Lower bounds on the approximation ratios of leading heuristics for the single machine total tardiness problem
Texte intégral
(2) .
(3) .
(4)
(5) . . . .
(6) . .
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16) .
(17)
(18)
(19)
(20)
(21)
(22) .
(23)
(24) .
(25)
(26)
(27)
(28)
(29)
(30)
(31)
(32)
(33)
(34)
(35)
(36)
(37)
(38)
(39)
(40)
(41) !
(42) "
(43)
(44) ! O(n7 /) ∀ > 0 #
(45) $## %& '( )*+,-. /0 12-
(46)
(47) !
(48)
(49)
(50)
(51)
(52)
(53)
(54)
(55) #,)3,## #,)3$## #,)3%& #,)3'(.
(56)
(57) .
(58) 4.
(59)
(60)
(61)
(62)
(63) !
(64)
(65)
(66) !
(67) .
(68)
(69)
(70)
(71)
(72)
(73) .
(74) . . 5
(75) 6"
(76)
(77)
(78)
(79)
(80) ! "
(81)
(82) +
(83)
(84) 7 "
(85)
(86) 5
(87) " "
(88)
(89) !
(90) ! 5 ! " O(n7 /ε) ( 5 5
(91)
(92) 65
(93) " $## %& '( )*+,-. /0 12-
(94)
(95) "
(96)
(97) !
(98)
(99) %
(100)
(101) 5 " #,)3,## #,)3$## #,)3%&
(102) #,)3'(. 4
(103)
(104)
(105) !
(106)
(107)
(108) !
(109)
(110) . .
(111)
(112) . 1|| Tj N = {1, 2, . . . , n} n j pj dj !. " S ∗ = (1, 2, . . . , n) n n ∗ T (N, S ) = i=1 Ti = i=1 max{Ci − di , 0} Ci = ni=1 pi ! 1|| Tj # $%&' ( )
(113) # # ) ! ) %&
(114) *++ , # # )
(115) % & ) # O(n7 /ε) # # )
(116) $ )
(117) % * - . / 0 +&' ! 1 #2 ) 1|| Tj 3
(118) ,. S N 1|| Tj # TA (N, S) S # rA (N, S) ) TA (N, S)/T (N, S ∗ ) S ∗ 1|| Tj N rA
(119) rA (N, S)
(120) N " #
(121) . 454 456 45789 45( # ) # 2 ! 2 : 8 1 ; 8 .
(122) <# ) ; # 8 1. .
(123) !
(124) 1 3
(125) N = {1, 2, . . . , n} (1, 2, . . . , n) 87! " $ i < j
(126) pi = pj di dj ' = ([1], [2], . . . , [n]) 4 " $ [i] < [j]
(127) di = dj pi pj ' , 1 2 = p(B) = k∈B pk = Bj Aj
(128) # j " # ej lj j " ! ej = p(Bj )+pj lj = p(N −Aj ) ! 1 .
(129) !.
(130) . !. . j → i di + pi > lj . i j i < j i → j di max{dj , ej } . n
(131)
(132) [k] n h k
(133) ! ! k
(134) " #"
(135) $ Bn = {[1], [2], . . . , [k − 1], [k + 1], . . . , [h]} An = {[h + 1], . . . , [n]} h % & Cn (h) = j=1 p[j] n [k] h k ' . . !. Cn (h) d[h+1] h < n( Cn (h) < d[h] + p[h] h > k ( Cn (h) d[r] + p[r]
(136) $ r = k, . . . , h − 1 .
(137) ># ) 7 # = %& # # ) # O(n4 pi ) , # "
(138) % & # # ) ) # O(n7 /ε)
(139) $%* - .&' # . " #
(140)
(141) . !
(142) ) 4 " ( n . "
(143) ! rEDD n " ! N 4 " SEDD N # Tmax (N, SEDD )
(144) ) SEDD # S ∗ 1|| Tj N , 4 2 1||Tmax
(145) Tmax (N, SEDD ) Tmax (N, S ∗ ) T (N, S ∗ ) > T (N, SEDD ) nTmax (N, SEDD ) nT (N, S ∗ )
(146) (
(147)
(148) ) $E1 ': N = {1, 2, . . . , n} p1 = m p2 , . . . , pn = 1 d1 = 0 d2 , . . . , dn = ε ! " S ∗ = (2, . . . , n, 1) T (N, S ∗ ) = n(n + 1)/2 + m − 1 − (n − 1)ε ! 4 " SEDD = (1, 2, . . . , n) T1 = m Ti = m + i − 1 − ε i = 2, . . . , n ! T (N, SEDD ) = nm + n(n − 1)/2 − (n − 1)ε ? m ε
(149) rEDD (N, SEDD ) ≈ n
(150) . !
(151) # ! )
(152) @ 1 # > <# )
(153) 1|| Tj 1 . # #
(154) 6 $%&': t i j max{t+pi , di } < max{t+pj , dj } max{t+pi , di } = max{t + pj , dj } pi < pj 789: 87! " ; ) .
