J Intell Manuf

DOI 10.1007/s10845-015-1086-9

A two-phase approach to solve the synchronized bin–forklift

scheduling problem
Nizar El Hachemi1,4 · Mohammed Saddoune2,3,5 · Issmail El Hallaoui3,5 ·
Louis-Martin Rousseau3,4

Received: 18 July 2014 / Accepted: 27 April 2015

© Springer Science+Business Media New York 2015

Abstract In this paper, we propose a two-phase approach show that our approach is efficient in terms of both solution
to solve a combined routing and scheduling problem that quality and computational time.
occurs in the textile industry: fabrics are dyed by dye-jets
and transported by forklifts. The objective is to minimize Keywords Scheduling · Column generation ·
the cost of the unproductive activities, i.e., the dye-jet setup Mixed-integer programming · Textile industry
times and the forklift waiting time. The first phase solves
an integer linear program to assign jobs (fabrics) to dye-
jets while minimizing the setup cost; we compare an arc- Introduction
based and a path-based formulation. The second phase uses
a mixed-integer linear program for the dye-jet scheduling and The textile industry is primarily concerned with the design
both the routing and scheduling of forklifts. Experiments are and manufacture of clothing and the distribution and use of
performed on real data provided by a major multinational textiles. This industry occupies an important place in the
company, and larger test problems are randomly generated economy of countries such as China, India, Turkey, and
to assess the algorithm. The tests were conducted using Cplex Morocco. In Morocco, it is one of the leading sectors, and
12.6.0 and a column generation solver. The numerical results its production represents about 15 % of the total value of
the processing industries. Morocco benefits from a quali-
B Nizar El Hachemi fied workforce, competitive wage rates, and proximity to the
nizar.el-hachemi@polymtl.ca European markets, which increases its efficiency and respon-
Mohammed Saddoune siveness. Ready-made garments (shirts, skirts, jackets, etc.)
mohammed.saddoune@polymtl.ca and knitwear production are the most dynamic subsectors.
Issmail El Hallaoui They represent 43 % of the total production and more than
issmail.el-hallaoui@polymtl.ca 70 % of the total Moroccan exports. Morocco manufactures
Louis-Martin Rousseau for well-known brands such as GAP, Banana Republic, Levi
louis-martin.rousseau@cirrelt.net Strauss, and Marks & Spencer. Our industrial partner is a
1 Département de génie industriel, École Mohammadia
major player in this field.
d’Ingénieurs, Université Mohammed V Agdal, Rabat, Textile operations consist of activities such as spinning,
Morocco weaving, dyeing, drying, and printing. The dyeing section
2 Faculté des Sciences Techniques de Mohammedia, Université in our partner’s plant can be broken down into four primary
Hassan II, Casablanca, Morocco processes as follows: (1) dyeing, (2) wringing, (3) drying,
3 Département de mathématiques et génie industriel, École and (4) compacting. Dyeing is the process of coloring fab-
Polytechnique de Montréal, Montreal, Canada rics using computerized jet dyeing equipment. After they are
4 CIRRELT, Université de Montréal, C.P. 6128, Succ. dyed, the fabrics are transported to the PAD machine to be
Centre-ville, Montreal, QC H3C 3J7, Canada wrung. It is also necessary to dewater the fabric, to apply soft-
5 GERAD, C.P. 6128, Succ. Centre-ville, Montreal, ener to it in order to maintain good color, and to overstretch
QC H3C 3A7, Canada it. It is then preshrunk to the required finished width and

