Page - 25 - in Algorithms for Scheduling Problems
Image of the Page - 25 -
Text of the Page - 25 -
Algorithms 2018,11, 66
Lemma2. Atmost two jobs in the blockBrāBmayhave thenon-emptyoptimality segments inanyļ¬xed
permutationĻkāS.
Proof. DuetoDeļ¬nition3, the inclusion [bLr ,bUr ]ā [pLi ,pUi ]holds foreach job JiāBr. Thus,due to
Deļ¬nition2,only job Jki,which is theļ¬rstone in thepermutationĻk amongthe jobs fromtheblock
Br, andonly job Jkv āBr,which is the lastone in thepermutationĻk amongthe jobs fromtheblock
Br,mayhavethenon-emptyoptimalitysegments. It is clear that thesenon-emptysegments lookas
follows: [pLki,u ā
ki ]and [lākv,p U
kv ],whereuāki ⤠bLr andbUr ⤠lākv.
Lemma3. IfOB(Ļk,T) =ā
, then any two jobs Jv ā Br and Jw ā Bs,which are ļ¬xed indifferent blocks,
r< s,mustbeordered in thepermutationĻkāSwith the increasingof the left bounds (and the rightbounds
aswell) of the coresof theirblocks, i.e., thepermutationĻk looksas follows:Ļk=(. . . , Jv, . . . , Jw, . . .),where
bLr < bLs .
Proof. For any two jobs Jv ā Br and Jw ā Bs, r< s, the condition [pLv,pUv ]ā© [pLw,pUw] = ā
holds.
Therefore, thesamepermutationĻk=(. . . , Jv, . . . , Jw, . . .) isobtainedif jobs JvāBr and JwāBs are
orderedeither in the increasingof the leftboundsof thecoresof theirblocksor in the increasingof
therightboundsof thecoresof theirblocks.WeproveLemma3byacontradiction. Let thecondition
OB(Ļk,T) =ā
hold.However,weassumethat thereareļ¬xed jobs JvāBr and JwāBs, r< s,which
are located inthepermutationĻkāS in thedecreasingorderof the leftboundsof thecoresof their
blocksBr andBs.Note thatblocksBr andBs cannotbevirtual.
Due toourassumption, thepermutationĻk looksas follows: Ļk = (. . . , Jw, . . . , Jv, . . .),where
bLr < bLs . UsingDeļ¬nition3,onecanconvince that thecoresofanytwoblockshavenocommonpoint.
Thus, the inequalitybLr < bLs implies the inequalitybUr < bLs . The inequalities pLv ⤠pv⤠bUr < bLs ā¤
pLw⤠pwholdforanyfeasibleprocessingtime pv andanyfeasibleprocessingtime pw,where pāT.
Hence, the inequality pv< pwholdsaswell, andduetoTheorem1, thepermutationĻkāScannot
beoptimal foranyļ¬xedscenario pāT. Thus, theequalityOB(Ļk,T)=ā
holdsduetoDeļ¬nition2.
Thiscontradictionwithourassumptioncompletes theproof.
Next, we assume that blocks in the set B = {B1,B2, . . . ,Bm} are numbered according to the
increasing of the left bounds of their cores, i.e., the inequality bLv < bLu implies v < u. Due to
Deļ¬nition3,eachblockBr={Jr1, Jr2, . . . , Jr|Br|}mayincludejobsofthefollowingfourtypes. IfpLri = bLr
and pUri = b U
r ,we say that job Jri is a core job in theblockBr. LetB ā
r bea set of all core jobs in the
blockBr. If pLri< b L
r ,wesaythat job Jri isa left job in theblockBr. If p U
ri > b U
r ,wesaythat job Jri isa
right job in theblockBr. LetBār (B+r )beasetofall left (right) jobs in theblockBr.Note that some job
Jri āBrmaybe left-right job in theblockBr, since it ispossible thatB\{BārāŖBār āŖB+r } =ā
.
Twojobs Jv and Jw are identical ifbothequalities pLv = pLw and pUv = pUw hold.Obviously, if the
setBrāB is a singleton, |Br|= 1, then theequalityBr=Bār holds. Furthermore, the latterequality
cannotholdforavirtualblockBtāB sinceanytrivialblockmustcontainat least twonon-ļ¬xed jobs.
Theorem3. For theproblem1|pLi ⤠pi⤠pUi |āCi, anypermutationĻkā Shasanemptyoptimality box
OB(Ļk,T)=ā
, if andonly if for eachblockBrāB,eithercondition |Br|= |Bār |ā„2holdsorBr=Bār āŖB+r ,
all jobs in the setBār (in the setB+r ) are identical and the following inequalitieshold: |Bār |ā„2and |B+r |ā„2.
Proof. Sufļ¬ciency. It is easy to prove that there is no virtual block in the considered set B. First,
weassumethat foreachblockBrāB, thecondition |Br|= |Bār |ā„2holds.
DuetoLemma3, in thepermutationĻkāSwiththenon-emptyoptimalityboxOB(Ļk,T) =ā
,
all jobs must be located in the increasing order of the left bounds of the cores of their blocks.
However, in the permutation Ļk ā S, the following two equalities hold for each block Br ā B:
minJiāBr p U
i =maxJiāBr p U
i andmaxJiāBr p L
i =minJiāBr p L
i . Since |Br| ā„ 2, there is no job Jv ā Br,
25
back to the
book Algorithms for Scheduling Problems"
Algorithms for Scheduling Problems
- Title
- Algorithms for Scheduling Problems
- Authors
- Frank Werner
- Larysa Burtseva
- Yuri Sotskov
- Editor
- MDPI
- Location
- Basel
- Date
- 2018
- Language
- English
- License
- CC BY 4.0
- ISBN
- 978-3-03897-120-7
- Size
- 17.0 x 24.4 cm
- Pages
- 212
- Keywords
- Scheduling Problems in Logistics, Transport, Timetabling, Sports, Healthcare, Engineering, Energy Management
- Categories
- Informatik
- Technik