Anda di halaman 1dari 24

Flex Serv Manuf J

DOI 10.1007/s10696-012-9143-6

Using Genetic Algorithms to solve scheduling problems


on flexible manufacturing systems (FMS): a literature
survey, classification and analysis
Moacir Godinho Filho Clarissa Fullin Barco
Roberto Fernandes Tavares Neto

 Springer Science+Business Media, LLC 2012

Abstract This paper reviews the literature regarding Genetic Algorithms (GAs)
applied to flexible manufacturing system (FMS) scheduling. On the basis of this
literature review, a classification system is proposed that encompasses 6 main
dimensions: FMS type, types of resource constraints, job description, scheduling
problem, measure of performance and solution approach. The literature review
found 40 papers, which were classified according to these criteria. The literature was
analyzed using the proposed classification system, which provides the following
results regarding the application of GAs to FMS scheduling: (1) combinations of
GAs and other methods were relatively important in the reviewed papers; (2)
although most studies deal with complex environments concerning both the routing
flexibility and the job complexity, only a minority of papers simultaneously consider the variety of possible capacity constraints on an FMS environment, including
pallets and automated guided vehicles; (3) local search is rarely used; (4) makespan
is the most widely used measure of performance.
Keywords Scheduling  Flexible manufacturing systems (FMS) 
Genetic Algorithms  Literature review

M. Godinho Filho  C. F. Barco (&)  R. F. Tavares Neto


Department of Industrial Engineering, Federal University of Sao Carlos,
Via Washington Lus, Km 235, Sao Carlos, SP 13565-905, Brazil
e-mail: cfbarco@gmail.com
M. Godinho Filho
e-mail: moacir@dep.ufscar.br
R. F. Tavares Neto
e-mail: tavares@dep.ufscar.br

123

M. Godinho Filho et al.

1 Introduction
Flexible manufacturing systems (FMS) are important because they have the
capacity to quickly respond to the dynamics of the market (Chan and Chan 2004).
MacCarthy and Liu (1993a, b) state that FMS aim to achieve both high productive
flexibility and high productivity to meet present competitive needs.
FMS have been broadly studied over the last 25 years. In the last decade, they
became important elements in the success of enterprises (Chan and Chan 2004).
These manufacturing systems are composed of numeric control (NC) or computer
numeric control (CNC) machines connected by an automated material handling
system. This system configuration must be operated by an automated computer
system and is responsible for the large initial investment required to implement an
FMS. However, Balogun and Popplewell (1999) state that this initial investment
generates a set of benefits that include lead time reduction, increased throughput,
decreased inventory, and other benefits that combine to assure the economic
viability of the system. An analysis of those benefits easily shows that only a few of
them can be achieved without production scheduling. Morton and Pentico (1993)
state that scheduling is the process of organizing, assigning and temporizing
available resources to fulfill a set of activities required to process a set of jobs.
Scheduling must consider a set of constraints and one or multiple objectives. A large
number of studies have devoted attention to the roles of the activities performed
during production planning and control.
The first approaches to scheduling problems included optimal methods and
problem-specific heuristics and were restricted to manufacturing environments with
limited complexity. Classical algorithms are inadequate to treating FMS environments, as they respect a set of premises derived from less complex problems. Thus,
implementing an FMS requires the development of specific methods that consider
all of the assumptions and constraints describing the system. Of the proposed set of
solution methods, artificial intelligence techniques (e.g., Specialist Systems, Genetic
Algorithms, Neural Networks) have proven to be adequate strategies.
Genetic Algorithms (GAs) were proposed by John H (1975). Holland in 1975 and
can be understood as a process for optimizing complex functions based on the
mechanisms of genetics and natural evolution. According to Goldberg (1989), GAs
can be adapted to treat the complexity levels required to provide adaptive search at
the required robustness. Ponnambalam et al. (2001) state that GAs are the most
popular type of evolutionary algorithms. The good results provided by this class of
algorithms have led many researchers to use it to solve scheduling problems,
including those in FMS environments.
The main objective of this paper is to perform a bibliographic review,
classification and analysis of the application of GAs to solving scheduling problems
in FMS. The first step was to search the Compendex, Science direct and Google
Scholar databases. Then, a classification scheme for the resulting papers found in
step 1 was proposed. This classification scheme encompasses six criteria. In a third
phase, the papers were read and classified. Finally, the results were analyzed.
This paper is structured as follows: Sect. 2 presents the required concepts used in
this paper, and Sect. 3 describes the proposed classification scheme. Section 4

123

Using Genetic Algorithms

presents the bibliographic survey, structured according to the presented classification scheme. Section 5 contains a general analysis of the reviewed papers themes.
Section 6 presents the final conclusions and suggestions for future works.

2 An overview of the concepts used in this paper


The specialized literature contains several definitions of the term FMS (e.g., see
Kaltwasser et al. (1986), Byrkett et al. (1988), O Keefe and Kasirajan (1992), and
others). MacCarthy and Liu (1993a, b) believed that an FMS must be classified
according to its operational and control characteristics and proposed that an FMS is
a production system composed of some CNC and/or NC machines connected by a
material handling system, is capable of producing a set of products with some
degree of variability and is operated by an automated computer-controlled system.
MacCarthy and Liu (1993a b) state that an FMS is composed of three subsystems:
(1) a processing system composed of a set of CNC machines with automatic toolchanging capabilities; (2) a material handling and storage system composed of
robots, AGVs (Automated Guided Vehicles) and other systems that allow flexibility
in the movement of the processing material; (3) a computer control system that
automatically operates the system.
According to Balogun and Popplewell (1999), an FMS contains a large number
of variables and constraints that change over time, and these characteristics justify
the use of dynamic scheduling algorithms. Balogun and Popplewell (1999) note six
approaches to solving scheduling problems in FMS: (a) combinatory optimization
methods; (b) artificial intelligence; (c) simulation; (d) heuristics; (e) multi-criteria
decision making; (f) hybrid solutions. The field of Artificial Intelligence provides
other methods including specialist systems, Genetic Algorithms, and Neural
Networks. This paper focuses on Genetic Algorithms.
According to Haupt and Haupt (1998), a GA is a process to optimize highly
complex functions utilizing the mechanisms of natural evolution and genetics.
Goldberg (1989) stated that a GA is a stochastic search technique that is based on
selection and evolution. Carvalho et al. (2003) define GAs as evolutionary programs
that are based on natural selection and heredity theories. These definitions all
include the idea that in a given population, individuals with beneficial genetic
characteristics have better chances of survival, which promotes the reproduction of
the fittest individuals and the extinction of the least fit.
When Genetic Algorithms are applied to problems, each individual, or
chromosome, represents a viable solution to the problem. The algorithm, according
to Zhou et al. (2001), identifies the best solutions and combines them to generate
new individuals and renew the population.
Carvalho et al. (2003) explain the behavior of GAs as follows: initially, a
population of individuals (a set of viable solutions to the problem) is generated.
During the evolutionary process, this population is evaluated, and the fitness of each
individual is calculated and stored. A subset of the individuals with the highest
fitness scores is saved, and more individuals are generated using mutation and
crossover operations.

123

M. Godinho Filho et al.

3 The proposed classification system and the bibliographic survey


The system for classifying the literature surrounding the application of Genetic
Algorithms to scheduling problems is proposed and presented in this section. The
system is based on six criteria: (A) type of fms; (B) type of resources and capacity
constraints; (C) job characteristics; (D) scheduling problem approached; (E) measure
of performance; (F) utilized approach.
The first criterion, Type of FMS, is based on the results found by Maccarthy and
Liu (1993a) and classifies the FMS environment into five subtypes:

SFM (single flexible machine): a single production unit composed of a CNC


with tool-changing capabilities and a material handling and storage system;
FMC (flexible manufacturing cell): an FMS environment composed of a group
of SFMs sharing a single material handling system;
MMFMS (multi-machine flexible manufacturing system): contains some SFMs
connected by an automatic material handling system composed of two or more
material handling sub-systems, allowing it to serve two or more machines
simultaneously;
MCFMS (multi-cell flexible manufacturing system): an FMS composed of
several FMCs and possibly SFMs, all connected by an automatic material
handling system.
Not available: environments in which the number of machines, system
characteristics and/or physical configuration were not specified.

