Computers & Industrial Engineering 73 (2014) 115–123

Contents lists available at ScienceDirect

Computers & Industrial Engineering

journal homepage: www.elsevier.com/locate/caie

An Integrated Quay Crane Assignment and Scheduling Problem

Ali Diabat ⇑, Effrosyni Theodorou
Department of Engineering Systems & Management, Masdar Institute of Science & Technology, Abu Dhabi, United Arab Emirates

a r t i c l e i n f o a b s t r a c t

Article history: As maritime container transport is developing rapidly, the need arises for efficient operations at container
Received 12 June 2013 terminals. One of the most important determinants of container handling efficiency is the productivity of
Received in revised form 17 October 2013 quay cranes, which are responsible for unloading and loading operations for container vessels. For this
Accepted 20 December 2013
reason, the Quay Crane Assignment Problem (QCAP) and the Quay Crane Scheduling Problem (QCSP) have
Available online 20 March 2014
received increasing attention in the literature and the present paper deals with the integration of these
interrelated problems. A formulation is developed for the Quay Crane Assignment and Scheduling Prob-
Keywords:
lem (QCASP), which accounts for crane positioning conditions and a Genetic Algorithm (GA) is developed
Quay crane assignment
Quay crane scheduling
to solve the QCASP. Both the model formulation and the solution methodology are presented in detail and
Genetic algorithms computational analysis is conducted in order to evaluate the performance of the proposed GA. The results
Container terminals obtained from the GA are compared with the results from an exact technique, thus providing complete
Maritime operations information about the performance of the heuristic in terms of solution quality.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction QCs are responsible for the unloading and loading of containers
from and to the vessel. The planning of QC operations is part of
An increase in the number of transshipped container goods has the quayside operations of a container terminal and consists of
been marked over the recent decades, due to globalization. Con- the QCAP and QCSP. These problems are frequently integrated, as
tainer terminals are called to meet the challenge of accommodat- they are interrelated.
ing very large vessels, which are capable of carrying 10,000– The QCAP is basically an assignment problem which considers
12,000 twenty-foot equivalent container units (TEUs). A systematic additional parameters, such as service agreements contracted with
approach to container terminal optimization therefore becomes vessel operators, dictating a minimum or maximum number of
necessary in order to overcome this challenge. cranes that can be assigned to a vessel, the available QCs at the
As far as container terminal operations are concerned, they can port, the number of vessels berthed within a given planning hori-
be divided into quayside and yard side operations, as illustrated in zon, the container workload on each vessel, and whether or not
Fig. 1. On the one hand, quayside operations involve allocating cranes are allowed to perform handling operations on more than
berths to arriving ships, known as the Berth Allocation Problem one ship within a planning horizon.
(BAP), the assignment of cranes to ships, known as the Quay Crane The QCSP is a scheduling problem, more complicated than the
Assignment Problem (QCAP) as well as the sequencing of quay QCAP, as it decides upon the sequencing of the QCs’ handling tasks
crane operations, known as the Quay Crane Scheduling Problem and the points in time at which these are performed. An important
(QCSP). On the other hand, yard side operations include the alloca- aspect of the QCSP is the fact that positioning conditions must be
tion of containers to certain storage locations, the scheduling of enforced at all times. More specifically, since cranes travel on a sin-
container transporting vehicles and the scheduling of yard cranes gle rail, they are not allowed to cross one another. These are known
for optimal container storage sequence. as the non-crossing constraints. Furthermore, assuming that cranes
Because Quay Cranes (QCs) are the most expensive equipment are indexed based on their position, middle-indexed cranes cannot
utilized at container terminals, their performance largely affects serve end bays, because again this would violate the non-crossing
the container throughput and handling efficiency (Meisel, 2011). conditions. In several models clearance conditions are also
QCs move on a single rail track alongside the quay of the port, as accounted for, in order to prevent adjacent cranes from being posi-
depicted in Fig. 1. As soon as a ship is positioned at the berth, tioned too close to one another. Yard congestion constraints are
also considered in certain cases, where it is important to ensure
⇑ Corresponding author. Tel.: +971 2 810 9101; fax: 971 2 810 9901. that there will not be traffic at the yard storage areas at any point
E-mail address: [email protected] (A. Diabat). in time.

http://dx.doi.org/10.1016/j.cie.2013.12.012
0360-8352/Ó 2014 Elsevier Ltd. All rights reserved.
116 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123

Fig. 1. Schematic representation of container terminal layout. Source: ‘The Quay Crane Scheduling Problem with Time Windows’, Frank Meisel.

