You are on page 1of 21

OR Spectrum (2008) 30:675–695

DOI 10.1007/s00291-007-0111-6

REGULAR ARTICLE

An optimal layout of container yards

Kap Hwan Kim · Young-Man Park · Mi-Ju Jin

Published online: 24 November 2007


© Springer-Verlag 2007

Abstract This study presents a method for designing the layout of container yards
where transfer cranes and yard trucks are used for stacking containers in the yard and
delivering them between the yard and the quay. A method to determine the layout
type, the outline of the yard, and the numbers of vertical and horizontal aisles is
suggested. For estimating the effects of the design variables on operational cost terms,
formulas for estimating the number of relocations by yard cranes and the expected
travel distance of trucks are derived. Numerical examples are provided to illustrate
the design procedure. It was also shown that the layout in which blocks are laid out
parallel to the side with the gate or the quay outperforms that in which blocks are laid
out perpendicular to the side with the gate or the quay.

Keywords Facilities planning and design · Container terminals · Yard layout

1 Introduction

The layout of a container yard is an influential factor in the productivity of the container
handling operations. For designing the layout of a container yard, the layout type, the
outline of the yard, and the number of aisles (equivalently, rows and columns of blocks)
must be determined. The effects of the design variables on the expected travel distance
of trucks and the expected number of relocations are analyzed.

K. H. Kim (B) · M.-J. Jin


Department of Industrial Engineering, Pusan National University,
Changjeon-dong, Kumjeong-ku, Busan 609-735, South Korea
e-mail: kapkim@pusan.ac.kr

Y.-M. Park
Department of Management Science, Korea Naval Academy,
Jinhae 645-797, South Korea

123
676 K. H. Kim et al.

Related to the expected travel distance, Larson and Odoni (1981) proposed
methods—based on geometrical probability—for deriving the probability distribu-
tion of travel distance between two points randomly located on a line or plane.
Bozer and White (1984); Foley and Frazelle (1991), and Eynan and Rosenblatt (1994)
analyzed the travel time of automated storage and retrieval machines in automated
storage and retrieval systems. The travel distance models in this paper are different
from those of previous studies in that some aisles are restricted to be unidirectional
and the number of aisles is considered in the models.
Bassan et al. (1980) and Hall (1993) studied the layout of shelves or racks in two-
dimensional space. However, characteristics of the design problems for warehouses
were different from those of container yards in their methods of estimating the expected
travel distances and in that the latter considers the relocation cost while the former did
not.
Related to container yards, Castilho and Daganzo (1993) analyzed various ope-
rational strategies for import container yards. They proposed a method to estimate
the number of relocations for picking up a container from import container stacks.
Taleb-Ibrahimi et al. (1993) analyzed various space allocation methods for export
containers. Kozan (2000) proposed a network model to describe the container flows
in port container terminals. It was attempted to determine the flows of different types
of containers in a way that the total handling cost is minimized. Kim and Kim (2002)
discussed a method to simultaneously determine the amount of space and the number
of transfer cranes (TCs) for import container yards. However, most studies on stacking
yards in container terminals are related to operational problems (Zhang et al. 2003; Lee
et al. 2006; Dekker et al. 2006) and no research on designing stacking yards is known
to the authors. Although there have been many researches on the efficient operation of
vehicles in container terminals (Vis and Harika 2004; Yang et al. 2004; Grunow et al.
2004, 2006; Briskorn et al. 2006), all the studies assumed that the layout of the yard
is given.
This study considers two typical layout types for container yards: the parallel layout
(refer to Fig. 1) in which blocks are laid out parallel to the gate or the berth; the

outer lane gate block

bay transfer lane TC


YT

driving
a module lane

berth

Fig. 1 An illustration of a container yard with a parallel layout of blocks

123
An optimal layout of container yards 677

e w/2

Fig. 2 Lanes and container stacks under TCs

perpendicular layout in which blocks are laid out perpendicular to the gate or the berth.
This study addresses container yards where TCs are the primary handling equipment
in the container yard. Figure 2 illustrates lanes and container stacks when TCs are
used. Most of container terminals with TCs in the yard usually have parallel layouts.
However, we can find examples of parallel layouts like the yard in Deltaport of TSI
System Inc. in Vancouver.
The handling operations involves loading and unloading operations associated with
containerships, receiving/delivery operations for outside road trucks, and storage ope-
rations in a yard. During the discharging operations, quay cranes (QCs) unload contai-
ners from the ship to internal yard trucks. Then, the yard trucks deliver the inbound
containers to TCs that pick them up and stack them in the yard at some positions.
Then, the empty yard trucks move back to the QCs to receive the next discharged
containers. For the loading operation, the process is carried out in the opposite direc-
tion. Yard trucks repeat a loaded travel and an empty travel alternately (called single
cycle operation) between a quay crane and TCs, because they are usually assigned to
a single QC (dedicated assignment) and QCs start the loading operations only after
all the unloading operations are completed. However, when yard trucks are shared by
more than one QC (pooled dispatching) or a QC mixes the discharging and the loading
operations alternately (dual cycle QC operation), both of which can be found rarely
in practice, empty travels may be significantly reduced.
During receiving and delivery operations, when an outside road truck arrives at a
container terminal, it is routed to a transfer point of the yard. When the outside road
truck arrives at a transfer point of the yard, a TC either receives a container from the
road truck, which is called the “receiving operation,” or delivers a container to the
road truck, which is called the “delivery operation.”
Three unique characteristics of travels by trucks in container yards can be summa-
rized as follows:
(1) The aisles for trucks in the container yard are classified into two types: lanes
for the container transfer and lanes for driving. A truck moves into a transfer
lane, which is unidirectional, only when the truck needs to transfer a container

123
678 K. H. Kim et al.

