19 3 0 5 5
S5 11 7 0 5 1

Step2:Subtr
actmini
mum val
ueineachcolumni nr
educedcosttabl
e1from each
valuei
nthecolumn.Testf
oropti
mali
tybydrawingl
inestocoverzer
os.Theseare
shownintabl
ebelow(i
ntheReducedcostTabl
e2)

ReducedcostTabl
e2
Di
str
ict
Sal
esman D1 D2 D3 D4 D5
S1 8 0 0 7 0
S2 0 14 12 14 4
S3 0 12 8 6 4
S4 19 1 0 0 5
S5 5 0 0
11 1

Sincethenumberofl
i
nescover
ingal
lzer
osi
sfewert
hann,wesel
ectuncov
eredcel
l
value,
whichequal
s4.Wi
ththi
s,wecanmodi
fythet
ableasgi
veni
ntheReducedCost
Table3.

St
eps4,
5,6:
Findi
mpr
ovedsol
uti
on.Testf
oropt
imal
i
tyandmake

assi
gnments.ReducedcostTabl
e3
Di
str
ict
Sal
esman D1 D2 D3 D4 D5
S1 12 0 0X 7 0X

S2 0 10 8 10 0X
S3 0X 8 4 2 0
S4 23 1 0 0X 5
S5 15 5 0X 0 1

Mor
ethanoneopt
imal
assi
gnmenti
spossi
blei
nthi
scasebecauseoft
he
 exi
stenceofmul t
iplezer
osindiffer
entrowsandcolumns.Thepossi
ble
assi
gnment sare:
S1- D2,
S2-D5, S3-
D1,S4-D3,S5-D4;
or
S1-D2,S2-D1,S3-D5,S4-
D3,S5-D5;or
S1-D2,S2-D1,S3-D1,S4-
D4,S5-D3;or
S1-D2,S2-D1,S3-D5,S4-
D4,S5-D
Eachoft heseassignmentpat t
ernswouldleadtoexpectedaggr
egatedsal
esequalt
o
231thousandbi r
rs.

Cer
tai
nmat
chesar
eundesi
rabl
eornotal
l
owed

Ithappenssomet imest hatawor kercannotperform acer

tainjobori snott obe
assignedapar t
icularjob.Tocopewi ththissi
tuati
on,thecostofper f
ormingt hat
j
obbysuchper soni stakentobeext r
emelylar
ge( whi
chiswritt
enasM) .Thent he
soluti
ontotheassi gnmentpr obl
em proceedsi
nt hemannerdiscussedear l
i
er.The
eff
ectofassigningpr ohi
bit
ivecosttosuchperson-j
obcombinationsisthattheydo
notfigur
einthef i
nalsoluti
on.

Ex
ampl
e:
Youar egi v
ent hei nf
ormationaboutt hecostofperformi
ng dif
ferentj
obsby
dif
ferentpersons.Thejob-personmarki
ngXi ndi
catest
hattheindi
viduali
nvol
ved
cannotper f
orm thepar t
icularjob.Usi
ngthisinf
ormati
on,stat
e( i
)t heopti
mal
assignmentofjobs,and(i
i)thecostofsuchassi
gnment,

Job
J1 J2 J3 J4 J5
P1
27 18 X 20 21
son

P2
31 24 21 12 17
Per

P3
20 17 20 X 16
P4
22 28 20 16 27
Sol
uti
on:-Bal
anci
ngtheprobl
em notassigni
ngahi ghcostt
othepai
ri
ngsP1-
J3
andP3-J4,wehavet
hecostgiveni
nthetablebel
ow.

Job
J1 J2 J3 J4 J5
P1 27 18 M 20 21
P2 31 24 21 12 17
P3 20 17 20 M 16
son

P4 22 28 20 16 27
Per

P5 0 0 0 0 0
dummy
Now wecander i
vethereducedcosttableasshowni nt abl
eshownbel ow.Note
thatthecellswithprohi
bit
edassignmentscont i
nuetobeshownwi ththecost
elementM, si
nceM isdefi
nedtobeextremelylar
gesot hatsubtr
acti
onoraddit
ion
ofv al
uedoesnotpracti
cal
lyaff
ecti
t.Totestopti
mali
ty,l
inesaredrawntocoveral
l
zeros.Si
ncethenumberoflines
coveri
ngallzerosi
slessthann,weselectthelowestuncoveredcell
,whi
chequals
4.Wi t
hthisval
ue,wecanobt ai
ntherevi
sedreducedcostt abl
e,showninthefi
nal
table.
ReducedMat
ri
x
Job
J1 J2 J3 J4 J5
P1 9 0 M 2 3
P2 19 12 9 0 5
P3 4 1 44 M 0

son
P4 6 12 4 0 11

per
P5 0 0 0 0 0
dummy
ReducedMa
r
rt
ix
Job
J1 J2 J3 J4 J5
P1 9 0 M 2 3
P2 19 12 9 0 5
P3 4 1 4 M 0
son

P4 6 12 4 0 11
per

P5 0 0 0 0 0
dummy

Assi
gnmentof
Job
Job
J1 J2 J3 J4 J5
P1 9 0 M 2 3
P2 15 8 9 0 1
P3 4 1 4 M 0
son

P4 2 8 0 0X 7
per

P5 0 0X 0X 0 0X
dummy

Thenumberofl
inescov
eri
ngzerosisequalt
o5(=n),
hencet
heoptimalassi
gnment
canbemade.Theassignmentis:P1-
J2,P2-J4,P3-
J5,P4-
J3,whil
ejobJ1woul d
remai
nunassi
gned.Thi
sassignmentpatter
nwouldcost18+12+16+20=66inthe
aggr
egat
e.

