Manager’s Preferences Modeling within Multi-Criteria Flowshop Scheduling Problem: A Metaheuristic Approach

Mohamed Anis Allouche
International Journal of Business Research and Management (IJBRM), Volume (1): Issue (2) 33
Manager’s Preferences Modeling Within Multi-Criteria Flowshop
Scheduling Problem: A Metaheuristic Approach
Mohamed Anis Allouche anis.allouche@fsegs.rnu.tn
Unité de Recherche en Gestion Industrielle et Aide à la Décision,
Faculté des Sciences Economiques et de Gestion, Université de Sfax,
CP 1088-3018, Sfax. Tunisia.
Abstract
This paper proposes a metaheuristic to solve the permutation flow shop
scheduling problem where several criteria are to be considered, i.e., the
makespan, total flowtime and total tardiness of jobs. The proposed metaheuristic
is based on tabu search algorithm. The Compromise Programming model and
the concept of satisfaction functions are utilized to integrate explicitly the
manager’s preferences. The proposed approach has been tested through a
computational experiment. This approach can be useful for large scale
scheduling problems and the manager can consider additional scheduling
criteria.
Keywords: Permutation Flowshop, Multi-Criteria Scheduling, Compromise Programming, Satisfaction
Functions, manager’s Preferences.
1. INTRODUCTION
The aim of the scheduling theory is the allocation of a set of limited resources to process a given
number of jobs (MacCarthy and Liu,[20]). More specifically, a scheduling problem consists of
finding the sequence of a set of jobs (tasks) to be processed on different machines, so that
technological constraints are satisfied and one or several performance criteria are optimized
(T’kindt and Billaut, [23]). The scheduling literature reveals that several criteria should be
considered to present more realistic solutions to the production manager. However, it is generally
impossible to find a sequence that optimizes simultaneously different conflicting scheduling
criteria. Thus, the manager must consider the sequence of the best compromise. Hence, the
manager needs to make some trade-offs between the scheduling criteria. Thus, the obtained
solution will be a satisfactory solution.
Several approaches and models are proposed to solve the scheduling problem, namely discrete
variable mathematical programming, simulation techniques and network analysis. Specific
algorithms and heuristics have been utilized to deal with the scheduling problem. The choice of
an appropriate approach depends on the complexity of the problem, the number of the jobs to be
scheduled, the configuration of the machines, the production system and the nature of the job
arrivals (static or dynamic). For example, Gangadhran and Rajendran [10] have applied the
Simulated Annealing technique to minimize the makespan and the total flow time (ΣCi). Kondakci
et al. [17] have utilized the shortest processing time and the earliest due date rules to minimize
the total flow time and the maximum tardiness penalties. We also notice many other algorithms
have been developed to deal with multi-criteria scheduling problems.
Gupta et al. [16] have proposed some heuristics to solve a bi-criteria scheduling problem. Arroyo
and Armentano [5] have proposed a partial enumeration heuristic for multi-objective flowshop
scheduling problem where they provide the manager with approximate Pareto optimal solutions.