The second criterion refers to the types of resources and constraints on the system. This
criterion is based on the work of MacCarthy and Liu (1996), which identifies resource
types using the following notation: machines (M), storage buffers (SB), material handling
devices (HD), tool-changing devices (TD), fixtures (FX) and pallets (PL). The quantity of
each kind of resources limits the system capacity. In this paper these constraints are
represented by the notation: lim: resource (quantity of this resource). For example, if a
scenario has five machines and only one material handling device system, the second
criterion for this environment will be denoted by: lim: M (5); HD (10). Its important to
note that constrains of a system justify the scheduling requirement. According to
MacCarthy and Liu (1996), if the capacity of a specific resource is unlimited, then it does
not represent a constraint on the studied scheduling problem.
The third criterion is also based on the work of MacCarthy and Liu (1996) and
classifies the jobs according to complexity. The measure of complexity is based on
the number of operations in each job. Two options are possible:

JC1 (job complexity = 1): each job contains just one operation;
JC? (job complexity [1): some or all jobs contain two or more operations;

The third criterion also classifies the jobs according to routing flexibility, which
can assume two values:

RF1: there is only one machine enabled to perform a single operation;


RF ? : there are two or more machines allowed to perform one or more
operations.

123

Using Genetic Algorithms

The fourth criterion considers the scheduling problem that the paper focuses on.
Because the scheduling literature contains numerous methods for classifying
scheduling problems, this paper adopts some aspects of the nomenclature used by
Chan and Chan (2004) and by Balogun and Popplewell (1999). Then, in this paper
scheduling problems are classifies as follows:
(1)
(2)
(3)
(4)
(5)

sequencing problem for jobs and operations;


positioning and routing problems for the AGVs;
loading problem (allocation of operations and tools required for a machine to
perform an operation);
routing problem (routes used by each job in the FMS);
others.

It is important to note that this research analyzed papers that treated sequencing
problems in FMS environments, which explains why all of the reviewed papers
contain sequencing problems in their scope.
The fifth criterion is the measure of performance used in the paper. The present
literature review admits that these measures can contain one (mono-criterion) or
more criteria (multi-criteria). Following some classical references such as Sipper
and Bulfin (1997) and the bibliographic review itself, Table 1 contains all of the
criteria used in this paper.
Finally, the sixth criterion used in the proposed classification system is the
approach used in the paper to solve the FMS scheduling problem. Table 2 presents
the approaches found in the reviewed papers. Some authors used only one approach
to solving the problem (denoted in this paper by pure), while others considered
hybrid techniques combining two or more approaches (denoted in this paper by
hybrid).
Table 1 Criteria and the
codification used in this paper

Criteria

Code

Idle time

T idle

Length of the AGVs route

Route length

Number of backtrackings of each AGV

Backtrackings

Total flowtime

Mean flow time


Makespan

F medium
Lmax
Cmax

Tardiness

Maximum tardiness

Tmax

Maximum lateness

Due date

Dd

Cost for tardiness and earliness; production cost,


penalty cost

Cost

Throughput

Work in process

WIP

Machine utilization

Maximum utilization of the machines

Umax

123

M. Godinho Filho et al.


Table 2 Approaches found in the reviewed papers
Approach

Description

Genetic Algorithm (GA)

GAs are methods for moving from one population of chromosomes


(candidate solutions to a problem) to a new population by using a kind
of natural selection together with the genetically inspired operators
of crossover, mutation, and inversion (Mitchell 1998)

Particle swarm optimization


(PSO)

PSO is aimed at producing computational intelligence by exploiting


simple analogs of social interaction. Problem-solving is a populationwide phenomenon that emerges from the individual behaviors and
interactions of particles. Populations are organized according to some
sort of communication structure or topology, often thought of as a
social network (Poli et al. 2007)

Ant colony optimization


(ACO)

ACO is a meta-heuristic approach inspired by the pheromone trail and


following behavior of real ants, who use pheromones as a
communication medium. The pheromone trails in ACO serve as
numerical information that the ants use to probabilistically construct
solutions to the problem (Dorigo and Stutzle 2003)

Simulated Annealing (SA)

SA is a stochastic computational technique evolved from statistical


mechanics for discovering near-global-minimum-cost solutions to
large optimization problems (Kamboj and Sengupta 2009)

Tabu search (TS)

Tabu Search is a meta-heuristic created to tackle large, hard


combinatorial optimization problems. It is based on the principle that
intelligent search must embrace more efficient and systematic forms
of direction, such as memorizing and learning (Kamboj and Sengupta
2009)

Memetic algorithms (MAs)

Memetic algorithms (MAs) are search strategies that use a populationbased approach in which a set of cooperating and competing agents
are engaged in periods of individual improvement to the solution
while sporadically interacting. MAs denote a family of metaheuristics
whose central theme is hybridization and are intrinsically concerned
with exploiting all available knowledge about the problem under study
(Moscato and Cotta 2003)

Simulation/Petri Nets

Petri Nets allow for the modeling of states, events, conditions,


synchronization, parallelism, choice, and iteration and descriptions of
real processes that tend to be complex and extremely large

Fuzzy

A fuzzy set is a class of objects with a continuum of membership grades.


Such a set is characterized by a membership function that assigns each
object a membership grade ranging between zero and one. Fuzzy
Logic provides a natural way of dealing with these sets (Zadeh 1965)

Priority rules/dispatching rules

Priority rules are rules used to schedule operations and jobs. The basic
idea is to choose the job with the highest priority according to the rule
if there is more than one job waiting to be processed by the same
machine (Sipper and Bulfin 1997)

Earliest due date (EDD)


First come, first served
(FCFS)
Shortest processing time
(SPT)
Critical ratio (CR)
Neural Networks

123

A neural network is a massively parallel, distributed processor that has a


natural propensity for storing experimental knowledge and making it
available for use. Neural Networks resemble the brain in that
knowledge is acquired through a learning process and interneuron
connection strengths, known as synaptic weights, are used to store the
knowledge (Aleksander and Morton 1990)

Using Genetic Algorithms


Table 2 continued
Approach

Description

Deterministic dynamic
programming (DDP)

Programming used to solve optimization problems. The main feature of


DDP is the decomposition of the original problem into a sequence of
smaller and simpler problems (Arenales et al. 2007)

Other heuristics

Heuristics or heuristic algorithms are techniques that often quickly


provide a good solution, but not necessarily the optimal solution, i.e.,
the best solution (Fernandes and Godinho Filho 2010)

Note that, according to the scope of the present work, all of the reviewed papers
use the GA technique. This criterion is still useful to differentiate the studies that use
only GAs from the ones that combine GAs with other AI techniques.
Table 3 presents the 40 reviewed papers classified according to the six proposed
criteria.

