• Aucun résultat trouvé

Lower bounds on the approximation ratios of leading heuristics for the single machine total tardiness problem

N/A
N/A
Protected

Academic year: 2021

Partager "Lower bounds on the approximation ratios of leading heuristics for the single machine total tardiness problem"

Copied!
9
0
0

Texte intégral

(1)Laboratoire d'Analyse et Modélisation de Systèmes pour l'Aide à la Décision CNRS UMR 7024. CAHIER DU LAMSADE 193 Avril 2002. Lower bounds on the approximation ratios of leading heuristics for the single machine total tardiness problem F. Della Croce, A. Grosso, V. Th. Paschos.

(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)

Références

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