Their heuristic offers a set of feasible solutions and the manager’s preferences are partially
considered according to a posteriori articulation.
Armentano and Arroyo [4] have proposed a new tabu search algorithm for multi-objective
combinatorial problems with the aim of obtaining a good approximation of the Pareto-optimal or
efficient solutions. A nice feature of this multi-objective algorithm is that it introduces only one
additional parameter, namely, the number of paths. This algorithm is applied to the permutation
flowshop scheduling problem in order to minimize the criteria of makespan and maximum
tardiness. For instances involving two machines, the performance of the algorithm is tested
against the Branch-and-Bound enumeration algorithm, and for more than two machines it is
compared with that of a tabu search algorithm and a genetic local search algorithm, both from the
literature. Computational results show that the heuristic proposed by Armentano and Arroyo [4]
yields to a better approximation than these algorithms.
Allahverdi [1] was interested in a machine flowshop problem with the objective of minimizing a
weighted sum of makespan and maximum tardiness. Varadharajan and Rajendran [26] presented
a multi-objective simulated annealing algorithm (MOSA) for permutation flowshop scheduling to
minimise the makespan and total flowtime for jobs. The MOSA seeks to obtain non-dominated
solutions through the implementation of a simple probability function that attempts to generate
solutions on the Pareto-optimal front. Framinan and Leisten [7] tackle the problem of total
flowtime and makespan minimisation in a permutation flowshop. The authors have introduced a
multi-criteria iterated greedy search algorithm. Their algorithm iterates over a multi-criteria
constructive heuristic approach to yield a set of Pareto-efficient solutions.
Loukil et al. [19] have adapted the MOSA to solve a multi-criteria flowshop scheduling problem.
Lemesre et al. [18] proposed an exact method to solve a bi-criteria scheduling problem named
the parallel partitioning method. This method allows the generation of efficient solutions.
According to Lemesre et al [18], their method requires less CPU time comparatively to the two
phases method of Ulungu and Teghem [25]. In their book, T’kindt and Billaut [23] present a quite
complete literature review regarding multi-criteria scheduling theory.
Gagné et al. [9] have proposed a generic approach to finding compromise solutions for multiple-
objective scheduling problems using metaheuristics. As an illustration, they present a new hybrid
Tabu Search/Variable Neighbourhood Search (Tabu-VNS) application of this approach for the
solution of a bi-objective scheduling problem. Through numerical experiments they demonstrate
its efficiency and effectiveness. They have confirmed that compromise programming with the
Tabu-VNS metaheuristic generates solutions that approach those of the known reference sets.
Gagné et al. [8] presented an adjustment of an Ant Colony Optimization for an eventual use in a
generic research procedure of compromise solutions for single machine scheduling problem.
Our literature review of multi-criteria scheduling problems shows that a large number of the
proposed approaches to solve the flow shop scheduling problems do not take into account
explicitly the manager’s preferences. However, Aouni et al. [3] and Allouche et al. [2] have
developed an aggregation procedure that considers three different criteria to obtain the best
sequence in a flowshop production environment. The authors utilize the compromise
programming model and the concept of satisfaction functions to integrate explicitly the manager’s
preferences. The satisfaction functions measure the intensity of preference regarding the
deviations between the achievement and the aspiration levels of the following criteria: makespan,
total flow time and total tardiness. Their procedure is easy to apply and it requires few parameters
(thresholds) to be provided by the manager. The satisfaction functions thresholds have a specific
economic interpretation that the manager can understand and provide the values. The proposed
model can be extended to introduce additional criteria. This model provides the best scheduling
sequence that satisfies the manager’s preferences. However, this approach is sensitive to the
size of the scheduling problem to be solved. It requires a large computational time for the large
scale problems. To deal with such difficulty, we recommend the use of metaheuristics which is
the purpose of this paper.

The aim of this paper is to propose a metaheuristic based on the tabu search algorithm to solve a
multi-criteria scheduling problem. The new proposed approach will explicitly incorporate the
structure of manager’s preferences with the use of satisfaction function concept.
This paper is organised as follows. The description of the proposed metaheuristic is given in the
second section. In fact, this metaheuristic is based on three components which are the
compromise programming model, the concept of satisfaction functions and the tabu search
algorithm. The third section presents the different steps of the proposed metaheuristic. These
steps are useful for obtaining the sequence of the best compromise. The computational
experiments and results are summarized in the fourth section.
2. METAHEURISTIC COMPONENTS
Within this section, we will describe the proposed metaheuristic. This metaheuristic is based on
the three following components: a) the Compromise Programming model, b) the concept of
Satisfaction Functions, and c) the tabu search algorithm.
2.1. Compromise Programming Model
The Compromise Programming model (CP) was introduced first by Zeleny [29]. This model is
based on the minimization of the distance between the achievement level ( ( )qf x ) of objective q
and the ideal value (
*
qg ) of this objective. This model is based on the Zeleny’s axiom of choice
where the solutions that are closer to the ideal points (
*
qg ) are preferred to those that are farther
(Zeleny, [27], [28]).
2.2. Satisfaction Function in Compromise Programming Model
In this section we will utilize the concept of the satisfaction functions to formulate a scheduling
model where the manager’s preferences are explicitly incorporated. Martel and Aouni [21] have
introduced the concept of satisfaction functions in the Goal Programming (GP) model. Through
this concept, the manager can explicitly express his/her preferences regarding the unwanted
deviations between the achievement and the aspiration levels associated to the different
objectives. Figure 1 illustrates the general shape of the satisfaction functions.
1
( )q qF δ
q dα q oα q vα qδ
FIGURE1: General shape of satisfaction function
where:
( )q qF δ : value of satisfaction function for deviation qδ ,
qdα : indifference threshold,
qoα : nil satisfaction threshold,
qvα : veto threshold.