2.7m e

Track for TCs

6..36m

w
Fig. 3 The layout and dimensions associated with each module

C B C B F
D

E
A A G
(a) (b)
Fig. 4 An illustration of travels by a yard truck

from/into a block in either side of the transfer lane. Trucks usually move on a
driving lane, which is bi-directional, to travel from one location to another. In
Fig. 1, horizontal lanes between two adjacent blocks are transfer lanes, while
vertical lanes and outer lanes in the yard are driving lanes.
(2) Trucks can transfer containers only at one side of a block because of the design
of TCs (see Fig. 2). And, each module contains two blocks that are separated by
transfer lanes (see Figs. 1, 2, 3). A module is the basic unit of the layout. TC is
the most popular type of yard crane in Asian countries, and similar approaches
may be applied to the other types of yard cranes.
(3) A U-turn is impossible for a truck, even in driving lanes, due to the large turning
radius of a truck, compared with the width of the lanes.
Because yard trucks have to move around blocks, the expected travel distance
becomes longer, as the size of blocks become larger. For example, suppose that a yard
truck is assigned to a QC positioned at A. Then, it starts the travel from position A,
picks up a container at position B, and returns to position A again. On the travel path
in Fig. 4a, the yard truck travels along the route A → C → B → D → E → A, while
it moves along A → C → B → F → G → A on the travel path in Fig. 4b. The travel
distance in Fig. 4b is shorter than the one in Fig. 4a, although the locations of point A
and B are the same in both the figures.
Figures 5 and 6 illustrate travel routes of trucks for the parallel layout and the
perpendicular layout, respectively. The notation “v” represents the width of a driving
lane in the parallel layout, while it represents the width of a transfer lane in the

123
An optimal layout of container yards 679

(a) A module (b) driving lane

transfer lane

v A
A
h

Gate Berth
B B

Fig. 5 Travels on parallel layouts. a Travel between the gate and the yard of the parallel layout. b Travel
between the berth and the yard of the parallel layout

(a) A module transfer lane


(b)

v
A A

h driving lane

Gate
B Berth
B
Fig. 6 Travels on perpendicular layouts. a Travel between the gate and the yard of the perpendicular
layout. b Travel between the berth and the yard of the perpendicular layout

perpendicular layout. The notation “h” represents the width of a transfer lane in the
paralleled layout, while it represents the width of the driving lane in the perpendicular
layout.
Road trucks travel between the storage yard and the gate which is assumed to be
located in the center of a side of the rectangular yard. In container terminals, during the
loading and the unloading operations, internal yard trucks travel between the storage
yard and the berth. Although there may be cases where these assumptions about the
location of the gate and the configuration of the berth are not valid, an analysis similar
to the procedure described in this study can be followed to obtain similar results.

123
680 K. H. Kim et al.

This study proposes ways to decide which layout type and what sized blocks should
be used for a given size of a rectangular shaped yard. These decisions about the layout
will affect the throughput rate of the ship operation and the delivery and receiving
operation as well as the storage capacity of the container yard.
The next section defines the layout problems and suggests a decision procedure for
the layout problem. Section 3 provides various formulas for estimating the expected
number of relocations and the expected travel distance for each type of a yard layout.
Section 4 provides numerical examples of the layout problems. The final section
provides concluding remarks.

2 Optimizing the layout of the yard

The decision issues for the layout planning are as follows: (1) to decide the layout type
of the yard; (2) to determine the outline of the yard, which is represented by the lengths
of two sides of the rectangular yard; and (3) to determine the number of driving aisles
(lanes), equivalently, the length of blocks in the yard. As the length of blocks becomes
shorter, meaning that the number of aisles increases, the expected travel distance of
trucks is reduced. However, in this case, because the area for the aisles becomes larger,
the area for stacking containers decreases. To stack the same amount of containers in a
smaller area, the stacking height must be increased, which results in a higher number
of relocations during retrieval operations. Thus, there must be an economical length
of a block that minimizes the total handling cost.
The following assumptions are introduced for formulating the layout problem:
1. Yard trucks in container terminals deliver containers in a single-cycle mode, in
which they repeat a loaded travel and an empty travel, alternately, between transfer
positions under QCs and storage positions in the yard.
2. Each outside truck has a single delivery or receiving order, which implies that
each truck moves from the gate to a designated position in the yard and moves
back directly to the gate.
3. The entire area of the yard is uniformly used for the storage of containers. For
the berth, the transfer operation by a QC occurs uniformly over the entire range
of the berth. When the number of berths and the size of the yard become large,
then the yard is decomposed into several partitions and each partition is assigned
to each berth for the storage of containers. Even in the case, results in this paper
can be applied by evaluating the expected travel distance of trucks in Sect. 4 over
the partition instead of the entire yard.
This paper considers costs of trucks and TCs for determining the optimal confi-
guration of a layout. The costs of trucks or TCs include the labor cost, the fuel cost,
the maintenance cost, and the overhead. For constructing the objective function, we
considered the effects of the layout decision variables on the operation time of TCs as
well as that of trucks.
In container yards as assumed in this paper, a TC can move from a block to another,
which is the common case in Asian ports. Thus, each TC is deployed to a limited area
in the yard for serving trucks. Travel distances of TCs are not sensitive to the size of