In the current paper, we propose an integrated model for the span and the total completion time, but the drawback of their for-
QCAP and the QCSP, namely the Quay Crane Assignment and mulation in terms of constraints is that they do not consider crane
Scheduling Problem (QCASP). The purpose of the model is to assign clearance conditions, i.e. constraints that ensure that cranes will be
cranes to ships that are berthed within a given planning horizon. positioned at least certain bays apart, and they only consider the
Furthermore, the model specifically decides which crane is allo- single-ship case. Clearance conditions were added to the model
cated to which bays and it aims to minimize the time required of Kim and Park (2004) by Moccia, Cordeau, Gaudioso, and
for the completion of the handling of the latest ship, i.e. the ship Laporte (2005), rendering the formulation more robust. Both for-
carrying the largest number of containers, which is expected to mulations have since been used in numerous works.
take the most time at the berth. This article presents the imple- Legato, Mazza, and Trunfio (2008) minimize the maximum
mentation of a Genetic Algorithm (GA) for solving the QCASP and makespan required to complete the number of existing tasks. Their
reports the results of the computational studies performed for cer- Mixed Integer Programming (MIP) formulation allocates quay
tain problem instances. cranes to tasks and time slots, without considering ships or bays.
The main contribution of this paper is in the integration of the Jung, Park, Lee, and Kim, and Ryu (2006) employ time windows
QCAP and QCSP, two interrelated problems that have mostly been in which cranes can be assigned to perform a task; the single ship
dealt with independently in the literature. Furthermore, the devel- case is only considered and the authors assume ship-clusters
oped model holds the advantage of simplicity, while at the same which containers cannot be loaded into or unloaded from simulta-
time it considers realistic circumstances, as it accounts for all posi- neously. Lee, Qiu Wang, and Miao (2008) develop a MIP formula-
tioning constraints in order to generate practical solutions. The dis- tion for the QCSP with handling priority for each ship bay. The
advantage of the large number of variables is overcome through objective function is therefore a minimization function of the
the use of a GA specifically developed for this problem. Although sum of the weighted completion times of every ship bay.
heuristics have been largely implemented in the literature, the Liu, Wan, and Wang (2006) formulate an MIP whose objective is
present paper thoroughly evaluates the performance of the GA, to minimize the maximum relative tardiness of vessel departures.
since it compares the solution with a solution generated through In terms of model assumptions, the authors consider the aggre-
an exact approach. gated workload of each bay, which is the product of the number
The present paper is structured as follows: Section 2 provides a of containers to be handled in the bay and the average processing
literature review on the QCSP, focusing on the models built and the time per container. As is commonly the case, each vessel is parti-
solution approaches developed for these problems. Section 3 con- tioned into bays, and so is the berth. Clearance constraints between
tains the detailed problem description and its mathematical for- adjacent cranes are considered, while productivity is assumed
mulation, while Section 4 introduces the GA that was developed identical for all quay cranes and interference amongst them is
to solve this specific problem. Section 5 reports the results of the ignored.
computational analysis and evaluates the performance of the pro- The early work of Daganzo (1989), who first introduced the
posed heuristic, while Section 6 concludes the article with the crane scheduling problem, develops an exact MIP for the loading
important findings of this work, as well as directions and recom- of ships, assigning cranes to bays for certain time slots, such that
mendations for future research on this topic. the overall workload is well balanced for cranes. The objective is
to minimize the aggregate cost of delay incurred on the vessels,
unlike what became prevalent thereafter. Another note on the
2. Literature review author’s formulation is that several important factors are not con-
sidered, such as crane interference and crossing. The work of Zhang
Several models are proposed in the literature for the QCSP and a and Kim (2009) also introduces a different objective than that seen
very useful classification of these models can be found in the work prevalent in the literature: the aim of their MIP formulation is to
of Bierwirth and Meisel (2010). As far as the problem formulation minimize the total number of cycles of QC activities on one ship,
is concerned, the prevalent objective is the minimization of the rather than to minimize the makespan. Tavakkoli-Moghaddam,
makespan required to complete tasks. In the work of Kim and Makui, Salahi, Bazzazi, and Taheri (2009) integrate the quay crane
Park (2004) the authors minimize the weighted sum of the make- scheduling and allocation problem (QCSAP) in their work. They
 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123 117