123
 J Intell Manuf

dried. Finally, it goes through compactors that carefully fix it dyeing and finishing operations. They consider a flexible
so that it can maintain the required width when in use. Most job shop with sequence-dependent setups, and they mini-
companies have automated jet dyeing tanks. These rotate mize the maximum lateness (completion time − due date).
the fabric automatically through the dyeing process for a Zhou et al. (2010) present a short-term scheduling MILP
specified period depending on the color. There are several model, based on continuous time, for the printing and dye-
operational challenges, especially in the dyeing process. Our ing industry; they solve it using the ILOG-CPLEX solver. For
partner wishes to develop optimized production schedules more details about scheduling dyeing operations, see Cho
for the dyeing and finishing operations. (2004).
Textile firms must optimize their operations because of The problem addressed in this study is similar to many
global competition and customer demand for improved qual- problems encountered in manufacturing contexts such as the
ity, customization, and shorter delivery lead times. To remain navigation of autonomous mobile robots and the use of auto-
competitive, decision makers and managers must optimize matic guided vehicles (AGV). Many research groups have
the entire flow of the materials in the supply chain. investigated the use of autonomous mobile robots in flexi-
Currently, most companies manually generate a color ble manufacturing, giving special attention to the scheduling
sequence on each dye-jet and perform a manual routing and component (see Giralt and Chatila 1987; O’Kane 2000; Felix
scheduling of the forklifts. These solutions are generally inef- et al. 2004). Hu and Gu (2000) propose a navigation algo-
ficient since they generate large (unproductive) setup and rithm that simultaneously locates the robots and updates
waiting times, and more powerful solution methods are there- the landmarks in a manufacturing environment. Current
fore needed. The aim of this study is to present a two-phase research into mobility in the context of flexible manufac-
approach that considers the synchronization of the forklift turing systems involves the use of AGVs for the internal
routing and scheduling and the dye-jet scheduling. and external transport of materials. Some researchers have
The paper is organized as follows. Second section presents investigated the operational issues of dispatching, routing,
a literature review focusing on the scheduling problem and scheduling AGVs. For examples, Hao et al. (1996)
encountered in the dyeing house. Third and fourth sections describe the development of a prototype neural network
present respectively the synchronized bin–forklift schedul- model for the free-ranging AGV route planning problem.
ing problem and our solution approach. The experimental Bruno et al. (2000) pressent two fast and effective heuris-
setting is described in fifth section, where our computational tics which dynamically determine the home positions. Lin
results are also reported. Final section provides concluding et al. (2006) propose an effective evolutionary approach for
remarks. solving a kind of AGV’s problems in which the objective
is to minimize the time required to complete all jobs and
the number of AGVs, simultaneously. Singh et al. (2009)
Literature review present the scheduling of AGVs for efficient and uniform
material distribution from a truck-dock to machining units.
Several models have been developed in the literature to They introduce the notion of zones with comparable demands
address production planning and scheduling in a dyehouse. for AGVs, and they assign one AGV to each zone so that each
Morales et al. (1996) and Maldonado et al. (2000) study AGV can operate independently. Recently, Confessore et al.
models to determine a color sequence that minimizes the (2011) consider AGV dispatching, modeling the problem as
setup (cleaning) time between two colors. Nara Lace Co., a minimum-cost network flow problem. For more details see
Ltd. (2003) has developed a decision support system (DSS) Vis (2006).
called Asprova for the scheduling of the dyeing process. A few researchers have focused on the transportation prob-
The manager’s manual scheduling takes three to four hours, lems encountered in the textile industry. Brahmadeep and
whereas Asprova reduces this to half an hour. It eliminates Thomassey (2014) present a model in which all the processes
the need to write instructions, efficiently handles customer are simulated. They minimize the costs and delays by con-
due dates, and increases the accuracy of the scheduling. sidering all the parameters and constraints that explain the
Saydam and Cooper (2002) report the development and production flow and the distribution logic of bobbins for the
implementation of a DSS for scheduling jobs on multi-port rewinding process in a yarn dyeing factory. Surprisingly, the
dyeing machines. They present a two-phase approach. In synchronization issue has received little attention in the liter-
the first phase, they use a linear programming model to ature. To the best of our knowledge, El Hachemi et al. (2013)
maximize machine utilization when allocating dye-orders is the first study to synchronize dye-jets and forklifts in the
to dye machines. In the second phase, they develop a textile sector.
heuristic approach to sequence fabric rolls while minimiz- The goal of this project is to build a model in which all
ing the manual operations on the plant floor. Laoboonlur the processes can be simulated and all the parameters and
et al. (2006) discuss a detailed production schedule for constraints are considered.

