Analysis of evolutionary techniques applied in the programming of

operations

Mariano Frutos 1,3, Fernando Tohmé 2,3, Daniel A. Rossit 1,3, Elizabeth Garmendia 1,3 and
Barbara Damiani 1,3
1
Department of Engineering
Av. Alem 1253, Bahía Blanca (B8000CPB), Argentina
[email protected]
Phone: +54 0291 4595135
Fax: +54 0291 4595136
2
Department of Economics
12 de Octubre y San Juan, Bahía Blanca (B8000CPB), Argentina
[email protected]
Phone: +54 0291 4595138
Fax: +54 0291 4595139
3
Universidad Nacional del Sur

Abstract
Planning, in production activities, means to design, coordinate, manage and control all the operations involved in the
production system. This framework generates many MOPs (Multi-Objective Optimization Problems). They require the
optimization of several functions that are usually very complex which makes the search for solutions very expensive.
Multi-objective optimization seeks Pareto-optimal solutions for these problems. Given the dimensionality of the
problems and the combinatorial nature of the algorithms used makes it difficult to find efficient solutions. In the last
decades, many MOPs have been applied on production problems. They have used, successfully, techniques based on
Genetic Algorithms (GA). In this work, we evaluate, in the realm of non-standardized planning processes, NSGAII
(Non-dominated Sorting Genetic Algorithm II) and SPEAII (Strength Pareto Evolutionary Algorithm II) as well as their
predecessors, NSGA and SPEA.

Keywords: Multi-objective Optimization, Pareto Frontier, Genetic Algorithm, Job-Shop Scheduling Problem.

Resumen
La planificación, en el ámbito productivo, se encarga de diseñar, coordinar, administrar y controlar todas las operaciones
que se hallan presentes en la explotación de los sistemas productivos. En este marco de trabajo, aparecen numerosos
MOPs (Multi-Objective Optimization Problems). Estos constan de varias funciones que suelen ser complejas y
evaluarlas puede ser muy costoso. La optimización multi-objetivo es la disciplina que trata de encontrar las soluciones,
denominadas Pareto óptimas, a este tipo de problemas. La compleja resolución de los MOPs es debida a las dimensiones
propias del problema, al carácter combinatorio de los algoritmos y a la naturaleza de los objetivos los cuales están
vinculados a la eficiencia del sistema. En las últimas décadas muchos MOPs vinculados a la producción han sido
tratados con éxito con técnicas de resolución basadas en Algoritmos Genéticos (GAs, Genetic Algorithms). En este
trabajo se evalúa a NSGAII (Non-dominated Sorting Genetic Algorithm II), SPEAII (Strength Pareto Evolutionary
Algorithm II) y a sus antecesores, NSGA y SPEA, en el proceso de planificación de la producción no estandarizada.

Palabras clave: Optimización Multi-objetivo, Frontera de Pareto, Algoritmo Genético, Job-Shop Scheduling Problem.
 Introduction

To schedule production in a Job-Shop environment means to allocate adequately the available resources. To do that it is
necessary to rely on efficient optimization procedures. In fact, the Job-Shop Scheduling Problem (JSSP) is a NP-Hard
problem [29], so ad-hoc algorithms have to be applied to its solution [14]. This is similar to other combinatorial
programming problems [23], [8]. Most instances of the Job-Shop Scheduling Problem involve the simultaneous
optimization of two usually conflicting goals. This one, like most multi-objective problems, tends to have many
solutions. The Pareto frontier reached by an optimization procedure has to contain a uniformly distributed number of
solutions close to the ones in the true Pareto frontier. This feature facilitates the task of the expert who interprets the
solutions [20]. In this paper we present a Genetic Algorithm linked to a Simulated Annealing procedure able to schedule
the production in a Job-Shop manufacturing system [5], [8],[17], [28], [30].