123
An optimal layout of container yards 681

blocks but depend on the total number of TCs deployed to the yard and the total length
of blocks, the former of which is not an issue of this paper and the latter of which can
be considered to be constant.
Consider a delivery operation by a truck. The operation time of the truck consists
of the round-trip travel time between the gate and a TC, the waiting time under the TC,
and the transfer time for receiving a container from the TC. Also, the operation time
of a TC consists of the traveling time to the transfer position of the next container,
the waiting time for the arrival of the corresponding truck, the relocation time for
containers on the top of the next container, and the transfer time of the next container
to the truck.
The transfer time of a container by a TC to a truck is independent of the layout of
blocks in the yard. The waiting time of the truck or the TC depends on the number
of TCs and the arrival rate of trucks, which are not the main issues of this paper. The
average travel distance of a TC does not directly depend on the number of columns of
blocks. Thus, among all the time elements of trucks and TCs, only two time elements—
the travel time of trucks between the gate and TCs and the time for TCs to relocate
containers on the top of the next container—depend on the yard layout. For other types
of operations, similar arguments can be provided.
The following notations are used to describe the procedures for optimizing yard
layouts:

tt The travel time of yard trucks per meter (s).


tr The time required for the relocation of a container (s).
ct The cost of a truck per second. This cost includes the overhead cost and
the operating cost (including costs for labor, fuel, and maintenance).
cr The cost of a TC per second. This cost also includes the overhead cost
and the operating cost of a TC.
a The area allocated for the container yard.
e The width of a block (m) (refer to Fig. 2). e = 16.2 m in the numerical
example of Sect. 5.
w The width of a module (m) (refer to Fig. 3). w = 50.88 m in the numerical
example of Sect. 5.
h The width of a horizontal aisle between blocks (m). 20 m (parallel layout)
and 26 m (perpendicular layout) were used in the numerical examples.
Refer to Figs. 5a and 6a.
v The width of a vertical aisle. 26 m (parallel layout) and 20 m (perpendi-
cular layout) were used in the numerical examples. Refer to Figs. 5a and
6a.
s Space required for a TEU, including the allowance between containers
(m2 ). 17.172 m2 was used in the numerical example.
c The average number of containers stacked in the yard (TEU).
P The available ground space for stacking containers (m2 ).
G The total number of ground slots in a container yard (TEU).
A The vertical length of the yard (m) (a decision variable)
B The horizontal length of the yard (m) (a decision variable)

123
682 K. H. Kim et al.

N The number of columns of blocks in the layout (a decision variable).


N = 3 in Figs. 5 and 6.
M The number of rows of blocks in the layout (a decision variable) M = 4
and 3 in Figs. 5 and 6, respectively.
T The average number of tiers per stack, i.e., the average number of contai-
ners standing on top of each other. It is fixed when values of M and N
are determined. Figure 2 illustrates stacks whose value of T is 3.5.
dg (M, N ) The expected round-trip travel distance of road trucks between the gate
and a random position in the yard for given values of M and N
db (M, N ) The expected round-trip travel distance of yard trucks between a random
position of the berth and a random position in the yard for given values
of M and N

2.1 Optimizing the layout of off-dock container yards (ODCYs)

This paper addresses the layout design problem not only for port container terminals
but also for ODCYs, container yards which are located outside port container terminals
and used for storing temporarily containers before they are moved to port container
terminals or delivered to customers by road trucks.
The area allocated for the yard is assumed to be given and thus either of the vertical
length, A (or equivalently M) or the horizontal length, B (or equivalently N ) of the yard
is determined, the value of the other decision variables is automatically determined.
Thus, the number of the decision variables in this problem is one.
One of the cost terms is the cost of trucks for one container, which is incurred by
the travel of trucks in the yard. The travel cost is proportional to the travel distance of
trucks. ODCYs are assumed to have just one gate. Trucks travel from the gate to the
yard and then return to the gate.
The other cost term is the cost of transfer cranes resulting from the relocations
of containers. Although trucks need to travel twice—once for delivery and once for
retrieval—to a position in the yard for a container, the relocation by a TC occurs once
for a container. Thus, the relocation cost term was multiplied by 0.5 in the objective
function. Thus, the problem can be stated as follows:

Minimize 0.5cr tr R(α |M, N ) + ct tt dg (M, N )


M,N
subject to A × B = a, (1)

where R(α |M, N ) is the expected number of relocations when the number of rows
of the block is α for given values of M and N .
In the following, it will be discussed how to estimate the values of R(α |M, N ) and
dg (M, N ) given values of M and N . Once the methods are provided, the values of M
and N minimizing the objective function (1) can be easily obtained by enumerating
all the possible combinations of M and N .
In the parallel layout, the value of M indicates the number of blocks in the vertical
direction of the yard. Because blocks are laid out in the unit of a module, the number
of blocks will be an even number. When the value of M is fixed, the value of A