develop a MIP formulation whose objective is to minimize the total the solution improvement phase, the constructed solution is
cost of the proposed model. locally improved by an exchange operation until no further
Bierwirth and Meisel (2009) use the MIP developed by Moccia improvement is possible. Results demonstrate that the computa-
et al. (2005), which is the corrected version of the model developed tional time is far too high to be used in practice, with the improve-
by Kim and Park (2004). Later work of Meisel (2011) utilizes a dif- ment phase being most time-consuming, with a marginal
ferent approach for quay crane scheduling, where the availability improvement in the quality of the solution.
of cranes at a vessel is restricted to certain time windows, as pre- Liu et al. (2006) propose a heuristic decomposition approach to
scribed by a given crane assignment. The problem is formulated as break down the problem into two smaller, linked models, namely
a MIP with the objective of minimizing the vessel handling time, the vessel-level model and the berth-level model. Both models
determined by the latest completion time of all tasks. Choo, are formulated as MINLPs, but the authors claim that they require
Klabjan, and Simchi-Levi (2009) propose a MIP for the crane less computational effort when compared to the MIP with many
sequencing problem with entirely new set-covering formulations variables. Computational experiments show that the proposed
which allow the authors to solve large-scale instances. They approach is effective and efficient. Zhang and Kim (2009) develop
include clearance and yard congestion constraints, the latter hav- a heuristic algorithm, by decomposing the QC scheduling into an
ing seldom been included in the QCSP in previous literature. intra-stage optimization, which finds the optimal sequencing of
In terms of solution methodologies, heuristics are widely all stacks in one bay, and an inter-stage optimization, which
employed due to the large number of variables and high complex- sequences all the bays. Numerical experiments reveal that the pro-
ity of the discussed problems. Exact techniques have also been posed approach found the optimal solution in most of the cases
implemented in addition to heuristics. Such is the case in the work and greatly outperformed the scheduling methods used by port
of Kim and Park (2004). The authors propose a Branch-and-Bound operators.
(B&B) method to obtain the optimal solution of the MIP for the Lee et al. (2008) develop a GA to reach near-optimal solutions
QCSP, as well as a heuristic search algorithm, the Greedy Random- due to the large size of the problem. Based on the 43 computational
ized Adaptive Search Procedure (GRASP), in order to overcome the experiments, the authors conclude that the performance of the GA
difficulty of the B&B method. Daganzo (1989) also proposes both is satisfactory in solving large size instances. Tavakkoli-
an exact and an approximate solution method for crane Moghaddam et al. (2009) also employ a GA, due to the complexity
scheduling. of their model and in order to solve real-size instances of the prob-
In the work of Choo et al. (2009), the authors solve the MIP sin- lem. The obtained results showed a reasonable gap between the
gle-ship proposed model with the help of a heuristic approach that optimal solutions found by a commercial solver and the GA
produces good results. The model is then reformulated as a gener- (between 1.9% and 3.5%), while the GA reaches the near-optimal
alized set covering problem and solved exactly by Branch-and- solution in reasonable time.
Price (B&P). For multiship sequencing, the yard congestion con- Chung and Choy (2012) base their formulation on the one
straints are relaxed in the spirit of Lagrangian relaxation, so that developed by Kim and Park (2004) and they propose a modified
the problem decomposes by vessel into smaller problems that genetic algorithm for solving the model. They compare their
can be solved by B&P at each iteration of the subgradient algo- results with a set of benchmarking data comprised of 43 instances.
rithm. The authors observe that computational results vary consid- Their analysis proves that the proposed GA can achieve results very
erably but in all cases, the proposed Lagrangian relaxation close to the best known solutions by other approaches that were
technique is the recommended approach in terms of computa- implemented for these instances, such as Tabu Search (TS) and
tional efficiency and solution quality for the multiship problems. Branch-and-Cut (B&C). Especially for smaller problem sizes, the
For the single-ship case, Guan, Yang, and Zhou (2010) also gap between the proposed GA and the best known solution is 0%.
develop both heuristic and exact solution approaches to the prob- For larger problems the performance of the GA is compromised
lem. More specifically, for small-sized problems the authors in terms of optimal results, but remains attractive due to the lower
employ a time-space network flow formulation with non-crossing computational time required.
constraints and apply an exact solution approach to obtain the Bierwirth and Meisel (2009) propose a heuristic solution proce-
optimal solution. For medium-sized problem instances they dure for the model based on that of Moccia et al. (2005), the B&B
develop a Lagrangian relaxation approach in order to obtain tight algorithm being at its core. It is applied for searching a subset of
bounds and near-optimal solutions, while for large-sized instances above average quality schedules and it takes advantage of efficient
they develop heuristic techniques. In terms of the Lagrangian criteria for branching and bounding the search with respect to the
relaxation they adopt the subgradient method to obtain the impact of crane interference. The authors noted results that were
Lagrangian bound and develop a Lagrangian relaxation heuristic very satisfactory, even compared to recent competing approaches.
to guarantee feasibility. The authors concluded that this approach For instances of a set of benchmarks with a small number of cranes
can provide tighter lower bounds and obtain a feasible solution (two and three) the best known solutions were found in every case.
within a reasonable amount of time. Also, the computational effort was cut down to fractions as com-
The optimization framework used by Legato et al. (2008) is the pared to other methods. Even in larger instances, although not as
A-BEES search method, which searches for the optimal solution satisfactory, the algorithm performed well. The authors attributed
through the feasible region by balancing two phases: exploration the heuristic’s fast performance to the exclusive consideration of
and exploitation. The authors also apply simulated annealing, unidirectional schedules, i.e. the movement of all cranes in a single
which is adapted to the context of the QCSP, in terms of possible direction alongside the quay. In later work, Meisel (2011) proposes
moves which respect precedence and availability constraints. In a solution method for the Quay Crane Scheduling Problem with
terms of results, the A-BEES metaheuristic seems promising in Time Windows (QCSPTW), that searches the solution space of uni-
quickly providing cost effective solutions to the practical problem directional schedules, based on his findings from previous work
of determining the sub(optimal) assignment schedule of quay (Bierwirth & Meisel, 2009). In this paper however, he concludes
cranes to vessel tasks. Jung et al. (2006) employ in the first phase that the best unidirectional schedule is not necessarily an optimal
of their heuristic, a Greedy Randomized Adaptive Search Procedure solution for a QCSPTW instance, but the solutions are still of high
(GRASP), an iterative randomized sampling technique which gen- average quality. Computational results for a large set of test
erates a solution at each iteration. The second phase involves an instances reveal that the proposed heuristic delivers near optimal
improvement of the solution constructed in the first phase. In solutions within short times.
118 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123