4 Structuring the literature review using the proposed method


This section presents the structure of the review of literature pertaining to Genetic
Algorithms applied to FMS scheduling. This structure is based on two of the
classification criteria proposed in the previous section: the scheduling problem and
the work-focused measure of performance utilized by the authors. Thus, the sections
below treat two classes of papers: papers that addressed only the sequencing
problem and articles that addressed problems beyond sequencing, which may
include the allocation and AGV routing problems, the loading problem, or others.
Within these two classes, the 40 papers were sub-classified according to the number
of measures of performance used (single-criterion or multi-criteria).
4.1 Papers that address only the sequencing problem
In this section, we present the work that addressed the sequence of operations as a
single scheduling problem. Of the 40 studies in this literature review, 32 belong to
this category. Of the 32 papers presented in this section, 19 used only one measure
of performance (single-criterion), and the remaining 13 are classified as multicriteria.
4.1.1 Papers that use only one measure of performance (single-criterion)
Jawahar et al. (1998a) propose knowledge-based scheduling schemes (Work cell
attribute-oriented dynamic schedulers WCAODSs) to control the flow of parts in
real-time for FMS in which the part-mix varies continually with the planning
horizon. The work uses Genetic Algorithms to schedule an FMS with 1020
machines to minimize the makespan. A comparison made with a GA-based
scheduling methodology shows that the proposed approach provides solutions
nearer to the optimum.

123

Lim:M(5); HD(1)

Lim: M(6); HD(2)

Not available

Not available

Not available

MMFMS

MCFMS

MMFM

FMC

Holsapple et al. (1993)

Rabelo et al. (1994)

123

Fujimoto et al. (1995)

Chiu and Fu (1997)

Erkmen et al. (1997)

Ulusoy et al. (1997)

Jawahar et al. (1998a)

Not available

MCFMS

FMC

FMC

Hsu et al. (2002)

Sankar et al. (2003)

Haq et al. (2003)

Keung et al. (2003)

Lim: M(3)

Not available

Not available

Yang (2001)

Reyes et al. (2001)

FMC

Not available

Keung et al. (2001)

Saitou et al. (2002)

Lim:M(5)

FMC

Rossi and Dini (2000)

Lim: M(5); HD(1)

Lim: M(3); HD(1)

Lim: M(9); HD(1)

Lim: M(16) HD(11)

Lim: M(6)

Lim: M(10)

Lim: M(4)

Lim: M(4)

Lim: M(2); HD(1)

Lim: M(16); HD(1);


B(16)

Lim: M(210); HD(1)

MCFMS

FMC

Chung et al. (1998)

Jawahar et al. (1998b)

Lim:M(3); HD (1)

Lim:M(5); HD(1)

Lim: M(3); HD(3)

Lim:M(6)

Not available

Lim: M(3)

Lim: M(5); HD (1);


SB (1)

FMC

Rabelo et al. (1993)

Criterion B

Criterion A

Author/publication year

Table 3 Classification and coding of the studied papers

JC1; RF?

JC? ; RF?

JC? ; RF?

JC1; RF?

i, iii

i, ii

i
i

JC? ; RF?

i, ii

Criterion D

JC? ; RF1

JC? ; RF?

Not available

JC? ; RF?

JC? ; RF?

JC? ; RF1

JC? ; RF?

JC? ; RF?

JC? ; RF1

JC? ; RF?

Not available

Not available

JC? ; RF?

JC1; RF1

Criterion C

Hybrid (PN_GA_dispatching
rules)

GA

hybrid (GA_heuristic) and


hybrid (SA_ heuristic)

Multi (Cmax; route length;


backtrackings)
Multi (costs)

GA

WIP
Multi (Cmax; T idle)

GA

Hybrid (PN_GA)

Cmax
Multi (Cmax; cost)

GA

GA

Cmax

Hybrid (GA_DDP)

Hybrid (PN_GA)
GA

Multi (Cmax; Dd; Umax)


Cmax

Multi (costs)
Cmax

Hybrid (GA_dispatching rules_


RN)

GA

Hybrid(Fuzzy_GA)

Hybrid(PN_GA)

Cmax

Due date
Cmax

Hybrid (PN_simulation_GA)
Hybrid (GA_simulation)

Hybrid (GA_heuristic)

Cmax
Multi (Cmax; F medium)
Cmax

Hybrid (RN_GA)

Multi (F medium; Tmax)

WIP

Criterion F

Criterion E

M. Godinho Filho et al.

Not available

MCFMS

MMFMS

Not available

Jerald et al. (2006)

Reddy and Rao (2006)

Kim et al. (2007a)

Not available

Chan et al. (2005)

Chan et al. (2006b)

MCFMS

Jerald et al. (2005)

MCFMS

FMC

Sankar et al. (2004b)

Not available

Not available

Nearchou (2004)

Sankar et al. (2005)

MCFMS

Sankar et al. (2004a)

Chan et al. (2006a)

Not available

FMC

Chiang and Fu (2004)

Not available

Honghong and Zhiming


(2003)

Gang and Wu (2004)

Criterion A

Author/publication year

Table 3 continued

Lim: M(5)

Lim: M(6); HD(2)

Lim: M(16); HD(11)

Lim: M(30)

Lim: M(15)

Lim: M(3)

Lim: M(30)

Lim: M(15)

Lim: M(3)

Lim: M(16);HD(2)

Lim: M(33)

Lim: M(3)

Lim: M(16); HD(11)

Lim:M(15); HD(11)

Lim: M(8); HD(11)

Lim: M(4); HD(1)

Lim:M(20)

Lim: M(10)

Lim:M(5)

Lim: M(16); HD (2)

Lim: M(3); B(1)

Lim: M(20)

Lim: M(33); M(6)

Criterion B

JC?; RF?

JC?; RF1

JC?; RF1

JC?; RF?

i;ii

i, ii

i
i

JC?; RF1

JC?; RF?

JC?; RF?

i; ii, iv
i

JC?; RF1

JC?; RF1

JC?; RF1

Criterion D

JC?;RF1

JC?; RF1

JC?; RF?

JC?; RF?

Criterion C

Multi (cost; T idle)


Multi (Cmax; F medium; T)
Multi (Cmax, F)

Cmax

Multi (U; T)
Cmax

Cmax

Multi (costs; T idle)

Cmax

Multi (cost; T idle)


Cmax

GA

Hybrid (GA_heuristic)

GA

GA

GA

GA

GA

GA, SA, MA, PSO

Hybrid (GA_Simulation)

Hybrid (GA_SA)

GA

Hybrid (PN_GA)

Hybrid (GA_heuristic)

GA

Multi (cost; Cmax)


Due date
Cmax

Criterion F

Criterion E

Using Genetic Algorithms

123

Criterion A

FMC

Not available

Not available

Not available

Not available

FMC

Not available

Author/publication year

Turkcan et al. (2007)

Kim et al. (2007b)

Chan et al. (2008)

Hsu et al. (2008)

Choudhury et al. (2009)

Taghavifard et al.
(2009)

MohammadPour et al.
(2010)

Table 3 continued

123

Lim: M(4)

Lim: M(2); HD(1)

Lim: M(4)

Lim: M(6)

Lim:M(10)

Lim:M(33)

Lim:M(3)

Lim: M(10)

Lim:M(2);HD(3)

Criterion B

JC?; RF?

JC?; RF?

JC?; RF?

i, iii

JC?; RF?

JC?; RF1

i, iii

Criterion D

JC?; RF?

JC?; RF?