1.1 JSSP Treatments: State of the Art

The huge literature on the topic presents a variety of solution strategies that go from simple priority rules to sophisticated
parallel branch-and-bound algorithms. A particular variety of scheduling problem is the JSSP. Muth and Thompson’s 1964
[22] book Industrial Scheduling presented the JSSP, basically in its currently known form. Even before, Jackson in 1956
[18] generalized the flow-shop algorithm of Johnson (1954) [19] to yield a job-shop algorithm. In 1955, Akers and
Friedman [2] gave a Boolean representation of the procedure, which later Roy and Sussman (1964) [25] described by
means of a disjunctive graph, while Egon Balas, already in 1969 [4], applied an enumerative approach that could be better
understood in terms of this graph. Giffler and Thompson (1960) [15] presented an algorithm based on rule priorities to
guide the search. For these reasons, the problem was already part of the folklore in Operations Research years before its
official inception. The JSSP generated a huge literature. Its resiliency made it an ideal problem for further study. Besides,
its usefulness made it a problem worth to scrutinize. Due to its complexity, several alternative presentations of the problem
have been tried [6], [26], [10], in order to apply particular algorithms like Clonal Selection [9], Taboo Search [3], Ant
Colony Optimization [21], Genetic Algorithms [31], Priority Rules [24], Shifting Bottlenecks [1], etc. The performance of
these meta-heuristic procedures varies, and some seem fitter than others.

1.2 Multi-Objective Optimization: Basic Concepts

Our goal in this section is to characterize the general framework in which we will state the Job-Shop problem. We
assume, without loss of generality, that there are several goals (objectives) to be minimized. Then, we seek to find a
 
vector x * = [x1* ,..., x *n ]T of decision
  variables, satisfying q inequalities g i (x) ≥ 0, i = 1,..., q as well as p equations
   T
h i (x) = 0, i = 1,..., p , such that f (x) = [f1 (x),..., f k (x)] , a vector of k functions, each one corresponding to an objective,
defined over the decision variables, attains its minimum. The class of the decision vectors satisfying the q inequalities
 
and the p equations is denoted by Ω and each x ∈ Ω is a feasible alternative. A x * ∈ Ω is Pareto optimal if for any
 *  
x ∈ Ω and every i=1,…,k, f i (x ) ≤ f i (x) . That is, if there is no x that improves some objectives without worsening the
   
others. To simplify the notation, we say that a vector u = [u1 ,..., u n ]T dominates another, v = [v1 ,..., v n ]T (denoted u  v )
if and only if ∀i∈ {1,...,k} , u i ≤ vi ∧ ∃i ∈ {1,..., k} : u i < v i . Then, the set   of Pareto optima is
    
P* = {x ∈ Ω ¬ ∃ x ' ∈ Ω, f (x ' )  f (x)} while the corresponding Pareto frontier is FP* = {f (x), x ∈ P* } .

 Flexible Job-Shop Scheduling Problem

The JSSP can be described as that of organizing the execution of n jobs on m machines. We assume a finite number of
tasks, {J j }nj=1 . These tasks must be processed by a finite number of machines {M k }km=1 . To process a task J j in a
machine M k is denoted by Oijk , where i indicates the order in which a class of operations {S j }nj=1 is applied on a task J j .
Oijk requires the uninterrupted use of a machine M k for a period τijk (the processing time) at a cost υijk . In the
particular case of Flexible JSSP, the allocation of Oijk on M k is undifferentiated, which means that each Oijk can be
processed by any of the machines in {M k }km=1 . After allocating the operations, we obtain a finite class E of groupings of
the Oijk s on the same machine. We denote each of these groupings as E k , for k = 1,..., m . A key issue here is the
scheduling of activities, i.e. the determination of the starting time t ijk of each Oijk . (see Table 1).

2
 MF01 / Problem 3 × 4 with 8 operations (flexible)

M1 M2 M3 M4
Jj Oijk
τ i
j1 υi
j1 τ i
j2 υ
i
j2 τ i
j3 υ
i
j3 τ i
j4 υij4

O11k 1 10 3 8 4 6 1 9

J1 2 3 4 8 2 2 10 1 12
O1k
3 3 8 5 4 4 6 7 3
O1k
1 4 7 1 16 1 14 4 6
O 2k

J2 O 22k 2 10 3 8 9 3 3 8

O32k 9 3 1 15 2 10 2 13

O13k 8 6 6 8 3 12 5 10
J3
2 4 11 5 8 8 6 1 18
O 3k

Table 1. Flexible Job-Shop Scheduling Problem

The Flexible JSSP demands a procedure to handle its two sub-problems: the allocation of the Oijk s on the different M k s
and their sequencing, guided by the goals to reach. That is, to find optimal levels of Processing Time (PT) (1) and Total
Operation Costs (TOC) (2).