The satisfaction functions measure the intensity of the manager’s preferences regarding the
unwanted deviations between the achievement level ( )qf x and the ideal value
*
qg ( q Q∀ ∈ ).
The intensity of preference ( )q qF δ for each objective is defined on the interval[ ]0; 1 . Thus,
there is no need for computing the nadir point that usually used for the normalization procedure in
the CP model. In their paper, Martel and Aouni [21] propose different shapes that the manager
can adopt to elucidate explicitly his/her preferences. The manager can choose or adapt one of
functions presented by figure2.
This list of satisfaction functions neither exhaustive nor restrictive. The manager will adopt the
one that reflects better and accurately his/her preferences. For the purpose of illustration, we will
utilize the satisfaction function of type (d) (Fig. 2). This function will be applied to the three criteria
(makespan, total flowtime and total tardiness) that we are considering in our computational
experiment.
(a) (b)
(c) (d)
FIGURE2: Different shapes of satisfaction functions
2.3. The tabu search algorithm
The tabu search algorithm has been developed by Glover [11] and it was the first general
framework for modern heuristic search. This algorithm is considered as a general iterative
approach of combinatorial optimization. The details of the tabu search algorithm are available in
the following references: Glover ([12],[13]), Glover et al. [15] and Glover and Laguna [14]. In this
paper, through a new metaheuristic based on tabu search algorithm, we look for a best
scheduling sequence that takes into account the manager’s preferences. The details of each step
of this metaheuristic will be presented in the next section.
3. METAHEURISTIC STEPS
In this section, we will present the different steps of the proposed metaheuristic for multi-criteria
scheduling problem. The first step determines the lower bounds of each criterion and the second
step consists of determining the best compromise sequence.
0 αid αiv δi
Fi(δi)
1
0 αio αiv δi
Fi(δi)
1
0 αid αio αiv δi
Fi(δi)
1
0 αid αio αiv δi
Fi(δi)
1
0.5

3.1. Step 1: Finding the Best Values
We consider the best values as the ideal points obtained by optimizing each criterion separately.
For this, a metaheuristic based on a tabu search algorithm was developed and has the following
characteristics:
- Initial solution: randomly generated;
- Neighbourhood structure: we consider all permutation as a solution. The permutation’s
neighbourhood is created by a set of moves. Given a sequence s, let i and j be two positions
in this sequence s. A neighbour of s is obtained by interchanging the jobs in positions i and j
which are randomly selected.
- Selection of the best neighbour: the objective function is to minimize the makespan, the
total flowtime and the total tardiness. In fact, we define best by reference to the objective
function’s value. The best neighbour has to be non tabu or tabu and satisfying the aspiration
criteria.
- Tabu list: the size of tabu list is fixed at 7. This means that the tabu list contains 7 prohibited
moves. Note that this list is managed in a circular manner.
- Aspiration criteria: we used the simplest form of aspiration criterion which is stated as
follows: a tabu move is accepted if it produces a solution better than the best obtained so far.
- Stopping criteria: the algorithm is stopped when it reaches a number of iterations without
improvement of the evaluation function. This parameter is fixed in advance.
The developed metaheuristic was tested through a set of problems used by Taillard [24] and are
in “OR library” [22]. These problems have different sizes such as 20 jobs (5-10-20 machines), 50
jobs (5-10 machines) and 100 jobs (-5 machines). For each size of problem, we have tested 10
different instances.
In order to evaluate the performance of the developed metaheuristic, we are used Taillard’s
benchmarks for the makespan criterion. The obtained results are quiet similar to those presented
by Taillard’s benchmarks. The platform of our computational experiments is personal computer
“Pentium Centrino: Dell Latitude D810” with a 1.73 GHz processor. These set of problems have
been also used to determine the best values of the total flowtime and total tardiness. So, tables 1
and 2 summarize the best values of makespan and total flowtime criteria for the problem size 20
jobs – 20 machines and the CPU time. In the other hand, in order to obtain the best value of total
tardiness criterion, we are, first, referred to the Daniels and Chambers [6] technique to generate
due date of jobs. The due date (dj) of jobs is randomly generated within the following interval:
j
R R
d ABP 1-T- , ABP 1-T+
2 2
    
