A Survey of The State-Of-The-Art of Common Due Date Assignment and Scheduling Research

European Journal of Operational Research 139 (2002) 125
www.elsevier.com/locate/dsw
Invited Review
A survey of the state-of-the-art of common due date assignment

and scheduling research
Valery Gordon
a
a,*
, Jean-Marie Proth b, Chengbin Chu
Institute of Engineering Cybernetics, National Academy of Sciences of Belarus, Surganova 6, 220012 Minsk, Belarus
b
SAGEP Project, INRIA-Lorraine, Ile du Saulcy, 57012 Metz, France
c
Universit
e de Technologie de Troyes, Troyes, 12, rue Marie Curie 10010, France
Received 24 March 1999; accepted 27 April 2001
Abstract
We aim at providing a unied framework of the common due date assignment and scheduling problems in the
deterministic case by surveying the literature concerning the models involving single machine and parallel machines.
The problems with due date determination have received considerable attention in the last 15 years due to the introduction of new methods of inventory management such as just-in-time (JIT) concepts. The common due date model
corresponds, for instance, to an assembly system in which the components of the product should be ready at the same
time, or to a shop where several jobs constitute a single customers order. In the problems under consideration, the
objective is to nd an optimal value of the common due date and the related optimal schedule in order to optimize a
given criterion based on the due date and the completion times of jobs. The results on the algorithms and complexity of
the common due date assignment and scheduling problems are summarized. 2002 Elsevier Science B.V. All rights
reserved.
Keywords: Scheduling; Sequencing; Due date assignment; Common due date
1. Introduction
The scheduling problems involving due dates
are of permanent interest. One of the rst scheduling investigations undertaken by Jackson (1955)
*
Corresponding author. Tel.: +375-172-842-120; fax: +375172-318-403.
E-mail addresses: gordon@newman.bas-net.by (V. Gordon), Jean-Marie.Proth@loria.fr (J.-M. Proth), chengbin.chu@univ-troyes.fr (C. Chu).
was the starting point of this type of research. The

subsequent publications of survey papers on specic aspects of due date scheduling problems as,
for instance, the works of Sen and Gupta (1984),
Raghavachari (1988), Cheng and Gupta (1989),
Baker and Scudder (1990), Koulamas (1994),
conrm this interest. The problems with the objective functions involving lateness, tardiness and
earliness with respect to the given due dates take
up a signicant place in the recent reviews on
scheduling problems presented by Gupta and
Kyparisis (1987), Kawaguchi and Kyan (1988),
0377-2217/02/$ - see front matter 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 7 - 2 2 1 7 ( 0 1 ) 0 0 1 8 1 - 3
V. Gordon et al. / European Journal of Operational Research 139 (2002) 125
Cheng and Sin (1990), Lawler et al. (1993), Hoogeveen et al. (1997), Chen et al. (1998). The
scheduling books which have been published in the
1990s also contain extensive lists of references for
the scheduling and sequencing problems with due
dates as, for instance, Blazewicz et al. (1993),
Morton and Pentico (1993), Tanaev et al.
(1994a,b), Brucker (1998, 1st ed. in 1995), Chretienne et al. (1995), Pinedo (1995), Blazewicz et al.
(1996). But there is only one survey by Cheng and
Gupta (1989) where the due date assignment decisions are of primary interest. Some aspects of
due date assignment are considered by Baker and
Scudder (1990) who give a comprehensive review
of scheduling problems with non-regular performance measure. Both reviews cover only the results published before the 1990s.
In this paper, we aim at providing a unied
framework of the common due date assignment
problems in the deterministic case by surveying the
literature concerning the models involving single
machine and parallel machines. We pay particular
attention to the single and parallel machine shops
since there are quite few results on open, ow and
job shop due date assignment (see, for example,
Jozefowska et al. (1994); Sarper (1995); Della
Croce et al. (2000); Blazewicz et al. (2000) for a
common due date two-machine ow and open
shop problems).
We focus on static production settings in
which a xed set of jobs is available for processing as opposed to dynamic production settings where jobs continuously arrive in the
system and should be scheduled on-line. In
particular, we consider the models which concern
the scheduling of n jobs (the activities) on m
machines (the resources), where a machine can
process at most one job at a time, and the objective is to optimize some function of the job
completion times and the due dates.
In early papers, the computer simulation techniques were applied to determine the better due
dates (Conway, 1965; Eilon and Chowdhury,
1976; Weeks and Fryer, 1979). An overview of the
studies which examine (using simulation) the relative performance of simple heuristic rules for the
due date assignment can be found in the survey by
Cheng and Gupta (1989).
Seidmann et al. (1981) and Panwalkar et al.

(1982) were the rst who considered the optimal
due date assignment problems together with
scheduling decisions and tackled the problems
analytically. They noted that job scheduling can be
approached either from the viewpoint of the shop
manager or from the viewpoint of the customer
placing the job orders. When the shop viewpoint is
taken, the objective of the scheduling is usually
related to the minimization of one or more cost
functions associated with the state of the shop
(work-in-process, machine load, etc.), while it is
more likely related to the due date from the customer viewpoint.
The due date assignment problems make practical sense when a rm oers a due date to its
customers during sale negotiations and has to oer
a price reduction when the due date is far away
from the expected one. In many instances, due
dates are negotiated rather than simply dictated by
the customers. The later the due dates are xed,
the higher the probability that the product will be
completed or delivered on time. In order to
maintain a good image among the customers,
many companies tolerate reasonable holding costs
in favour of keeping the established due dates.
Thus, the decision maker has to balance the losses
resulting from the holding costs and the advantages of fullling the orders in time.
The problems with due date determination have
received considerable attention in the last 15 years
due to the introduction of new methods of inventory management such as just-in-time (JIT) concepts. In JIT systems (Monden, 1983), jobs are to
be completed neither too early nor too late which
leads to the scheduling problems with both earliness and tardiness costs and assigning due dates.
T.C.E. Cheng, who contributed a lot to the due
date assignment and the related scheduling approaches, remarks that completing a job early
means to bear the costs of holding unnecessary
inventories, while nishing a job late results in
contractual penalty and loss of customer goodwill (Cheng, 1988a).
Let dj denote the due date of job j (to be assigned). Several models of assigning due dates are
considered in the scheduling literature; some of
them are mentioned below. The simplest is the
model in which all jobs have a common due date

(i.e., dj d; j 1; . . . ; n). Such a model corresponds, for instance, to an assembly system in
which the components of the product should be
ready at the same time, or to a shop where several
jobs constitute a single customers order. More
generally, this model corresponds to any system in
which, for some reason (appointment, technical
constraints, etc.), several tasks should be completed at the same time. This method of due date
assignment is known in the scheduling literature as
CON model, where CON stands for constant ow
allowance. It is also called the common due date
model.
In the SLK due date assignment (SLK stands
for slack), a ow allowance that reects equal
waiting time or slack, denoted by q, is assigned to
the jobs. Thus, dj pj q. In this model, the goal
is to nd an optimal value of the slack q with respect to the criterion to be optimized.
In the TWK due date assignment (TWK stands
for total work), due dates are based on total work
content and are equal to a multiple of the job
processing times, i.e., dj kpj . The optimal due
date assignment consists of nding the value of the
common multiplier k.
The PPW due date assignment (processing plus
wait) combines SLK and TWK due date assignment rules. In this model, dj kpj q. The optimal due date assignment consists of nding the
optimal values of k and q.
In the NOP (number of operations) model, due
dates are determined on the basis of the number of
operations nj to be performed on job j: dj knj .
The goal, in this case, is to nd a value of k which
yields optimum to the objective function.
In the class of generalized due date scheduling
problems identied by Hall (1986a), Hall et al.
(1991b), the set of due dates is given, and the
problem consists in assigning the jobs to the due
dates. When k jobs should be performed by the kth
due date, such due dates may be called positional
due dates (Hoogeveen et al., 1997). For instance, if
three due dates d1 ; d2 ; d3 , where d1 < d2 < d3 , are
given for the set of three jobs, the objective would
be to complete at least one job by d1 , at least two
jobs by d2 , and all three jobs by d3 . This method of
due date assignment describes the situation where
what matters is how many jobs have been completed by any point in time, rather than which jobs
they are.
The survey of the results on the SLK, TWK,
NOP and PPW due date assignment models as
well as on the positional due dates is given by
Gordon et al. (1998).
In the RDM (random allowance) method, each
job receives a random due date, usually following a
probability distribution; in the JIQ (jobs in queue)
and in the JIS (jobs in system) methods, due dates are
determined based on, respectively, the current job
queue length and the number of jobs in the system.
These due date assignment rules are usually studied
by simulation models. Some recent results related to
these approaches may be found in Ramasesh (1990),
Vig and Dooley (1991, 1993), Gee and Smith (1993),
Smith et al. (1995), Lawrence (1995), Roman and
del Valle (1996), Philipoom et al. (1997), Tsai et al.
(1997).
This paper is organized as follows. In Sections 2
and 3, respectively, we consider the common due
date model for the single machine and for parallel
machines. We assume that all jobs are simultaneously available (i.e., the release date is the same
for each job) and all the jobs have a common due
date d which is to be assigned. The objective is to
nd an optimal value of this due date and the related optimal schedule in order to optimize a given
criterion based on the due date and the completion
times of jobs. The subdivision of each section reects the consideration of diverse objective functions. In Section 4, the results on complexity of the
common due date assignment and scheduling
problems are summarized.
The following notations are used in the rest of
the paper.
If Cj denotes the completion time of job j in a
certain schedule, then the earliness and the tardiness of job j are dened, respectively, as

Ej max 0; dj Cj and Tj max 0; Cj dj :
1
Lateness Lj of job j is dened as Lj Cj dj .
Hence, Ej maxf0; Lj g and Tj maxf0; Lj g.
All coecients (a; b; . . .) of the objective functions below are non-negative constants.
2. Single machine
Consider the problem of scheduling n jobs on a
single machine. A processing time pj is associated
with each job j, j 1; . . . ; n; and preemption is not
allowed. Let r 1 ; 2 ; . . . ; n be an arbitrary
sequence of the jobs where j is the jth job in r.
Let us introduce some additional notations
useful for further discussion. Assuming (throughout this section) that the jobs are labelled in the
non-decreasing order of their processing times, we
dene the ordered sets A and B as follows:
B fn; n 2; . . . ; 1g; A f2; 4; . . . ; n 1g when n
is odd, and B fn; n 2; . . . ; P
2g, A f1; 3; . . . ;
n 1g when n is even. Let D j2B pj .
decreasing order of the pj values (shortest processing time or SPT order).
2.2. Panwalkar, Smith and Seidmann due date

assignment problem
Panwalkar et al. (1982) were the rst to consider the MAD problem as a due date assignment
problem. Actually, they consider the problem with
a more general objective function than MAD,
namely, the problem of nding a sequence r and a
common due date d which lead to the minimal
value of
f d; r
2.1. Mean absolute lateness