PT → Min : máx (t ijk + τijk ) (1)

TOC → Min : ∑∑∑ j i k

x ijk υijk (2)

Where x ijk = 1 if Oijk ∈ E k and 0 otherwise. On the other hand ∑ k x ijk = 1 . Besides, t ijk = máx (t (ijh−1) + τ(ijh−1) , t spk + τspk , 0)
(i −1) (i −1)
máx (t jh +τ jh , t spk + τspk , 0) for each pair Oijh−1 , Ospk ∈ E k and all machines M k , M h and operations Si , Ss .

 Hybrid Genetic Algorithm

Due to its many advantages, evolutionary algorithms have become very popular for solving multi-objective optimization
problems [33], [7]. Among the evolutionary algorithms used, some of the most interesting are Genetic Algorithms (GA)
[16]. To represent the individuals, we use a variant of [30]. Since the Flexible JSSP has two subproblems, the Hybrid
Genetic Algorithm (HGA) presented here operates over two chromosomes. The first one represents the allocation A ijk of
each Oijk to every M k . We denote with values between 0 and (m-1) the allocation of each M k , that is, for m = 4, we
might have something like 0→M1, 1→M2, 2→M3 and 3→M4. The second chromosome represents the sequencing of the
Oijk already assigned to each of the M k (∀Oijk ∈ E k ) . We denote with values between 0 and (n!-1) the sequence of J j in
each M k . That is, for n = 3, we may have 0→J1J2J3, 1→J1J3J2, 2→J2J1J3, 3→J2J3J1, 4→J3J1J2 and 5→J3J2J1 (see Table
2).
The algorithm NSGAII (Non-Dominated Sorting Genetic Algorithm II) [11], creates an initial population, be it random
or otherwise. NSGAII uses an elitist strategy joint with an explicit diversity mechanism. Each individual candidate
solution i is assumed to have an associated rank of non-dominance ri and a distance d i which indicates the radius of the
area in the search space around i not occupied by another solution (3). A solution i is preferred over j if ri < rj . When i
and j have the same rank, i is preferred if d i > d j . Let Yi be an ordered class of individuals with same rank as i and f ji+1
the value for objective j for the individual after i, while f ji−1 is the value for the individual before i. f jmáx is the maximal
value for j among Yi while f jmín is the minimal value among Yi . The distances consider all the objective functions and
attach an infinite value to the extreme solutions in Yi . Since these yield the best values for one of the objective functions
on the frontier, the resulting distance is the sum of the distances for the N objective functions.

3
 MF01 / Problem 3 × 4 with 8 operations (flexible)

M1 M2 M3 M4
Jj
3 3 0 5

O11k 2

J1 2 1
O1k
3 0
O1k
1 1
O 2k

J2 O 22k 2

O32k 3

O13k 0
J3
2 3
O 3k

0→J1J2J3, 1→J1J3J2, 2→J2J1J3, 3→J2J3J1, 4→J3J1J2 y 5→J3J2J1

0→MB1B, 1→MB2B, 2→MB3B, 3→MB4

Table 2. Chromosome encoding process

Starting with a population Pt a new population of descendents Q t obtains. These two populations mix to yield a new
one, R t of size 2N (N is the original size of Pt ). The individuals in R t are ranked with respect the frontier and a new
population Pt +1 obtains applying a tournament selection to R t .

d = ∑ (f − f ) (f
N

i j
i +1
j
i −1 máx
j
−f )mín
j
(3)
j=1

After experimenting with several genetic operators we have chosen the Uniform Crossover (UX) for the crossover and
Two-Swap for mutation [13]. After the individuals have been affected by these operators and before allowing them to
become part of a new population we apply an improvement operator. This operator has been designed following the
meta-heuristic Simulated Annealing (SA) [12]. For the change of structure of both chromosomes we select a gen
randomly to change its value. This process, which complements the genetic procedure follows the guidelines of SA. The
general layout of the whole procedure is depicted in Fig. 1.

Number of Jobs Number of Machines

(n) (m)

Set of sequences
Inicialization Algorithm

Set of integer numbers Set of integer numbers Set of machines

{J1J2...J(n-1)Jn, J1J2...JnJ(n-
{0, 1, …, (n!-1)} {0, 1, …, (m-1)} {M1, M2, …, Mm}
1),…}

Relating sets Relating sets

0→J1J2...J(n-1)Jn 0→M1
1→J1J2...JnJ(n-1) 1→M2
... ...