∈     
    
,
( )
_
ABP= 1 Pn m+ − and
_
1 1
P
P
n m
ij
i j
n m
= =
=
×
∑∑
,
where:
ABP: the Average Busy Period that serves as an approximation of the achievement time of the
job in the sequence.
T: delay factor or average percentage of overdue jobs, { }T 0.4; 0.6; 0.8∈ ,
R: factor controlling the extent of due dates, { }R 0.2; 0.6; 1∈ ,
P : Mean processing time,
n: number of jobs,
m: number of machines.

Problems Best founded value Time (seconds)
20_20_1 2297 16.794
20_20_2 2099 19.027
20_20_3 2326 19.805
20_20_4 2223 33.380
20_20_5 2291 21.391
20_20_6 2226 31.716
20_20_7 2273 39.386
20_20_8 2200 17.355
20_20_9 2237 22.001
20_20_10 2178 31.436
Average value of CPU time: 25.229
TABLE 1: Best founded values of 20PF/20/Cmax
Problems Best founded value Time (seconds)
20_20_1 33816 41.157
20_20_2 31674 34.032
20_20_3 33920 31.281
20_20_4 31722 20.766
20_20_5 34557 48.234
20_20_6 32753 34.250
20_20_7 32922 35.516
20_20_8 32444 21.641
20_20_9 33623 42.046
20_20_10 32262 40.204
The average value of the CPU time: 34.912
TABLE 2: Best founded values of 20PF/20/ ΣCi
Note that the due date of each job was computed with T equal to 0.4 and R equal to 0.6. Table 3
represents the best values of the tardiness criterion. In this context, the obtained values will help
the manager to obtain a sequence which better reflects his/her preferences. In the second step,
we propose a new approach that takes into account the manager’s preferences by utilizing the
concept of satisfaction function. We believe that this will give more flexibility to the manager to
express explicitly his preferences.
Problems Best founded values Time (seconds)
20_20_1 11019 37.782
20_20_2 10435 44.391
20_20_3 12095 15.819
20_20_4 9509 48.343
20_20_5 13526 55.297
20_20_6 10402 43.253
20_20_7 12509 39.703
20_20_8 9334 38.062
20_20_9 10138 39.012
20_20_10 10478 40.204
The average of the CPU time: 40.1866
TABLE3: Best founded value of 20PF/20/ ΣTi

3.2. Step 2: Determining the best compromise solution
A new metaheuristic based on tabu search algorithm has been developed to generate the
sequence of the best compromise. The manager’s preferences are expressed using the
satisfaction functions.
a) Basis Concepts
The basic concepts of the proposed algorithm are presented as follows:
- Initial sequence: it can be selected from the set of three sequences obtained by optimizing
each criterion as it has been presented in the previous step. The selected sequence will be
stored in memory.
- Neighbourhood structure: the retained neighbourhood consists on a permutation of two
jobs selected randomly.
- Selection of the best neighbour: the best neighbour is the sequence which offers to the
manager the highest satisfaction level. To do so, we evaluate the neighbourhood and we
choose the best one non-tabu or tabu and satisfy the aspiration criteria.
- Tabu list: the tabu list is managed in a circular manner. In this list, 7 prohibited moves can
appear.
- Aspiration criteria: we used the simplest form of aspiration criterion which is stated as
follows: a tabu move is accepted if it produces a sequence better than the best obtained so
far.
- Stopping criteria: the algorithm is stopped when it reaches a number of iterations without
improvement of the evaluation function. This number is fixed in advance.
In addition to these concepts, the set of Pareto optimal solutions is used. It contains all non-
dominated sequences. In fact, the principle of dominance concerns only the value of the
optimized criteria and not the value of the objective function.
b) The algorithm structure
The proposed algorithm is as follows:

Initialisation
• Initial sequence so
• sn= so, f
*
=f(so)
• LT = Φ, PE = {so}
Iterative Processes
1. – Generate the neighborhood of the current sequence sn
- Select the best neighbour of sn, s
*
∈SV(sn)
2. A) if s
*
is a non-dominate sequence
- if the move sn to s
*
is non tabu
» » update the tabu list LT
- Update the set of Pareto optimal sequence PE
- sn+1← s
*
- if f
*
< f(s
*
) :
- initialize the iteration counter
- f
*
← f(s
*
)
- go to 3
Otherwise, go to C)
B) if s
*
is a dominated solution
- if the move sn to s
*
is non tabou
- Update the tabu list LT
- sn+1← s
*
- go to C)
- Otherwise:
- chose s’
*
such that f(s’
*
) = Max { f(si), si∈SV(sn) and si non-tabu},
- sn+1← s’
*
C) i = i+1, (increment the iterations counter),
3. if the iterations counter is less than the number of iterations without improvement,
go to 1..
Otherwise, End.
Where:
so: initial sequence;
sn: current sequence;
sn+1: new current sequence;
LT: tabu list;
PE: set of the Pareto optimal sequences;
SV(sn): neighbourhood of the current sequence.
This algorithm is based on two stages which are: a) generating an initial sequence, and b)
generating a neighbourhood sequence. The initial sequence is chosen from among the three
calculated sequences in step 1. In this stage, the tabu list is empty and the set of the Pareto
optimal sequence contains only the initial sequence which later becomes the current sequence of
the second phase. The neighbourhood sequence is generated in order to choose the best that
has the highest satisfaction level. Dominance tests will be established followed by an update of
the set of Pareto optimal sequences and hence the current sequence. For each iteration, we
proceed in the same way until we reach the algorithm’s stopping condition.
c) The algorithm description
The initial sequence (so) is generated from step 1. At this stage, we obtain three sequences which
correspond to the three criteria optimized alternately, so that the choice of the initial sequence is
made in an arbitrary manner. This sequence is stored in memory as the best sequence found (sn)
and will be assigned to the set of Pareto optimal sequences. Thus, it is classified as a current

sequence. Starting from this sequence, the iterative process of the algorithm starts by obtaining a
new sequence at each iteration. Indeed, a search in the neighbourhood is made to find the best
neighbour (s*). The quality of the sequence is evaluated through the objective function value, but
without taking into consideration the value of the objective function of (sn). The best neighbour of
the current sequence thus obtained will be tested for dominance over all the set of Pareto optimal
sequences. This test is done by comparing the values obtained for each criterion. In such context,
two cases appear:
- The best neighbour of the current sequence is not dominated:
This means that there are no sequences to be considered, belonging to the set of Pareto optimal
sequences, which dominates. Therefore, this sequence will be chosen as the new current one if it
is not subject to a tabu status, or through the application of the aspiration criterion which revokes
that tabu status. An update of the list of the Pareto optimal sequences will be done by eliminating
any soequence dominated by (s*). At this stage, we conduct a comparison between the best
objective function value of the old sequence f* and the current sequence f (s*). If the latter case is
better than f*, its value will be stored in memory and the counter of iterations will be initialized
again to make another iteration. Therefore, the iterative process stops after a certain number of
iterations without improving the objective function value.
- The best neighbour of the current sequence is dominated:
In such case, we will check the status of the sequence. If the movement from which the sequence
was obtained is not tabu, the best neighbour is used to explore the neighbourhood in search of
other sequences. It will serve as a starting point for the next iteration. Nevertheless, if this
movement is tabu, the new current sequence of the next iteration is the neighbour who is not tabu
and maximizes the objective function value. This process is repeated for each iteration until
stopping the algorithm.
This metaheuristic was tested through a computational experience that we will present the results
in the next section.
4. COMPUTATIONAL RESULTS
Several tests were conducted to check the performance of the proposed metaheuristic. We have
used several problems, presented by Taillard [24] with different sizes such as: 20 jobs-5
machines; 20 jobs-10 machines; 20 jobs-20 machines; 50 jobs-5 machines; 50 jobs- 10 machines
and 100 jobs-5 machines. For each problem, we took 10 different instances. Table 4 shows the
date file of one scheduling problem characterized by 20 jobs-5 machines
4.1. Parameters file
In this file, we inscribe the information about the best values obtained for each criterion, type of
satisfaction function used, different thresholds, weights of criteria, tabu list length, neighbourhood
list length and the number of iterations without improvement.