In the case where the due date is given, the
problem usually considered in the literature aims
at minimizing the mean absolute deviation of
completion times about a common due date
(MAD) or, which is the same, at minimizing the
mean absolute lateness. The MAD problem is
dened as follows: nd a sequence r and a
starting time P
t to minimize the penalty function
given by 1=n nj1 jCj dj. Without loss of generality, we may remove the constant 1=n and consider the objective
Pn function of the MAD problem
as MADr j1 jCj dj. This objective function may be P
easily rewritten in the following way:
n
MADr j1 Ej Tj , where Ej maxf0;
d Cj g, Tj maxf0; Cj dg.
Kanet (1981)
P shows that, for a given common
due date d P nj1 pj , the schedule B; A with the
starting time t d D is optimal for the MAD
problem.
Bagchi et al. (1986) show that Kanets result
holds for any due date d P D and gave a procedure for nding all optimal solutions of the
problem. Raghavachari (1986) shows that for
any value of the due date d, an optimal sequence
for this problem is V-shaped. A V-shaped sequence is such that a subset of jobs placed in the
non-increasing order of the pj values (longest
processing time order, known as LPT) is followed by the remaining jobs in the non-
n
X
aEj bTj cd:
j1
When c 0 and a b 1 we have the MAD

problem where due date is to be assigned.
It is easy to show (Panwalkar et al., 1982) that
in the case where c P b, the SPT order is optimal
and d 0. If c < b, the following properties
(Panwalkar et al., 1982) lead to a polynomial-time
algorithm for this problem.
Property 1. For any specified sequence r, there
exists an optimal value of d which coincides with a
completion time of one of the jobs in r.
Property 2. For any specified sequence r, there
exists an optimal due date equal to CK , where K is
the smallest integer value greater than or equal to
nb c=a b, and exactly K jobs will be nontardy.
So, the total penalty is equal to f CK ; r and,
introducing CK p1 p2 pK , we get
f CK ; r
K
X
nc j 1ap j
j1
n
X
bn 1 jp j
jK1
n
X
j1
aj p j ;
where aj is equal to nc j 1a for j 6 K and to

bn 1 j for j > K. To nd the optimal sequence, we need to nd the minimal penalty
among all sequences r. The term aj may be viewed
as the positional penalty for the job in position j
in r or, in other words, the penalty resulting from
the job in position j.
Pn
Property 3. Quantity
j1 aj pj is minimized by
matching the smallest value of a with the largest
value of p, the next smallest value of a with the next
largest value of p, and so on.
Note that this property follows from the wellknown result in linear algebra (Hardy et al., 1934)
and leads to an On log n algorithm (Panwalkar
et al., 1982). It nds rst the number of non-tardy
jobs (K), and then calculates the positional penalties a, the optimal sequence r and the optimal due
date d . The sequence generated by this algorithm
is V-shaped. The optimal schedule starts at time
t 0 and has no inserted idle periods.
Cheng (1988b, 1989) notes that if K
nb c=a b, it is not necessary to constrain
d to be equal to one of the job completion times.
The optimal due date can be set at d CK
dpK1 ; 0 6 d < 1. Thus, in this case, there exists an
innite number of optimal due dates. For example,
if c 0, a b and n is even, then K n=2 and any
value between Cn=2 and Cn=21 is an optimal due
date.
Cheng et al. (1996) generalize the results of
Panwalkar et al. (1982) to the case of compressible
processing times, where the normal processing
time of each job can be compressed by an amount
xj P 0 with the unit cost bj of compression
for job
Pn
j. The problem of minimizing
aE
bTj
j
j1
cd bj xj is solved by formulating it as an assignment problem.
Biskup and Jahnke (2001) P
consider the problem
of minimizing f d; r nj1 aEj bTj cd
with jointly reducible processing times: all processing times are reduced by the same amount, and
the cost kx for operating the machine is introduced which is a monotonous increasing function
in the proportion of the processing time reduction
x. They propose a procedure to nd optimal values
of d and x using the results of Kanet (1981) and
Panwalkar et al. (1982) for nding the optimal

schedule.
2.3. MAD problem with due date assignment

Returning to the MAD problem with due date
assignment (the special case of the considered
problem with c 0, a b 1), we derive from
Property 2 that an optimal due date is equal to the
completion time of the Kth job, where K is the
smallest integer value greater than or equal to n=2,
i.e., K n=2 if n is even, and K n 1=2 if n is
odd.
The positional penalty a for the job in position j
for the MAD problem is j 1 if j 6 K and
n 1 j if j > K. Hence, the penalties are the
same for positions 2 and n, for 3 and n 1, for 4
and n 2, and so on. Therefore, according to the
Property 3, there are two ways of matching values
of a with processing times and hence of assigning
jobs to the positions of the optimal sequence.
Consequently, the total number of optimal sequences (when the processing times are unique) is
2r , where r n 1=2 if n is odd, and r n=2 if n
is even (Bagchi et al., 1986; Hall, 1986b; Karacapilidis and Pappis, 1995a). Each of these
schedules (sequences) has its own set of optimal
due dates.
So, there may be dierent optimal due dates
and dierent optimal sequences. Moreover, increasing d by a given value t > 0 and increasing
the starting times of the jobs in the optimal sequence by the same value t lead to another optimal
solution.
Bagchi et al. (1987a) nd that the minimal
completion time of the Kth job in the optimal
schedule (i.e., the minimum sum of processing
times of non-tardy jobs) is D. So, the minimal
value of optimal due date is d D.
The complexity of the MAD problem changes
drastically when the admissible due dates are upper bounded by D with D < D. Such a problem is
called restricted common due date problem as
opposed to the unrestricted problems where d P D
(note that, in the unrestricted problem, the optimal
cost cannot decrease when the common due date
increases). For the restricted problem, Properties
13 do not hold any longer, and the optimal solution may contain a straddling job which starts
before d and is completed after d (but there still
exists an optimal schedule which is V-shaped). The
following useful property holds for this problem
(Hall et al., 1991a).
Property 4. Either there exists a job in an optimal
schedule that completes at the due date or the
schedule starts at time zero.
Considering the problems without due date
assignment, Hall et al. (1991a) and Hoogeveen and
van de Velde (1991) prove that the restricted MAD
problem is NP-hard.
For this problem, Bagchi et al. (1986) propose
an algorithm (enumerative in nature, which is
eective at solving problems up to 15 jobs in
size) that leads to an optimal solution under the
assumption that the starting time of the schedule
is zero. Note that the introduction of this assumption is restrictive since, due to Property 4,
the best solution may include the delayed start.
A sucient condition (depending on the processing times and due date) for a schedule to be
optimal when starting at time zero is given by
Szwarc (1989), as well as a branch-and-bound
procedure to deal with the problems up to 25
jobs in size. In the case of a zero starting time,
Sundararaghavan and Ahmed (1984) propose a
heuristic
P procedure. Hall et al. (1991a) give an
On pj algorithm without the zero starting
time assumption. This pseudopolynomial-time
algorithm allows to deal with the problems up to
1000 jobs. Ventura and Weng (1995) show that
two out of the three subroutines in the algorithm
proposed by Hall et al. (1991a) can be eliminated and, consequently, that the total computational eort can be reduced. Hoogeveen et al.
(1994) present a branch-and-bound algorithm
based on Lagrangean lower and upper bounds
(which is eective at solving problems up to 50
jobs in size) and an On log n 4/3-approximation
algorithm for the MAD problem. Performance
guarantee 4/3 means that, for any instance, the
approximation algorithm produces a solution
with objective function value at most 4/3 times
the optimal value.
Hall et al. (1991a) also give an On log n algorithm for the special case of the restricted MAD
problem,
where there exists l P dn=2e satisfying
Pl
p
<
d and pl1 ; . . . ; pn > 2d. Another special
j1 j
case solvable in On log n time is obtained when d
is so small that, whichever the schedule considered,
no job is early (Bagchi et al., 1986).
2.4. Weighted sum of absolute deviation
Let us now consider the problem (derived from
the one studied by Panwalkar et al. (1982) when
c 0), where a and b may be dierent. The goal is
to minimize the weighted sum of absolute deviation (WSAD) of the completion times with regard
to a common due date. The objective function for
the WSAD problem may be written as
X
X

C j d
WSADd; r a
jCj dj b
j2E
X
j2T
d Cj b
j2E
X
Cj d
j2T
n
X
aEj bTj :
j1
Here and hereafter, E fjjCj 6 dg is a set of jobs

that are not tardy and T fjjCj > dg is a set of
jobs that are tardy.
Like the MAD problem in Section 2.3, this
problem has dierent optimal due dates as well as
an innite number of optimal solutions obtained
by increasing the optimal due date and the starting
times of the jobs by the same value. Bagchi et al.
(1987a) give an On log n algorithm alternative to
the matching procedure of Panwalkar et al. (1982).
This algorithm yields a unique sequence with the
minimal value of the due date among the optimal
due dates.
Panwalkar and Rajagopalan (1992) consider
the WSAD problem with controllable processing
times. The processing time pj of job j may be
chosen in the interval tj0 ; tj , where pj tj is the
normal time but pj can be reduced as long as it
remains greater than tj0 . In this case, the cost incurred by such reduction is kj per time unit. If xj is
the actual amount of time by which job j is compressed, then kj xj represents the processing cost of
job j. The objective is to nd a set of processing

times, a sequence and the smallest due date to
minimize the cost function
n
X
f d; r
aEj bTj kj xj :
j1
Panwalkar and Rajagopalan (1992) show that there

exists an optimal sequence with pj 2 ftj0 ; tj g (that is,
without partial compression), and propose an On3
algorithm to nd the optimal solution by formulating the problem as an assignment problem.
The restricted WSAD problem (when the values of due date have a given upper bound D and
D < D) is NP-hard due to NP-hardness of the
restricted MAD problem. For this problem,
Bagchi et al. (1987a) propose branching procedure which is suitable for the problems up to 15
jobs in size and gives an optimal solution only
under the assumption that the processing starts at
time zero.
Liman et al. (1998) consider the WSAD problem in which jobs have a common due window
d; d D; d P 0; D P 0, such that any job completed on or before d is early with earliness
Ej maxf0; d Cj g, and any job completed after
d D is tardy with tardiness Tj maxf0;
Cj d Dg. An On log n algorithm is proposed
for nding the optimal size and location of the
window as well as an optimal sequence to minimize the cost function
f d; D; r
n
X
aEj bTj dd cD:
j1
2.5. Total weighted earliness and tardiness

A natural generalization of MAD and WSAD
problems is the problem which consists in minimizing the total weighted earliness and tardiness
(TWET) about a common due date. This problem
may be formulated as a due date assignment and
scheduling problem in the following way: nd a
sequence r and a common due date d to minimize the objective function
n
X
TWETd; r
aj Ej bj Tj ;
j1
where for job j, aj is a unit earliness penalty