123
J Intell Manuf

The synchronized bin–forklift scheduling problem Let G = (N , A) be a network where the set N of nodes
consists of the start of the sequence (node s), the end of
We consider a set J of jobs (fabrics), a set D of dye-jets, the sequence (node p), and all the jobs, i.e., N = {s} ∪
and a set F of forklifts. Each job j ∈ J is characterized by {d j , j ∈ J } ∪ { p}. For each arc (i, j) ∈ A, we define a
its color c j and its size w j expressed in terms of bins (one cost ci j as well as a cleaning time ti j when jobs i and j
bin is 529 lbs of fabric). Each dye-jet d ∈ D has a specified are sequentially treated by the same dye-jet. If the nodes i
capacity Q d . We must generate an optimal plan where each and j (∈ N) correspond respectively to jobs i and j, then
fabric is dyed on a dye-jet and transported by a forklift to ci j = r j + ti j where r j represents the treatment duration of
the wringing section. Each day, the forklifts must begin and job j. Finally, we define csi = 0 and ci p = ri .
finish their work at the dyeing section. They can carry only Let L represent the set of types of dye-jets. For each l ∈ L,
one bin of a fabric at a time, traveling back and forth between Ml represents the number of available dye-jets of type l. We
the dye-jets and the wringing section. denote by Sl the set of feasible sequences for dye-jets of type
The synchronized bin–forklift scheduling problem l (on graph G). A sequence of jobs is feasible for a dye-jet
(SBFSP) consists in constructing a schedule for the dye-jets l if (1) it starts at source s and ends at sink p, (2) the size
and a transportation plan for the forklifts such that all the of each job in the sequence is less than the capacity of the
jobs are dyed and transported to the wringing section at a associated dye-jet, and (3) the total working time (processing
minimum total cost, and the dye-jet capacity constraints are and cleaning times) for the sequence does not exceed the
respected. In the context of the textile industry, the various length of the given horizon H . For each dye-jet type l ∈ L
sections (dyeing, wringing, drying, and compacting) of the and each feasible sequence s ∈ Sl , we define a binary variable
plant are not large, and the distances between the dye-jet xsl that equals 1 if sequence s is chosen for dye-jet type l,
pads are similar. Consequently, the transportation costs are and 0 otherwise. We also define for each s ∈ Sl and each
independent of the scheduling decisions. The objective func- job j ∈ J a parameter as j such that as j = 1 if sequence s
tion minimizes the unproductive activities (setup and waiting contains job j, and as j = 0 otherwise. Finally, csl is the total
times) of the dye-jets and forklifts. cost of the cleaning time.

The SBFSP is closely related to routing problems encoun- Min csl xsl (1)
tered in other industries, in particular, the so-called pick-up l∈L s∈Sl
and delivery problems. For general surveys of the vehicle 
subject to: as j xsl = 1, ∀ j ∈ J (2)
routing problem (VRP) and the pick-up and delivery prob-
l∈L s∈Sl
lem with time windows (PDPTW), see the book edited by 
Toth and Vigo (2001). xsl ≤ Ml , ∀l ∈ L (3)
s∈Sl
xsl ∈ {0, 1}, ∀l ∈ L ∀s ∈ Sl (4)
Solution approach
The objective function (1) minimizes the total cleaning
We develop a two-phase approach to the SBFSP. The first time of the dye-jets. Constraints (2) ensure that each job is
phase determines the sequence of jobs to assign to each assigned to one dye-jet (or one sequence). Constraints (3)
dye-jet while minimizing the total setup (cleaning) times limit the number of available dye-jets of each type. Finally,
of the dye-jets. The second phase schedules the jobs and constraints (4) are integrality constraints.
the transportation while minimizing the total waiting times Column generation (CG) is an iterative method that can
(associated with jobs). be used to solve the linear relaxation of model (1)–(4), which
is called the master problem. It iterates between a restricted
First phase: assigning jobs to dye-jets master problem (RMP) and several subproblems. The RMP
is derived from the master problem by considering a sub-
In this phase we assign jobs to dye-jets while respecting the set of its variables that is updated at each iteration. The
capacity constraints and the planning horizon. We have devel- RMP is solved by a linear programming solver. Given an
oped both a column generation method and a mixed integer optimal dual solution for the current RMP, the role of the
program (MIP). subproblems is to identify negative-reduced-cost columns
(variables) to be added to the RMP before starting another
Branch and price approach iteration. If no such columns exist, the solution process stops,
and the optimal primal solution of the current RMP is also
The problem of determining the sequence of jobs to assign optimal for the master problem. The subproblems corre-
to each dye-jet is equivalent to solving a VRP, where the spond to resource-constrained shortest path problems that are
vehicles and customers are dye-jets and jobs, respectively. solved by dynamic programming. All the constraints related