Criterion C

Hybrid (PN_GA)

Cmax

GA and SA
Hybrid (dispatching rules_GA)

Multi (costs; U)
Cmax

GA

GA

Multi(Cmax; T)

WIP

GA

Hybrid (GA_heuristic)

Criterion F

Multi (cost, T)
Cmax

Criterion E

M. Godinho Filho et al.

Using Genetic Algorithms

Rossi and Dini (2000) also used a pure approach but focused on solving the
problem of dynamic scheduling in FMS. The proposed technique is based on
optimizing the genetic complexity and reducing the time required to generate a new
schedule. This Real-Time Genetic Algorithm (REGAL) is applied to an FMS with
16 machines and 1 material handling device to obtain the sequencing of parts
resulting in the lowest makespan. The results are compared with other techniques,
such as the rule-oriented algorithm (ROA) and the Generic Genetic Algorithm
(GGA), and the proposed method produced the best result.
Hsu et al. (2002) propose a pure GA approach to solve the problem of cyclical
scheduling in FMS environments. The authors validate the algorithm by applying it
to a flexible manufacturing system with 6 machines to achieve the optimal
production speed while minimizing the work in process (WIP). The algorithm
proposed by Hsu et al. (2002) was validated on 5 FMS cyclic scheduling problem
tests.
Chan et al. (2005), Chan et al. (2006a) and Chan et al. (2006b) presented a work
sequence using a GA that incorporates the Dominant Genes technique. In Chan et al.
(2005), a Genetic Algorithm with Dominant Genes (GADG) is proposed to deal
with FMS problems with alternative production routing in two FMS environments:
one with 3 machines and one with 33 machines. The objective was to minimize the
makespan. The results were compared with other techniques such as Ant Colony
Optimization (ACO) and Petri Nets (PNs). The proposed algorithm (GADG)
exhibited better performance in the analyzed scenarios. Chan et al. (2006a) use the
same approach (GADG) to deal with distributed FMS scheduling problems subject
to machine maintenance constraints. In the two scenarios studied (one with one
factory and another with two factories), the results of the GADG approach were
compared with results from other approaches such as PNs, ACO and a simple
Genetic Algorithm. The proposed approach presented the lowest makespan in the
two considered environments. In Chan et al. (2006a), the same approach is applied
to an environment with four plants containing three machines each and again aims
to minimize the makespan. The authors highlight that the idea of dominant genes
is to identify and record the critical genes in the chromosome to enhance the
performance of the genetic search. The GADG results again exhibit better
performance than other approaches such as SGA, Petri Nets and ACO.
Kim et al. (2007b) use the ASMEA (symbiotic evolutionary asymmetric
multileveled algorithm) to solve the scheduling problem in an FMS. The algorithm
was applied to minimize the makespan in an FMS with 10 machines, and four types
of flexibility were considered: machines, tools, process and sequencing flexibilities.
The results were compared with other approaches: HEA (hierarchical evolutionary
algorithm), TEA (traditional evolutionary algorithm) and SEA (Symbiotic Evolutionary Algorithm). The solution quality and the speed of convergence demonstrate
the superiority of the proposed algorithm (ASMEA).
The most recent work in this category is the work of Hsu et al. (2008), who, like
Hsu et al. (2002) address the problem of sequencing cyclic tasks (the cyclic
scheduling problem) and aim to reduce the Work In Process (WIP). After modeling
the problem using Petri Nets, the authors use a GA to obtaining the sequence of
tasks for a flexible manufacturing cell. The authors conclude that in 75 % of cases,

123

M. Godinho Filho et al.

the performance achieved by the proposed algorithm was equivalent to the best
heuristics in the literature.
Among the reviewed papers that only address the sequencing problem with only
one measure of performance are some that used a hybrid approach. Unlike the work
presented above, the following works use Genetic Algorithms combined with
another approach. Holsapple et al. (1993), for example, propose a hybrid approach
to solve the problem of static scheduling of an FMS with 3 machines. In addition to
a GA, the authors use the problem-knowledge processing (PPK) heuristic. The
benefits of this hybrid approach are seen in several tests aimed at reducing the
makespan. The proposed method exhibited good performance and also demonstrated the importance of addressing the more complex problem of simultaneously
satisfying multiple scheduling criteria.
Rabelo et al. (1994) use a hybrid architecture that utilizes Neural Networks,
Simulation, Genetic Algorithms and induction mechanisms to solve the problem of
sequencing operations in an FMS. The results (in terms of WIP) of the proposed
approach are compared against those from using a single dispatching rule at all
times (a technique currently used in industry).
Three other studies (Chiu and Fu (1997), Reyes et al. (2001), and Gang and Wu
(2004)) use a hybrid approach combining Genetic Algorithms with Petri Nets. Chiu
and Fu (1997) first develop a Petri-net model of an FMS composed of two submodels: a transportation model and a process-flow model. An embedded GA search
method is applied on the basis of the full PNs model. The studied environment
consists of 3 machines and 3 material handling devices. The evaluated measure of
performance is the makespan. The authors conclude that the makespan obtained
with this hybrid approach demonstrate that it is a good alternative to other
techniques within this class of problems. Reyes et al. (2001) propose a hybrid FMS
scheduling methodology that combines Petri Nets and GA techniques. Experiments
on a three-machine FMS were presented to illustrate the degree of effectiveness of
the proposed scheme. The operation sequencing seeks to provide the lowest
makespan. The performance of the proposed method compared favorably with other
concurrent work integrating PNs and heuristic search techniques. The authors
suggest that further research into incorporating GA operators (crossover and
mutation) into the method is required.
Gang and Wu (2004) use such a hybrid approach (GA and PN) to solve
sequencing problems in an FMS with three machines. To demonstrate the
effectiveness of the proposed method, the authors present several examples in
which they consider different buffer allocations and the corresponding makespans
obtained with such allocations.
Erkmen et al. (1997) also use a hybrid approach, but they combine Fuzzy Logic
with Genetic Algorithms to solve FMS sequencing problems. Aiming to optimize
due dates, the method is applied in an environment containing multiple cells with 5
numerical control machines and a material handling device. They consider a number
of factors including machine availability and process time.
The FMS sequencing problem is also addressed by Jawahar et al. (1998a), who
approach a general shop scheduling and rescheduling problem with alternative route
choices in an FMS environment. Their GA-based heuristic uses priority dispatching

123

Using Genetic Algorithms

rules and is applied in a 5-machine scenario and in a three-machine scenario, both


with a material handling system. The comparison in terms of makespan and
computational time indicates that the GA-based heuristic search process is suitable
for FMS scheduling problems.
Also using makespan as the only performance measure, Yang (2001) proposes
the use of a DDP-GA hybrid approach that uses Genetic Algorithms and discrete
dynamic programming to solve sequencing problems in flexible manufacturing
systems. The proposed approach is applied in an FMS with 5 machines to generate a
sequence that results in the smallest makespan. The results from their research
support the strategy of combining traditional algorithmic procedures with heuristics
or adaptive search techniques to develop hybrid FMS scheduling approaches.
Chiang and Fu (2004), in turn, address the sequencing problem in FMS and
propose two hybrid approaches based on the critical ratio rule. The problem is
divided into two sub-problems: job sequencing, for which the authors propose the
ECR (enhanced critical ratio) rule, and dispatching, for which the MCR (minimum
critical ratio) rule is proposed. The objective concerned is maximizing the meetdue-date rate. The results show that the rules perform well, and the simplicity of the
rule ensures good user acceptance according to the authors. Finally, to improve the
performance of the ECR solution, the authors apply a Genetic Algorithm called
the TCGA (time-constrained Genetic Algorithm), which was able to reduce the
computational time required by the proposed approach.
The hybrid method proposed by Nearchou (2004) is used to solve sequencing
problems in flow shop environments. This approach involves using Genetic
Algorithms and Simulated Annealing and is applied to minimize the makespan in
environments with 5, 10 and 20 machines. The solutions are of comparable quality
to those produced by the best algorithms available in the heuristics literature. The
authors conclude that hybridizing Simulated Annealing Algorithms with features
from both global and local search techniques results in a robust optimization tool
capable of producing high-quality solutions to flow shop scheduling problems.
The final reviewed study that uses a hybrid approach to solving the sequencing
problem considering only one measure of performance is the work of MohammadPour et al. (2010). The authors propose a hybrid method that uses Petri Nets,
Genetic Algorithms and Tabu Search to solve the addressed problem. The algorithm
is applied to minimize the makespan of a flexible manufacturing system with 4
machines. The performance of the proposed algorithm is compared with two other
approaches (TPN and PN-GA-GA) and exhibits a higher measure of performance
than the others. However, MohammadPour et al. (2010) observed that the proposed
algorithm requires a longer computational time than the other two approaches
because of the use of Tabu Search as the local search technique.
4.1.2 Papers that used more than one measure of performance (multi-criteria)
Among the studies that use more than one performance measure, the following
works use only Genetic Algorithms: Keung et al. (2001), Sankar et al. (2003),
Sankar et al. (2004b), Sankar et al. (2005), Honghong and Zhiming (2003), Kim
et al. (2007a) and Chan et al. (2008).