(weight) and bj is a unit tardiness penalty (weight).
If aj bj for each j 1; . . . ; n, the earliness and
tardiness penalty weights are said to be symmetric.
Hall and Posner (1991) prove that the TWET
problem is NP-hard even in unrestricted setting
withP symmetric weights. They propose an
On pj dynamic programming algorithm that
solves to optimality large instances of the problem.
De et al. (1990a) propose two dynamic programP
ming procedures
to solve the problem in On pj
P
and On aj time and present a 01 quadratic
programming formulation which leads
P to exact
and P
heuristic procedures. An On pj and an
On aj dynamic programs are proposed for this
problem by Jurisch et al. (1997). P
Hoogeveen and
van de Velde (1991) give an On2 pj algorithm
for the restricted variant of this problem. Since
these algorithms are pseudopolynomial, the
TWET problem with symmetric weights is NPhard in the ordinary sense.
For a given job sequence, the TWET problem
consists in determining a schedule by inserting idle
time between jobs and is polynomially solvable even
with dierent due dates either for aj bj or aj 6 bj
(Garey et al., 1988; Davis and Kanet, 1993).
Cheng (1985, 1987a), Quaddus (1987), Bector
et al. (1988), Baker and Scudder (1989), Hall and
Posner (1991) investigate the properties of the
unrestricted TWET problem. Most of these
properties are similar to Properties 1 and 2 in
Section 2.2 and can be formulated in the following
way (Cheng, 1985, 1987a).
Property 5. There exists an optimal schedule which
is:
(a) V-shaped in the sense that the jobs in set E
(non-tardy jobs) are sequenced in the non-increasing order of the ratio pj =aj and followed by the
jobs of set T (tardy jobs) in the non-decreasing order of the ratio pj =bj ,
(b) the optimal due date coincides with the completion time of the last non-tardy job in an optimalP
sequence, P
and
(c) j2T bj 6 j2E aj .
Property 5a is valid also for the restricted problem
(excluding the straddling job), however 5b and 5c
do not hold in the restrictive case.
Hall and Posner (1991) develop a fully polynomial approximation scheme for the unrestricted
TWET problem with symmetric weights when the
maximum weight is bounded by a polynomial
function of n. However, this special case of the
problem is established to be polynomially solvable
by Jurisch et al. (1997). For the TWET problem
without bounded maximum weight assumption, a
fully polynomial approximation scheme is proposed by Kovalyov and Kubiak (1999). This
scheme does not require any prior knowledge of
lower and upper bounds on the value of a complete optimal solution since it recursively computes
lower and upper bounds on the value of partial
optimal solutions.
Cheng (1990a) presents an On1=2 2n local
search algorithm for determining an optimal sequence and the corresponding common due date
for the unrestricted TWET problem with symmetric weights. Quaddus (1987) considers the
separate earliness and tardiness penalties for each
job (aj 6 bj ) and shows that, once the job sequence is given, an optimal due date can easily be
assigned. To nd both the optimal due date and
the optimal sequence in the case when aj 6 bj ,
branching procedures (Dileepan, 1993; De et al.,
1994a) and heuristic algorithms (Gupta et al.,
1990; Dileepan, 1993; De et al., 1994a) are proposed. Among them, the most convincing results
of computational experiment are due to De et al.
(1994a), who propose a greedy randomized
adaptive search heuristic procedure. The procedure consists in two phases: an initial solution is
constructed in the rst phase through controlled
randomization, and is improved in the second
phase through steepest descent neighbourhood
search.
For the unrestricted TWET problem with
asymmetric weights, Van den Akker et al. (2000)
present a combined column generation and Lagrangean relaxation algorithm that solves instances with up to 125 jobs to optimality.
James (1997) proposes to use tabu search for
both restricted and unrestricted TWET problems.
Hao et al. (1996) consider the common due date
determination and sequencing using tabu search
for the TWET problem with symmetric weights;
for asymmetric weights, Lee and Kim (1995) pro-
pose a genetic algorithm. Note that James (1997)

and Lee and Kim (1995) restrict their search to
sequences starting at time zero, and it can lead to
excluding optimal schedules from the search process. Biskup and Feldmann (2001) and Feldmann
and Biskup (1999) construct heuristics for the
TWET problem and develop a problem generator
to obtain the benchmark instances of the problem
(changing a parameter to calculate more or less
restrictive common due date).
Lee et al. (1991) present a pseudopolynomial
time procedure for the special case of the TWET
problem with asymmetric weights when the jobs
have agreeable ratios, which means that
pi =ai < pj =aj implies pi =bi < pj =bj for all i and j.
Szwarc (1996) proposes several improvements of
this procedure.
Hall and Posner (1991) nd some polynomially
solvable special cases of the unrestricted TWET
problem. In particular, an On algorithm solves
the problem with aj bj pj , j 1; . . . ; n, and an
On log n algorithm solves the problem with
aj bj ; pj 1, j 1; . . . ; n. Dependence of the
weights on processing times relates the signicance
of a job to the production time spent to complete
it. Pappis and Adamopoulos (1993) propose an
algorithm in case of aj kpja , bj kpjb , where
k > 0, a and b are the non-negative integers. For
the problem where weights are proportional to the
processing times of the corresponding jobs
(aj bj kpj , k > 0, j 1; . . . ; n), Karacapilidis
and Pappis (1995b) propose an algorithm to nd
r!n r! alternative optimal sequences, where r is
a position of job whose completion time coincides
with the optimal due date. The optimal due date
and each sequence are found in polynomial time.
The unrestricted TWET problem with almost
common due date (where the due dates of the jobs
are dierent but lie between a large constant d and
d pj for each job j) is considered by Hoogeveen
and van de Velde (1997) who propose an On2
dynamic programming algorithm.
Hoogeveen and van de Velde (1991) study the
following polynomially solvable special cases for
the restricted TWET problem: (a) pj p for each
job j, j 1; . . . ; n; (b) pj wj for each job j,
j 1; . . . ; n. For both cases, On log n algorithms
are proposed.
2.6. Tolerance intervals

Cheng (1988a) was the rst who considered
common due date assignment and scheduling
problem by introducing a completion time deviation allowance e. This problem models the real life
situation when penalty does not occur if the deviation of job completion time from the due date is
small enough. Assuming that parameter e satises
the condition 2e < minfpj j1 6 j 6 ng, Cheng
(1988a) proposes an On log n algorithm to nd
an optimal due date and an optimal sequence for
the objective function
f d; e; r
n
X
Ej U Ej e Tj U Tj e;
j1
where U x e 0 if x 6 e, and U x e 1
otherwise.
Dickman et al. (1991) show that there exist alternative optima which Cheng (1988a) did not nd
out, i.e., for an optimal due date there exist interval solution Cn1=2 e; Cn1=2 e if n is
odd, and two solutions Cn=2 e and Cn=21 e
if n is even.
Weng and Ventura (1994) and Wilamowsky
et al. (1996) relax Chengs restrictive requirement
on parameter e and proposed polynomial-time
algorithms in the case when no penalty costs are
incurred for job completed within limits of any
prespecied size around the common due date.
So, the tolerance intervals of any size are considered instead of Chengs small due date tolerance allowance where at most one job can avoid
the penalty. Unlike Cheng (1988a), more than
one job can avoid the penalty in this case.
Wilamowsky et al. (1996) consider also a case
when tolerance intervals for earliness and tardiness are not identical.
Unlike Cheng (1988a) where, outside the tolerance interval, the earliness and tardiness penalties are calculated from the due date rather than
from the end of the tolerance interval, Weng and
Ventura (1994) consider a penalty reduced by
d e compared to the one proposed by Cheng. As
pointed out by Baker and Scudder (1990), such an
earlinesstardiness measurement is more conventional and consistent.
Baker and Scudder (1990) generalize Chengs

approach to nd the properties of an optimal
schedule for the TWET due date assignment problem when completion time deviation allowances are
dierent: job j avoids penalty if its completion time
falls into the interval d ej ; d tj with the requirement that pj tj ej > 0 for all pairs of jobs
i; j. The last inequality presupposed that the due
date tolerance interval is relatively small compared
to the processing times of the jobs.
The period of time within which jobs incur no
penalty is often called a due window (with the earliest due date as the start of the window, and the
latest due date as the end of it). When location of the
common window is a decision variable, On log n
algorithms are proposed for the WSAD problem
(Kramer and Lee, 1993) and for the problem of
minimizing the weighted sum of earliness, tardiness
and due window location penalties (Liman et al.,
1996). Yeung et al. (2001) propose a pseudopolynomial dynamic programming algorithm for the
problem with the objective function which includes
the sum of weighted earlinesstardiness, weighted
number of early and tardy jobs and common due
window location penalty (when the ratios of the job
processing times to the earlinesstardiness penalties
are agreeable). Azizoglu and Webster (1997a) propose a two-step branch and bound algorithm for the
TWET problem with an unrestricted common due
window to nd optimal solutions for the problem
instances up to 30 jobs in size.
2.7. Quadratic penalty functions

Another way to take into account the situation,
when only small deviations of the completion time
from the due date are acceptable, is to consider the
problem of minimizing the mean squared deviation (MSD) of job completion times about a
common due date, i.e. the problem of minimizing
MSDd; r
n
X
j1
Cj d
n
X
Ej2 Tj2 :
j1
Being a quadratic penalty function, the MSD penalizes larger deviations at a higher rate.
10
For
Pn the unrestricted MSD problem (with
d P j1 pj ), where an increase in the due date
does not result in any further decrease in MSD,
Bagchi et al. (1987b) nd the property which followed from the results well known in statistics.
Property 6. The optimal due date that minimizes
the MSD for any given schedule
P is equal to the mean
completion time C 1=n nj1 Cj .
Therefore, the unrestricted MSD problem is
equivalent to the problem ofPminimizing complen
2
tion time variance CTV j1 Cj C which
was considered by Merten and Muller (1972). The
CTV problem is shown to be NP-hard by Kubiak
(1993). Therefore, the unrestricted MSD problem
is also NP-hard.
For the CTV problem, Gupta et al. (1993)
propose a heuristic procedure based on genetic
algorithms. De et al. (1992) and Kubiak (1995)
present pseudopolynomial-time dynamic programming procedures. Cai (1995) derives a fully
polynomial approximation scheme for the CTV
problem.
The CTV problem reduces to the problems of
nding a maximum cut in a complete weighted
graph with non-negative weights (Kubiak, 1995)
and of nding a minimum weighted clique
(MinClique) in a complete weighted graph with
negative weights and self-loops allowed (Jurisch
et al., 1997). Jurisch et al. (1997) propose a family
of heuristics to solve the MinClique, the CTV and
the symmetric TWET problems.
It is worth noting that, for the unrestricted
problem, MSD is invariant to any feasible linear
translation of a schedule with its due date (the
same as for the unrestricted MAD problem). Let
us denote the optimal MSD by z . Then there exists a minimal due date d for which an MSD equal
to z is the solution. Bagchi et al. (1987b) propose
a procedure based on the CTV problem to nd d .
The restricted MSD problem is the problem
where the values of due date have a given upper
bound D with DP< D. Kahlbacher (1989, 1993)
proposes an On pj algorithm for the restricted
problem to minimize
Pn a more general objective
function, that is, j1 aEjc bTjc for an arbitrary
positive c. Enumerative procedures are proposed
by Bagchi et al. (1987b) for the restricted MSD

problem and by Bagchi et al. (1987a) for the restricted problem to minimize
the weighted sum of
P
squared deviations nj1 aEj2 bTj2 with the assumption that schedule begins at time zero. This
unnecessary assumption is removed by De et al.
(1989, 1990b) in their enumerative procedure.
Using dynamic programming, De et al. (1993)
provide a solution methodology with a pseudopolynomial complexity
with
Pnfor 2the problems
P
n
2
objective functions
E
and
j
j
j1
j1
2
2
aEj bTj , as well as for the restricted MAD and
WSAD problems considered in Sections 2.3 and 2.4.
2.8. Arbitrary earlinesstardiness penalties
Cai et al. (1997) consider a general model, in
which earliness and tardiness penalties are arbitrary non-decreasing functions. In this case, the
problem consists in nding an optimal due date d
and the corresponding sequence r to minimize the
objective function
X
X
f d; r
ej aj hLj
tj bj gLj ;
j2E0
j2T
where g and h are the non-decreasing real-valued