The present paper formulates the integrated QCAP and QCSP, with travel times for cranes between bays are not considered. Finally,
the objective of minimizing the makespan of all vessels. The integra- the ships that will be handled within the planning horizon are
tion of the two interrelated models is expected to yield better results already berthed and only after their tasks have been completed
than considering these problems independently, since more informa- can a new profile of ships be assigned to the quay.
tion is utilized given the interdependence of these problems. All posi-
tioning conditions for quay cranes are considered; these are the end-
3.2. Example
bay positioning constraints and the non-crossing constraints, while
clearance conditions are also accounted for through modeling
In this section an example is used to demonstrate how solving
assumptions. This work considers a multiship rather than a single
the integrated QCASP can yield better results than the traditional
ship case, which can also lead to more useful results because it is
sequential approach, where the crane assignment is firstly solved
realistic and allows for an improved allocation of available resources.
and used as an input to solve the crane scheduling problem. The
Finally, the employed GA is benchmarked against an exact solution
graphic illustration of the problem is presented in Fig. 2. In this
which allows for clear evaluation of its performance.
problem instance there are two ships and three quay cranes avail-
able to handle these ships.
3. Modeling the Quay Crane Assignment and Scheduling In Table 1 to Table 4, the second row contains the indices for the
Problem bays of each ship separately, while the third row creates a single
index for all bays. This is possible because, as mentioned in the
This section provides a description for the QCASP, the character- model assumptions, the ships and bays are indexed sequentially
istics of the problem and how these are modeled in a mathematical based on their positions at the quay. Unifying the bay indices into
formulation. An example is provided to illustrate the problem and one index essentially transforms the problem into a single-ship case,
its solution. The final sub-section examines how the input data of thus simplifying the solution procedure without loss of generality. In
the problem is generated for all problem instances that will be row 6 the total number of containers handled per crane is shown,
tested and presented in the subsequent section. while the handling time required is shown in row 7, assuming a rate
of 25 containers per hour. Finally, the objective value is presented in
3.1. Problem description row 8, and it is equal to the greatest handling time required.
Assuming that the first ship carries 85 containers and the second
For a given number of ships I berthed within a certain planning carries 50 containers as summarized in Table 1, and given that there
horizon and given a number of available quay cranes K, the aim of are three available quay cranes for both ships, the simple quay crane
the QCASP is to allocate cranes to the bays j of every ship i, carrying assignment problem would allocate 2 cranes to the first ship and 1
a certain number of containers, cj, with an objective of minimizing crane to the second ship. Assuming a handling rate of 25 containers
the handling time of the latest ship. By reducing the makespan for per crane-hour, this would mean that the makespan is equal to 2 hour,
the cranes to perform all operations, it is possible to ensure an based on the result of the scheduling problem when taking the deci-
early completion time for all ships and an efficient distribution of sion of the assignment problem as an input, as shown in Table 1.
container workload amongst port resources. However, the integrated assignment and scheduling problem
Certain assumptions have been considered for the specific prob- would consider the container load of each bay, rather than each
lem. Firstly, all quay cranes are assumed identical, and they are ship. The optimal solution in this case would be to assign 2 cranes
indexed sequentially according to increasing positions alongside to the first ship and 2 cranes to the second ship, which implies that
the quay. The bays are also indexed sequentially and cranes begin one of the cranes would be assigned to both ships. This achieves a
handling the lowest-indexed bay out of the total bays they are makespan of 1.8 hour which is less than that obtained by sequen-
assigned to. This implies that cranes travel in a single direction, tially solving the two problems and it is depicted in Table 2. Even
identical for all, in order to enforce positioning constraints and though this may seem to be a small reduction, it becomes signifi-
avoid crane interference. Furthermore, by adopting a unidirec- cant for larger problems.
tional movement for cranes, clearance conditions are enforced Assuming that the QCAP is able to allocate one crane to more
(cranes will always be positioned certain bays away from other than one ship, another possible solution to the assignment prob-
cranes). Since cranes move on a single rail they are not allowed lem would be to assign 2 cranes to the first ship and 2 cranes to
to cross each other, which leads to the need for non-crossing con- the second ship. In this case, the scheduling problem would then
straints. Furthermore, it is possible for one quay crane to be produce the crane schedule shown in Table 3. However, maintain-
assigned to more than one ship, which means that cranes can move ing the same total workload in terms of containers but with a dif-
between ships. Only unloading operations are considered and all ferent distribution on the bays (case 2), would mean that this leads
containers are assumed identical. Container handling rates for again to a not optimal solution, as opposed to the integrated
cranes are identical, measured in containers per crane-hour, while QCASP, shown in Table 4. Solving the problems sequentially would

Fig. 2. Graphic Illustration of the QCASP with 2 ships and 3 QCS.

 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123 119

Table 1 Decision variables

Problem data and solution using sequential assignment and scheduling approach,

case 1. xjk 1; if crane k is assigned to bay j
1. Ship 1 2
0; otherwise
2. Ship bays 1 2 3 4 5 6 1 2 3 4 T makespan, latest completion time among ships
3. All bays 1 2 3 4 5 6 7 8 9 10
4. Containers 45 5 10 10 10 5 5 5 20 20
5. Cranes assigned to bays 1 2 2 2 2 2 3 3 3 3
6. Containers per crane 45 40 50 Parameters
7. Handling time per crane (h) 1.8 1.6 2
8. Makespan (h) 2 cj number of containers on bay j
r crane handling rate, in container per crane-hour
ji first bay of ship i
Bi last bay of ship i
Table 2
Problem data and solution using integrated assignment and scheduling approach,
case 1.

1. Ship 1 2 The MIP formulation of the QCASP can be stated as follows:

2. Ship bays 1 2 3 4 5 6 1 2 3 4
3. All bays 1 2 3 4 5 6 7 8 9 10
min T ð1Þ
4. Containers 45 5 10 10 10 5 5 5 20 20
5. Cranes assigned to bays 1 2 2 2 2 2 2 3 3 3 X
jKj
6. Containers per crane 45 45 45 xjk ¼ 1; j ¼ 1; . . . ; jBj ð2Þ
7. Handling time per crane (h) 1.8 1.8 1.8 k¼1
8. Makespan (h) 1.8
X
k1
xjk 6 0; k ¼ 2; . . . ; jKj ð3Þ
Table 3 j¼1
Problem data and solution using sequential assignment and scheduling approach,
case 2.
X
jBj