Satisfaction Functions/
______________________________
Path and name of the data file:>sat205-1.txt
Number of tasks :20 Number of machines :5
___________________________
Simulation parameters File:>excel205-1.txt
___________________________________
Lower bounds, SF type and Thresholds
Makespan:>1278.00
SF Type: 4 Weight:0.400
Threshold 1: 100.00 Threshold 2: 180.00 Veto: 200.00
Flow time:>14108.00
Total tardiness:>815.00
________________
Tabou parameters
Tabou list length : 7
Neighbourhood list length : 50
Number of iteration : 2000
Introduce initial solution Y/N (y/n) ? :>y
______________________
Initial solution file:>sat205-1.txt
Initial solution
9_15_11_3_13_17_14_19_6_5_1_4_2_18_7_8_16_10_20_12
CMAX:1278.000 ΣCi :15117.000 ΣTi:2045.000 Sat. level:0.4902
TABLE 4: Exemple of provided data
4.2. Initial sequence file
This file indicates which initial sequence will be used to start the iterative process of the algorithm.
This sequence is chosen among a set of three sequences obtained through the optimization of
only one criterion. We find also the values of the three optimized criteria and the value of the
satisfaction level. Ten (10) test problems were generated. Table 5 and 6 summarize the obtained
results. Based on the results we can conclude that the proposed approach is able to solve multi-
criteria permutation flowshop problems in different sizes and integrate explicitly the manager’s
preferences. In fact, the obtained sequences respect the manager’s preferences which may
explain the high level of the achieved satisfaction. In this context, the average of satisfaction
degree for the following problems (20-5), (20-10), (20-20), (50-5), (50-10) and (100-5) is equal to
1; 0.952; 0.9126; 1; 0.9508 and 1 respectively.
Based on the obtained results in Tables 5 and 6, some benefits of the proposed approach can be
identified; it is fast and flexible. The speed of the approach is measured by the computation time
required to solve different sizes of problems. The mean computation time is equal to 260.87
seconds for the 20 jobs -20 machines problems. This time may be acceptable. On its flexibility,
the proposed approach can solve different problems with different satisfaction levels. Similarly, in
its current version, this approach includes all necessary components to add other performance
criteria.