functions, E0 fjjCj < dg and Lj is a lateness of
job j. Here ej P 0 (or tj P 0) represents a xed cost
incurred when job j is early (or tardy).
Cai et al. (1997) show that the problem has a
W-shaped optimum when it is agreeably weighted,
in the sense that pi > pj implies ai 6 aj and bi 6 bj
for all i, j. A schedule is said to be W-shaped with
respect to the processing times if pj P pk when
Cj < Ck < d, and pj 6 pk when d 6 Sj < Sk , where
Sj Cj pj denotes the starting time of job j. In a
W-shaped schedule, it is possible to have a job m
with Sm < d 6 Cm , called straddling job, such that
pm > pm0 and pm > pm00 , where jobs m0 and m00 are,
respectively, the jobs sequenced immediately before and after job m.
As a motivation for the above agreeable
condition, Cai et al. (1997) indicate that, in
practice, a smaller job is charged at a higher
rate, which in turn makes it relatively more important, and thus the penalty of missing its due
date becomes heavier. They propose two dy-
namic programming algorithms for the problem

with the agreeable weight condition.PThe rst
one leads to an optimum in time On2 pj , and
the second
P one can generate a suboptimum in
time On pj . The second algorithm leads to
an optimal solution in the special cases when (a)
ai aj , bi bj , ei ej , ti tj for all i and j, or
(b) ai aj , bi bj for all i and j, and ej tj for
all j. An upper bound on the relative error of the
sub-optimal solution is derived, and it is shown
that, under certain conditions, the error tends to
zero as n increases.
Computational results by Cai et al. (1997)
demonstrate the eectiveness of the proposed algorithms for the agreeably weighted problems up
to 200 jobs in size, and show the possibility to use
these algorithms as heuristic approaches for the
general problem.
Federgruen and Mosheiov (1994) consider a
special case of the above problem when
ej tj 0, aj bj 1, j 1; . . . ; n, and propose
a greedy heuristic which requires On3 elementary operations and evaluations of the functions
g and h. Their numerical experiments suggest
that the greedy algorithm performs well under
non-convex as well as convex functions g and h.
For the same case of the problem, Kahlbacher
P
(1993)
proposes
an
On log n n n1
j1 pj
pseudopolynomial algorithm. Another pseudopolynomial-time algorithm is proposed by Federgruen and Mosheiov (1993) for the problem
where a non-decreasing function cd is added to
the objective function.
2.9. Other due date assignment models
n
X

aEj bTj c max 0; dj D ;
j1
where Ej and Tj are given by (1), and c > 0.

For this model, Seidmann et al. (1981) nd that
SPT sequence is optimal and
Pj that the optimal due
dates are given by dj P
i1 pi , j 1; . . . ; n, if
j
c 6 b, and dj minfD; i1 pi g, j 1; . . . ; n, otherwise.
De et al. (1991a) formulate another problem
involving the lead time D. The problem consists in
nding a due date d, common to all the jobs, and a
sequence r to minimize the following objective
function:
X
f d; r
wj c maxf0; d Dg;
j2T
where c > 0, wj > 0, j 1; . . . ; n. For this problem, the order in which jobs are sequenced in each
of the sets E (non-tardy jobs) and T is indierent.
Let M be the sum of the processing times of the
jobs j with wj =pj > c. Then, the problem is called
restricted if M > D and unrestricted if M 6 D. De
et al. (1991a) show that the restricted problem is
NP-hard, propose a solution based on a formulation of the problem as a knapsack problem, and
give a simple heuristic based on the continuous
relaxation of the knapsack problem. A polynomial-time algorithm was presented for the unrestricted problem.
A more general objective function is considered
by Kahlbacher and Cheng (1993): the problem is
to nd a common due date d and a sequence r to
minimize the objective function
X
X
wj
hd Cj gd;
f d; r
j2T
Seidmann et al. (1981) propose another model

of due date assignment where any due date dj may
be assigned to a job j, but there is a given common
due date D, which represents the lead time that
customers consider to be reasonable and expected.
If a due date dj is within this lead time, there is no
penalty. However, each additional unit of due date
beyond the reasonable lead time causes a penalty.
So, the problem is to nd the optimal values dj ,
(j 1; . . . ; n), and an optimal sequence r to
minimize the total penalty
11
j2E
where g and h are the non-decreasing functions,

wj > 0, j 1; . . . ; n. The problem is shown to be
NP-hard even when (a) hx 0 and gd is an
arbitrary function or (b) gd 0 and hx is an
arbitrary function. Kahlbacher and Cheng (1993)
propose polynomial-time algorithms in the following cases: (a) wj w, j 1; . . . ; n (an On log n
algorithm), (b) hx 0 and gd cd, c P 0 (an
On algorithm), (c) hx and gd are linear (an
On4 algorithm). It is interesting to note that for
12
an externally given common due date (i.e., when

the due date is not a decision variable), the problem with dierent wj values is NP-hard even if
hx 0 (due to Lawler and Moore, 1969, who
showed an equivalence between this problem and
the knapsack problem). As a consequence, the
restricted variant of due date assignment problem
(where the due date is restricted by a given upper
bound) is NP-hard.
Cheng (1990b)P
considers the problem of minin
mizing f d; r j1 wUj cd, where w is the
penalty cost for nishing a job late, Uj 1 if
Cj > d and 0 otherwise, and shows that the optimal solution is obtained if jobs are sequenced according to the SPT rule and the assigned due date
coincides with the completion time of bth job so
that inequality pb 6 w=nc 6 pb1 holds. Biskup and
Jahnke (2001) generalize this result to the case of
jointly reducible processing times.
Lee et al. (1991) consider the problem which
consists of nding a common due date d and the
associated sequence r to minimize the objective
function
f d; r
n
X
aj Ej bj Tj wj Uj :
j1
The problem diers from the TWET problem

considered in Section 2.5 by an additional xed
cost for each tardy job. The problem is considered
under the assumption that pi =ai < pj =aj implies
pi =bi < pj =bj for all i and j. A practical motivation
for this condition is the expectation that, if job j is
relatively more important than job i, both its earliness and tardiness weights will be greater than
those of job i. Pseudopolynomial-time dynamic
programming algorithms
P are proposed for both
unrestricted (when d P
pj ) and restricted (when
P
due date is upper bounded by D and D <
pj )
problems, and an On2 time algorithm is proposed in the special case of unrestricted problem
with constant weights (aj a, bj b and wj w
for all j) (Lee et al., 1991).
The introduction of learning eects to job
processing times for the WSAD problem (with
additional completion time penalties) is considered
by Biskup (1999). It is assumed that normal processing time pj of job j is incurred if the job is
scheduled rst in a sequence. The processing times

of the following jobs are smaller than their processing times because of the learning eect, that is,
the processing time of job j if it is scheduled in
position r in a sequence is pjr pj rl , where l 6 0 is
the learning index, given as the logarithm to the
base 2 of the learning rate. For the problem with
f d; r
n
X
aEj bTj hCj ;
j1
it is shown that it can be solved in On3 time in

two steps. Firstly, an assignment problem which
leads to an optimal sequence of the jobs is solved.
Secondly, the optimal due date is determined like
in the problem without learning eects since introduction of these eects does not have an inuence on the number of early and tardy jobs.
Biskup and Cheng (1999a) consider the problem of nding an optimal sequence of the jobs, an
optimal due date and optimal processing times
which
jointly minimize the cost function
Pn
aE
bTj hCj gj xj consisting of the
j
j1
earliness, tardiness, completion time and compressing costs, where the job processing times are
compressible as a linear cost function. The actual
processing time of job j is pj p~j xj , where p~j is
the normal processing time which can be compressed by an amount xj with the per unit cost gj of
compressing the job. A polynomial-time algorithm
is proposed to nd an optimal sequence, an optimal compression vector and an optimal due date
by formulating the sequencing problem as an assignment problem.
Cheng (1991) considers the problem with constant due date assignment in a formulation which
diers from the problems considered above: a due
date dj for job j is dened as dj Sj d, where Sj
is the time at which the machine starts processing
job j, and d is the constant ow allowance. So, in a
sequence r 1 ; . . . ; n , the due date of the j th
job will be dj Sj d Cj1 d. For the
problem which consists in minimizing
f d; r
n
X
j1

a j C j d j cd;
Pn
where 0 6 aj 6 1; j1 aj 1, Cheng (1991) proposes a linear programming (LP) formulation
and, by considering the LP dual problem, shows
that the optimal ow allowance is independent
of the job sequence and equal to one of the job
processing times if 0 6 c < 1=n or equal to d 0
if c P 1=n.
2.10. Due date assignment with batching

In industry, the shipping costs are signicant
cost factors. Therefore, the production model
should include both the sequencing of the jobs and
the scheduling of the deliveries. When the jobs are
delivered to the customer in batches, it is useful to
consider problems where three types of decisions
are combined: scheduling, batching and due date
assignment.
Herrmann and Lee (1993) consider the
WSAD problem (see Section 2.4) with job
batching and batch delivery costs. Under the
similar introduction of job batching and batch
delivery costs, Chen (1996) considers the problem of Panwalkar et al. (1982) (see Section 2.2).
It is assumed that, after completion, the jobs are
delivered to the customer in batches. There is no
capacity limitation on the size of a batch (it may
also include only one job). All early jobs (i.e.,
the jobs with Cj < d) are delivered in one batch
at the common due date d without any delivery
cost. Each batch with a tardy job (let us call it
tardy batch) incurs a xed delivery cost h independent of the number of jobs in the batch. The
delivery date of the batch is equal to the largest
completion time among the jobs in this batch.
Let Dj be the delivery date of job j. Then, earliness and tardiness of job j are dened, respectively, as Ej Dj Cj and Tj Dj d. Let a
and b be, respectively, the earliness and the
tardiness penalty weights (identical for all the
jobs). It is assumed that a 6 b. So, for an early
job j, Ej d Cj and its penalty is aEj . A tardy
job may incur both earliness and tardiness penalties if it is not delivered immediately upon its
completion. The earliness and tardiness penalties
of a tardy job j are aEj aDj Cj and
bTj bDj d. The objective function is
f r; d; l cd hl
13
n
X
aEj bTj ;
j1
where l is the number of tardy batches. The goal is

to nd a due date d , a job sequence r and a
partition of the set of jobs into batches which
jointly minimize this objective function.
It can be easily shown that each tardy batch is a
set of consecutive tardy jobs in the job sequence r
and its delivery date is the completion time of the
job completed last.
When
P the common due date is restrictive, i.e.,
d < nj1 pj , the problem is NP-hard even if
h 0 and a b: it follows from the NP-hardness of the restricted MAD problem (Hall et al.,
1991a; see Section 2.3). Herrmann and Lee
(1993) propose a pseudopolynomial-time algorithm to nd an optimal solution to the restricted variant of the problem with h 6 0, c 0
and a 6 b. Chen (1996) proposes an On5 dynamic programming algorithm for the unrestricted variant of the problem with h 6 0, c 6 0
and a 6 b.
Cheng and Kovalyov (1996) consider a problem
with an objective function similar to the one considered by De et al. (1991a) (see Section 2.9) but
under an assumption that jobs are grouped in n
dierent sets. Each group j consists of qj P 1 identical jobs with the same processing time pj P 0 and
the same weight wj P 0. Each group will be partitioned into batches containing contiguously scheduled jobs. A set-up time sj is required before a batch
of group j is processed if it is processed rst on the
machine or immediately after a batch of another
group. The machine can handle only one job at a
time and cannot process any job when a set-up is
performed. The objective is to determine a value for
the common due date d and a schedule (i.e., the
number of jobs in each batch and a processing order
of the batches) so as to minimize
n
X
wj Uj0 c maxf0; d Dg;
j1
where Uj0 is the number of tardy jobs of group j,

and D is the lead time. The following property
(Cheng and Kovalyov, 1996) is useful to nd an
optimal solution.
14
Property 7. There exists an optimal schedule which

consists of at most one early batch (i.e., Cj 6 d for
all jobs in this batch) and at most one tardy batch
for each group.
rithm is proposed in the case when jobs among

families can be mixed together for processing.
2.11. Maximal weighted absolute lateness