123
An optimal layout of container yards 683

is fixed. As a result, from A × B = a, the value of B is also fixed. That is, A =


(e × M) + (0.5 × h × M) + 2h and B = a/A. For a given value of B, as the value of N
(the number of columns of blocks) increases, the number of vertical aisles increases,
which results in higher stacks and shorter travel distances in the yard.
In the perpendicular layout, the value of N indicates the number of blocks in the
horizontal direction of the yard. That is, from constraint A × B = a, B = (e × N ) +
(0.5 × v × N ) + 2v and A = a/B. Thus, for enumerating the solutions, the value of
N will be changed by two. For a given value of N (that is, for a given value of A),
various values of M, which will affect the number of aisles, will be enumerated.

2.2 Optimizing the layout of container terminals

Container terminals have both the gate and the berth unlike ODCYs which have only
a gate. The objective function of the layout problem for container terminals is the
same as that for ODCYs. However, the expected travel distance of trucks in container
terminals must be estimated for travels not only between the gate and the yard but also
between the berth and the yard.
However, the length of the berth in container terminals, which is the same as the
value of B, is usually determined by the average length of vessels and the number
of berths required to accommodate a given arrival rate of vessels. Thus, we assume
that B is given before we design the layout of the yard. Because we assume that the
total area of the yard is given, the value of A is also fixed. Thus, the only decision
variable to determine is N for the case of the parallel layout and M for the case of
perpendicular layout. Where β denotes the proportion of transshipment containers
among all containers, the problem may be formulated as follows:
(Parallel layout)

Minimize 0.5cr tr R(α |M, N ) + βct tt db (M, N ) + (1 − β)ct tt


N
{dg (M, N ) + db (M, N )}/2
for a given M, A, andB, (2)

(Perpendicular layout)

Minimize 0.5cr tr R(α |M, N ) + βct tt db (M, N ) + (1 − β)ct tt {dg (M, N )


M
+db (M, N )}/2 for a given N , A, and B. (3)

Note that the objective functions (1) and (2) assume that not only inbound containers
but also outbound containers are retrieved in random order. However, in many cases,
the retrieval sequence of outbound containers is carefully pre-determined for reducing
the number of relocations, in which case the coefficient of the first term in (2) and (3)
can be adjusted between 0.25 and 0.5.
When outside trucks belong to trucking companies which are not the terminal
operator, the travel cost for the outside trucks may be excluded from the objective
functions, (2) and (3). However, considering the short turnaround time of outside

123
684 K. H. Kim et al.

trucks in terminals is one of the important criteria for evaluating the competitiveness
of terminals, they may be included in the objective function even in that situation.

3 Estimating the expected number of relocations for picking up a container

This section provides formulas for evaluating storage capacities of container yards and
for estimating the number of relocations for a given storage requirement and layout
of a container yard.
The area of the ground on which containers can be stacked is expressed as (4) in
case of the parallel layout and (5) in case of the perpendicular layout. The area for
stacking containers decreases as the number of driving lanes (aisles) (N , in case of
the parallel layout; M, in case of the perpendicular layout) increases. From Figs. 3, 5,
and 6, it follows that

Pa = e × M × B − e × M × v × (N + 1) (4)
Pe = e × N × A − e × N × h × (M + 1) (5)

In (4), the first term is the total area in a parallel layout excluding the area for all the
horizontal aisles which is represented as dotted area in Fig. 7. The second term in (4)
is to subtract areas with slanting lines from the remaining area in Fig. 7. Equation (5)
can be derived in a way similar to that for (4). Also, the number of ground slots and
the expected number of tiers of stacks can be calculated as follows:

e v

N=2, M=3

Fig. 7 Partitions of the yard for calculating Pa

123
An optimal layout of container yards 685

Pa Pe
G= or (TEUs), (6)
s s
T = c/G (TEUs). (7)

As the height of stacks increases, the stacking capacity increases proportionally.


However, the number of relocations required for retrieving a container also increases.
The following discusses a method for estimating the number of relocations for retrie-
ving a container.
Suppose that the number of rows in a yard-bay is α. Because the average height
of stacks is T , the maximum height of stacks, which is the height of stacks just
before the retrieval operation begins, can be assumed to be 2T . As the retrieval of
containers progresses, the height of stacks decreases. As a result, the expected number
of relocations to pick up a random container from the yard-bay decreases. The average
value of the average number of relocations, which are required to pick up a container,
over all (2T × α) retrievals can be evaluated as follows (Kim 1997):

R(α|M, N ) = (2T − 1)/4 + (T + 1)/(8α), (8)

where T can be evaluated by (7).


Note that formula (8) can be derived by summing the expected number of relocations
to pick up an arbitrary container from a yard-bay with (2T × α) containers, that from
a bay with (2T × α) − 1 containers, that from a bay with (2T × α) − 2 containers,
. . ., and that from a bay with 1 container. For detailed derivation of (8), refer to Kim’s
paper (1997). This paper assumes that there is no pre-schedule to reduce relocations.
As the values of M and N increase, the area for stacking containers becomes smaller
because of the larger space consumed by aisles. For a given yard area, this means that
the stacking height must be increased, which results in a higher expected number of
relocations.