123

M. Godinho Filho et al.

Keung et al. (2001) use a genetic algorithm to solve sequencing problems from
the perspective of ETPSP (earliness/tardiness production scheduling and planning).
This method is applied to a flexible manufacturing cell with two numerical control
machines and one material handling device. The results using the chosen objective
function (minimizing the penalty cost) were compared with those from other two
methods in the literature. The authors conclude that using the GA provides a lower
cost penalty than the other two mentioned approaches.
Sankar et al. (2003) propose a mechanism to perform the scheduling for an FMS
based on Genetic Algorithms with two different GA coding schemes, namely, Pheno
style codification and Binary codification. The scheme is applied on one FMS with
16 machines, 9 robots and 2 AGVs, and the makespan and idle time are used as the
performance measures. The results from this combined objective function are
compared with results obtained by some dispatching rules (EDD, HP, SPT, among
others), and the proposed approach (GA) is shown to be superior.
Honghong and Zhiming (2003) propose an FMS rescheduling system using an
Adaptive Genetic Algorithm (AGA) and consider an environment with realistic
interruptions and the constraint of a time-restricted response to rescheduling. The
proposed rescheduling system is based on records from a dynamic database (DDB)
and is tested on two scenarios: one with 33 machines and another with 6 machines.
In these two tests, the performance measures are the weighted quadratic tardiness
and makespan, respectively. The results obtained by applying the proposed AGA
are compared with those obtained using a Simple Genetic Algorithm (SGA). The
proposed Adaptive Genetic Algorithm exhibits better scheduling performance in
less computational time than the SGA approach.
Sankar et al. (2004a) propose using the Niched Pareto Genetic Algorithm
(NPGA) to generate schedules for an FMS with multiple cells, a total of 16 CNC
machines and 2 material handling devices, with the goal of obtaining an operation
sequence that enables both lower cost and fewer idle machines. The results are
compared with other approaches such as Tabu Search (TS), Simulated Annealing
(SA) and dispatching rules. The results show that the proposed approach has a better
performance than the other analyzed techniques.
Sankar et al. (2005) use an algorithm called the multi-evolutionary algorithm
(MOEA) that generates a near-optimal schedule by simultaneously achieving two
contradictory objectives of a flexible manufacturing system. The approach is
applied to a flexible manufacturing system with 16 machines, two material handling
devices and two robots, with the objectives of maximizing machine utilization and
minimizing tardiness. The authors conclude that the heterogeneous population
concept introduced in the work allows MOEA to achieve genetic diversity more
effectively than the traditional methods.
Kim et al. (2007a) propose a multi-objective Genetic Algorithm (moGA) to solve
a multistage job processing schedule problem in an FMS environment. The
feasibility of the method is demonstrated through experimental results and a
comparison with other approaches (shortest average processing time- SAP; discrete
dynamic programming-DDP; GA-based discrete dynamic programming GADDP;
and GA without local search). In the analyzed environment, which contains 5
machines, a multi-objective function compares the makespan and total flow time of

123

Using Genetic Algorithms

solutions for all of the approaches. The conclusion is that the proposed moGA is
more efficient and more flexible in locating the Pareto solution than the competing
algorithms.
Finally Chan et al. (2008) propose using a Genetic Algorithm with Dominant
Genes (GADG) to solve sequencing problems in flexible manufacturing systems
with alternative production routing. The authors test the performance of the
proposed approach in three situations. In Example 1, the GADG is applied to an
FMS with three machines with the goal of finding the sequence with the smallest
makespan. The results are compared with those of Petri Nets and ant colony
optimization. In the second example, the environment has 33 machines, and the
objective is a sequence that achieves the lowest tardiness. In this case, the examples
are compared with Lagrangian Relaxation, a Simple Genetic Algorithm (SGA) and
an Adaptive Genetic Algorithm (AGA). Finally, for an FMS with 10 machines, the
GADG performance is compared with that of an SGA to obtain a schedule with the
lowest makespan. The authors conclude that the results obtained with the algorithm
in each of the three examples are superior to those of the other approaches.
Two works propose more than one approach to solving the problem of multicriteria sequencing. Although they are still pure GA approaches, Jerald et al. (2005)
and Choudhury et al. (2009) propose different approaches and compare them in
efforts to evaluate the performance of the proposed techniques. Jerald et al. (2005)
develop optimization procedures on the basis of four different approaches: Genetic
Algorithms, Simulated Annealing, memetic algorithms and particle swarm optimization. These approaches are implemented to solve the schedule optimization
problem of three FMS (with 8, 15 and 16 machines), and the utilized measure of
performance combines the total penalty cost and the machine idleness. In a
comparison of the obtained results, particle swarm optimization proved to be
superior to the other examined approaches and achieved the minimum combined
objective function. Choudhury et al. (2009) address the sequencing problem in
Flexible manufacturing systems through the use of Genetic Algorithms and
Simulated Annealing to optimize a multi-criteria objective function. The authors
apply the proposed approaches to an FMS with 4 machines to simultaneously
minimize the penalty costs and maximize the machine utilization. The results show
that GA scores better than SA in dealing with FMS scheduling under constrained
conditions. The GA performs better because the GA is more capable than the SA at
treating the complexity of the problem.
Four papers were found that use hybrid approaches to solve sequencing problems
using multiple performance measures. Fujimoto et al. (1995) propose a hybrid
intelligent approach to a production scheduling problem in Flexible Manufacturing
Systems. The authors evaluate the performance of the approach through two
examples in which performance measures including the makespan and mean flow
time are evaluated. The authors conclude that the proposed approach efficiently
seeks the best combination of dispatching rules to obtain an appropriate production
schedule given specific performance measures.
Rabelo et al. (1993) propose a scheme for scheduling FMSs that integrates
Neural Networks, parallel Monte-Carlo simulation, Genetic Algorithms and
machine learning. To evaluate the method, the authors apply it to sequencing a

123

M. Godinho Filho et al.

flexible manufacturing cell with 5 machines and 1 material handling device. The
mean flow time and maximum tardiness are used as performance measures.
Chung et al. (1998), in turn, propose a systematic Petri Net model of an FMS that
supports the use of an adaptive scheduling method incorporating Genetic Algorithm
search to solve the sequencing problems. The authors apply this technique to an
FMS with 5 machines and analyze 3 performance measures: makespan, due date
and machine utilization. According to the authors, the proposed method not only
generates an efficient schedule but also allows the user to set different priorities
among the jobs.
Saitou et al. (2002) also use Petri Nets with Genetic Algorithms, but they use
them with dispatching rules. The authors present a Genetic Algorithm coupled to a
specific dispatching rule (the shortest imminent operation timeSIO), which is
used to simultaneously find the near-optimal resource allocation and the eventdriven schedule of a Colored Petri Net. This mechanism is formulated as a multiobjective optimization problem that aims to minimize the production costs and the
reconfiguration costs due to changes in the production plan. To evaluate the
methods performance, the authors apply it to three different scenarios containing
from 4 to 10 machines. The results demonstrate the satisfactory performance of the
method as well as its ability to obtain a large number of feasible solutions to
sequencing problems.
4.2 Articles that discuss sequencing problems along with other scheduling
problems
Eight papers were found that use Genetic Algorithms to solve more than one
scheduling problem in flexible manufacturing systems. In addition to sequencing
problems, these works also addressed AGV routing, allocation and loading. In the
next section, this class of work is summarized and subdivided into single-criterion
and multi-criteria techniques.
4.2.1 Papers that used only one measure of performance (single-criterion)
Three papers address problems beyond the scheduling problem using only one
performance measure. All of these works use makespan as the performance
measure.
Ulusoy et al. (1997) apply Genetic Algorithms to the problem of simultaneous
scheduling of machines and a number of identical automated guided vehicles
(AGVs) in a flexible manufacturing system (FMS). The authors analyzed the
performance of the algorithm by applying it to three different layouts, all with six
workstations and two AGVs. The resulting performance measures (makespan) were
satisfactory, allowing the authors to conclude that the proposed algorithm is well
suited to the proposed type of problem.
Sankar et al. (2004b) deal with the simultaneous scheduling of incoming jobs,
machines and AGVs. The authors make use of a hybrid method that includes
optimization (using Genetic Algorithms) and computer simulation. The method was
applied to an FMS with four machines and one AGV. The results are evaluated by