Moreover, tardy batches can be scheduled in an
arbitrary order after the last early batch, and, since
there is a common due date, early batches can also
be scheduled in an arbitrary order.
Cheng and Kovalyov (1996) show that the
problem is NP-hard in the following cases: (a)
wj 1, pj 1; (b) wj 1, sj s; (c) qj 1, sj 0;
(d) qj 1, pj 0, where j 1; . . . ; n; and propose
two pseudopolynomial algorithms as well as a
fully polynomial approximation scheme. In the
special cases, when (a) wj 1, qj q, (b) wj 1,
pj p, sj s, or (c) qj q, pj p, sj s, where
j 1; . . . ; n; On log n algorithms are proposed.
Kovalyov (1997) shows that the problem with
sj s and pj p is NP-hard.
Azizoglu and Webster (1997b) consider unrestricted TWET problem (see Section 2.5) with the
additional assumptions that each job belongs to
one of the f families, and that set-up time si is required in order to process a job of family i after a
job in some other family. No set-up is required
between jobs from the same family. A branch-andbound algorithm is proposed which is eective at
solving problems up to 15 jobs in size, and a beam
search heuristic branch-and-bound procedure was
applied to the problem. In this procedure, at each
level of the branch and bound tree, only a limited
number of nodes are selected for branching. By
choosing this number of nodes (beam width), user
can inuence the trade-o between the time to
obtain a solution and its quality.
With the same assumption of belonging jobs to
f families, Chen et al. (1997) consider the problem
of nding an optimal due date and the associated
sequence of jobs to minimize the objective function
f d; r a
n
X
Uj cd;
j1
where Uj 0 if Cj 6 d and Uj 1 otherwise. An

On2 algorithm is proposed in the case when all
jobs of each family are to be processed in one
group one after another, and an On log n algo-
Consider the problem of minimizing the maximal weighted absolute lateness (MWAL), i.e., of
nding a due date d and the associated sequence
r to minimize the objective function

f d; r max wj Cj d :
16j6n
Let d 0 be the best possible due date for a given

sequence r. Then, there exist an early job i and a
tardy job k such that wk Ck d 0 wi d 0 Ci
max1 6 j 6 n fwj jCj d 0 jg since, otherwise, a d > 0
would exist such that either d 0 d or d 0 d is a
better due date. Note that wk Ck d 0
wi d 0 Ci implies d 0 wi Ci wk Ck =wi wk .
Thus, if we consider di; k wi Ci wk Ck =
wi wk , the value of d 0 could be determined by
checking all possible di; k in On2 time (Li and
Cheng, 1994).
Unfortunately, nding an optimal sequence is
dicult. Li and Cheng (1994) show that the
MWAL problem for a given d is NP-hard.
When wj 1 for all jobs, the problem of nding an optimal sequence becomes easy even for
dierent given due dates. Lakshminarayan et al.
(1978) propose an On log n algorithm to solve the
problem. For the common due date assignment,
this problem is considered by Cheng (1987b).
3. Parallel machines
Consider the problem of scheduling n jobs on m
parallel machines. Each job can be processed by
any of the m machines taking into account the
following constraints: a machine performs at most
one job at a time, and each job is performed at
most by one machine at a time. If the machines are
identical, they operate at the same speed, and
processing time for job j is pj . If the machines are
uniform, each machine i has its own speed si , and
processing time of job j on this machine will be
pij pj =si . If the machines are unrelated, the speed
of a machine is job-dependent, and processing time

of job j on machine i is pij pj =sij .
The goal is to determine an optimal due date
and to schedule the jobs on the machines, so as to
minimize an objective function. The type of objective function will be specied below for each of
the problems under consideration. In what follows, it is assumed that preemption is not permitted and all jobs become available for processing
simultaneously at time zero.
3.1. Mean absolute deviation of completion times

about a due date
In the case when the due date is given (i.e., is not a
decision variable) and parallel machines are identical, the problem to minimize the mean absolute
deviation (MAD) of completion times about a
common due date d is considered by Sundararaghavan and Ahmed (1984), Hall (1986b) and Emmons
(1987).
In this problem,
the objective function is
Pn
Pn
jC

dj
Tj . For any schedule S,

j
j
j1
j1
let n0i and n00i be, respectively, the number of early
jobs and tardy jobs on machine i, i 1; . . . ; m. We
denote by ji the jth early job on machine i in
schedule S, and by ji the jth (if counted from the
last job) tardy job on machine i in S; i 1; . . . ; m.
Then, according to Emmons (1987), the objective
function may be rewritten as
m
X
i1
0
@
n0i
X
j1
n00i
j 1pji
1
jpji A:
j1
Since at each machine the objective function is

formulated as a sum of pairwise products of job
processing times and of positional weights, we
have the similar situation as in the single machine
problem considered in Section 2.2. Therefore, the
algorithm for parallel machines will be the following. Consider the jobs in the LPT order. Each
job from the m jobs with the largest processing
times should be placed rst on each machine (the
order of assigning these jobs to machines is arbitrary); the next 2m jobs with the largest processing
times should be assigned in any order to the second and last positions; the next 2m jobs should
15
occupy the third and the penultimate positions,

etc. If the nal group of jobs is less than 2m, they
can be assigned to any subset of the next 2m positions.
Emmons (1987) proposes a similar approach
for the MAD problem with uniform machines and
for the problem which
Pn consists of minimizing the
objective function
j1 aEj bTj (the weighted
sum of absolute deviation (WSAD) of completion
times about a common due date) in the cases of
identical or uniform machines.
For parallel machines, Kubiak et al. (1990)
show the equivalence of the WSAD problem and
the problem of minimizing the mean ow time. On
this basis, when parallel machines are unrelated,
they propose an algorithm for the WSAD problem
by formulating the corresponding transportation
problem.
For parallel identical machines, Webster (1997)
studies the complexity of the MAD problem and
the problem of minimizing the total weighted
earliness and tardiness (TWET) with symmetric
weights under the assumption that jobs belong to
dierent families, and set-ups are required between
the jobs from dierent families.
3.2. MAD, WSAD and TWET problems with due

date assignment
Emmons (1987) was the rst to consider the
parallel machine scheduling problems with due
date assignment. For the MAD and WSAD
problems with identical and uniform parallel machines, he proposes On log n algorithms to nd
not only an optimal schedule but also an optimal
due date. For identical machines, alternative optima are generated, and for uniform machines the
algorithm gives a unique optimal solution where
the value of
common due date is equal to
0
Pnthe
i
max1 6 i 6 m f j1
pji =si g, where si is the speed of
machine i and other notations are the same as in
Section 3.1.
Emmons (1987) modies the algorithm for
identical machines introduced in Section 3.1 so as
to nd an optimal due date considering the due
date as a secondary criterion to be minimized. If
n 6 4m, this modication gives an optimal solution
16
in polynomial time, otherwise it is a heuristic

procedure.
The TWET problem, which is NP-hard in the
ordinary sense in the single machine case (see
Section 2.5), is strongly NP-hard for parallel
identical machines (in both restricted and unrestricted cases and even with symmetric weights)
(Webster, 1997).
Alidaee and Panwalkar (1993) consider the
MAD problem for the unrelated parallel machines
and proposed an On3 algorithm for nding an
optimal due date and the associated optimal
schedule (by reducing the problem to a transportation problem). This may be also derived from the
results of Kubiak et al. (1990) (see Section 3.1)
where the due date is not considered as a decision
variable.
Note that, for the above algorithms, the optimal due date is not always the minimal one. It is
necessary to distinguish the problem of nding the
optimal due date from the problem of nding the
minimal one among optimal due dates. Unlike in
the single machine case, these problems dier in
complexity in the case of parallel machines.
Alidaee and Ahmadian (1993) extend the results for the WSAD problem with controllable
processing times given by Panwalkar and Rajagopalan (1992) to the case of unrelated parallel
machines. They reduce the WSAD problem to a
transportation problem which is, as well known,
can be solved in On3 time. The reduction
procedure is similar to the results given by
Bruno et al. (1974) for parallel machine scheduling with the objective of minimizing total ow
time where the jobs have a xed processing time
on each machine.
De et al. (1994b) show that, for the identical
parallel machines, the problem of nding the
minimal due date and the associated optimal
schedule
Pn in order to minimize the objective function
j1 aEj bTj is NP-hard even if m 2,
and strongly NP-hard for an arbitrary m. For the
xed m, they propose a pseudopolynomial-time
algorithm with a worst-case complexity
Onm2pmax m , where pmax is the maximal processing time among the jobs. Alidaee and Panwalkar (1993) propose a heuristic procedure to
minimize makespan and to reduce the optimal
common due date among

Pn optimal schedules with
the objective function j1 Ej Tj .
3.3. Other due date assignment models
For identical parallel machines, Cheng and
Chen (1994) and De et al. (1994b) show that the
problem which consists of nding an optimal due
date d and the associated optimal schedule S
with respect to the objective function
f S; d
n
X
aEj bTj cd
j1
is NP-hard even if m 2. Remind that, for a single

machine, this problem is polynomially solvable
(see Section 2.2).
The following properties (Hall, 1986b; De et al.,
1994b) are valid for this problem (as well as for the
problems considered in Section 3.2).
Property 8. In an optimal schedule, there is no idle
time between jobs on any machine, and the sequence
of jobs on each machine is V-shaped about the due
date: jobs completed by the due date are in LPT
order, and jobs completed after the due date are in
SPT order.
Property 9. In an optimal schedule, there exists at
least one job j such that Cj d .
For any schedule S, let z S mind P 0
ff S; dg and d S mind P 0 fdjf S; d
z Sg. Given a schedule S, the value d S can
be easily computed. Let CK be the Kth smallest
completion time in S, n0 be the number of jobs
in S with completion time equal to CK , and
l nb c =a b. Then, d S CK , where
K satises the inequality l 6 K 6 l n0 . This result is similar to Property 2 in the single machine
case.
Cheng (1989) proposes a heuristic algorithm for
this problem, De et al. (1991b) characterize necessary conditions for an optimal schedule. For an
arbitrary m, the problem is shown to be strongly
NP-hard, while for the xed m a pseudopolynomial-time algorithm is proposed (De et al., 1994b).
In the following special cases of this problem,

polynomial-time algorithms are proposed: (a) c 0
(Emmons, 1987) (see Section 3.2), (b) pj p,
j 1; . . . ; n (Cheng and Chen, 1994), (c) c PP
b; in
this case the problem is equivalent to a P jj Cj
problem (according to the notations of Lawler
et al., 1993). In cases (a) and (c), the complexity of
the algorithms is On log n, while it is O1 in case
(b).
Adamopoulos and Pappis (1998) propose a
heuristic procedure for the problem with the same
objective function in case of unrelated parallel
machines.
For identical parallel machines, Biskup and
Cheng (1999b) present an NP-hardness proof for
the problem of nding a common due date and an
optimal schedule in order
Pnto minimize the objective function f S; d j1 aEj bTj hCj . A
heuristic algorithm is developed to obtain nearoptimal solutions. The most time consuming step
of the algorithm is the solution of the linear programming problem. In the special case of equal
processing times, a polynomial algorithm is presented to nd an optimal solution.
Kahlbacher and Cheng (1993) consider the
problem of assigning due date and scheduling jobs
on identical machines in order to minimize the
objective function
X
X
f d; S
hd Cj
wj gd;
j2E
j2T
where wj > 0 is a weight associated with job j,