Problems Satisfaction
level
level
level
Job_20_5_1 1 Job_20_10_1 0.9985 Job_20_20_1 1
Job_20_5_2 1 Job_20_10_2 0.6168 Job_20_20_2 1
Job_20_5_ 3 1 Job_20_10_ 3 1 Job_20_20_ 3 0.96
Job_20_5_4 1 Job_20_10_4 1 Job_20_20_4 1
Job_20_5_5 1 Job_20_10_5 0.93 Job_20_20_5 0.69
Job_20_5_6 1 Job_20_10_6 1 Job_20_20_6 0.834
Job_20_5_7 1 Job_20_10_7 1 Job_20_20_7 0.82
Job_20_5_8 1 Job_20_10_8 0.9814 Job_20_20_8 0.96
Job_20_5_9 1 Job_20_10_9 1 Job_20_20_9 0.894
Job_20_5_10 1 Job_20_10_10 1 Job_20_20_10 0.968
Average 1 Average 0.952 Average 0.9126
TABLE5: Obtained satisfaction level for problems (20-5), (20-10) and (20-20)
level
level
level
Job_50_5_1 1 Job_50_10_1 0.9955 Job_100_5_1 1
Job_50_5_2 1 Job_50_10_2 1 Job_100_5_2 1
Job_50_5_ 3 1 Job_50_10_ 3 1 Job_100_5_ 3 1
Job_50_5_4 1 Job_50_10_4 0.8919 Job_100_5_4 1
Job_50_5_5 1 Job_50_10_5 1 Job_100_5_5 1
Job_50_5_6 1 Job_50_10_6 1 Job_100_5_6 1
Job_50_5_7 1 Job_50_10_7 1 Job_100_5_7 1
Job_50_5_8 1 Job_50_10_8 0.9888 Job_100_5_8 1
Job_50_5_9 1 Job_50_10_9 0.848 Job_100_5_9 1
Job_50_5_10 1 Job_50_10_10 0.7837 Job_100_5_10 1
Average 1 Average 0.9508 Average 1
TABLE 6: Obtained satisfaction level for problems (50-5), (50-10) and (100-5)
5. CONCLUSION
In this paper, we have presented a new metaheuristics to solve multi-criteria permutation
flowshop problems taking into account the manager’s preferences. This metaheuristic is an
improvement of the model proposed by Allouche et al. [2] which can be now useful for large scale
multi-criteria scheduling problems. This metaheuristic is based on the tabu search algorithm. The
concept of satisfaction functions was utilized to integrate explicitly the manager’s preferences.
The obtained sequence can be qualified as the best compromise. We have considered three
scheduling criteria. However, this metaheuristic can be extended for additional criteria that the
manager may want to consider. We believe that our approach is easy to use and requires a small
number of parameters to be provided by the manager. This approach can be qualified as a good
tool for multi-criteria scheduling problems.
6. REFERENCES
1. Allahverdi, A., “A new heuristic for m-machine flowshop scheduling problem with bicriteria of
makespan and maximum tardiness”. Computers &Operations research, 31(2): 157-180, 2004.
2. Allouche, M.A., Aouni, B., Martel, J.-M., Loukil, T. and Rebaï, A., “Solving Multi-criteria
Schedulin flow shop Problem through Compromise Programming and satisfaction
Functions. European Journal of Operational Research, 192: 460-467, 2009.