4 Derivation of expected travel distances of trucks

This section derives formulas for the expected travel distances of yard trucks for
various container yard layouts. The following assumptions were introduced to derive
the formulas for the expected travel distance:
1. The outmost lane of the yard is bi-directional.
2. The width of aisles is ignored in the evaluation of the travel distance.
3. The gate is located in the middle of a side of a rectangular yard.
Trucks travel along horizontal and vertical aisles and thus the travel distance can be
represented by the sum of the horizontal travel distance and the vertical travel distance.
We will derive the expected values of the travel distance in the horizontal direction
(X N ), which is a function of the number of columns of blocks, and that in the vertical
direction (Y M ), which is a function of the number of rows of blocks in the yard. Let
f N (X N ) be the probability density function of X N and g M (Y M ) be the probability
density function of Y M .

123
686 K. H. Kim et al.

The expected travel distance of yard trucks can be calculated as in the following
subsections.

4.1 Travel distance between the gate and the yard of the parallel layout (AG)

Because X N and Y M are independent, the expected travel distance can be represented
by

dg (M, N ) = E(X N + Y M ) = E(X N ) + E(Y M ). (9)

If we ignore the discrete nature of the blocks for the simplification, then,
because the travel distance in the vertical direction follows a uniform distribution,
U (0, 2 A), g M (Y M ) = 1/(2 A) and E(Y M ) = A. And, formulas for f N (X N ) and
E(X N ) can be derived as follows:
Refer to routes in Fig. 5a. Suppose that the number of columns (N ) is odd. If the
transfer location in the yard is in the central module (k = 1), then X N = 2B/N and
the probability that a pickup is performed at the central module is 1/N . When the
transfer location in the yard is in the kth module (k = 2, 3, . . . , (N + 1)/2) from the
location of the gate, the travel distance becomes (2k − 1)B/N and the probability is
2/N . Note that there is one kth module on both the right- and the left-hand sides of
the location of the gate. Thus, finally, the following results may be obtained:

P{X N = 2B/N } = 1/N , (10)


and P{X N = (2k − 1)B/N } = 2/N , for k = 2, 3, 4, . . . , (N + 1)/2. (11)

N +1
 2  N +1  (N +1)2
Thus, E(X N ) = 2(2k−1)B
N2
= 2B
N2 2 ( N 2+1 + 1) − N +1
2 = 2N 2
B.
k=1
Finally,

(N + 1)2
dg (M, N ) = B + A. (12)
2N 2

When the number of columns is even, the following results may be obtained by
using a similar derivation procedure:

P(X N = 2k B/N ) = 2/N , for k = 1, 2, 3, . . . , N /2. (13)


(N + 2)
Thus, dg (M, N ) = B + A. (14)
2N

4.2 Travel distance between the berth and the yard of the parallel layout (AB)

The expected travel distance in the vertical direction for AB is the same as in the
case of AG. The expected travel distance in the horizontal direction can be derived as
follows: The berth is partitioned into N segments where N corresponds to the number

123
An optimal layout of container yards 687

of columns. In each segment of the berth, it was assumed that all the pickups and
deliveries occur at the center of the segment.
When N = 1, the travel distance is 2B. P(X N = 2B) = 1. When N > 1, the
probability that a randomly chosen segment is in the same column as a randomly chosen
module is 1/N , and the travel distance in this case is 2B/N . Thus, P(X N = 2B
N ) = N.
1

The probability that the difference between the columns of a randomly chosen
segment and a randomly chosen module will be equal to one is 2(N − 1)/N 2 and the
travel distance in this case is 3B/N . Thus, P(X N = 3B/N ) = 2(N −1)/N 2 . Similarly,
the following can be derived: P(X N = 5B/N ) = 2(N − 2)/N 2 , . . . , P(X N =
(2N − 1)B/N ) = 2/N 2 .
Thus, E(X 1 ) = 2B and

2B 1 3B 2(N − 1) (2N − 1)B 2 B


E(X N ) = + + ··· + = (2N 2 + 3N + 1).
N N N N2 N N 2 3N 2
(2N 2 + 3N + 1)
Thus, db (M, N ) = B + A. (15)
3N 2

4.3 Travel distance between the gate and the yard of the perpendicular layout (EG)

E(X N ) and E(Y M ) can be derived in a similar way to that in Sects. 4.1 and 4.2.
Detailed derivations are given in Appendix 1 and the results are
⎧ B

⎪ {M(4 + 3N 2 ) − N 2 − 4} when the number of modules
⎨ 2M N 2
in each row is odd;
E(X N ) = (16)

⎪ (3M − 1) B
, when the number of modules
⎩ 2M
in each row is even.
M +1
and E(Y M ) == A. (17)
M

4.4 Travel distance between the berth and the yard of the perpendicular layout (EB)

The expected vertical travel distance of EB is the same as that of E G. The expected
horizontal travel distance of EB can be also derived in a similar way to that for E G.
A detailed derivation is given in Appendix 2, and the result is

2B
E(X N ) = (1 + 2N 2 ) (18)
3N 2

4.5 Numerical examples

The following illustrates the expected travel distance of trucks for a round-trip. The
area was assumed to be 100 modules each of whose size is 50.88 × 50.88 m2 . Because
the total area is fixed to be 100 modules, in case of the parallel layout, the horizontal
length (B) of the yard decreases as the value of M increases. Thus, E(X N ) of the

123
688 K. H. Kim et al.