(155) " ; %/& (: # 5 ; %+& AB4C! $%*&':
(156) " S . # ) PIj $E
(157) . 7 # ': 1 dj p(S) + pj p(S ∪ E) − d j p(S) + pj < dj < p(S ∪ E) PIj = p(E) − pj 0 p(S ∪ E) dj ; .
(158) PIj /pj ! 1|| wj Tj ,D $%.&': nt i j ui > uj ui = exp[− max{di − t − pi , 0}/k p¯]/pi p¯ =. #
(159) 1 COVERT i=1 pi /n ! . 1|| wj Tj 1 . 7 # E>C: 4 1 # 7 # E > C ; %-&.
(160) !. rMDD = rPSK = rWI = rCOVERT n/2. ) # E2 : N = {1, 2, . . . , n + 1} p1 = n p2 , . . . , pn+1 = 1 d1 = n d2 , . . . , dn+1 = n + ε ! 6 t = 0 , $ ' ? 4 " S = (1, . . . , n + 1) T1 = 0 Ti = i − 1 − ε i = 2, . . . , n + 1 ! TMDD (N, S) = n(n + 1)/2 − nε ! " S ∗ = (2, . . . , n + 1, 1) T (N, S ∗ ) = n ? ε rMDD (N, S) ≈ n/2 , %/& %0& 789 ( # "
(161) 6 ) E2 i j pi = pj di = dj max{t + pi , di } = max{t + pj , dj } ( 789 ( 6 ) E2 # AB4C! 7 # 2 → 3 → · · · → n + 1
(162)
(163) , PI1 = 1 S = ∅ E = {1, 2} ! p(S) = 0 p(E) = n + 1 PI2 = (1 − ε)/n ? # " SCOVERT = (1, 2, . . . , n + 1) . .
(164) . ! rAU nk
(165) . . k 1. ) E3 : N = {1, 2, . . . , n + 1} p1 = n p2 , . . . , pn+1 = ε d1 = n d2 , . . . , dn+1 = n − 1 E p¯ = (n + nε)/n ≈ 1 n
(166) S ∗ = (2, . . . , n, n + 1, 1) T (N, S ∗ ) = nε ,# ,D t = 0 : u1 = 1/n i = 2, . . . , n ui = (1/ε)(exp[−(n − 1 − ε)/(k p¯)]) ε = n−k n
(167) ui < u1 ! $ ' " ) ! " S = (1, 2, . . . , n) T (N, S) = n − nε
(168) rAU (N, SAU ) ≈ 1/ε = nk .
(169) . ! rNBR n/6
(170) ) # E4 : n = 2m + 2 N = {1, 2, . . . , 2m + 2} p1 = m p2 = 1 p3 = · · · = pm+1 = ε pm+2 = · · · = p2m+1 = 1 p2m+2 = 2ε d1 = m d2 = m + (ε/2) di = m + (i − 2)ε i = 3, . . . , m + 1 dm+2 = m + 1 + (m − 1)ε dj = j + (m − 1)ε j = m + 3, . . . , 2m d2m+1 = d2m+2 = 2m + (m − 1)ε ! E>C # " S¯ i1 < i2 < · · · < ik ¯ S Tik < pik pi1 > pi2 > · · · > pik , i1 , . . . , ik−1
(171) ik ) $ %-& ' E. ) E4 E>C ) 1, 2m + 1, 2m + 2
(172) (m+1)+(m+1)ε ! " # E>C SNBR = (2, 3, . . . , 2m+2, 1) T (N, SNBR ) = (m+1)+(m+1)ε ! " .
(173) S ∗ = (2, 3, . . . , m + 1, 1, m + 2, m + 3, . . . , 2m + 2) T (N, S ∗ ) = 3 + (m + 1)ε C n = 2m + 2
(174) rNBR (N, SNBR ) = ((m + 1) + (m + 1)ε)/(3 + 2ε) m/3 ≈ n/6. ! $ %0 +& > 1 1|| Tj 454: ) 7 ; k . # 4
(175)
(176) k ;
(177) 45$6789(':
(178) 6 $ 789 (' 4. ! rDEC/EDD n/2
(179) ) # E5 : n = 2m + 1 N = {1, 2, . . . , 2m + 1} p1 = m − ε p2 = · · · = pm+1 = 1 pm+2 = · · · = p2m+1 = ε d1 = m di = m + i − 1 i = 2, . . . , m dm+1 = 2m − 1 di = 2m − 1 + ε i = m + 2, . . . , 2m + 1
(180) SEDD = (1, . . . , 2m + 1) F ! C1 (r) r m+r−1−ε r = 1, . . . , m + 1 C1 (r) = 2m + (r − m − 2)ε r = m + 2, . . . , 2m + 1. 7 r = 2, . . . , m + 1 # dr + pr > C1 (r); r = m + 2, . . . , 2m # C1 (r) > dr+1 F 2m + 1 " µ = (1, . . . , 2m + 1) θ = (2, . . . , 2m + 1, 1) T (N, µ) = (1 − ε) + εm(m − 1)/2 − (m − 2)ε + (m + 1) T (N, θ) = m + (m − 2)ε ? T (N, µ) > T (N, θ) 2m + 1 m + (m − 2)ε ! " (2, . . . , 2m) # # # 8 SDEC/EDD = θ > " S ∗ = (1, m + 2, . . . , 2m + 1, 2, . . . , m + 1) T (N, S ∗ ) = 1 + m(m − 1)ε ? ε rDEC/EDD (N, SDEC/EDD ) = (m + (m − 2)ε)/(1 + m(m − 1)ε) ≈ m ≈ n/2. ! rDEC/MDD = rDEC/PSK = rDEC/WI n/3.