123
 J Intell Manuf

Fig. 1 Example of a network for a given type of dye-jet

to sequences are managed at the subproblem level via the A series of subproblems, one for each dye-jet type l ∈ L, is
network structure that uses resource variables. For SBFSP, defined as follows:
there is only one resource constraint: it computes the total
working time of a dye-jet. 
(SPl ) : max c¯sl = csl − as j μ∗j − l∗ (5)
A subproblem is generally a shortest path problem with a
j∈J
number of resource constraints defined on a network that
s.t. s ∈ Sl . (6)
implicitly represents all the feasible columns. A resource
is a quantity that varies along a path and whose value is
restricted to fall within a given interval, called a resource If the optimal solution of each (SPl ), l ∈ L yields a nonpos-
window, at each node. Such subproblems can be solved itive objective value, then the current solution is optimal and
by a label-setting algorithm (see Desrosiers et al. 1995). the LP relaxation is solved. Otherwise, sequences (yielded
Every feasible column corresponds to a path from the by different subproblems) with positive reduced costs are
source to the sink node in this network. The feasibil- identified and added to the RMP and the process continues.
ity rules are treated during the path construction in the Let G l = (N l , Al ) be the network associated with sub-
label-setting algorithm via the use of constrained resource problem (SPl ). Each node i ∈ N l represents one job. Each
variables. arc ((i, j) ∈ A) means that jobs i and j are performed
In our case, testing optimality implies computing the sequentially. An example of a network is given in Fig. 1. This
reduced costs of all the feasible sequences in S. The network contains three types of nodes: the source, the sink,
reduced cost associated with sequence s ∈ Sl , l ∈ L is and the job. There is a single source node and a single sink
given by: node to represent the start and the end of the sequence, respec-
 tively. The network has three arc types: start of the sequence,
c¯sl = csl − as j μ∗j − l∗ , end of the sequence, and cleaning-time arcs. The source and
j∈J the sink nodes are linked to each job node. Cleaning-time
arcs link each pair of jobs together. In Fig. 1, we give an
where (μ∗j ) j∈J and (l∗ )l∈L are the optimal values of the dual example of a network for a given type of dye-jet where we
variables associated with constraints (2) and (3), respectively. consider 8 jobs, j − 1 to j − 8.

123
J Intell Manuf

The MIP formulation Second phase: transporting and scheduling jobs from
dye-jets to pads
We now present an integer programming approach for the
problem of assigning jobs to dye-jets, all parameters and We present in this subsection a mixed integer linear model
variables are defined below. that combines the transportation of the fabrics from dye-
H planning horizon, jets to pads with the scheduling of the dyeing and trip
C set of colors, tasks.
b jc binary parameter that equals 1 if c is the color of job j,
and 0 otherwise, Parameters
 setup (cleaning) time when two consecutive jobs of dif-
ferent colors are processed on the same dye-jet, d j time for dyeing job j ,
M large constant, t j total time to transport the bins of job j by a forklift (one
Ycd binary variable that equals 1 if color c is treated by bin at a time) from the dyeing house to a pad,
dye-jet d, and 0 otherwise,  j maximum time that job j can wait before being
S jd binary variable that equals 1 if job j is assigned to dye- processed on a pad,
jet d, and 0 otherwise, Jd set of jobs performed on dye-jet d according to the first
Wd integer variable that represents the minimum number of step,
setup times associated with dye-jet d. f i j binary parameter equals to 1 if the two jobs i and j have
 different colors, and 0 otherwise,
Min Wd (7) F set of Forklifts,
d∈D
 M  large constant.
subject to: S jd = 1, ∀ j ∈ J (8)
d∈D Variables
w j S jd ≤ Q d , ∀ j ∈ J ∀d ∈ D (9)
X jd start time of dyeing job j on dye-jet d,
 V j f start time of transportation of job j by forklift f ,
b jc S jd ≤ M.Ycd , ∀c ∈ C ∀d ∈ D (10) R j f binary variable equals to 1 if job j is transported by
j∈J forklift f , and 0 otherwise,
 Z j j  binary variable needed to represent disjunction that
t j S jd ≤ H − Wd , ∀d ∈ D (11) arise naturally in dyeing j and j  ,
j∈J 
Z j j  f binary variable needed to represent disjunction that
 arise naturally in transporting j and j  by f .
Ycd ≤ Wd + 1, ∀d ∈ D (12)
c∈C
It should be noted that if forklift f is not transporting job
Ycd ∈ {0, 1}, ∀c ∈ C ∀d ∈ D (13) j then V j f will be equal to 0.
 