750

550

350

14
150
12 10
14 16
18 6 v
20
u 22
24 2
28

Fig. 8 E(X N + Y M |M = v, N = u, EG) − E(X N + Y M |M = u, N = v, AG)

parallel layout changes as the value of M changes. The same argument is true for
E(Y M ) of the perpendicular layout.
Figure 8 compares the difference in the expected round-trip travel distance from the
gate to a random position of the yard between the parallel layout and the perpendicular
layout. Note that the same area for both types of layout is assumed. A comparison
was made between each pair of layouts in which the numbers of rows and columns
of blocks in the parallel layout are the same as the numbers of columns and rows in
the perpendicular layout. Figure 8 shows that, compared to the perpendicular layout,
the parallel layout results in a much shorter travel distance. Similar results have been
observed for the case of the berth, as shown in Fig. 9. Note that the conclusion on the
comparison is valid under the assumption that w = 50.88 m2 which is the width of
a module when rubber-tired gantry cranes are used in blocks with six rows, which is
the most popular case in Korean container terminals.

5 Numerical examples of optimal layouts

This section provides numerical examples of optimal layouts for both the cases of
ODCYs and container terminals.

5.1 Off-dock container yards

This study attempted to find N and M that minimize the objective function of (1). In the
following numerical experiment, it was assumed that tt = 0.20 s which is equivalent to
18 km/h, the average driving speed of trucks in terminals. And it was set tr = 74.19 s

123
An optimal layout of container yards 689

400

350

300

250

14
200 10
12 v
14 6
16
18
20
u 22 2
24
28

Fig. 9 E(X N + Y M |M = v, N = u, EB) − E(X N + Y M |M = u, N = v, AB)

which is the average relocation time collected by a stop watch time study in an actual
site. Because the annual equivalent amount of the purchasing and operating cost of
a TC is approximately 15.56 times larger than that of a truck, it was assumed that
ct = 1 and cr = 15.56. It was assumed that the throughput rate of an ODCY is 5,687
containers (20 ft) per day and a container stays at a yard for 5.7 days on average, which
are the data collected from a container terminal in 2006. We assumed a = 258,877 m2
which is the area of 100 modules. We set α = 6 which is the number of rows for
the most popular type of rubber-tired gantry cranes. It was assumed that v = 26 m in
the parallel layout, and h = 26 m in the perpendicular layout, which is the width of
traveling aisles collected from real terminals.
Figure 10 shows the curves of the expected total cost per container for the parallel
layout. The optimal layout was obtained when N = 2 and M = 10. The optimal
configuration of the container yard was A = 508.8 m and B = 508.8 m. The minimum
cost was 1023 per container. Figure 10 also compares the expected total cost of the
perpendicular layout with that of the parallel layout for various values of (N , M). In
the perpendicular layout, the minimum cost (1278) resulted when N = 14 and M = 1.
The optimal configuration of the yard was A = 712 m and B = 363 m. Note that the
cost of the parallel layout was lower than that of the perpendicular layout.

5.2 Container terminals

This subsection provides an example of a port container terminal. It was assumed that
a = 711,783 m2 , A = 593 m, and B = 1, 200 m. This configuration of the container

123
690 K. H. Kim et al.

7000

6000

Total cost per container 5000

4000

3000

2000

1000 10
8
0 6
12 14 v
16 4
18
u 20 22 2

Perpendicular layout Parallel layout

Fig. 10 The expected total cost of an ODCY of a parallel (M = u, N = v) and a perpendicular layout
(M = v, N = u)

2500
perpendicular layout parallel layout
2000
Cost per container

1500

1000

500

0
1 2 3 4 5 6 7 8 9 10 11 12
v

Fig. 11 The expected total cost of a container terminal of a parallel (N = v) and a perpendicular layout
(M = v)

yard is the same as that of a container terminal in Korea. It was assumed that β = 0.
There is no difference in the solution procedure even when β > 0.
Figure 11 compares the expected total cost of a container terminal for a parallel
layout with that for a perpendicular layout. The minimum cost was obtained when
N = 6 with the cost of 827 for the parallel layout. When N = 6, the length of a
block was 170 m. For the perpendicular layout, the optimal layout was obtained when
M = 2 and the minimum cost was 1,278 for the perpendicular layout. Note also that
the minimum cost of the perpendicular layout is higher than that of the parallel layout.
Figures 12 and 13 show the changes in N ∗ for different values of cr /ct and c. Note
that N ∗ remains the same when values of the parameters exceed some specific values.
The average length of actual blocks in the terminal, from which data in this numeri-
cal example were collected, was 172 m which is almost the same as the optimal value
(170 m) in this paper. Figure 14 shows the distribution of the length of blocks in actual