1. Ship 1 2 xjk 6 0; k ¼ 1; . . . ; jKj  1 ð4Þ

2. Ship bays 1 2 3 4 5 6 1 2 3 4 j¼ðjBjþ1ÞðjKjkÞ
3. All bays 1 2 3 4 5 6 7 8 9 10
4. Containers 5 5 10 10 10 45 20 5 20 5
X
jBj
5. Cranes assigned to bays 1 1 1 1 1 2 2 3 3 3
xjk 6 xl;kþ1 ; j ¼ 1; . . . ; jBj  1; k ¼ 1; . . . ; jKj  1 ð5Þ
6. Containers per crane 40 65 30
l¼jþ1
7. Handling time per crane (h) 1.6 2.6 1.2
8. Makespan (h) 2.6
X
j1
xjk 6 xl;k1 ; j ¼ 2; . . . ; jBj; k ¼ 2; . . . ; jKj ð6Þ
l¼1

lead to a solution of 2.6 hour, while the integrated problem leads to Bi

X
jKj X
a better solution of 2 hour. TP cj xjk ; i ¼ 1; . . . ; jIj ð7Þ
It becomes evident that the integrated approach outperforms k¼1 j¼ji
the sequential solution approach in most cases, as more informa-
tion is exploited and it is possible to reach a better solution. In cer- xjk ¼ 0; 1 j ¼ 1; . . . ; jBj; k ¼ 1; . . . ; jKj ð8Þ
tain cases, the solution from the two problems may be the same,
but this is mainly the case for very small problems, like the exam- TP0 ð9Þ
ple in this section. For larger problems, the improvement in the
solution is apparent and leads to a significant reduction in handling The objective function (1) aims to minimize the handling makespan
time. of the ship that requires the greatest time. Constraints (2) ensure
that every bay will be assigned exactly one quay crane, while con-
3.3. Optimization model straints (3) and (4) guarantee the end-bay positioning conditions for
quay cranes, ensuring that middle-indexed cranes will not be
We introduce the following notations: placed at end bays. Constraints (5) and (6) are the non-crossing con-
Sets straints, enforcing the condition that a lower-indexed crane will not
be positioned to the right of a higher-indexed crane (constraints
I Set of ships, indexed by i = 1, . . . , |I| (5)), and that a higher-indexed crane will not be positioned to the
B Set of bays, indexed by j = 1, . . . , |B| left of a lower-indexed crane (constraints (6)). Constraints (7) guar-
K Set of cranes, indexed by k = 1, . . . , |K| antee that the makespan will be equal to the latest completion time

Table 4
Problem data and solution using integrated assignment and scheduling approach, case 2.

1. Ship 1 2
2. Ship bays 1 2 3 4 5 6 1 2 3 4
3. All bays 1 2 3 4 5 6 7 8 9 10
4. Containers 5 5 10 10 10 45 20 5 20 5
5. Cranes assigned to bays 1 1 1 1 1 2 3 3 3 3
6. Containers per crane 40 45 50
7. Handling time per crane (h) 1.6 1.8 2
8. Makespan (h) 2
120 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123

amongst all cranes. Finally, constraints (8) and (9) respectively By examining the current model, an initial representation of the
define the integrity and non-negativity requirements. chromosome can be derived by assigning each of the binary deci-
sion variables, i.e. the crane to bay assignment variables, to a gene.
3.4. Data generation The integer variable, i.e. the makespan, can be directly calculated in
the objective function in order to provide a fitness value to the cor-
The data used for the problem instances, namely the number of responding chromosome. This would mean that the length of each
bays and containers on each bay, for a given number of ships and chromosome is equal to the number of bays, multiplied by the
cranes is randomly generated. The number of bays for each ship number of cranes, i.e. K  B, and it would be a binary representa-
is a randomly chosen integer for the interval [10, 100], while the tion. For the example presented in Section 3, the length of each
number of containers is chosen from the interval [1, 100], both chromosome would be 120 genes and the chromosome of the opti-
with a standard uniform distribution. mal solution would have the following structure:
The problem instances are classified as small, medium and
large, based on the number of variables and constraints for each x11 x21 x31 xjk x18,6 x19,6 x20,6
problem instance. More specifically, the small problem instances 1 1 0 ... 1 1 1
refer to up to 400 variables and approximately 400 constraints.
The medium problem instances have approximately 1000 variables
and 1100 constraints, while the large problems comprise of almost In this example, each gene that has a value of one implies that
1600 variables and 1800 constraints. there is an assignment of a certain crane to a certain bay. Specifi-
cally, the fact that x11 = x21 = 1 implies that quay crane 1 (second
4. The proposed genetic algorithm index) is assigned to bays 1 and 2 (first indices). This binary repre-
sentation is straight-forward, yet not efficient, as the length of the
Genetic Algorithms (GAs) have been extensively employed in chromosome would significantly increase for larger problem sizes,
combinatorial optimization problems, including sequencing and with a greater number of cranes, bays and ships. A more efficient
scheduling problems (Tavakkoli-Moghaddam et al., 2009). The representation would be to replace it with a chromosome of inte-
GA is a meta-heuristic approach that is based on the concept of ger value genes, from the interval [1 K], determining which crane
natural biological evolution of living organisms. It operates on a should be assigned to each bay. This implies that the length of
population of potential solutions and applies structured, yet ran- the chromosome is equal to the total number of bays, B. For the
domized information exchange in order to robustly explore and aforementioned example, this representation would lead to the fol-
exploit the solution space. Furthermore, it applies the principle of lowing chromosome:
survival of the fittest in order to continuously produce better
approximations to a solution. At each generation, a new set of Bay 1 2 3 j 18 19 20
approximations is created by the process of selecting individuals Crane 1 1 2 ... 6 6 6
according to their level of fitness in the problem domain and
breeding them together using operators borrowed from natural
genetics, such as crossover and mutation. This process leads to In this representation, it is easier to identify the cranes assigned to
the evolution of populations of individuals that are fitter with ref- bays. Crane 1 is assigned to bays 1 and 2, crane 2 is assigned to bay
erence to their environment than the individuals that they were 3 and possibly more, while crane 6 is assigned to bays 18, 19 and 20.
created from, as occurs in natural adaptation (Chipperfield & This representation can significantly reduce the time and complex-
Fleming, 1995). ity of the genetic algorithm, as compared to the initial representa-
tion. However, it holds the disadvantage of easily generating
4.1. Chromosome representation and population initialization infeasible solutions as a result of genetic operations.
The chromosome representation can be further simplified by
The individuals of the population are encoded as strings, defining the value of each gene as the number of bays that each
namely chromosomes, which are created based on a certain alpha- crane is assigned to. This implies that the length of the chromo-
bet, in a way that the genotypes or genes (chromosome values) some (the number of genes) is equal to the number of available
correspond to a unique position on the decision variable domain. cranes K.
The representation of chromosomes is the initial and most impor- Referring again to the previous example since, QC 1 is assigned
tant step of the GA (Chipperfield & Fleming, 1995; Tavakkoli- to 2 bays, QC 2 is assigned to 5 bays, QCs 3, 4 and 5 are each
Moghaddam et al., 2009). assigned to 3 bays and QC 6 is assigned to 4 bays, the chromosome
The structure of the initial population is also of crucial impor- corresponding to this solution is the following:
tance to the efficiency of the GA. Especially in the case of con-
strained problems, it must be decided whether or not the Crane 1 2 3 4 5 6
initially generated chromosomes will satisfy all equalities and Number of bays 2 5 3 3 3 4
inequalities. Several proposed algorithms allow for infeasible solu-
tions within the population with the purpose of exploring new fea-
sible regions from different directions. In these cases, it is Another advantage of this final chromosome representation,
necessary to implement techniques of controlling infeasible chro- aside from the fact that it is shorter in length, is that it always gen-
mosomes, given that the solution that the algorithm converges to erates feasible solutions, and therefore the need for a feasibility
must be feasible. Other algorithms generate an initial population restoration technique in the initial pool of individuals is not
of feasible chromosomes and maintain feasibility through evolu- required. The only requirement for feasibility with this new repre-
tion by carefully choosing the crossover and mutation variation sentation is that the sum of the values of genes is exactly equal to
operators. In the case where these operators generate infeasible the total number of bays. This chromosome representation will
chromosomes, a repair procedure is required in order to restore thus be employed in the present paper.
feasibility, with the disadvantage of added algorithmic complexity In order to randomly generate individuals, the value for the first
(Hiassat, 2012). (K  1) genes (corresponding to all cranes except the last one) is
 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123 121