j 1; . . . ; n; g and h are real-valued, non-decreasing functions dened on the set of real, non-negative numbers so that g0 and h0 are equal to 0.
This problem for an externally given due date is
NP-hard (Kahlbacher and Cheng, 1993) even if
hx 0 and wj 1, j 1; . . . ; n (it may be also
regarded as the restricted due date assignment
problem). When the due date can be arbitrarily
assigned, polynomial-time algorithms are proposed for linear function hx and gd 0 (complexity is On log n if wj w, j 1; . . . ; n, and
On4 if values of wj are dierent) (Kahlbacher and
Cheng, 1993). The problem is shown to be NPhard (even if wj w, j 1; . . . ; n) for (a) linear
gd and hx 0, and (b) arbitrary hx and
gd 0.
17
Cheng and Chen (1994) show that the problem

of nding a due date and the associated schedule
which minimize the objective function

f d; S max Ej ; Tj
j
is NP-hard for parallel identical machines. Remind

that, in a single machine case, this problem is
polynomially solvable (see Section 2.11).When
each job has a weight wj , the problem (which we
refer in Section 2.11 as the MWAL problem; it
consists in minimizing maximal weighted lateness)
becomes strongly NP-hard (Li and Cheng, 1994).
4. Summary
In this paper, we have presented a review on
scheduling jobs on a single machine and parallel
machines under a common due date which is a
decision variable. We did not consider the problems with given release dates and practically did
not touch upon the problems with unit processing
times. The SLK, TWK, NOP and PPW due date
assignment models and the assignment of positional due dates are considered in another paper
(Gordon et al., 1998).
The results for the common due date assignment and scheduling problems are summarized in
Table 1. We adopt the standard three-eld notation ajbjc used for scheduling problems (Lawler
et al., 1993), where a describes the machine environment and species the number of machines, b
describes the schedule and job characteristics, and
c describes the optimality criterion. We extend this
notation in the following way. The notation dj d
denotes the (unrestricted) problem with a common
due date assignment, and dj d res denotes the restricted version of this problem. The notation
dj de denotes the unrestricted version of the
problem with no penalty cost within interval
d e; d e around a common due date d. The
notation dj d min denotes the problem to nd
minimal due date among the optimal ones. For the
parallel machines, as usual, notations P, Q and R
denote, respectively, identical, uniform and unrelated machines, while Pm means that the number
of identical machines is xed and equal to m.
18
Table 1
CON due date assignment and scheduling
Problem
1 1 jdj dj
2 1 jdj d res j
4 1 jdj d res j
Ej Tj
Algorithm
On log n
Kanet (1981), Panwalkar et al. (1982), Bagchi et al. (1986),
Hall (1986b), Bagchi et al. (1987a)
PsP
Hall et al. (1991a), Ventura and Weng (1995), De et al. (1993)
Enumer.
Bagchi et al. (1986), Szwarc (1989)
Heurist.
Sundararaghavan and Ahmed (1984)
Approx.
Hoogeveen et al. (1994)
On log n
Panwalkar et al. (1982), Bagchi et al. (1987a)
PsP
De et al. (1993)
Enumer.
Bagchi et al. (1987a)
On log n
Panwalkar et al. (1982)
On log n
Seidmann et al. (1981)
PsP
Hall and Posner (1991) De et al. (1990a), Jurisch et al. (1997)
Enumer.
Cheng (1990a) Approx. Kovalyov and Kubiak (1999)
Heurist.
Hao et al. (1996)
PsP
Hoogeveen and van de Velde (1991)
Enumer.
Dileepan (1993), De et al. (1994a)
Heurist.
Gupta et al. (1990), Dileepan (1993), De et al. (1994a),
James (1997), Lee and Kim (1995)
Heurist.
James (1997)
On log n
Cheng (1988a), Weng and Ventura (1994), Wilamowsky et al. (1996)
Psp
Kahlbacher (1989), De et al. (1992, 1993)
Enumer.
NP-hard
Hall et al. (1991a), Hoogeveen
and van de Velde (1991)
aEj bTj
NP-hard (from 2)
aEj bTj
5 1 jdj dj
aEj bTj cd
6 1 jdj dj
aEj bTj c maxf0; dj Dg
7 1 jdj dj
aj Ej Tj
NP-hard
Hall and Posner (1991)
8 1 jdj d res j
9 1 jdj dj
aj Ej Tj
NP-hard (from 2)
aj Ej bj Tj
NP-hard (from 7)
aj Ej bj Tj
NP-hard (from 2)
10 1 jdj d res j
11 1 jdj dej
12 1 jdj dj
Ej Tj
Ej2 Tj2
NP-hard
Kubiak (1993)
3 1 jdj dj
Ej Tj
Complexity
Table 1 (Continued)
Problem
Complexity
13 1 jdj d res j
P
wj Uj c maxf0; d Dg
P
15 1 jdj d res j wj Uj c maxf0; d Dg
16 1 jdj dj
17 1 jdj dj
18 1 jdj dj
wj Uj
19 1 jdj dj
wj Uj gd
20 1 jdj dj
wj Uj cd
Uj
wj Uj
NP-hard (from 12)
P
NP-hard from
Lawler and Moore (1969)
De et al. (1991a)
P
hEj gd
aEj cd
hEj
21 1 jdj dj maxEmax ; Tmax

22 1 jdj dj maxj wj jLj j
NP-hard
Kahlbacher and Cheng (1993)
NP-hard
P
23 Q jdj dj
Ej Tj
NP-hard
Li and Cheng (1994)
P
24 Q jdj dj
aEj bTj
25 R jdj dj
P
Ej Tj
26 P 2 jdj d min j
27 Pm jdj d min j
P
aEj bTj
28 P jdj d min j
Ej Tj
Ej Tj
29 P jdj d; pj pj
P
aEj bTj cd
aEj bTj cd
NP-hard
De et al. (1994b)
NP-hard (from 26)
strongly NP-hard
De et al. (1994b)
P
NP-hard
On2
Cheng (1987b)
On log n
Emmons (1987)
On log n
Emmons (1987)
On3
Alidaee and Panwalkar (1993), Kubiak et al. (1990)
PsP
De et al. (1994b)
PsP
De et al. (1994b)
Heurist.
Alidaee and Panwalkar (1993)
O1
Cheng and Chen (1994)
PsP
19
30 P 2 jdj dj
Bagchi et al. (1987a)

Heurist.
Gupta et al. (1993), Jurisch et al. (1997)
PsP
Kahlbacher (1989, 1993), De et al. (1993)
Enumer.
Bagchi et al. (1987a,b), De et al. (1989, 1990a)
On
De et al. (1991a)
Enumer.
De et al. (1991a)
Heurist.
De et al. (1991a)
On
On log n
On4
14 1 jdj dj
P
aEj2 bTj2
Algorithm
38 P jdj dj maxj wj jLj j
Uj cd
36 P jdj dj
37 P jdj dj maxEmax ; Tmax
NP-hard
NP-hard
Cheng and Chen (1994)
strongly NP-hard
Li and Cheng (1994)
P
aEj
w j Uj
35 P jdj dj
P
aEj
P
P
P
Uj
34 P jdj dj
aEj bTj cd
33 Q jdj dj
strongly NP-hard
De et al. (1994b)
strongly NP-hard (from 32)
P
aEj bTj cd
32 P jdj dj
De et al. (1994b)
Cheng and Chen (1994),

De et al. (1994b)
NP-hard (from 30)
P
31 Pm jdj dj aEj bTj cd
Algorithm
Complexity
PsP
De et al. (1994b)
Heuristic.
Cheng (1989)
Heuristic.
Adamopoulos and Pappis (1998)
On log n
On4
Problem
Table 1 (Continued)
20
Table 1 is organized as follows. The single

machine problems are followed by parallel machine problems. In each group, the problems are
placed according to their objective functions: rst
minsum and then minmax problems which are
placed from the simple to more complicated objective functions. With the same objective functions, the unrestricted problems are followed by
the restricted ones.
In the column algorithm, notations PsP and
Enumer. denote algorithms which allow to nd the
optimal solutions and use, respectively, pseudopolynomial-time procedures (for example, dynamic programming algorithms for the problems
which are NP-hard but not in the strong sense) or
some kind of branch-and-bound procedures (by
implicit enumeration of all sets of feasible solutions through examination of increasingly smaller
subsets of such solutions); approx. and heurist.
denote algorithms which allow to nd near-optimal solutions, respectively, with or without estimation how the solutions may dier from the
optimal ones. In other words, we use the notation
approx. for the heuristic (suboptimal) algorithms
with analytically evaluated accuracy. In the column complexity, notation P denotes the polynomially solvable problem.
The references are given in the columns
complexity and algorithm to show, respectively,
where NP-hardness was proved or where the
algorithms were provided. As a rule, the most
general and the best results are included in the
table, although there are some of the results
which are included to emphasize the priority in
publications. We do not include in the table the
results related to the special cases of the problems and to the special additional conditions
(such as agreeable weights, processing in batches,
etc.), which are described above. The results for
the general model introduced in Section 2.8 are
not included in the table since the notations for
the objective function require additional explanations. Empty positions in the column algorithm show the most challenging and open
problems where algorithms were not proposed
or, for NP-hard problems, where it is not clear
whether the problem is NP-hard in the strong or
in the ordinary sense.
Acknowledgements
The authors would like to acknowledge the
anonymous referees for their helpful comments
and suggestions. This work was supported in part
by INTAS (Project 00-217). The rst author was
also partly supported by ISTC (Project B-104) and
by the Belarus Foundation for Basic Research
(Project F99-119). The paper was prepared when
the rst author was visiting INRIA (Metz) and
UTT (Troyes). Support of this visit from CIES is
gratefully acknowledged.
References
Adamopoulos, G.I., Pappis, C.P., 1998. Scheduling under a
common due-date on parallel unrelated machines. European Journal of Operational Research 105, 494501.
Alidaee, B., Ahmadian, A., 1993. Two parallel machine
sequencing problems involving controllable job processing
times. European Journal of Operational Research 70,
335341.
Alidaee, B., Panwalkar, S.S., 1993. Single stage minimum
absolute lateness problem with a common due date on nonidentical machines. Journal of the Operational Research
Society 44, 2936.
Azizoglu, M., Webster, S., 1997a. Scheduling about an unrestricted common due window with arbitrary earliness/
tardiness penalty rates. IIE Transactions 29, 10011006.
Azizoglu, M., Webster, S., 1997b. Scheduling job families
about an unrestricted common due date on a single machine. International Journal of Production Research 35,
13211330.
Bagchi, U., Chang, Y.L., Sullivan, R.S., 1987a. Minimizing
absolute and squared deviations of completion times with
dierent earliness and tardiness penalties and a common due
date. Naval Research Logistics 34, 739751.
Bagchi, U., Sullivan, R.S., Chang, Y.L., 1986. Minimizing
mean absolute deviation of completion times about a
common due date. Naval Research Logistics 33, 227240.
Bagchi, U., Sullivan, R.S., Chang, Y.L., 1987b. Minimizing
mean squared deviation of completion times about a
common due date. Management Science 33, 894906.
Baker, K.R., Scudder, G.D., 1989. On the assignment of
optimal due dates. Journal of the Operational Research
Society 40, 9395.
Baker, K.R., Scudder, G.D., 1990. Sequencing with earliness
and tardiness penalties: A review. Operations Research 38,
2236.
Bector, C.R., Gupta, Y.P., Gupta, M.C., 1988. Determination
of an optimal common due date and optimal sequence in a
single machine job shop. International Journal of Production Research 26, 613628.
21
Biskup, D., 1999. Single-machine scheduling with learning