Mul
ti
pleOpt
imal
Sol
uti
onsex
ist

Inthepr ocessofmaki ngassignment s,i

twasst atedear l
ierthatwesel ecta
row/ columnwi t
honl yasinglezer
ot omakeanassi gnment .Howev er,asit
uat i
on
maywher ei
nt hevariousrowsandcol umns, whereassignmenti syettobedone,
hav eallmul ti
plezeros.Insuchcases, wegetmul t
ipleoptimal solut
ionstot he
givenpr oblem.I nanyoft heprobl
emsdi scussedsof ar,wehav enotexperienced
suchasi tuati
on.Hence, eachoneoft hem hashadauni queopt imalsolut
ion.When
apr oblem hasauni queoptimalsolut
ion, i
tmeanst hatnoot hersoluti
ontot he
problem ex ist
swhi chy i
eldsthesameobj ecti
vefunctionvalue( cost,t
ime,pr of
it
etc.)ast heoneobt ainedf r
om theopt i
mal solut
ionder i
ved.Ont heotherhand, i
na
 pr
oblem wi
thmult
ipl
eopti
mal sol
uti
ons,t
her
eexi
stsmorethanonesol
uti
onwhi
ch
al
li
soptimalandequal
l
yattr
activ
e.Consi
dert
hefol
lowi
ngexample.

Ex
ampl
e:
Sol
vet
hef
oll
owi
ngassi
gnmentpr
obl
em andobt
aint
hemi
nimum costatwhi
chal
lthej
obscanbeper
for
med.

Job(
costi
n’00Br
.)
Wor
ker 1 2 3 4 5
A 25 18 32 20 21
B 34 25 21 12 17
C 20 17 20 32 16
D 20 28 20 16 27

Solut
ion:Thi
spr
obl
em i
sunbal
ancedsi
ncenumberofj
obsi
s5whi
let
henumberof
workersis4.Wefi
rstbal
anceitbyi
ntr
oduci
ngadummyworkerE,asshownin
tabl
ebelow.

Bal
anci
ngt
heAssignmentPr
oblem
Job(
costin’
00Br
.)
Worker 1 2 3 4 5
A 25 18 32 20 21
B 34 25 21 12 17
C 20 17 20 32 16
D 20 28 20 16 27
E 0 0 0 0 0

Step1:Obtai
nreducedcostval
uesbysubt
ract
ingthemi
nimum v
aluei
neachr
ow
fr
om ever
ycelli
ntherow.Thi
sisgi
veni
nTablebelow.

ReducedCost1
Job(
costi
n’00Br
.)
Wor
ker 1 2 3 4 5
A 7 0 14 2 3
B 22 13 9 0 5
C 4 1 4 16 0
D 4 12 4 0 11
E 0 0 0 0 0

Sincet hereisatl eastonezeroineachr ow andcolumn,wet estitforoptimali

ty.
Accor dingl
y,linesar edrawn.Allzer
osar ecoveredby4lines,whichislesst han5
(theor deroft hegi venmatrix
).Hence,wepr oceedtoimpr ov
et hesolution.The
leastuncov ered
v
alueis4.Subt ractingf rom ev
eryuncoveredvalueandaddi ngittoev er
yv alue
l
yi
ngatt hei nt
ersectionofl i
nes,wegettherevi
sedvaluesasshownbel ow.

Reducecost2andAssi
gnment
Job(
costi
n’00Br
.)
Worker
 1 2 3 4 5
A 7 0 14 6 3
B 18 9 5 0 1
C 4 1 4 20 0
D 0 8 0 0X 7
E 0X 0X 0 4 0X
Thesolut
iongiveninthereducedci st2t ablei
sopt i
malsincethenumberofl ines
cover
ingzer
osmat cheswi t
ht heorderoft hemat r
ix.Wecan,ther
efor
e,proceedt o
makeassignments.Tobegi nwi t
h,sinceeachoft hecolumnshasmul t
iplezer os,
wecannotstartmakingassignment sconsi der
ingcolumnsandhav e,t
herefore,to
l
ookthroughrows.Thef i
rstrowhasasi nglezero.Thus,wemakeassignmentA- 2
andcrossoutzeroatE-2.

Further,t
hesecondandthet hirdr owshaveonezeroeach.Wemakeassi gnments
B-4andC- 5,andcrossoutzer osatD-4andE- 5.Now,bot
ht herowsl
eftt
wozer os
eachandsohav eboththecol umns.Thisindi
catesexi
stenceofmult
ipl
eoptimal
soluti
ons.To obtainthesol utions,weselectzerosarbitr
ari
l
yand proceed as
discussedbelow.

i.SelectthezeroatD- 1,makeassi gnmentandcr ossoutzer osatD-3andE-1(as

both,workerDandj ob1, arenotav ail
abl
eanymor e)
.Next,assi
gnworkerEtojob
3,cor r
espondi
ngtot heonl yzerolef t
.Evi
dentl
y,select
ingthezeroatE-3ini
ti
all
y
woul dhavetheeff
ectofmaki ngsameassi gnment s.
i
i
.Sel ectthezeroatD- 3,makeassi gnmentandcr ossatD- 1andE-3.Next
,make
assignmentattheonl yzeroleftatE-1.Obvi
ously,sel
ecti
ngt hezer
oatE-1making
assignmentinthefi
rstplacewoul dl eadt
ot hesameassi gnments.
Toconcl
ude,
thepr
obl
em hast
woopt
imal
sol
uti
onsasgi
venbel
ow.

Sol
ution1 (’
00Br.
) Sol
ution2 (’
00Br.
)cost
Worker Job Cost Worker Job Cost
A 2 18 A 2 18
B 4 12 B 4 12
C 5 16 C 5 16
D 1 20 D 3 20
Jobleft 3 Jobleft 1
Tot
al 66 Tot
al 66

Thankyou!
!