generated between the interval [1, round(2  B/K)], where B is the

number of bays and K is the number of cranes. This is chosen in
order to ensure consistency in terms of the number of bays that
can be assigned to each crane and during the genetic operations
it is possible that the number of bays assigned to a crane increases
beyond the maximum number of the interval.
As far as population initialization is concerned, the choice is to In terms of the operation of mutation, care is required in order to
generate a population of feasible chromosomes. This population is not result in non-feasible chromosomes. For this reason, swap
randomly generated but ensures that the sum of genes will be mutation will be employed for this problem. This basically means
equal to the total number of bays. This guarantees that every bay the shuffling of genes within the chromosome. Because the primary
will be assigned exactly one crane, thus ensuring feasibility. concern for feasibility is maintaining the sum of genes equal to the
number of bays, shuffling the genes, i.e. changing their positions in
4.2. Design of genetic operators the chromosome, will not lead to infeasibility. Furthermore, it can
achieve what mutation is intended to achieve, that is the constant
The operators of crossover and mutation in the present paper diversification of the population. Swap mutation is performed
have been designed to ensure that feasibility will be maintained between two genes in the specific problem; however, it could even
throughout the evolution process, in order to prevent the need be applied to more genes, as this would still maintain feasibility.
for a computationally complex repair procedure. Assume mutation is performed on Child 1. This will result in the fol-
In order to demonstrate how the operation of crossover is performed, lowing chromosome:
the following steps are presented, accompanied by an example:
Child 1 4 3 5 3 3 2
Step 1: The parents are chosen to be paired in order to produce
the offspring. Assume the following two chromosomes are the
first that will be paired.
4.3. Parent selection strategy and insertion of offspring

The parent selection strategy plays a significant role in regulat-

ing the bias in the reproduction process, because it determines the
Parent 1 5 3 2 3 4 3
number of times a particular individual from the population is cho-
Parent 2 1 6 4 4 3 2
sen for reproduction and, thus, the number of offspring that an
individual will produce.
Several selection techniques implement a ‘roulette wheel’
Step 2: For every pair, a random number of genes is selected for mechanism to probabilistically select individuals based on a cer-
both chromosomes, to be exchanged between them, assume 3 tain measure of their performance. Assuming that the fitness value
genes. is the chosen measure of performance, individuals are mapped into
Step 3: After the number of genes is chosen, the indices corre- contiguous intervals in the range between 0 and the sum of their
sponding to genes are also randomly selected, assume: 2, 3, fitness values. The select an individual, a random number is gener-
and 5. The genes that will be crossed over are highlighted: ated within this range and the individual whose segment includes
the random number is selected. The process is repeated, until the
desired number of individuals has been selected. This technique
is generally employed when it is preferred that better solutions
have a greater chance of being chosen for creating offspring
(Chipperfield & Fleming, 1995; Tavakkoli-Moghaddam et al., 2009).
This technique is implemented in the present paper as a parent
Step 4: At this point, the genes that correspond to the indices of selection strategy. Referring once again to the aforementioned
Step 3 are crossed over between the parent chromosomes, example and considering the following population of chromo-
resulting in two new chromosomes, namely the children somes, the roulette wheel from which they would be sampled is
chromosomes. depicted in Fig. 3.
As seen in Table 5, Chromosome 2 has the lowest objective
value, which corresponds to the highest fitness value, and there-
fore holds the greater share of the wheel depicted in Fig. 3. Chro-
mosome 5 has the highest objective value, or lowest fitness
value, and so it has the lowest probability of being selected as a
parent since it is assigned the smallest share of the wheel.
After selecting the parents and recombining them to produce
the new population, the fitness values of the individuals from the
new populations must be determined. If fewer individuals are pro-
duced than the initial size of the population, then this fractional
Step 5: The sum of genes in the offspring will most likely not be difference between the old and new population sizes is called the
equal to the number of bays, therefore constituting the off- generation gap. In order to maintain the size of the original popu-
spring infeasible solutions. If the new chromosome presents a lation, the new individuals are reinserted into the old population.
surplus of bays, this will be subtracted from the gene with the The most apparent strategy of replacing members of the old popu-
highest number of bays (in order to ensure that there will be lation by new members is to replace the least fit members deter-
no negative genes) and similarly, in the chromosome with ministically. Another strategy replaces the oldest members of the
fewer bays, the required extra bays will be randomly added to population and is considered a successful replacement scheme. In
one of the genes. fact, it resembles the natural evolution of individuals and it ensures
122 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123