Initial population
Multi-Objective Memetic Algorithm

↓↑
Random initialization Individual Random initialization
Mk→Value ↓ Oikj→Value
(Chromosomes of Decoding (Chromosomes of
Sequences) ↓ Allocations)
Fitness

NSGAII * Simulated Annealing + Decoding

End Algorithm

Fig. 1. Lay-out of the Hybrid Genetic Algorithm

4
 Experiences and results

The parameters and characteristics of the computing equipment used during these experiments were as follows: size of
the population: 200; number of generations: 500; type of cross-over: uniform; probability of cross-over: 0.90; type of
mutation: two-swap; probability of mutation: 0.01; type of local search: simulated annealing (T i: 850, Tf: 0.01, α: 0.95,
ω: 10); probability of local search: 0.01; CPU: 3.00 GHZ and RAM: 1.00 GB. We compare now our algorithm with
other three Multi-objective Evolutionary Algorithms (MOEAs).

4.1 Comparing NSGAII to other algorithms

The MOEAs considered in our exercise were: Non-dominated Sorting Genetic Algorithm (NSGA) [27], Strength Pareto
Evolutionary algorithm (SPEA) [32] and Strength Pareto Evolutionary algorithm II (SPEAII) [34]. NSGA classifies the
individuals in layers or fronts, by grouping all the non-dominated individuals in a single front, with the same value of
fitness for each individual. This value is proportional to the size of the population as to provide reproduction potential
for all the individuals in the front. This procedure is repeated on the remaining individuals (those outside the non-
dominated front) and so on until all the individuals in the population are classified. Since the ones in the first front have
higher fitness they get more attention than the rest of the individuals. SPEA is an algorithm that at each generation
memorizes the non-dominated individuals and deletes from memory those that became dominated. For each individual
in the external system, a strength value is computed, proportional to the number of solutions in which it is dominant. In
SPEA, the fitness of a member of the current population is computed by adding the strengths of the external non-
dominated solutions that dominate it. SPEAII instead, has a fine-tuning procedure according to which the fitness of an
individual obtains as a balance between the number of solutions that are dominated by it and the number that dominate
the individual. Besides, it uses the “nearest neighbor” for valuing the density of feasible solutions and thus leading to a
more efficient search.

4.2 Experiments and results

Initially we present four solutions, two dominated solutions (see Table 3 and Table 5) and two undominated solutions
(see Table 4 and Table 6) for the problem MF01 [14]. For the problems MF01, MF02, MF03, MF04 and MF05 [14], we
show the results for the multi-objective analysis based on Processing Time and Total Operation Costs. They were
obtained by running each algorithm 10 times. For each algorithm the sets of undominated solutions P 1, P2,…, P10 were
obtained as well as the superpopulation PT = P1 ∪ P2 ∪ ... ∪ P10 . From each superpopulation a class of undominated
solutions was extracted, constituting the Pareto frontier for each algorithm. These are shown in Fig. 2 (MF01), Fig. 3
(MF02), Fig. 4 (MF03), Fig. 5 (MF04) and Fig. 6 (MF05).

Scheduling (MF01)
Jj Oijk
A i
jk t ijk t ijk + τijk υijk
M1→ J1 J3 J2 M3→ J3 J2 J1
O11k M4 0 1 9

J1 (→) 2
O1k M4 1 2 12

3
O1k M1 2 5 4

1
O 2k
M2 0 1 14
M2→ J2 J1 J3 M4→ J1 J3 J2
J 2 (→) O 22k M2 1 4 3

O32k M4 4 6 13

O13k M3 0 3 5
J 3 (→)
2
O 3k
M4 3 4 18

PT: 6, TOC: 78

Table 3. Scheduling (HGA) (PT: 6, TOC: 78)

5
 Scheduling (MF01)
Jj Oijk
A ijk t ijk t ijk + τijk υijk
M1→ J2 J1 J3 M3→ J3 J1 J2
O11k M1 0 1 4

J1 (→) 2
O1k M4 1 2 12

3
O1k M1 3 6 4

O12k M2 0 1 4
M2→ J2 J1 J3 M4→ J1 J3 J2
J 2 (→) O 2
2k
M1 1 3 9

O32k M2 3 4 10

O13k M3 1 4 5
J 3 (→)
2
O3k M4 4 5 18

PT: 6, TOC: 66

Table 4. Scheduling (HGA) (PT: 6, TOC: 66)

