Page - 29 - in Algorithms for Scheduling Problems
Image of the Page - 29 -
Text of the Page - 29 -
Algorithms 2018,11, 66
OB(Ļk,T)with the largest length of the relative perimetermaybe obtained as theCartesianproduct of the
optimalityboxes constructed for the subproblemsP1 andP2.
Proof. Since theblocksBr andBr+1 haveno common jobs, the inequality pUu < pLv holds for each
pair of jobs Ju āārk=1Bk and Jv āāmk=r+1Bk. Due toTheorem2, any job Ju āārk=1Bk dominates
any job Jv āāmk=r+1Bk. Thus, due toDeļ¬nition1, there is nooptimalpermutationĻk ā S for the
instance1|p|āCiwithascenario pāT suchthat job Jvprecedes job Ju in thepermutationĻk. Dueto
Deļ¬nition2,anon-emptyoptimalityboxOB(Ļk,T) ispossibleonly if job Ju is locatedbefore job Jv in
thepermutationĻr. TheoptimalityboxOB(Ļk,T)with the largest relativeperimeter for theproblem
1|pLi ⤠pi⤠pUi |āCimaybeobtainedastheCartesianproductof theoptimalityboxeswiththe largest
lengthof the relativeperimeters constructed for the subproblemsP1 andP2,where the set of
jobsār
k=1Bk andthesetof jobs ām
k=r+1Bk areconsideredseparatelyonefromanother.
LetBādenoteasubsetofallblocksof thesetB,whichcontainonlyļ¬xed jobs. LetJādenotea
setofallnon-ļ¬xed jobs in thesetJ . Let thesetB(Ju)āBādenoteasetofallblockscontaining the
non-ļ¬xed job JuāJā. Theorem5,Lemma6andtheconstructiveproofof the followingclaimallows
us todevelopanO(n logn)-algorithmforconstructingthepermutationĻkwiththe largestvalueof
PerOB(Ļk,T) for the special caseof theproblem1|pLi ⤠pi ⤠pUi |āCi. TheproofofTheorem6 is
basedonProcedure4.
Theorem6. Let eachblockBrāBcontainatmost onenon-ļ¬xed job. ThepermutationĻkāSwith the largest
valueof the relativeperimeterPerOB(Ļk,T) is constructed inO(n logn) time.
Proof. There isnovirtualblock in thesetB, sinceanyvirtualblockcontainsat least twonon-ļ¬xed
jobs while each block Br ā B contains at most one non-ļ¬xed job for the considered problem
1|pLi ⤠pi⤠pUi |āCi. UsingO(n logn)-Procedure1describedin theproofofLemma1,weconstruct
thesetB={B1,B2, . . . ,Bm}ofallblocks.UsingLemma3,weorderthejobsinthesetJ \Jāaccording
to the increasingof the leftboundsof thecoresof theirblocks.Asaresult, allnā|Jā| jobsare linearly
orderedsinceall jobsJ \Jā areļ¬xed in theirblocks. Suchanorderingof the jobs takesO(n logn)
timeduetoLemma1.
Since each block Br ā B contain at most one non-fixed job, the equality B(Ju)ā©B(Jv)=ā
holds for each pair of jobs Ju ā B(Ju) ā B \Bā and Jv ā B(Jv) ā B \Bā, u = v.
Furthermore, theproblem1|pLi ⤠pi⤠pUi |āCimaybedecomposedintoh= |Jā|+ |Bā|subproblems
P1,P2,. . . ,P|Jā|,P|Jā|+1,. . . ,Ph such that the condition of Lemma 4 holds for each pair Pr and Pr+1
of the adjacent subproblems, where 1 ⤠r ⤠hā1. Using Lemma4,wedecompose the problem
1|pLi ⤠pi⤠pUi |āCi intohsubproblems{P1,P2,. . . ,P|Jā|,P|Jā|+1,. . . ,Ph}=P. ThesetP ispartitioned
into twosubsets,P=P1āŖP2,where thesubsetP1= {P1,P2,. . . ,P|Jā|}containsall subproblemsPf
containingallblocksfromthesetB(Jif),where Jif āJāand |P1|= |Jā|.
ThesubsetP2={P|Jā|+1, . . . ,Ph} containsall subproblemsPd containingoneblockBrd fromthe
setBā, |P2|= |Bā|. If subproblemPdbelongs to thesetP2, thenusingO(n)-Procedure3described in
theproofofTheorem5,weconstruct theoptimalityboxOB(Ļ(d),T(d))withthe largest lengthof the
relativeperimeter for thesubproblemPd,whereĻ(d)denotes thepermutationof the jobsBrd āBā and
T(d)āT. It takesO(|Brd|) timetoconstructasuchoptimalityboxOB(Ļ(d),T(d)).
If subproblem Pf belongs to the setP1, it is necessary to consider all |B(Jjf)| ļ¬xings of the
job Jif in theblockBfj fromthe setB(Jjf) = {Bf1,Bf2, . . . ,Bf|B(Jjf )|}. Thus,wehave to solve |B(Jjf)|
subproblemsPgf ,where job Jif isļ¬xed in theblockBfg ā B(Jjf)and job Jif isdeleted fromallother
blocksB(Jjf)\{Bfg}.WeapplyProcedure3forconstructingtheoptimalityboxOB(Ļ(g),T(f))with
the largest length of the relative perimeterPerOB(Ļ(g),T(f)) for each subproblemPgf , where g ā
{1,2, . . . , |B(Jjf)|}. Then,wehave tochoose the largest lengthof therelativeperimeteramong |B(Jif)|
constructedones:PerOB(Ļ(fā),T(f))=maxgā{1,2,...,|B(Jjf )|}PerOB(Ļ (g),T(f)).
29
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