(181) ) E6 : n = 3m/2 N = {1, 2, . . . , 3m/2} p1 = m2 pi = 2m i = 2, . . . , m/2 pj = 2 j = (m/2) + 1, . . . , 3m/2 d1 = m2 di = m2 + 2(i − 1)m i = 2, . . . , (m/2) − 1 dm/2 = 2m2 − 2m − ε dj = 2m2 − 2m + ε j = (m/2) + 1, . . . , 3m/2 F 3m/2 # ( + 6 " (1, 2, . . . , 3m/2)
(182) m(m + 1) − (m − 1)ε = m2 + m − (m − 1)ε ( m2 # m2 ? ε ( SDEC/MDD = (2, . . . , 3m/2, 1) T (N, SDEC/MDD ) = m2 ?
(183) " ) S ∗ = (1, 2, . . . , (m/2)−1, (m/2)+1, . . . , 3m/2, m/2)
(184) 2m+ε C n = 3m/2
(185) rDEC/MDD (N, S) = m2 /(2m + ε) ≈ m/2 = n/3 , # 7
(186) 45789 45( 456 ) E7 . .
(187) $ . (
(188)
(189) ) 1|| Tj !
(190) #
(191)
(192) ) 1
(193) #
(194) # # ) # 2 E ) E5 E6
(195)
(196) %*& %-& . %
(197) . %& F F G ! = $00+' H6 2 E7IJ ) $ *
(198) +
(199) * / I0* %& 82 , 3 $++' H, ) J ,"
(200) " ! 0 I+ % & 4 = = $0/' H, # # ) J *
(201) +
(202)
(203) +.I+/ %& 9 C >1 F 6 > $0/' H, # # J ,"
(204) *
(205) )! $ .I %*& $0-*' H? " J 7 6 ( ! # 6, %-& F 4 ? 1 C 6 C $00' H, " 2 J ,"
(206) *
(207) +
(208) * I- %.& ! 4 6 C 6 C , B $0/' H,. # J 38(, 1 7 -/ / 6 D
(209) # 7, %/& 8 8 71 6 = 8 7 9 $00 ' H, J "
(210) ,"
(211) *
(212) +
(213) .+ +I + %0& E 7 = E B
(214) $00' H8 " J --
(215) 0 I+/ %+& = F 1 F ( $0.' H,
(216) 1J .--
(217) K 0I* %& ? 4 $0-0' HA " 2 J *
(218) +
(219) . .+I.* %& 4 = = $0..' H, # " 2 J .
(220) ) $ I % & E 7 = E B
(221) $0/' H, J *
(222) +
(223)
(224) * ..I/ %& 82 $00 ' H8
(225) J G ( $ '
(226) / -/ / .
(227) )! $ @ >1 +.I0 *.
(228) %*& 8 @ = 3 ! G $00*' HA J *
(229) +
(230)
(231) . I0 %-& C ! 7 >. , 3 $00/' H, J ,"
(232) *
(233) + 0
(234) 0 +I+- %.& 82 8 6 1## $00-' H J *
(235)
(236)
(237)
(238) 0 I*+ %/& > , 8 3 $00.' H, "
(239) 2 J "
(240) ,"
(241) *
(242) +
(243) 0- *I*. %0& ! $00' H, "
(244) 1 ( 6 " J $"
(245) - "
(246) !
(247) 0 ! - I-- %+& 7 9 $00' H! :
(248) ) J *
(249) +
(250) +*I+. -.
(251)
Documents relatifs
In the first call to MERGE RIGHT NODES, the left merge is applied to the first element of rnodes (line 2), all the children nodes of nodes in rnodes not involved in right nor
Dominance rules for the parallel machine total weighted tardiness scheduling problem with release dates.. Antoine Jouglet,
In this paper, we introduced several combinatorial optimization methods which can be used to get lower bounds for the Vehicle Routing Problem with Time Windows (VRPTW)..
A multi-start Dynasearch algorithm for the time dependent single- machine total weighted tardiness scheduling problem... single-mahine total weighted tardiness
Guohua and Benjamin [13] consider a single machine scheduling problem with dual cri- teria, i.e., the minimization of the total weighted earliness subject to minimum number of
All jobs of the same batch start and are completed simultaneously, i.e., at the starting time (respectively at the completion time) of the batch. On a serial batching machine,
For the general problem with asymmetric and task-dependent costs, we use a convexity property of the cost function of an allo- cated block and a strong necessary condition on
Study, Vol. CROWDER, Validation of Subgradiant Optimization, Math. LAURIÈRE, Un Algorithme pour les Problèmes de Recouvrements, R.A.I.R.O., Vol. KARP, The Travelling Salesman