Scheduling (MF01)
Jj Oijk
A i
jk t ijk t ijk + τijk υijk
M1→ J2 J1 J3 M3→ J1 J2 J3
O11k M3 0 4 6

J1 (→) 2
O1k M2 4 12 2

3
O1k M4 12 19 3

1
O 2k
M4 19 23 6
M2→ J1 J2 J3 M4→ J1 J2 J3
J 2 (→) O 22k M3 23 32 3

O32k M1 32 41 3

O13k M1 41 49 6
J 3 (→)
2
O 3k
M3 49 57 6

PT: 57, TOC: 35

Table 5. Scheduling (HGA) (PT: 57, TOC: 35)

Scheduling (MF01)
Jj Oijk
A i
jk t ijk t ijk + τijk υijk
M1→ J3 J2 J1 M3→ J1 J3 J2
O11k M3 0 4 6

J1 (→) 2
O1k M2 4 12 2

3
O1k M4 12 19 3

1
O 2k
M4 0 4 6
M2→ J3 J2 J1 M4→ J2 J1 J3
J 2 (→) O 22k M3 16 20 3

O32k M1 20 29 3

O13k M1 0 8 6
J 3 (→)
2
O 3k
M3 8 16 6

PT: 29, TOC: 35

Table 6. Scheduling (HGA) (PT: 29, TOC: 35)

6
 450 80

80
NSGAII
HGA
HGA
400 60
NSGA
GRASP
GRASP

350 SPEAII
TS
TS
40

ACO
SPEA
ACO
60300 Approximate Pareto
20
Approximate Pareto Frontier
Frontier

250 0
0 10 20 30 40

200
40
TOC 150

100

20 50

0
0 20 40 60 80 100 120 140

0
0 10 20 30 40

PT
Fig. 2. Solutions for MF01

450 80
200
NSGAII
HGAHGA
400 60
NSGA
GRASP
GRASP

350 SPEAII
TS TS 40

ACO
SPEA
ACO
150
300 Approximate Pareto Frontier
20
Approximate Pareto
Frontier

250 0
0 10 20 30 40

200
100
TOC 150

100

50
50

0
0 20 40 60 80 100 120 140

0
0 20 40 60 80

PT
Fig. 3. Solutions for MF02

450 80
450
NSGAII
HGA
HGA
400 60

400 NSGA
GRASP
GRASP

350 SPEAII
TS
TS 40

350 ACO
SPEA
ACO

300 Approximate Pareto

20
Approximate Pareto Frontier
Frontier
300
250 0
0 10 20 30 40

250
200

TOC 200
150

100
150

50
100

0
50
0 20 40 60 80 100 120 140

0
0 20 40 60 80 100 120 140

PT
Fig. 4. Solutions for MF03

7
 450 80
350
NSGAII
HGA
HGA
400 60
NSGA
GRASP
GRASP
300
350 SPEAII
TS
TS
40

ACO
SPEA
ACO

300
250 Approximate Pareto
20
Approximate Pareto Frontier
Frontier

250 0
0 10 20 30 40
200
200

TOC 150
150

100
100
50

500
0 20 40 60 80 100 120 140
0
0 20 40 60 80

PT
Fig. 5. Solutions for MF04

450 80
900
NSGAII
HGA
HGA
400 60

800 NSGA
GRASP
GRASP

350 SPEAII
TS
TS
40

700 ACO
SPEA
ACO

300 Approximate Pareto

20
Approximate Pareto Frontier
Frontier
600
250 0
0 10 20 30 40

500
200

TOC 400
150

300
100

50
200

0
100
0 20 40 60 80 100 120 140
0
0 20 40 60 80 100 120 140 160

PT
Fig. 6. Solutions for MF05

To obtain an approximation to the true Pareto front (Approximate Pareto Frontier (Fig. 2, Fig. 3, Fig.4, Fig. 5 and
Fig.6)), we take the fronts of each algorithm, from which all the dominated solutions are eliminated. for each algorithm
we compute the percentage of solutions in this front (see Table 7).