Min Vj f − X jd (16)
S jd ∈ {0, 1}, ∀ j ∈ J ∀d ∈ D (14) j∈J f ∈F j∈J d∈D

subject to: R j f = 1, ∀ j ∈ J (17)
Wd ∈ N, ∀d ∈ D (15) f ∈F
V j f ≤ (H − t j )R j f , ∀ j ∈ J, ∀ f ∈ F (18)
The objective function (7) minimizes the total cleaning 
V j f ≥ d j + X jd , ∀d ∈ D, ∀ j ∈ J (19)
time of the dye-jets. Constraints (8) ensure that each job is f ∈F
assigned to one dye-jet. Constraints (9) express the fact that
X jd ≥ X j  d + d j  − M Z j j  + f j j  ,
each dye-jet has a limited capacity. Constraints (10) ensure
that color c is treated by dye-jet d only if there is a job of ∀d ∈ D, ∀ j = j  ∈ Jd (20)
color c dyed by d. Constraints (11) ensure that the total work- X j  d ≥ X jd + d j − M(1 − Z j j  ) + f j j  ,
ing time (processing and cleaning times) does not exceed the ∀d ∈ D, ∀ j = j  ∈ Jd (21)
length of the horizon H . Constraints (12) compute the mini-
Vj f ≥ Vj f + t j R j f − M Z j j  f 
− M (1 − R j f ),
mum number of setups when a set of colors is processed by a

dye-jet. Finally, constraints (13), (14), and (15) are integrality ∀ j = j ∈ J, ∀ f ∈ F (22)
constraints. V j  f ≥ V j f + t j R j f − M(1 − Z j j  f ) − 
M (1 − R j  f ),

123
 J Intell Manuf

Table 1 Comparison of the

Criterion Approach Instances
first-phase approaches
I1 I2 I120%,1 I120%,2 I140%,1 I140%,2

Gap (%) MIP 0 0 0 0 0 0

CG 0 0 0 0 0 0
CPU (s) MIP 0.04 0.16 0.06 0.05 0.07 0.16
CG 6 7 10 8 29 1406
No. of nodes MIP 0 0 0 0 0 4
CG 5 2 4 2 16 56

∀ j = j  ∈ J, ∀ f ∈ F (23) Table 2 Second-phase results


V j f + t j − d j − X jd ≤  j , ∀d ∈ D, ∀ j ∈ Jd Instance C PU (s) Total waiting
f ∈F time (min)
(24) I1 0.48 0
R j f ∈ {0, 1}, ∀ j ∈ J ∀ f ∈ F (25) I2 3.07 144
Z j j  ∈ {0, 1}, ∀ j, j  ∈ J ∀ f ∈ F (26) I120%,1 0.71 0
Z j j  f 
∈ {0, 1}, ∀ j, j ∈ J ∀d ∈ D (27) I120%,2 0.61 0