considerations. European Journal of Operational Research
115, 173178.
Biskup, D., Cheng, T.C.E., 1999a. Single-machine scheduling
with controllable processing times and earliness, tardiness
and completion time penalties. Engineering Optimization
31, 329336.
Biskup, D., Cheng, T.C.E., 1999b. Multiple-machine scheduling with earliness, tardiness and completion time penalties.
Computers and Operations Research 26, 4557.
Biskup, D., Feldmann, M., 2001. Benchmarks for scheduling
on a single-machine against restrictive and unrestrictive
common due dates. Computers and Operations Research
28, 787801.
Biskup, D., Jahnke, H., 2001. Common due date assignment for
scheduling on a single machine with jointly reducible
processing times. International Journal of Production Economics 69, 317322.
Blazewicz, J., Ecker, K., Pesch, E., Schmidt, G., Weglarz, J.,
1996. Scheduling Computer and Manufacturing Processes.
Springer, Berlin.
Blazewicz, J., Ecker, K., Schmidt, G., Weglarz, J., 1993.
Scheduling in Computer and Manufacturing Systems.
Springer, Berlin.
Blazewicz, J., Pesch, E., Sterna, M., Werner, F., 2000. Total late
work criteria for scheduling problems. In: Inderfurth, K.
et al. (Eds.) Operations Research Proceedings 1999, Springer, Berlin, pp. 354359.
Brucker, P., 1998. Scheduling Algorithms. Springer, Berlin.
Bruno, J.L., Coman, E.G., Sethi, R., 1974. Scheduling
independent tasks to reduce mean nishing time. Communications of the ACM 17, 382387.
Cai, X., 1995. Minimization of agreeably weighted variance in
single machine systems. European Journal of Operational
Research 85, 576592.
Cai, X., Lum, V.Y.S., Chan, J.M.T., 1997. Scheduling about a
common due date with job-dependent asymmetric earliness
and tardiness penalties. European Journal of Operational
Chen, B., Potts, C.N., Woeginger, G.J., 1998. A review of
machine scheduling: Complexity, algorithms and approximability. In: Du, D.-Z., Pardalos, P.M. (Eds.), Handbook of Combinatorial Optimization, Kluwer, Dordrecht,
pp. 21169.
Chen, D., Li, S., Tang, G., 1997. Single machine scheduling
with common due date assignment in a group technology
environment. Mathematical and Computer Modelling 25,
8190.
Chen, Z.-L., 1996. Scheduling and common due date
assignment with earlinesstardiness penalties and batch
delivery costs. European Journal of Operational Research
93, 4960.
Cheng, T.C.E., 1985. A duality approach to optimal due-date
determination. Engineering Optimization 9, 127130.
Cheng, T.C.E., 1987a. An algorithm for the CON due-date
determination and sequencing problem. Computers and
Operations Research 14, 537542.
22
Cheng, T.C.E., 1987b. Minimizing the maximum deviation of

job completion time about a common due-date. Computers
and Mathematics with Applications 14, 279283.
Cheng, T.C.E., 1988a. Optimal common due-date with limited
completion time deviation. Computers and Operations
Research 15, 9196.
Cheng, T.C.E., 1988b. An alternative proof of optimality for
the common due-date assignment problem, European
Journal of Operational Research 37, 250253; Corrigendum:
European Journal of Operational Research 38 (1989) 259.
Cheng, T.C.E., 1989. A heuristic for common due-date assignment and job scheduling on parallel machines. Journal of
the Operational Research Society 40, 11291135.
Cheng, T.C.E., 1990a. A note on a partial search algorithm for
the single-machine optimal common due-date assignment
and sequencing problem. Computers and Operations Research 17, 321324.
Cheng, T.C.E., 1990b. Common due-date assignment and
scheduling for a single processor to minimize the number
of tardy jobs. Engineering Optimization 16, 129136.
Cheng, T.C.E., 1991. Optimal constant due-date determination
and sequencing of n jobs on a single machine. International
Journal of Production Economics 22, 259261.
Cheng, T.C.E., Chen, Z.-L., 1994. Parallel-machine scheduling
problems with earliness and tardiness penalties. Journal of
the Operational Research Society 45, 685695.
Cheng, T.C.E., Gupta, M.C., 1989. Survey of scheduling
research involving due date determination decisions. European Journal of Operational Research 38, 156166.
Cheng, T.C.E., Kovalyov, M.Y., 1996. Batch scheduling and
common due-date assignment on a single machine. Discrete
Applied Mathematics 70, 231245.
Cheng, T.C.E., Oguz, C., Qi, X.D., 1996. Due-date assignment
and single machine scheduling with compressible processing
times. International Journal of Production Economics 43,
2935.
Cheng, T.C.E., Sin, C.C.S., 1990. A state-of-the-art review of
parallel-machine scheduling research. European Journal of
Operational Research 47, 271292.
Conway, R.W., 1965. Priority dispatching and job lateness in a
job shop. Journal of Industrial Engineering 16, 228237.
Chretienne, P., Coman, Jr., E.G., Lenstra, J. K., Liu, Z.
(Eds.), 1995. Scheduling Theory and its Applications. Wiley,
Chichester.
Davis, J.S., Kanet, J.J., 1993. Single-machine scheduling with
early and tardy completion costs. Naval Research Logistics
40, 85101.
De, P., Ghosh, J.B., Wells, C.E., 1989. A note on the
minimization of mean squared deviation of completion
times about a common due date. Management Science 35,
11431147.
De, P., Ghosh, J.B., Wells, C.E., 1990a. CON due-date
determination and sequencing. Computers and Operations
De, P., Ghosh, J.B., Wells, C.E., 1990b. Scheduling about a
common due date with earliness and tardiness penalties.
Computers and Operations Research 17, 231241.
De, P., Ghosh, J.B., Wells, C.E., 1991a. Optimal delivery time
quotation and order sequencing. Decision Science 22, 379
390.
De, P., Ghosh, J.B., Wells, C.E., 1991b. On the multiplemachine extension to a common due-date assignment and
scheduling problem. Journal of the Operational Research
Society 42, 419422.
De, P., Ghosh, J.B., Wells, C.E., 1992. On the minimization of
completion time variance with a bicriteria extension. Operations Research 40, 11481155.
De, P., Ghosh, J.B., Wells, C.E., 1993. On the general solution
for a class of early/tardy problems. Computers and Operations Research 20, 141149.
De, P., Ghosh, J.B., Wells, C.E., 1994a. Solving a generalized
model for CON due date assignment and sequencing.
International Journal of Production Economics 34, 179
185.
De, P., Ghosh, J.B., Wells, C.E., 1994b. Due-date assignment
and early/tardy scheduling on identical parallel machines.
Naval Research Logistics 41, 1732.
Della Croce, F., Gupta, J.N.D., Tadei, R., 2000. Minimizing
tardy jobs in a owshop with common due date. European
Journal of Operational Research 120, 375381.
Dickman, B., Wilamowsky, Y., Epstein, S., 1991. Optimal
common due date with limited completion time. Computers
and Operations Research 18, 125127.
Dileepan, P., 1993. Common due date scheduling problem with
separate earliness and tardiness penalties. Computers and
Eilon, S., Chowdhury, I.J., 1976. Due dates in job shop
scheduling. International Journal of Production Research
14, 223237.
Emmons, H., 1987. Scheduling to a common due date on
parallel uniform processors. Naval Research Logistics
Quarterly 34, 803810.
Federgruen, A., Mosheiov, G., 1993. Simultaneous optimization of eciency and performance balance measures in
single-machine scheduling problems. Naval Research Logistics 40, 951970.
Federgruen, A., Mosheiov, G., 1994. Greedy heuristics for
single-machine scheduling problems with general earliness
and tardiness costs. Operations Research Letters 16, 199
208.
Feldmann, M., Biskup, D., 1999. Single-machine scheduling for
minimizing earliness and tardiness penalties by metaheuristc approaches, University of Bielefeld Discussion
Paper No. 425, 1-28; International Journal of Applied
Intelligence, to appear.
Garey, M.R., Tarian, R.T., Wilfong, G.T., 1988. One-processor
scheduling with symmetric earliness and tardiness penalties.
Mathematics of Operations Research 13, 330348.
Gee, E.S., Smith, C.H., 1993. Selecting allowance policies for
improved job shop performance. International Journal of
Production Research 31, 18391852.
Gordon, V.S., Proth, J.-M., Chu, C., 1998. A state-of-the-art
survey of due date assignment and scheduling research:
SLK, TWK, and other due date assignment models. INRIA

Rapport de Recherche 3537, 137; Production Planning and
Control, to appear.
Gupta, M.C., Gupta, Y.P., Kumar, A., 1993. Minimizing ow
time variance in a single machine system using genetic
algorithms. European Journal of Operational Research 70,
289303.
Gupta, S.K., Kyparisis, J., 1987. Single machine scheduling
research. OMEGA 15, 207227.
Gupta, Y.P., Bector, C.R., Gupta, M.C., 1990. Optimal
schedule on a single machine using various due date
determination methods. Computers in Industry 15, 245253.
Hall, N.G., 1986a. Scheduling problems with generalized due
dates. IIE Transactions 18, 220222.
Hall, N.G., 1986b. Single- and multiple-processor models for
minimizing completion time variance. Naval Research
Logistics Quarterly 33, 4954.
Hall, N.G., Kubiak, W., Sethi, S.P., 1991a. Earlinesstardiness
scheduling problems, II: Deviation of completion times
about a restrictive common due date. Operations Research
39, 847856.
Hall, N.G., Posner, M.E., 1991. Earlinesstardiness scheduling
problems, I: Weighted deviation of completion times about
a common due date. Operations Research 39, 836846.
Hall, N.G., Sethi, S.P., Sriskandarajah, C., 1991b. On the
complexity of generalized due date scheduling problems.
European Journal of Operational Research 51, 100109.
Hao, Q., Yang, Z., Wang, D., Li, Z., 1996. Common due date
determination and sequencing using tabu search. Computers and Operations Research 23, 409417.
Hardy, G.H., Littlewood, J.E., Polya, G., 1934. Inequalities.
Cambridge University Press, New York.
Herrmann, J.W., Lee, C.-Y., 1993. On scheduling to minimize
earlinesstardiness and batch delivery costs with a common
due date. European Journal of Operational Research 70,
272288.
Hoogeveen, J.A., Lenstra, J.K., van de Velde, S.L., 1997.
Sequencing and scheduling: An annotated bibliography,
Memorandum COSOR 97 - 02, Eindhoven University of
Technology, pp. 117.
Hoogeveen, J.A., Oosterhout, H., van de Velde, S.L., 1994.
New lower and upper bounds for scheduling around a small
common due date. Operations Research 42, 102110.
Hoogeveen, J.A., van de Velde, S.L., 1991. Scheduling around a
small common due date. European Journal of Operational
Hoogeveen, J.A., van de Velde, S.L., 1997. Earlinesstardiness
scheduling around almost equal due date. INFORMS
Journal on Computing 9, 9299.
Jackson, J.R., 1955. Scheduling a production line to minimize
maximum tardiness. Management Science Research Project,
UCLA.
James, R.J.W., 1997. Using tabu search to solve the common
due date early/tardy machine scheduling problem. Computers and Operations Research 24, 199208.
Jozefowska, J., Jurisch, B., Kubiak, W., 1994. Scheduling shops
to minimize the weighted number of late jobs. Operations
Research Letters 16, 277283.
23
Jurisch, B., Kubiak, W., Jozefowska, J., 1997. Algorithms for