Population Size = 180

1.4 6

1.2 5
1

Time (seconds)
4

Gap (%)
0.8
3
0.6
2
0.4
Time
0.2 1
Gap
0 0
20 70 120 170
Fig. 3. Roulette wheel for chromosomes of Table 5. Number of Generations

Fig. 5. Sensitivity analysis of time and gap, with respect to number of generations,
for a fixed population size.
Table 5
Example of chromosomes.

Objective value
Chromosome 1 4 4 4 2 3 3 9 Table 6
Chromosome 2 3 2 3 4 3 5 8.36 Input data for tested problem instances.
Chromosome 3 5 3 4 3 3 2 9.2
Chromosome 4 2 3 4 4 3 4 9.72 Tested Number Number of Total number Total number of
Chromosome 5 1 3 3 4 6 3 14.76 problems of ships cranes of bays containers
Problem instance 1 2 2 131 1475
Problem instance 2 3 2 200 3300
Problem instance 3 3 2 103 2000
Problem instance 4 4 3 230 4225
Number of Generations = 100 Problem instance 5 5 3 400 6275
0.8 25 Problem instance 6 5 4 385 5525
Time Problem instance 7 6 3 210 2975
0.7
Gap 20 Problem instance 8 6 3 516 7175
0.6 Problem instance 9 6 4 392 5125
Time (seconds)

Problem instance 10 7 4 320 4100

Gap (%)

0.5 15 Problem instance 11 8 3 400 6450

0.4 Problem instance 12 8 5 289 3325
0.3 10 Problem instance 13 9 6 220 2700

0.2
5
0.1
0 0
trade-off between time and optimality gap is noted for a popula-
20 70 120 170 tion size of 180.
Population Size Using this population size, tests were run now for a ranging
number of generations, and results are reported in Fig. 5. The cho-
Fig. 4. Sensitivity analysis of time and gap, with respect to population size, for a sen number of generations for tests was chosen to be 180. Given
fixed number of generations. that the sensitivity analysis was conducted on a very large problem
instance, the results are expected to be better for regular problem
sizes.
that for an individual to survive successive generations, it must be
After setting the population size to 180 individuals and the
sufficiently fit to achieve propagation into future generations
number of generations to 180, testing was performed on several
(Chipperfield & Fleming, 1995). The present paper employs a sim-
problem instances, ranging from small to large ones. Again, each
ple strategy of maintaining in the population the chromosomes
problem instance was run 10 times and average values were taken,
with the highest fitness value.
in order to reduce the level of randomness in the results. The
parameters of each problem (input data) are summarized in
5. Computational analysis Table 6, while the results are presented in Table 7.
The optimal objective values produced by the GA were com-
In this section the results of the computational analysis are pre- pared to the exact optimal value, wherever possible. This optimal
sented. Several problem instances were solved using the proposed value was obtained by solving the problem using the commercial
GA and these solutions were compared to those produced by an software GAMS, with default solvers. Where it was not possible
exact technique. The proposed GA was constructed and run in to reach the optimal value, the results from the GA were compared
MATLAB. An initial sensitivity analysis was conducted before to the lower bound obtained from an exact-based heuristic. This
deciding on the population size and number of iterations to be was the case in problem instances 3, 6, 10, 11, 12 and 13. The
used. Firstly, for a fixed number of 100 generations (iterations) of gap between the upper bound given by the GA and the lower
the algorithm, the time and gap of the generated solution were bound given by the exact-based heuristic is shown in Table 7. Fur-
reported, for a large problem instance. These values were the thermore, the time required by the GA is presented alongside the
acquired averages over 10 runs, in order to try and eliminate any time required to solve the problem exactly.
bias due to random initial population generation. Fig. 4 demon- It can be seen from Table 7 that the proposed GA achieves opti-
strates the results, which indicate the increase in time and reduc- mal solutions, with a gap of 0% for small and medium sized prob-
tion in gap with the increase of population size. A reasonable lems (problem instances 1–9), while there is a noticeable gap for
 A. Diabat, E. Theodorou / Computers & Industrial Engineering 73 (2014) 115–123 123

Table 7
Summary of computational results for tested problem instances.