X jd ≥ 0, ∀ j ∈ J, ∀d ∈ D (28) I140%,1 0.89 0

V j f ≥ 0, ∀ j ∈ J, ∀ f ∈ F (29) I140%,2 1.42 0

In this step, the objective function (16) minimizes the

overall waiting time (where the waiting time of a job is the loaded or unloaded. The time to clean a dye-jet is about 8 min.
time between the end of the dyeing task and the start of the A fabric must not wait more than 2 h between the end of the
transportation task). Constraints (17) ensure that each job is dyeing process and the end of the transportation process. The
transported by one forklift. Constraints (18) ensure that the sizes of the fabrics range from 1 to 12 bins, and the capac-
planning horizon is respected. Constraints (19) ensure that ities of the dye-jets range from 2 to 12 bins. From instance
transportation by a forklift starts only if the dyeing process I1 we randomly generated four instances, two (I120%,1 and
is finished. To schedule jobs on dye-jets or on forklifts, we I120%,2 ) with 20 % more fabrics and two (I140%,1 and I140%,2 )
have to add disjunctive constraints. Thus, constraints (20) and with 40 % more fabrics. We note that the added fabrics are
(21) ensure that when two jobs j and j  are performed on the randomly selected from the fabrics of the real instance, and
same dye-jet d, one job precedes the other. Similarly, the dis- the horizon is set to 30 h for the random instances. We set
junctive constraints (22) and (23) apply to two jobs j and j  the discretization step to be equal to a back-and-forth trip
transported by the same forklift f . For quality reasons, we between the dye-jet and the pad areas, i.e., 4 min. The tests
add constraints (24) ensuring that the waiting time of job j were run on a cluster of Itanium II 1.5 GHz processors.
does not exceed  j . Constraints (25)–(27) are the integrality As noted, we developed two models for the first phase. We
constraints. It should be noted that for each disjunctive con- solve the column generation model using a CG solver and the
straint we define a binary variable. Finally, constraints (28) MIP model using Cplex 12.6.0. The optimal solution does not
and (29) ensure that variables X jd and V j f are nonnegative. have any setup (cleaning) time since each dye-jet processes
fabrics of the same color. To assess the two approaches, we
report in Table 1 the gap, the number of branching nodes, and
Experimental results the computational time. Table 1 shows that the MIP is much
faster than the CG approach, especially for large instances.
We were provided with two case studies by our industrial This is because the daily instances processed by the company
partner. Case I1 involves 18 dye-jets, 2 forklifts, and 33 fab- are easy to solve (Cplex finds the optimal solution at node
rics to dye with 11 different colors, with a 1-day horizon. 0). However, the manager’s manual solution contains setup
Case I2 is the same except that there are 40 fabrics. Our (cleaning) time. We believe that the CG solver will provide
partner also provided approximate processing times for each good solutions compared to those of the MIP solver for fairly
operation: the dyeing time varies between 3 and 6 h depend- large instances.
ing on the color of the fabric. Each forklift takes about 2 min We solved the MIP model of the second phase using Cplex
to move between the dye-jet and the PAD areas, whether it is 12.6.0; Table 2 gives the computational time. The overall