123

Using Genetic Algorithms

comparing the obtained makespan with the results obtained by another algorithm in
the literature (Kangaroo Algorithm). The results demonstrate that the proposed
procedure performs better.
Taghavifard et al. (2009) also use a hybrid approach consisting of dispatching
rules and Genetic Algorithms to simultaneously schedule machines and AGVs. The
authors apply the algorithm to a flexible manufacturing cell consisting of two
machines and one material handling device with the objective of minimizing the
makespan. The authors compare the results with the optimal values found in the
literature and highlight the strong performance of the proposed method in terms of
efficiency and solution quality.
4.2.2 Papers that use more than one measure of performance (multi-criteria)
Keung et al. (2003) propose an intelligent hierarchical control technique for a
flexible manufacturing system. The control model comprises four levels: selection
of the production plan, master planning, sequencing and task control. At the
sequencing level, the goal is to maximize machine utilization and balance the tool
magazine capacity. The authors use Genetic Algorithms at all levels and evaluate
the model using two criteria: earliness and tardiness penalty costs. The authors
conclude that the intelligent hierarchical planning, scheduling and control model
provides a systematic way to effectively allocate resources over different time
horizons.
Jerald et al. (2006) also make use of a pure approach (Adaptive Genetic
Algorithm) to address the problem of simultaneously scheduling parts and AGVs in
an FMS with 16 machines and 11 material handling devices. The authors use a
combination of penalty costs and machine idle time as the performance measure.
The schedule obtained from the Adaptive Genetic Algorithm is compared with that
produced by a Genetic Algorithm, and experimental results have indicated that the
proposed adaptive Genetic Approach is very effective.
Three papers make use of hybrid approaches. Haq et al. (2003) deal with
multilevel scheduling decisions for a flexible manufacturing system. The authors
initially use the Giffler and Thompson (GT) algorithm with six different dispatching
rules for operation sequencing to optimize the makespan. The result of this
optimization is used to optimize a second objective function related to the routing of
AGVs. This second objective function includes both the distance traveled by the
AGVs and the number of backtrackings. The performance of Genetic Algorithms
and Simulated Annealing are compared, and the results show that the hybrid
approach using GA gives superior results.
Reddy and Rao (2006) address simultaneous scheduling of machines and AGVs
in FMSs using a multi-objective hybrid GA to minimize the makespan, mean flow
time and mean tardiness objectives. The hybrid approach is a combination of
Genetic Algorithms and heuristics and is applied to an FMS with six machines and
two AGVs. The results are analyzed, and the authors conclude that the proposed
algorithm presents many diverse solutions.
Turkcan et al. (2007) propose using a genetic algorithm called the PSGA
(Problem Space Genetic Algorithm) to simultaneously solve the problems of

123

M. Godinho Filho et al.

loading, sequencing and tool management in a flexible manufacturing system. The


authors apply the method to an FMS with two CNC machines, two robots, and one
conveyor and aim to minimize both the manufacturing costs and the total weighted
tardiness. The performance of the proposed algorithm was compared with that of a
sequential algorithm (PI) that sequentially solves the loading problem and the
sequencing problem. Finally, the authors point out the superiority of the proposed
algorithm as well as the significant interaction between the tool management and
sequencing decisions that are considered in the proposed method.

5 Analysis
In this section, a general analysis based on the proposed classification system is
performed on the literature review. In Table 4, we present the frequencies of
occurrence of the possible criteria values. It is important to note that Table 4
considers the possible criteria values presented in the Sect. 3. However, for better
visualization some criteria present the combination or grouping of some criteria
values.

Table 4 Number of articles by each classification criterion


Criteria values

Number
of papers

Frequency
(%)

Criterion A

Criteria values

Number
of papers

Frequency
(%)

Criterion B

SFM

0.0

JC1; RF1

2.5

FMC

11

27.5

JC1; RF?

5.0

MMFMS

7.5

JC?; RF1

12

30.0

MCFMS

17.5

JC?; RF?

22

55.0

Not available

Total

40

Not available

19

47.5

Total

40

100.0

Machines (M)

18

45.0

Machines(M) e buffer (B)

2.5

Machines (M) e material


handling Devices (HD)

18

45.0

M. B e HD

Not available

Total

40

100.0

Mono-criterion

22

55.0

Pure

20

50.0

Multi-criteria

18

45.0

Hybrid

20

50.0

Total

40

100.0

Criterion B

Criterion D
Sequencing

32

80.0

Sequencing and AGV


Routing

10.0

Sequencing and loading

7.5

5.0

Sequencing, loading
and Routing

2.5

2.5

Total

40

Criterion E

123

7.5
100.0

100.0

Criterion F

Total

100.0

Using Genetic Algorithms

Regarding Criterion B (resources and their constraints), 18 papers contain the


number of machines and material handling devices that exist in the FMS
environment in their constraints. One paper does not indicate the type of resources
or the capacity constraints. It is also noted that only 2 reviewed papers consider the
capacity constraints (machines, buffers and material handling device) in a
simultaneously way. This analysis shows that there is an opportunity for future
research to consider more complex scenarios that contain more constraints (such as
those found in machines, storage buffers and material handling systems). When
these constraints are considered, the studied problem becomes closer to reality,
allowing the generation of practical scenarios for testing algorithms such as Genetic
Algorithms.
Analyzing Criterion C (job complexity and routing flexibility), it is notable that
most of the reviewed papers treated the most complex scenario type (JC? ; JF?).
The environment JC? ; RF1 was studied in 12 reviewed papers. It also becomes
evident that there exists a relation between the complexity of the environment and
the approach used to solve the problem. Meta-heuristics such as GAs are, according
to the observed results, a viable solution for higher complexity environments.
Figure 1 compares Criteria D and E. Most of the papers (32) focus only on the
scheduling of operations or jobs. Of this total, 19 papers use only one objective
function. Of these 19 papers, 14 use the makespan as the performance measure.
Analyzing all of the papers studied in the review reveals that the makespan is used
in 24 papers in total, either as the sole component of the objective function or in
combination with other performance measures.
The FMS environment is an important component of the complexity level of the
studied scheduling problem, and Fig. 2 presents the frequency of each type of FMS
(Criterion A) in this literature review. Figure 2 reveals that the FMC environment is
studied most often. A simpler environment, the SFM, was not treated by any
reviewed papers. However, this analysis is somehow impaired, as 19 researched
papers did not specify the target environment.
Table 4 indicates that the two approaches (pure GA or hybrid) are treated in an
equal number of papers. However, in several papers in which the pure GA approach

Fig. 1 Relation between Criteria D and E

123

M. Godinho Filho et al.

Fig. 2 Types of productive environment studied

Fig. 3 Approaches used in the reviewed papers

was used, the authors considered the importance of analyzing combinations of GAs
with other approaches presented in the literature. Figure 3 presents the distribution
of the approaches found in the papers that use hybrid methods to solve the proposed
problem. Of the 20 papers that use the hybrid approach, 10 use simulations or Petri
Nets combined with GAs, 5 use other heuristics, and 3 use dispatching rules. Neural
Networks and Simulated Annealing were each used in 2 papers, and DDP and Fuzzy
Logic were each used in one paper.