Tested problems Objective from GA (h) Exact solution objective (h) Gap (%) GA time (s) Exact solution time (s)
Problem instance 1 29.52 29.52 0 0.24 1.77
Problem instance 2 63.81 63.81 0 0.24 2.61
Problem instance 3 40.95 40.95 0 0.25 3.79
Problem instance 4 82.86 82.86 0 0.27 3.79
Problem instance 5 32.38 32.38 0 0.27 4.2
Problem instance 6 101.9 101.9 0 0.31 9.27
Problem instance 7 45.71 45.71 0 0.33 3.28
Problem instance 8 95.24 95.24 0 0.5 4.4
Problem instance 9 48.57 48.57 0 0.65 5.66
Problem instance 10 34.76 34.29 1.37 0.79 6.83
Problem instance 11 82.38 80.95 1.77 0.82 9.91
Problem instance 12 30.48 28.57 6.69 0.95 37.3
Problem instance 13 18.1 16.19 11.8 1.23 38.28

larger problem instances. For problem instances 3 and 6 it was not tional circumstances should be considered, such as a non-constant
possible to obtain the exact solution from the aforementioned crane productivity rate, to account for situations in which handling
exact-based heuristic. However, the fact that the upper bound pro- tasks must be performed faster in order to meet an agreed ship
vided by the GA and the lower bound from this technique are departure time. Also, the model could be extended to account for
equivalent, implies that this is in fact the optimal solution. Compu- both loading and unloading operations, as well as to incorporate
tational time is very low, <1 second for almost all problem sizes, crane travel times between bays. Another case would be to evalu-
which is the major advantage of the proposed GA, especially when ate whether or not unidirectional crane schedules generate better
compared to the time required for the exact solution, and compen- solutions than cases where cranes can travel in both directions.
sates for the slight increase in the optimality gap for larger prob-
lem sizes. References

Bierwirth, C., & Meisel, F. (2009). A fast heuristic for quay crane scheduling with
6. Conclusions interference constraints. Springer.
Bierwirth, C., & Meisel, F. (2010). A survey of berth allocation and quay crane
In this paper we developed a mathematical formulation for the scheduling problems in container terminals. European Journal of Operational
Research, 202(3), 615–627.
integrated QCASP which considers quay crane positioning condi- Chipperfield, A.J. & Fleming, P.J., (1995). The MATLAB genetic algorithm toolbox.
tions, such as end-bay positioning and non-crossing constraints. Applied Control Techniques Using MATLAB, IEE Colloquium on. (pp. 10/1–10/4),
Furthermore, a Genetic Algorithm (GA) was developed to solve http://dx.doi.org/10.1049/ic:19950061.
Choo, S., Klabjan, D., & Simchi-Levi, D. (2009). Multiship crane sequencing with yard
the problem and the chromosome representation, the genetic
congestion constraints. Transportation Science, 44(1), 98–115.
operators as well as the selection and insertion strategies are Chung, S. H., & Choy, K. L. (2012). A modified genetic algorithm for quay crane
presented. scheduling operations. Expert Systems with Applications, 4213–4221.
The most important characteristic of the developed model is the Daganzo, C. F. (1989). The crane scheduling problem. Transportation Research Part B:
Methodological, 23(3), 159–175.
integration of the assignment and scheduling problem for quay Guan, Y., Yang, K.-H., & Zhou, Z. (2010). The crane scheduling problem: Models and
cranes, which yields better results than solving these problems solution approaches. Annals of Operations Research, 203(1), 119–139.
independently. Furthermore, rather than significantly increasing Hiassat, A. (2012). An integrated location-inventory-routing problem: Metaheuristics
and environmental extensions an integrated location-inventory-routing problem:
the model complexity as expected during integration, this formu- Metaheuristics and environmental extensions. Masdar Institute of Science and
lation remains simpler than the scheduling problem alone. For Technology.
the GA it was possible to develop problem-specific chromosome Jung, D. H., Park, Y., Lee, B. K., Kim, K. H. (2006). A quay crane scheduling method
considering interference of yard cranes in container terminals (pp. 461–471).
representation and genetic operators in order to lead to an efficient Kim, K. H., & Park, Y.-M. (2004). A crane scheduling method for port container
algorithm. The performance of the GA in terms of solution quality terminals. European Journal of Operational Research, 156(3), 752–768.
was compared to the exact solution. Lee, D.-H., Qiu Wang, H., & Miao, L. (2008). Quay crane scheduling with handling
priority in port container terminals. Engineering Optimization, 40(2), 179–189.
Computational results indicate that the GA produces solutions Legato, P., Mazza, R. M., Trunfio, R. (2008). Simulation-based optimization for the
with a small gap from the optimal solution. This is true especially quay crane scheduling problem. In Proceedings of the 2008 winter simulation
in the case of small and medium-sized problems, where optimal conference.
Liu, J., Wan, Y., & Wang, L. (2006). Quay crane scheduling at container terminals to
solutions are obtained. In the case of one of the larger problem
minimize the maximum relative tardiness of vessel departures. Naval Research
instances, the gap reaches 11%, but the fact that computational Logistics, 53(1), 60–74.
time remains very low renders the proposed GA attractive even Meisel, F. (2011). The quay crane scheduling problem with, time windows, May, 2011.
in these cases. Moccia, L., Cordeau, J., Gaudioso, M., Laporte, G. (2005). A branch-and-cut algorithm
for the quay crane scheduling problem in a container terminal (pp. 1–37).
The main conclusion of this paper is that the GA constitutes a Tavakkoli-Moghaddam, R., Makui, a., Salahi, S., Bazzazi, M., & Taheri, F. (2009). An
successful approach for the QCASP, generating near-optimal solu- efficient algorithm for solving a new mathematical model for a quay crane
tions for small and medium-sized problems and performing extre- scheduling problem in container ports. Computers and Industrial Engineering,
56(1), 241–248.
mely efficiently in terms of time for all problem sizes. It is highly Zhang, H., & Kim, K. H. (2009). Maximizing the number of dual-cycle operations of
recommended that future work pursues the implementation of quay cranes in container terminals. Computers and Industrial Engineering, 56(3),
GAs for the integrated QCASP. To further advance the matter, addi- 979–992.