3. Aouni, B., M.A. Allouche et J.-M. Martel “ Ordonnancement multicritère à l’aide du Compromise
Programming”. In: Proceedings of The Annual Conference of the Administrative Sciences
Association of Canada, Production and Operations Management, 21(7): 1-11, 2000.
4. Armentano, V.A and J.E.C. Arroyo “An Application of a Multi-Objective Tabu Search Algorithm
to a Bicriteria Flowshop Problem”. Journal of Heuristics, 10: 463-481, 2004.
5. Arroyo, J.E.C and V.A. Armentano “A partial enumeration heuristic for multi-objective flowshop
scheduling problems”. The Journal of the Operational Research Society. Oxford: 55(9):
1000-1007, 2004.
6. Daniels, R.I. and R.J. Chambers, “Multiobjective flowshop Scheduling”. Naval Research
Logistics, 37: 981-995, 1990.
7. Framinan, J.M. and R. Leisten “A multi-objective iterated greedy search for flowshop
scheduling with makespan and flowtime criteria”. OR Spectrum, 30(4): 787-804, 2008.
8. Gagné, C., M. Gravel and W.L.Price " Optimisation Multi-Objectifs à l’aide d’un Algorithme de
Colonie de Fourmis”. Information systems and Operational Research (INFOR), 42(1):
23-42, 2004.
9. Gagné, C., M. Gravel and W.L.Price “Using Metaheuristic Compromise Programming for the
Solution of Multiple Objective Scheduling Problems”. Journal of the Operational Research
Society, 56(6): 687-698, 2005.
10. Gangadhran, R. and C. Rajendran “A Simulated Annealing Heuristic for Scheduling in a Flow-
Shop with Bicriteria”. Computers and Industrial Engineering, 27 (1-4): 473-476, 1994.
11. Glover, F. “Futur Paths for Integer Programming and Links to Atificial Intelligence”. Computers
and Operations Research, Vol.5, 533-549, 1986.
12. Glover, F. “Tabu search – Part I”. ORSA Journal on computing, 1: pp.190-206, 1989.
13. Glover, F. “Tabu search – Part II”. ORSA Journal on computing, 2: pp.4-32, 1990.
14. Glover, F., and M. Laguna. “Tabu search, in: C.R. reeves (ed.) Modern Heuristic Techniques
for Combinatorial Problems”, Blackwell Scientific Publications. Oxford, pp.70-150, 1989.
15. Glover, F., Taillard, E., and de Werra, D. “A user’s guide to tab search”. Annals of Operations
Research, 41: pp. 3-28. 1993.
16. Gupta, J.N.D., Neppall, V.R. and F. Werner. “Minimizing total flow time in a two-machine flow-
shop problem with minimum Makespan”. International Journal Of Production Economics,69:
323–338, 2001.
17. Kondakci, S.K., Azizoglu, M. and M. Ko¨ksalan, “Bicriteria scheduling for minimizing flow time
and maximum tardiness”. Naval Research Logistics, 43: pp. 929–936, 1996.
18. Lemesre, J., Dhaenens, C. and E.G. Talbi. “Parallel partitioning method (PPM): A new exact
method to solve bi-objective problems”. Computers and Operations Research, 34: 2450–
2462, 2007.
19. Loukil, T., J. Teghem and D. Tuyttens. “Solving multi-objective production scheduling using
met heuristics”. European Journal of Operational Research, 161(1): 42-61, 2005.

20. MacCarthy, B.L. and J. Liu. “Addressing the gap in scheduling research: A review of
optimization and Heuristic methods in production scheduling”. International Journal of
Production Research, 31 (1): 59–79, 1993.
21. Martel, J.-M and B. Aouni. “Incorporating the Decision-Maker’s Preferences in the Goal-
Programming Model”. Journal of Operational Research Society, 41: 1121-1132, 1990.
22. OR-Library: http://people.brunel.ac.uk/~mastjjb/jeb/orlib/flowshopinfo.html
23. T’kindt, V. and J.-C. Billaut. “Multicriteria Scheduling Theory, Models and Algorithms”.
Springer Verlag, 2002.
24. Taillard, E. “Benchmarks for Basic Scheduling Problems”. European Journal of Operational
Research, 64: 278-285, 1993.
25. Ulungu, E. and J. Teghem. “The two phases method: An efficient procedure to solve bi-
objective combinatorial optimization problems”. Foundations of Computing and Decision
Sciences 20(2): 149–165, 1995.
26. Varadharajan, T.K., and C. Rajendran. “Amulti-objective simulated-annealing algorithm for
scheduling in flowshop to minimize the makespan and total flowtime of jobs”. European
Journal of Operational Research. 167(3): 772-795, 2005.
27. Zeleny, M. “Multiple Criteria Decision Making”, Springer-Verlag, Berlin( 1976).
28. Zeleny, M. “Multiple Criteria Decision Making”. McGraw-Hill, New York (1982).
29. Zeleny, M. “Compromise Programming”. In: Multiple Criteria Decision Making, Cochrane, J.L.
And M. Zeleny (Eds) University of South Carolina Press, Columbia, 262-301.1973.

Manager’s Preferences Modeling within Multi-Criteria Flowshop Scheduling Problem: A Metaheuristic Approach

Recommended

More Related Content

What's hot (19)

Viewers also liked (18)

Similar to Manager’s Preferences Modeling within Multi-Criteria Flowshop Scheduling Problem: A Metaheuristic Approach (20)

More from Waqas Tariq (20)

Recently uploaded (20)

Manager’s Preferences Modeling within Multi-Criteria Flowshop Scheduling Problem: A Metaheuristic Approach