6 Final remarks
This paper presents a literature review on the application of Genetic Algorithms to
solving FMS scheduling problems. The survey utilizes a classification system with
six criteria: type of FMS, resources and constraints, job characteristics, scheduling
problem approached, measure of performance and solution approach. All papers
found in the literature review were classified according to these criteria and
presented according to their characteristics.
Some aspects of the review, classification and analysis can be highlighted:

The solution approaches employed exhibit high diversity. There are several
combinations of GAs and other techniques among the hybrid techniques. A

123

Using Genetic Algorithms

combination of GAs and Petri Nets is a representative approach. Several authors,


such as Chiu and Fu (1997), Gang and Wu (2004), Chung et al. (1998), Reyes
et al. (2001), Rabelo et al. (1994), Saitou et al. (2002) and MohammadPour
et al.(2010) use Petri Networks to construct the problem and increase the
efficiency of the GA. However, a few authors, like MohammadPour et al.
(2010), use mechanisms for local search. There are opportunities for future
research in this area.
Although all of the researched papers studied FMS environments, only a few
authors explicitly state all of the characteristics of the environment they
approached in their research: Jawahar et al. (1998a), Chiu and Fu (1997), Gang
and Wu (2004), Rabelo et al. (1993), Chung et al. (1998), Rossi and Dini (2000),
Jawahar et al. (1998a), Keung et al. (2001), Erkmen et al. (1997), Taghavifard
et al. (2009), Turkcan et al. (2007), Jerald et al. (2005,2006), Haq et al. (2003),
Ulusoy et al. (1997), Sankar et al. (2003), Sankar et al. (2004a, 2004b, 2005),
Reddy and Rao (2006), Hsu et al. (2008), Keung et al. (2003). Because of this
lack of information, classification by Criterion A is a difficult task. We
understand that the studied scenarios allow the reader a greater comprehension of
the complexity of the problem and the application of the developed techniques.
The quantitative analysis demonstrated that most of the studies use the same
measure of performance (the makespan). It appears that there exist opportunities
to use further measures, either alone or in combinations, as is done by Sankar
et al. (2003), Sankar et al. (2004b, 2005), Rabelo et al. (1993), Jerald et al. (2005,
2006), Keung et al. (2001), Choudhury et al. (2009), Turkcan et al. (2007), Haq
et al. (2003) and Keung et al. (2003). We also considered that, in practical
applications, the simultaneous use of manufacturing costs and makespan as
performance measures could provide more information to the decision process.
Although 100 % of the reviewed papers approached scheduling problems, only a
few of them present a Gantt Chart with the final solution, which could be helpful
when comparing different studies and aiming at providing a more practical
solution for managers in the shop floor.
Most of the papers use the following: (1) problems already approached by
another, already published paper or (2) analysis that compares, for the problem
analyzed, the results obtained with the proposed approach and approaches that
have been studied in literature.

Finally, it can be stated that FMS scheduling continues to represent a great


research opportunity for application of GAs and other Artificial Intelligence
approaches, specially nowadays where FMS became widespread in a lot of
manufacturing environments.

References
Aleksander I, Morton H (1990) Neural computing. Champman & Hall, London
Arenales MN, Armentano VA, Morabito Neto R, Yanasse HH (2007) Pesquisa operacional: modelagem e
algoritmos. Elsevier, Rio de Janeiro

123

M. Godinho Filho et al.


Balogun OO, Popplewell K (1999) Towards the integration of flexible manufacturing system scheduling.
Int J Prod Res 37(15):33993428
Byrkett DL, Ozden MH, Patton JM (1988) Integrating flexible manufacturing systems with traditional
manufacturing, planning and control. Journal of Production and Inventory Management 29:1521
Carvalho ACPLF, Braga A, Ludermir T (2003) Computacao Evolutiva. In: Rezende Solange (ed)
Sistemas inteligentesfundamentos e aplicacoes. Editora Manole, Barueri, pp 135
Chan FTS, Chan HK (2004) A comprehensive survey and future trend of simulation study on FMS
scheduling. J Intell Manuf 15:87102
Chan FTS, Chung SH, Chan PLY (2005) An introduction of dominant genes in genetic algorithm for
scheduling of FMS. International Symposium on Intelligent, Control, pp 14291434
Chan FTS, Chung SH, Chan LY, Finke G, Tiwari MK (2006a) Solving distributed FMS scheduling
problems subject to maintenance: genetic algorithm approach. Robotic and Computer-Integrated
Manufacturing 22:493504
Chan FTS, Chung SH, Chan PLY (2006b) Application of Genetic Algorithms with dominant genes in a
distributed scheduling problem in flexible manufacturing systems. Int J Prod Res 44(3):523543
Chan FTS, Chung SH, Chan LY (2008) An introduction of dominant genes in genetic algorithm for FMS.
Int J Prod Res 46(16):43694389
Chiang TC, Fu LC (2004) Solving the FMS scheduling problem by critical ratio- based heuristics and the
genetic algorithm. Int Conf Robot Autom 3131336
Chiu YF, Fu LC (1997). A GA embedded dynamic search algorithm over a Petri Net model an FMS
scheduling. Proceedings of the international conference on robotics and automation 513518
Choudhury BB, Biswall BB, Mishra D, Mahapatra RN (2009) Appropriate evolutionary algorithm for
scheduling in FMS. World Congr Nat Biologically Inspired Comput 11391144
Chung YY, Fu LC, Lin MW (1998) Petri Net based modeling and GA based scheduling for a flexible
manufacturing system. Proceedings of the IEEE Conference on Decision and Control 4:43464347
Dorigo M, Stutzle T (2003) The Ant Colony Optimization Metaheuristic: Algorithms, Applications, and
Advances. In: Glover F, Kochenberg GA (eds) Handbook of metaheuristics. Kluwer Academic
Publishers, London, pp 251285
Erkmen AM, Erbudak M, Anlagan O, Unver O (1997) Genetically tuned fuzzy scheduling for flexible
manufacturing systems. Int Conf Robot Autom 951956
Fernandes FCF, Godinho Filho M (2010) Planejamento e controle da producao: dos fundamentos ao
essencial, 1st edn. Editora Atlas, Sao Paulo
Fujimoto H, Yasuda K, Tanigawa Y, Iwahashi K (1995) Applications of genetic algorithm and simulation
to dispatching rule-based FMS scheduling. Int Conf Robot Autom 190195
Gang X, Wu Z (2004) Deadlock-free scheduling strategy for automated production cell. IEEE Trans Syst
Man Cybern Part A Syst Hum 34(1):113122
Goldberg DE (1989) Genetic Algorithms in search, optimisation, and machine learning. Addison-Wesley,
California
Haq AN, Karthikeyan T, Dinesh M (2003) Scheduling decisions in FMS using a heuristic approach. Int J
Adv Manuf Technol 22:374379
Haupt RL, Haupt SE (1998) Pratical genetic algorithm. John Wiley & Sons, New York
Holland J (1975) Adaptation in natural and artificial systems. University of Michigan Press, Ann Arbor
Holsapple CW, Jacob VS, Pakath R, Zaveri JS (1993) A Genetics-based hybrid scheduler for generating
static schedules in flexible manufacturing contexts. IEEE Transactions on Systems, Man, and
Cybernetics 23(4):953972
Honghong Y, Zhiming W (2003) The application of Adaptive Genetic Algorithms in FMS dynamic
rescheduling. Int J Comput Integr Manuf 6(16):382397
Hsu T, Dupas R, Goncalves G (2002) A genetic algorithm to solving the problem of flexible
manufacturing system cyclic scheduling. Proceedings of the IEEE International Conference on
Systems, Man, and Cybernetics 3:438443
Hsu T, Korbaa O, Dupas R, Goncalves G (2008) Cyclic scheduling for FMS: modeling and evolutionary
solving approach. Eur J Oper Res 191:464484
Jawahar N, Aravindan P, Ponnambalam SG (1998a) A genetic algorithm for scheduling flexible
manufacturing Systems. Int J Adv Manuf Technol 14:588607
Jawahar N, Aravindan P, Ponnambalam SG, Raghavendra LN (1998b) Knowledge-based worcell
attribute oriented dynamic schedulers for flexible manufacturing systems. Int J Adv Manuf Technol
14:514538