123
An optimal layout of container yards 691

20
18 ODCY
16

The number of aisle


14 Container
Terminal
12
10
8
6
4
2
0
1 5 10 15 20 25 30
c r /c t

Fig. 12 The changes in the optimal N for different values of cr /ct in a parallel layout

22
20 ODCY
The number of aisles

18
16 Container
14 Terminal
12
10
8
6
4
2
0
0
00

00

00

00

00

00

00

00

00

00
10

20

30

40

50

60

70

80

90

10

c
Fig. 13 The changes in the optimal N for different values of c in a parallel layout

8
Number of terminal

6
5
4
3
2
1

0
100-125 126-150 151-175 176-200 201-225 226-250 251-275 276-300 301-
Average length of blocks

Fig. 14 The distribution of the length of blocks in real terminals

21 terminals of Asian countries, which utilize rubber-tired gantry cranes for blocks of
six rows. Although the length of blocks are scattered over a wide range, blocks of the
length between 176 and 200 m were the most popular and the average length of all the
21 cases was 183 m.

123
692 K. H. Kim et al.

According to Figs. 8 and 9, for every combination of a and b, we found that the
expected travel distance of trucks in the parallel layout with N = v and M = u is
shorter than that in the perpendicular layout with N = u and M = v. Note that in both
the layouts, the area for storing containers is the same, which means that the expected
number of relocations to pick up a container is the same in both the layouts. Thus, it
naturally follows that the parallel layout outperforms the perpendicular layout, which
could be observed in Figs. 10 and 11.

6 Conclusions

This study proposes a method for determining the layout of container yards. Two types
of layouts are considered: the layout where blocks are laid out parallel to the gate or the
berth (parallel layout); the layout where blocks are laid out to the direction of the gate
or the berth (perpendicular layout). For evaluating layouts of a yard, the travel cost
of trucks and the relocation cost of transfer cranes are considered as objective terms.
Procedures for optimizing layouts of container yards are suggested. It was discussed
how to determine the layout type of container yards, the outline of the rectangular
yard, and the number of aisles in the yard.
For evaluating the objective function, this study proposed a method for estimating
the expected number of relocations for picking up a container from a given layout of
the yard. This study also presented formulas for estimating the expected travel distance
of yard trucks for storages and retrievals in a given layout of a container yard.
Illustrations of numerical evaluations were provided using data from actual contai-
ner terminals. The results of the analysis showed that the parallel layout results in a
shorter expected travel distance for the same layout parameters as well as for the best
set of parameters. The total cost, which includes both the travel cost and the relocation
cost, was lower in the parallel layout than in the perpendicular layout.
This study assumed that the shape of the yard is rectangular. However, other confi-
gurations of yards may be observed in practice. Also, it was assumed that the gate is
located in the middle of the rectangle, which is an ideal assumption. The issues such
as the effects of the location of the gate and various yard conditions may be addressed
in future studies.

Acknowledgments This work was supported by the Korea Research Foundation Grant funded by the
Korean Government (MOEHRD) (The Regional Research Universities Program/Research Center for
Logistics Information Technology).

Appendix 1: Derivation of the expected travel distance for the case of E G

Refer to Fig. 6a. Trucks must always go to the end first, then “up” to the appropriate row,
because they cannot travel vertically on the transfer lanes. Let rY be the Y coordinate
of a random location in a yard.
1. Suppose that the number of modules in each row is odd.
(the case N = 2)
When 0 ≤ rY ≤ A/M, X N = B and P(0 ≤ rY ≤ A/M) = 1/M.

123
An optimal layout of container yards 693

When A/M < rY ≤ A, X N = 2B and P(A/M < rY ≤ A) = (M − 1)/M.


(the case N > 2)
When 0 ≤ rY ≤ A/M, X N = B and P(0 ≤ rY ≤ A/M) = 1/M.
When A/M < rY ≤ A, P(A/M < rY ≤ A and X N = B + {(4k − 2)B/N } =
4(M − 1)/(M N ) for k = 1, 2, . . . , (N − 2)/4 and P(A/M < rY ≤ A, X N =
2B) = 2(M − 1)/M N . Thus,

⎧ ⎫

⎨ 4
N −2


1 M − 1 4B  4k − 2 ⎬ 4(M − 1)
E(X N ) = B+ 1+ + B
M M ⎪ ⎩N N ⎪
⎭ MN
k=1

B
= {M(4 + 3N 2 ) − N 2 − 4}.
2M N 2

2. Suppose that the number of modules in each row is even.


When 0 ≤ rY ≤ A/M, X N = B and P(0 ≤ rY ≤ A/M) = 1/M.
 A/M < rY ≤ A, X N =AB + {(4k −2)B/N
When } and
P X N = B + (4k − 2) NB , M  rY  A = 4 (M−1)M N , for k = 1, 2, . . . , N /4.
Thus,

4  
N
B (M − 1) 4  B B
E(X N ) = + B + (4k − 2) = (3M − 1) .
M M N N 2M
k=1

The expected travel distance in the vertical direction can be derived as follows:
When (k − 1)A/M ≤ rY ≤ k A/M, Y = 2k A/M, for k = 1, 2, . . . , M. Thus,