Percentage of solutions in the Approximate Pareto Frontier

NSGAII NSGA SPEAII SPEA

MF01 83,33% 27,78% 66,67% 66,67%

MF02 85,71% 21,43% 82,14% 53,57%

MF03 95,45% 31,82% 86,36% 63,64%

MF04 92,59% 55,56% 88,89% 62,96%

MF05 90,32% 48,39% 77,42% 77,42%

Table 7. Comparing NSGAII, NSGA, SPEAII and SPEA

8
 Conclusions and Future Work

We presented a Hybrid Genetic Algorithm (HGA) to solve Flexible Job-Shop Scheduling Problem (Flexible JSSP)
based on NSGAII and Simulated Annealing. The application of HGA required the calibration of parameters, in order to
yield valid values for the problem at hand, which are also a reference for similar problems. Furthermore, we consider
alternative hybrid algorithms in which NSGAII is replaced by other multi-objective evolutionary algorithms (MOEAs).
In most of the problems the quality of the solutions found by NSGAII is equal or higher than the results of SPEAII. On
the other hand, NSGAII definitely improves over NSGA and SPEA. We can conclude that the HGA here is a good
approach to solving Flexible JSSPs. We are currently experimenting with other techniques of local search in order to
achieve a more aggressive exploration. We are also interested in evaluating the performance of the procedure for other
kinds of problems to see whether it saves resources without sacrificing precision in the convergence.

Acknowledgements

We would like to thank the economic support of the Consejo Nacional de Investigaciones Científicas y Técnicas
(CONICET) and the Universidad Nacional del Sur (UNS) for Grant PGI 24/JO39. We want also thank Dr. Ana C.
Olivera for her constant support and help during this research.

Reference

[1] Adams J., Balas E. and Zawack D., The Shifting Bottleneck Procedure for job shop scheduling, Management
Science, Vol. 34 (3), pp 391-401, 1998.
[2] Akers S. B. and Friedman J., A Non-Numerical Approach to Production Scheduling Problems, Operations
Research, Vol. 3 (4), pp 429-442, 1955.
[3] Armentano V. and Scrich C., Taboo search for minimizing total tardiness in a job-shop, International Journal of
Production Economics, Vol. 63, pp 131-140, 2000.
[4] Balas E., Duality in Discrete Programming: The Quadratic Case, Management Science, Vol. 16 (1), pp 14-32,
1969.
[5] Binato S., Hery W. J., Loewenstern D. M. and Resende M. G. C., A grasp for job shop scheduling, Essays and
Surveys in Meta-heuristics, pp. 59-80, 2001.
[6] Cheng C. C. and Smith S. F., Applyng constraint satisfaction techniques to job shop scheduling, Annals of
Operations Research, Vol. 70, pp 327-357, 1997.
[7] Coello Coello C. A., Van Veldhuizen D. A. and Lamont G. B., Evolutionary Algorithms for Solving Multi-
Objective Problems, Kluwer Academic Publishers, New York, 2002.
[8] Cortés Rivera D., Coello Coello C. A. and Cortés N. C., Job shop scheduling using the clonal selection principle,
ACDM'2004, UK, 2004.
[9] Cortés Rivera D., Coello Coello C. A. and Cortés N. C., Use of an Artificial Immune System for Job Shop
Scheduling, In Proceeding Second International Conference on Artificial Immune Systems, Edinburgh, Scotland.
Springer-Verlag, Lecture Notes in Computer Science, Vol. 2787, pp 1-10, 2003.
[10] Crawford J. M. and Baker A. B., Experimental Results on the Application of Satisfiability Algorithms to Scheduling
Problems, Computational Intelligence Research Laboratory, 1994.
[11] Deb K., Pratap A., Agarwal S. and Meyarivan T., A Fast and Elitist Multi-objective Genetic Algorithm: NSGAII,
IEEE Transactions on Evolutionary Computation, Vol. 6 (2), pp 182-197, 2002.
[12] Dowsland K. A., Simulated Annealing, Modern Heuristic Techniques for Combinatorial Problems, Ed. C. R.
Reeves, Blackwell Scientific Pub, Oxford, 1993.