123

Using Genetic Algorithms


Jerald J, Asokan P, Prabaharan G, Saravanan R (2005) Scheduling optimisation of flexible manufacturing
systems using particle swarm optimisation algorithm. Int J Adv Manuf Technol 25:964971
Jerald J, Asokan P, Saravan R, Delphin Carolina Rani A (2006) Simultaneous scheduling of parts and
automated guided vehicles in an FMS environment using adaptative genetic algorithm. Int J Adv
Manuf Technol 29:584589
Kaltwasser J, Hercht A, Lang R (1986) Hierarchical control of flexible manufacturing systems. IFAC
Information Control Problems in Manufacturing Technology, Suzdal, USSR, pp 3744
Kamboj MS, Sengupta J (2009) Comparative analysis of Simulated Annealing and tabu search channel
allocation algorithms. International Journal of Computer Theory an Engineering 1(5):17938201
Keung KW, Ip WH, Chan CY (2001) An enhanced MPS solution for FMS using GAs. Integrated
Manufacturing Systems 12(5):351359
Keung KW, Ip WH, Yuen D (2003) An intelligent hierarchical workstation control model for FMS.
Journal of Material Processing Technology 139:134139
Kim KW, Lee D, Jeong IJ (2007a) Multi-objective genetic algorithm for multistage-based job processing
schedules in fms environment. Int Conf Control Autom 17051709
Kim YK, Kim JY, Shin KS (2007b) An asymmetric multileveled symbiotic evolutionary algorithm for
integrated FMS scheduling. J Intell Manuf 18(6):631645
Maccarthy BL, Liu J (1993a) A new classification scheme for flexible manufacturing systems. Int J Prod
Res 31(2):299309
Maccarthy BL, Liu J (1993b) Addressing the gap in scheduling researcha review of optimization and
heuristic methods in production scheduling. Int J Prod Res 31(1):5979
Maccarthy BL, Liu J (1996) The classification of FMS scheduling problems. Int J Prod Res
34(3):647656
Mitchell M (1998) An introduction to genetic algorithms. MIT Press, Cambridge
MohammadPour T, Yadollahi M, Haghighat AT (2010) HPA-PN: a new algorithm for scheduling FMS
using combinational genetic algorithm and Timed Petri Net. Second Int Conf Comput Modeling
Simul 1518
Morton TE, Pentico DW (1993) Heuristic scheduling systems. John Wiley and Sons, New York
Moscato P, Cotta C (2003) A gentle introduction to memetic algorithms. In: Glover F, Kochenberger G
(eds) Handbook of metaheuristics. Kluwer Academic Publishers, Boston, pp 105144
Nearchou AC (2004) A novel metaheuristic approach for the flow shop scheduling problem. Eng Appl
Artif Intell 17:289300
O Keefe RM, Kasirajan T (1992) Interaction between dispatching and next station selections rules in a
dedicated flexible manufacturing system. Int J Prod Res 30(8):17531772
Poli R, Kennedy J, Blackwell T (2007) Particle swarm optimizationan overview. Swarm Intelligence
1:3357
Ponnambalam SG, Aravindan P, Sreenivasa Rao P (2001) Comparative evaluation of Genetic Algorithms
for job-shop scheduling. Production Planning & Control 12(6):560574
Rabelo L, Yih Y, Jones A, Tsai J (1993) Intelligent scheduling for flexible manufacturing systems.
Proceedings of International Conference on Robotics and Automation 3:810815
Rabelo LC, Jones A, Yih Y (1994) Development of a real-time learning scheduler using reinforcement
learning concepts. International Symposium on Intelligent, Control, pp 1618
Reddy BSP, Rao CSP (2006) A hybrid multi-objective GA for simultaneous scheduling of machines and
AGVs in FMS. Int J Adv Manuf Technol 31:602613
Reyes A, Yu H, Lloyd S (2001) An evolutionary hybrid scheduler based in Petri Net structures for FMS
scheduling. International Conference on Systems, Man, and Cybernetics 4:25162521
Rossi A, Dini G (2000) Dynamic scheduling of FMS using a real-time genetic algorithm. Int J Prod Res
38(1):120
Saitou K, Malpathak S, Qvam H (2002) Robust design of flexible manufacturing systems using, colored
Petri Net and genetic algorithm. J Intell Manuf 13:339351
Sankar SS, Ponnanbalam SG, Rajendran C (2003) A multi objective genetic algorithm for scheduling a
flexible manufacturing system. International Journal of Manufacturing Technology. 22:229236
Sankar SS, Ponnambalam SG, Rathinavel V, Gurumarimuthu M (2004a) A pareto based multi-objective
genetic algorithm for scheduling of FMS. Conference on Cybernetics and Intelligent Systems
699704
Sankar SS, Ponnambalam SG, Rajkumar R, Gurumarimuthu M (2004b) An intelligent integrated
scheduling model for flexible manufacturing system. Conference on Robotics, Automation and
Mechatronics, pp 10951100

123

M. Godinho Filho et al.


Sankar S, Ponnamabalam SG, Gurumarimuthu M (2005) Scheduling flexible manufacturing systems
using parallelization of multi-objective evolutionary algorithms. Int J Adv Manuf Technol
30:279285
Sipper D, Bulfin RL Jr (1997) Production: planning, control and integration. McGraw- Hill, New York
Taghavifard MT, Heydar M, Mousavi SS (2009) A genetic algorithm for scheduling flexible
manufacturing cells. Journal of Applied Sciences 9(1):97104
Turkcan A, Akturk MS, Storer RH (2007) Due date and cost-based FMS loading, scheduling and tool
management. Int J Prod Res 45(5):11831213
(1997) A genetic algorithm approach to the simultaneous scheduling of
Ulusoy G, Serifoglu FS, Bilge U
machines and automated guided vehicles. Comput Oper Res 24(4):335351
Yang JB (2001) GA-based discrete dynamic programming approach for scheduling in FMS environments. IEEE Trans Syst Man Cybern B Cybern 31(5):824835
Zadeh LA (1965) Fuzzy sets. Inf Control 8:338353
Zhou H, Feng Y, Han L (2001) The hybrid heuristic genetic algorithm for job shop scheduling. Comput
Ind Eng 40:191200

Author Biographies
Moacir Godinho Filho had done his postdoctoral research at North Carolina State University at Raleigh
(USA), postdoctoral research at University of Wisconsin-Madison (USA), PhD in Industrial Engineering
at Federal University of Sao Carlos, and M.S. in Industrial Engineering at Federal University of Sao
Carlos. He is a Professor at Federal University of Sao Carlos.
Clarissa Fullin Barco is a M.Sc. student in Industrial Engineering at Federal University of Sao Carlos
and received Bachelors Degree in Industrial Engineering at Federal University of Sao Carlos.
Roberto Fernandes Tavares Neto had done his PhD in Industrial Engineering at Federal University of
Sao Carlos, M.S. in Industrial Engineering at Catholic University of Parana, and Bachelors Degree in
Electrical Engineering at Federal University of Parana. He is a Professor at Federal University of Sao
Carlos.

123

Anda mungkin juga menyukai