2A 
M
M +1
E(Y M ) = k= A.
M2 M
k=1

Appendix 2: Derivation of the expected travel distance for the case of EB

The expected horizontal travel distance can be estimated as follows: note that there
are N /2 discrete positions of vertical aisles on the horizontal axis where pickups or
deliveries may occur in the yard. And, suppose that the berth is partitioned into N
segments, each of which corresponds to a width of a block and pickups or deliveries
occur at one of the center of n segments. Then, the probability that a pickup or delivery
occurs at one of the possible positions in the berth is 1/N , and the probability that it
occurs at one of the possible positions in the yard is 2/N .
Let X N coordinates of transfer points in a yard and the berth be k y and ks , respec-
tively. If k y + ks ≤ B, then a vehicle travels through the left-most aisle. However, if
k y + ks > B, then a vehicle travels through the right-most aisle. For each combination
of k y and ks , the probability of occurrence is 2/N 2 , and the travel distance must be
twice the one-way travel distance. Thus, the one-way travel distance is multiplied by

123
694 K. H. Kim et al.

0 ky B
yard
B ( N − 1) B
N N

0 ks B

berth
0.5 B 1.5 B ( N − 0.5) B
N N N
Fig. 15 Position of pickups and deliveries in EB

4/N 2 to obtain the expected round-trip distance. Thus, the expected travel distance in
the horizontal direction can be expressed as follows (Fig. 15):
⎧ ⎫
(N −1)B
 N ⎪

(N −0.5)B
N −k y
(N −0.5)B
 N ⎪

4
E(X N ) = 2 (k y + ks ) + (2B − (k y + ks ))
N ⎪
⎩ k = 0.5B (N −0.5)B


k =B
y N s N
B
ks = N −k y + N
(19)

Values that ks and k y can be as follows:

0.5B 1.5B 2.5B (N − 0.5)B


ks = , , ,...,
N N N N
B 3B 5B (N − 1)B
ky = , , ,...,
N N N N
Suppose that notations k y and ks are replaced by y and s for simplifying the ex-
pression as follows:

B N
k y = (2y − 1) y = 1, 2, . . . , ,
N 2
b
ks = (s − 0.5) s = 1, 2, . . . , N .
N
Then, Eq. (19) can be rewritten as follows:
N
⎧ ⎫
2 ⎨ N −2y+1
4    B⎬
N
B
(2y + s − 1.5) + (2B − (2y + s − 1.5)
N 2 ⎩ N N⎭
y=1 s=1 s=N −2y+2

The following equation can be easily obtained:

2B
E(X N ) = (1 + 2N 2 )
3N 2

123
An optimal layout of container yards 695

References

Bassan Y, Roll Y, Rosenblatt MJ (1980) Internal layout design of a warehouse. AIIE Trans 12(4):317–322
Bozer YA, White JA (1984) Travel-time models for automated storage/retrieval systems. IIE Trans
16(4):329–338
Briskorn D, Drexl A, Hartmann S (2006) Inventory-based dispatching of automated guided vehicles on
container terminals. OR Spectrum 28(4):611–630
Castilho B, Daganzo CF (1993) Handling strategies for import containers at marine terminals. Transport
Res Part B 27(2):51–166
Dekker R, Voogd P, van Asperen E (2006) Advanced methods for container stacking. OR Spectrum
28(4):563–586
Eynan A, Rosenblatt MJ (1994) Establishing zones in single-command class-based rectangular AS/RS. IIE
Trans 26(1):38–46
Foley RD, Frazelle EH (1991) Analytical results for miniload throughput and the distribution of dual com-
mand travel time. IIE Trans 23(3):273–281
Grunow M, Günther H-O, Lehmann M (2004) Dispatching multi-load AGVs in highly automated seaport
container terminals. OR Spectrum 26(2):211–236
Grunow M, Günther H-O, Lehmann M (2006) Strategies for dispatching AGVs at automated container
terminals. OR Spectrum 28(4):587–610
Hall RW (1993) Distance approximations for routing manual pickers in a warehouse. IIE Trans 25(4):76–87
Kim KH (1997) Evaluation of the number of rehandles in container yards. Computers Ind Eng 32(4):
701–711
Kim KH, Kim HB (2002) The optimal sizing of the storage space and handling facilities for import contai-
ners. Transport Res Part B 36:821–835
Kozan E (2000) Optimizing container transfers at multimodal terminals. Mathe Computer Modell 31:
235–243
Larson RC, Odoni AR (1981) Urban operations research. Prentice-Hall, New York, pp 80–85
Lee LH, Chew EP, Tan KC, Han Y (2006) An optimization model for storage yard management in trans-
shipment hubs. OR Spectrum 28(4):539–562
Taleb-Ibrahimi M, Castilho B, Daganzo CF (1993) Storage space vs handling work in container terminals.
Transport Res Part B 27(1):13–32
Vis IFA, Harika I (2004) Comparison of vehicle types at an automated container terminals. OR Spectrum
26(1):117–143
Yang CH, Choi YS, Ha TY (2004) Simulation-based performance evaluation of transport vehicles at auto-
mated container terminals. OR Spectrum 26(2):149–170
Zhang C, Liu J, Wan Y-W, Murty KG, Linn RJ (2003) Storage space allocation in container terminals.
Transport Res Part B 37:883–903

123

You might also like