123
J Intell Manuf

waiting time (as defined in “Solution approach” section) of Desrosiers, J., Dumas, Y., Solomon, M. & Soumis, F. (1995). Time con-
the fabrics is equal to zero, except for instance I2 , for which strained routing and scheduling. In M. Ball, et al. (Eds.), Network
Routing, Handbooks of Operations Research and Management
the value is 144 min. Science (Chap. 2, Vol. 8, pp. 35–139). Elsevier Science B.V.
For the transportation planning, the manager did not have El Hachemi, N., Saddoune, M., El Hallaoui, I., & Rousseau, L.-M.
a complete view of the horizon. Sometime, different dye-jets (2013). Production scheduling and routing problem in the tex-
finished larger fabrics almost simultaneously, and as a result, tile industry. In Industrial engineering and systems management
(2013), Proceedings of international industrial engineering and
the two forklifts were unable to transport all the orders on systems management conference.
time. The company plans to adopt our algorithm to solve its Felix, T., Chan, S., & Chan, H. K. (2004). A comprehensive survey and
scheduling problems. future trend of simulation study on FMS scheduling. Journal of
Intelligent Manufacturing, 15(1), 87–102.
Giralt, G., & Chatila, R. (1987). Task programming and motion control
for autonomous mobile robots in manufacturing. In IEEE interna-
Conclusion tional conference in robotics automation (1987).
Hao, G., Shang, J. S., & Vargas, L. G. (1996). A neural network model
We have described a two-phase approach for the SBFSP. for the free-ranging AGV route-planning problem. Journal of Intel-
ligent Manufacturing, 7(3), 217–227.
The first phase determines the optimal assignment of fabrics
Hu, H., & Gu, D. (2000). Landmark-based navigation of industrial
to dye-jets so as to minimize the overall cleaning time. We mobile robots. International Journal of Industry Robots, 27(6),
developed two models for this phase. For the second phase 458–467.
we presented an MIP model that combines the transportation Laoboonlur, P., Hodgson, T. J., & Thoney, K. A. (2006). Production
scheduling in a knitted dyeing and finishing process. Journal of
of fabrics from dye-jets to PADs and the scheduling of the the Textile Institute, 97(5), 391–399.
dyeing and the forklift trips. Lin, L., Shinn, S. W., Gen, M., & Hwang, H. (2006). Network model
Future research will focus on solving the problem in a and effective evolutionary approach for AGV dispatching in man-
single phase using an integrated MIP model; dealing with ufacturing system. Journal of Intelligent Manufacturing, 17(4),
465–477.
additional constraints related to the PAD area; and producing Maldonado, F., Ciurlizza, A., Radillo, R., & Ponce, E. (2000). Opti-
robust solutions by including stochasticity in the machine misation of the colour sequence in the dyeing process: Industrial
operations and the forklift travel times. We plan to make applications. JSDC, 116, 359–362.
initial decisions based on the deterministic context and to Morales, L., Maldonado, F., Radillo, R., & Ciurlizza, A. (1996). Opti-
misation of colour sequence in the process of fabric dyeing. JSDC,
observe the outcomes on a continuous, rolling-horizon basis. 112, 361–363.
If stochastic information becomes available or an unexpected Nara Lace Co., Ltd. (2003). Dye manufacturer chooses Asprova for high
event (machine breakdown) occurs, it is important to use this speed scheduling process and user-friendly GUI environment.
information in dynamic planning. We believe that doing so O’Kane, J. F. (2000). A knowledge-based system for reactive scheduling
decision-making in FMS. Journal of Intelligent Manufacturing,
would solve the assignment problems in most cases. 11(5), 461–474.
Saydam, C., & Cooper, W. D. (2002). A decision support system
for scheduling jobs on multi-port dyeing machines. International
References Journal of Operations & Production Management, 22(9), 1054–
1065.
Singh, N., Sarngadharan, P. V., & Pal, P. K. (2009). AGV scheduling
Brahmadeep, & Thomassey, S. (2014). A simulation based comparison:
for automated material distribution: A case study. Journal of Intel-
Manual and automatic distribution setup in a textile yarn rewinding
ligent Manufacturing, 22(2), 219–228.
unit of a yarn dyeing factory. Simulation Modelling Practice and
Toth, P., & Vigo, D. (2001). The vehicle routing problem. Philadelphia:
Theory, 45, 80–90.
Society for Industrial and Applied Mathematics.
Bruno, G., Ghiani, G., & Improta, G. (2000). Dynamic positioning of
Vis, I. F. A. (2006). Survey of research in the design and control of auto-
idle automated guided vehicles. Journal of Intelligent Manufac-
mated guided vehicle systems. European Journal of Operational
turing, 11(2), 209–215.
Research, 170, 677–709.
Cho, E. (2004). Scheduling supply chains with batchwise fabric dyeing
Zhou, X., Chen, C., Wu, P., & Zheng, J. (2010). Optimized scheduling
operations. Ph.D. thesis, North Carolina State University.
of production process based on continuous-time in printing and
Confessore, G., Fabiano, M., & Liotta, G. (2011). A network flow based
dyeing industry. CIESC Journal, 61(8), 1877–1881.
heuristic approach for optimizing AGV movements. Journal of
Intelligent Manufacturing, 24(2), 405–419.

123