9
[13] Fonseca C. M. and Fleming P. J., Multi-objective genetic algorithms made easy: Selection, sharing and mating
restriction, GALESIA, pp 45-52, 1995.
[14] Frutos M., Olivera A. C. and Tohmé F., A Memetic Algorithm based on a NSGAII Scheme for the Flexible Job-
Shop Scheduling Problem. Annals of Operations Research, Vol. 181, pp 745-765. 2010.
[15] Giffler B. and Thomson G. L., Algorithms for Solving Production Scheduling Problems, Operations Reseach, Vol.
8, pp 487-503, 1960.
[16] Goldberg, D., Genetic Algorithms in Search, Optimization, and Machine Learning, Addison Wesley, 1989.
[17] Heinonen J. and Pettersson F., Hybrid ant colony optimization and visibility studies applied to a job-shop
scheduling problem, Applied Mathematics and Computation, pp 989-998, 2007.
[18] Jackson J. R., An Extension of Johnson's Results on Job Lot Scheduling, Naval Research Logistics Quarterly, Vol.
2, pp 201-203, 1956.
[19] Johnson S. M., Optimal two- and three-stage production schedules with setup times included, Naval Research
Logistics Quarterly, Vol. 1, pp 61-68, 1954.
[20] Kacem I., Hammadi S. and Borne P., Approach by Localization and Multi-Objective Evolutionary Optimization
for Flexible Job-Shop Scheduling Problems, IEEE Trans. Syst. Man Cybernetics, Vol. 32, 2002.
[21] Merkle D. and Middendorf M., A new approach to solve permutation scheduling problems with ant colony optimization, In
Proceedings Applications of Evolutionary Computing, EvoWorkshops 2001, Vol. 2037, pp 484-494, 2001.
[22] Muth J. F. and Thompson G. L., Industrial Scheduling, Prentice-Hall Inc, 1964.
[23] Olivera A. C., Frutos M. y Casal R., Métodos para determinar secuencias de producción en un ambiente productivo
complejo, In Proceeding XIII Congreso Latino-Iberoamericano de Investigación Operativa, Uruguay, 2006.
[24] Panwalker S. and Iskander W., A survey of scheduling rules, Operations Research, Vol. 25 (1), pp 45-61, 1977.
[25] Roy B. and Sussman B., Les problèmes d'ordonnancement avec contraintes disjonctives, SEMA, 1964.
[26] Sadeh N. M. and Fox M. S., Variable and value ordering heuristics for the Job Shop scheduling constraint
satisfaction problem, Tecnical report CMU-RI-TR-95-39, Artificial Intelligence Journal, 1995.
[27] Srinivas N., Multi-Objetive Optimization Using Nondominated Sorting in Genetic Algorithms, Master Thesis, Indian
Institute of Technology, Kuanpur, India, 1994.
[28] Tsai C. F. and Lin F. C., A new hybrid heuristic technique for solving job-shop scheduling problem, Intelligent
Data Acquisition and Advanced Computing Systems, Second IEEE International Workshop, 2003.
[29] Ullman J. D., NP-complete scheduling problems”. Journal of Computer System sciences, Vol. 10, pp 384-393,
1975.
[30] Wu C. G., Xing X. L., Lee H. P., Zhou C. G. and Liang Y. C., Genetic Algorithm Application on the Job Shop
Scheduling Problem, Machine Learning and Cybernetics, International Conference, Vol. 4, pp 2102- 2106, 2004.
[31] Zalzala A. M. S. and Flemming P. J., Genetic Algorithms in engineering systems, London Institution of Electrical
Engineers, 1997.
[32] Zitzler E. and Thiele L., Multiobjective Evolutionary Algorithms: A Comparative Case Study and the Strength
Pareto Approach, IEEE Transactions Evolutionary Computation, Vol. 3 (4), 257-271, 1999.
[33] Zitzler E., Deb K., Thiele L., Coello Coello C. A. and Corne D., Evolutionary multi-criterion optimization, In
Lecture Notes in Computer Science, Vol. 1993, Springer, Berlin, 2001.
[34] Zitzler E., Laumanns M. and Thiele L., SPEAII: Improving the Strength Pareto Evolutionary Algorithm for
Multiobjective Optimization, In Giannakoglou, Tsahalis, Periaux, Papailiou, and Fogarty (eds), Evolutionary
Methods for Design, Optimisations and Control, 19-26, 2002.

10