minclique scheduling problems. Discrete Applied Mathematics 72, 115139.
Kahlbacher, H.G., 1989. SWEAT a program for a scheduling
problem with earliness and tardiness penalties. European
Kahlbacher, H.G., 1993. Scheduling with monotonous earliness
and tardiness penalties. European Journal of Operational
Kahlbacher, H.G., Cheng, T.C.E., 1993. Parallel machine
scheduling to minimize costs for earliness and number of
tardy jobs. Discrete Applied Mathematics 47, 139164.
Kanet, J.J., 1981. Minimizing the average deviation of job
completion times about a common due date. Naval
Research Logistics Quarterly 28, 643651.
Karacapilidis, N.I., Pappis, C.P., 1995a. Form similarities of
the CON and SLK due date determination methods.
Journal of the Operational Research Society 46, 762
770.
Karacapilidis, N.I., Pappis, C.P., 1995b. Optimization algorithms for a class of single machine scheduling problems
using due date determination methods. Yugoslavian Journal
of Operations Research 5, 289297.
Kawaguchi, T., Kyan, S., 1988. Deterministic scheduling in
computer systems: A survey. Journal of the Operational
Research Society of Japan 31, 190217.
Koulamas, C., 1994. The total tardiness problem: Review and
extensions. Operations Research 42, 10251041.
Kovalyov, M.Y., 1997. Batch scheduling and common due date
assignment problem: An NP-hard case. Discrete Applied
Mathematics 80, 251254.
Kovalyov, M.Y., Kubiak, W., 1999. A fully polynomial
approximation scheme for the weighted earlinesstardiness
problem. Operations Research 47, 757761.
Kramer, F.-J., Lee, C.-Y., 1993. Common due-window
scheduling. Production and Operations Management 2,
262275.
Kubiak, W., 1993. Completion time variance minimization on a
single machine is dicult. Operations Research Letters 14,
4959.
Kubiak, W., 1995. New results on the completion time
variance minimization. Discrete Applied Mathematics 58,
157168.
Kubiak, W., Lou, S., Sethi, R., 1990. Equivalence of mean ow
time problems and mean absolute deviation problems.
Operations Research Letters 9, 371374.
Lakshminarayan, S., Lakshmanan, R., Papineau, R.L., Rochette, R., 1978. Optimal single-machine scheduling with
earliness and tardiness penalties. Operations Research 26,
10791082.
Lawler, E.L., Lenstra, J.K., Rinnooy Kan, A.H.G., Shmoys,
D.B., 1993. Sequencing and scheduling: Algorithms and
complexity. In: Graves S.C., Zipkin P.H., Rinnooy
Kan A.H.G. (Eds.), Logistics of Production and Inventory; Handbooks in Operations Research and Management Science, vol. 4, North-Holland, Amsterdam, pp. 445
522.
24
Lawler, E.L., Moore, J.M., 1969. A functional equation and its

application to resource allocations and sequencing problems. Management Science 16, 7784.
Lawrence, S.R., 1995. Estimating owtimes and setting duedates in complex production systems. IIE Transactions 27,
657668.
Lee, C.-Y., Danusaputro, S.L., Lin, C.-S., 1991. Minimizing
weighted number of tardy jobs and weighted earlinesstardiness penalties about a common due date. Computers
Lee, C.-Y., Kim, S.J., 1995. Parallel genetic algorithms for the
earliness tardiness job scheduling problem with general
penalty weights. Computers and Industrial Engineering 28,
231243.
Li, C.-L., Cheng, T.C.E., 1994. The parallel machine minmax
weighted absolute lateness scheduling problem. Naval
Research Logistics 41, 3346.
Liman, S.D., Panwalkar, S.S., Thongmee, S., 1996. Determination of common due window location in a single machine
scheduling problem. European Journal of Operational
Research 93, 6874.
Liman, S.D., Panwalkar, S.S., Thongmee, S., 1998. Common
due window size and location determination in a single
machine scheduling problem. Journal of the Operational
Research Society 49, 10071010.
Merten, A.G., Muller, M.E., 1972. Variance minimization in a
single machine sequencing problems. Management Science
18, 518528.
Monden, Y. 1983. Toyota Production Systems. Industrial
Engineering and Management Press. Norcross, GA.
Morton, T.E., Pentico, D.W., 1993. Heuristic Scheduling
Systems. Wiley, New York.
Panwalkar, S.S., Rajagopalan, R., 1992. Single machine
sequencing with controllable processing times. European
Panwalkar, S.S., Smith, M.L., Seidmann, A., 1982. Common
due date assignment to minimize total penalty for the one
machine scheduling problem. Operations Research 30,
391399.
Pappis, C.P., Adamopoulos, G.I., 1993. Scheduling under the
due date criterion with varying penalties for lateness.
Yugoslavian Journal of Operations Research 3, 189198.
Philipoom, P.R., Wiegmann, L., Rees, L.P., 1997. Cost-based
due-date assignment with the use of classical and neuralnetwork approaches. Naval Research Logistics 44, 2146.
Pinedo, M.L., 1995. Scheduling: Theory, Algorithms, and
Systems. Prentice-Hall, Englewood Clis, NJ.
Quaddus, M.A., 1987. A generalized model of optimal due-date
assignment by linear programming. Journal of the Operational Research Society 38, 353359.
Raghavachari, M., 1986. A V-shape property of optimal
schedule of jobs about a common due date. European
Raghavachari, M., 1988. Scheduling problems with non-regular
penalty functions a review. Operations Research 25, 144
164.
Ramasesh, R., 1990. Dynamic job shop scheduling: A survey of

simulation research. OMEGA 18, 4357.
Roman, D.B., Valle, A.G., 1996. Dynamic assignation of duedates in an assembly shop based in simulation. International
Journal of Production Research 34, 15391554.
Sarper, H., 1995. Minimizing the sum of absolute deviations
about a common due date for the two-machine ow shop
problem. Applied Mathematics Modelling 19, 153161.
Seidmann, A., Panwalkar, S.S., Smith, M.L., 1981. Optimal
assignment of due-dates for a single processor scheduling
problem. International Journal of Production Research 19,
393399.
Sen, T., Gupta, S.K., 1984. A state-of-art survey of static
scheduling research involving due dates. OMEGA 12, 63
76.
Smith, C.H., Minor, E.D., Wen, H.J., 1995. Regression-based
due date assignment rules for improved assembly shop
performance. International Journal of Production Research
33, 23752385.
Sundararaghavan, P.S., Ahmed, M.U., 1984. Minimizing the
sum of absolute lateness in single-machine and multimachine scheduling. Naval Research Logistics Quarterly 31,
325333.
Szwarc, W., 1989. Single-machine scheduling to minimize
absolute deviation of completion times from a common
due date. Naval Research Logistics 36, 663673.
Szwarc, W., 1996. The weighted common due date single
machine scheduling problem revisited. Computers and
Tanaev, V.S., Gordon, V.S., Shafransky, Y.M., 1994a. Scheduling Theory: Single-Stage Systems. Kluwer, Dordrecht.
Tanaev, V.S., Sotskov, Y.N., Strusevich, V.A., 1994b. Scheduling Theory: Multi-Stage Systems. Kluwer, Dordrecht.
Tsai, C.-H., Chang, G.-T., Li, R.-K., 1997. Integrating order
release control with due-date assignment rules. International Journal of Production Research 35, 33793392.
Van den Akker, M., Hoogeveen, H., van de Velde, S., 2000.
Combining column generation and Lagrangean relaxation:
an application to a single-machine common due date
scheduling problem. In: Brucker, P., Heitman, S., Hurink,
J., Knust S. (Eds.), Seventh International Workshop on
Project Management and Scheduling (PMS 2000). Exrended
Abstracts, Osnabruck, pp. 3538.
Ventura, J.A., Weng, M.X., 1995. An improved dynamic
programming algorithm for the single-machine mean absolute deviation problem with a restrictive common due date.
Operations Research Letters 17, 149152.
Vig, M.M., Dooley, K.J., 1991. Dynamic rules for due date
assignments. International Journal of Production Research
29, 13611377.
Vig, M.M., Dooley, K.J., 1993. Mixing static and dynamic
owtime estimates for due-date assignment. Journal of
Operations Management 11, 6779.
Weeks, J.K., Fryer, J.S., 1979. A methodology for assigning
minimum cost due-dates. Management Science 23, 872
881.

Webster, S.T., 1997. The complexity of scheduling job families
about a common due date. Operations Research Letters 20,
6574.
Weng, M.X., Ventura, J.A., 1994. Scheduling about a large
common due date with tolerance to minimize mean absolute
deviation of completion times. Naval Research Logistics
Quarterly 41, 843851.
25
Wilamowsky, Y., Epstein, S., Dickman, B., 1996. Optimal

common due-date completion time tolerance. Computers
Yeung, W.K., Oguz, C., Cheng, T.C.E., 2001. Single-machine
scheduling with a common due window. Computers and

A Survey of The State-Of-The-Art of Common Due Date Assignment and Scheduling Research

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

A Survey of The State-Of-The-Art of Common Due Date Assignment and Scheduling Research

Hochgeladen von

Copyright:

Verfügbare Formate

European Journal of Operational Research 139 (2002) 125

A survey of the state-of-the-art of common due date assignment

, Jean-Marie Proth b, Chengbin Chu

was the starting point of this type of research. The

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

Seidmann et al. (1981) and Panwalkar et al.

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

model in which all jobs have a common due date

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

decreasing order of the pj values (shortest processing time or SPT order).

2.2. Panwalkar, Smith and Seidmann due date

2.1. Mean absolute lateness

aEj bTj cd:

When c 0 and a b 1 we have the MAD

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

where aj is equal to nc j 1a for j 6 K and to

Panwalkar et al. (1982) for nding the optimal

2.3. MAD problem with due date assignment

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

Here and hereafter, E fjjCj 6 dg is a set of jobs

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

job j. The objective is to nd a set of processing

Panwalkar and Rajagopalan (1992) show that there

aEj bTj dd cD:

2.5. Total weighted earliness and tardiness

where for job j, aj is a unit earliness penalty

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

pose a genetic algorithm. Note that James (1997)

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

2.6. Tolerance intervals

Baker and Scudder (1990) generalize Chengs

2.7. Quadratic penalty functions

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

by Bagchi et al. (1987b) for the restricted MSD

where g and h are the non-decreasing real-valued

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

namic programming algorithms for the problem

2.9. Other due date assignment models

where Ej and Tj are given by (1), and c > 0.

Seidmann et al. (1981) propose another model

where g and h are the non-decreasing functions,

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

an externally given common due date (i.e., when

The problem diers from the TWET problem

scheduled rst in a sequence. The processing times

aEj bTj hCj ;

it is shown that it can be solved in On3 time in

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

2.10. Due date assignment with batching

where l is the number of tardy batches. The goal is

wj Uj0 c maxf0; d Dg;

where Uj0 is the number of tardy jobs of group j,

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

Property 7. There exists an optimal schedule which

rithm is proposed in the case when jobs among

2.11. Maximal weighted absolute lateness

where Uj 0 if Cj 6 d and Uj 1 otherwise. An

Let d 0 be the best possible due date for a given

V. Gordon et al. / European Journal of Operational Research 139 (2002) 125

of a machine is job-dependent, and processing time

3.1. Mean absolute deviation of completion times

Tj . For any schedule S,

Since at each machine the objective function is