Transportation Research Part E 83 (2015) 158–169

Contents lists available at ScienceDirect

Transportation Research Part E

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

Empty container exchange among liner carriers

Jianfeng Zheng a,⇑, Zhuo Sun a, Ziyou Gao b
a
Transportation Management College, Dalian Maritime University, Dalian 116026, China
b
School of Traffic and Transportation, Beijing Jiaotong University, Beijing 100044, China

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

Article history: In an attempt to reduce the empty container repositioning costs, this paper studies an
Received 22 June 2015 empty container allocation problem considering the coordination among liner carriers.
Received in revised form 21 September We further measure the perceived values of empty container at different ports. The per-
2015
ceived values of empty container at the surplus (deficit) ports are described by the profits
Accepted 22 September 2015
(empty container exchange costs paid) for delivering empty containers. To solve our prob-
lems, we propose a two-stage optimization method. In stage I, liner carriers are guided to
pursue a centralized optimization solution of empty container allocation for all related
Keywords:
Empty container allocation
liner carriers. In stage II, the inverse optimization technique is used to determine the empty
Value of empty container container exchange costs, which are paid to liner carriers for exchanging empty containers
Two-stage optimization method and following the centralized optimization solution. The profits at the surplus ports are cal-
Inverse optimization culated with respect to the empty container exchange costs at the deficit ports. Finally,
numerical experiments on an Asia–Europe–Oceania shipping service network are
discussed.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction

In the liner shipping industry, empty container repositioning is a challenge for liner carriers due to the high costs. Since
1993, empty container movements have constituted about 20% of the total ocean container movements (Song and Dong,
2011). In 2003, the repositioning cost was up to $11 billion (Bonney, 2004), and that in 2010 was about $23.4 billion
(Drewry, 2011; Tran and Haasis, 2015). Song et al. (2005) estimated that the repositioning cost accounts for 27% of the total
world fleet running cost. Because of the trade imbalances between the major trading regions, empty container movements
cannot be avoided completely. However, minimizing these costly activities would considerably reduce the operating costs of
liner carriers. In an attempt to reduce the cost on repositioning empty containers, this paper proposes an empty container
allocation problem considering the coordination among liner carriers, where empty containers of any single liner carrier can
be repositioned to serve the needs of other liner carriers. Hence, empty containers are exchanged among liner carriers.
Similar to slot exchange agreements between liner carriers in a strategic alliance, empty container exchange agreements
can be signed between liner carriers to reduce the repositioning costs of liner carriers. Generally, empty container move-
ments do not generate revenue for liner carriers. In order to motivate liner carriers to follow a centralized optimization solu-
tion of empty container allocation for all related liner carriers, the extra incentives should be introduced to guide liner
carriers. Generally, these incentives can be described by the profits (costs paid) for delivering empty containers from (to)

⇑ Corresponding author.
E-mail address: jfzheng@dlmu.edu.cn (J. Zheng).

http://dx.doi.org/10.1016/j.tre.2015.09.007
1366-5545/Ó 2015 Elsevier Ltd. All rights reserved.
 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169 159

the surplus (deficit) ports, which have a surplus (deficit) of empty containers. This motivates us to measure the perceived
values of empty container at different ports.
There have been many studies related to the empty container allocation problem or the empty container reposition-
ing issue. Crainic et al. (1993) developed two dynamic deterministic formulations and a stochastic formulation for empty
container allocation in a land distribution and transportation system. Cheung and Chen (1998) investigated the dynamic
empty container allocation problem, which is formulated as a two-stage stochastic network model. A stochastic quasi-
gradient method and a stochastic hybrid approximation procedure were applied to solve the problem. Erera et al. (2009)
developed a robust optimization framework for dynamic empty repositioning problems modeled using time–space net-
works. Li et al. (2004) discussed empty container management at a port and derived the optimal pairs of critical policies,
(U, D) for this port. Namely, if the number of empty containers at this port is less than U, empty containers are imported
up to U, or empty containers are exported down to D if the number of empty containers at this port is larger than D.
Later, Li et al. (2007) extended this problem by considering multiple ports. Jula et al. (2006) studied empty container
movements by optimizing the empty container reuse. The dynamic empty container reuse was modeled and optimiza-
tion techniques were developed to optimize the empty container operations. Lam et al. (2007) presented an approximate
dynamic programming approach, in order to obtain the effective empty container relocation strategies. Feng and Chang
(2008) investigated the repositioning of empty containers for intra-Asia liner shipping. Song and Carter (2009) studied
general empty container balancing strategies depending on whether shipping lines are coordinating the container flows
over different routes and whether they are willing to share container fleets. Moon et al. (2010) proposed an empty
container repositioning problem considering leasing and purchasing. To address this problem, they presented a
mixed-integer linear optimization model and developed a genetic algorithm to solve it. Shintani et al. (2010) analyzed
the possibility to save the container fleet management costs in repositioning empty containers by using foldable con-
tainers. Later, Moon et al. (2013) compared the foldable containers with the standard containers on the cost for reposi-
tioning empty containers. Numerical experiments demonstrated the economic feasibility of foldable containers. Di
Francesco et al. (2009) addressed an empty container repositioning problem under uncertainty, where the historical data
were inappropriate for estimating uncertain parameters. In order to solve this problem, a time-extended multi-scenario
optimization model was developed. Later, Di Francesco et al. (2013) studied an empty container repositioning problem
under uncertain port disruptions. Long et al. (2012, 2015) investigated an empty container repositioning problem with
uncertainties, by using a sample average approximation method. Song and Dong (2011) discussed an empty container
repositioning policy with flexible destination ports. Bell et al. (2011, 2013) proposed two types of container assignment
models (i.e., a frequency-based container assignment model and a cost-based container assignment model), in which
both laden containers and empty containers were considered. Recently, Wang et al. (2015) extended these two container
assignment models by proposing several profit-based container assignment models. In addition, readers can refer to the
references in two reviews on empty container repositioning (Braekers et al., 2011; Song and Dong, 2015).
Furthermore, some researchers explored the combined optimization problems in liner shipping with empty container
repositioning. Shintani et al. (2007), Meng and Wang (2011), Song and Dong (2013) investigated the liner shipping network
design problem, as well as considering the repositioning of empty containers. Dong and Song (2009) addressed the joint
problem of container fleet sizing and empty container repositioning. Brouer et al. (2011) studied the cargo allocation prob-
lem with the repositioning of empty containers. Song and Dong (2012) investigated cargo routing, together with empty con-
tainer repositioning. For more optimization problems related to ship routing and scheduling in liner shipping, please refer to
some review papers (Christiansen et al., 2004, 2013; Meng et al., 2014).
To the best of our knowledge, there is currently no optimization model on the empty container allocation problem con-
sidering the coordination among liner carriers. Le (2003) studied a related problem based on a neutral Internet-based
information exchange platform, which may facilitate empty container reuse and sharing empty containers among liner
carriers. Theofanis and Boile (2009) mentioned that empty containers of a liner carrier can be used to match the needs
of other liner carriers. Both of these works discussed the empty container repositioning strategies from a qualitative point
of view.
Obviously, the coordination among liner carriers increases the flexibility of an empty container repositioning system
by exchanging empty containers among liner carriers and offers an opportunity to reduce the repositioning costs of liner
carriers. However, the coordination formation among liner carriers is challenging. Generally, the goal of a liner carrier is
the maximization of its own profit (or minimization of its own cost). Furthermore, some liner carriers may collude to obtain
a larger profit (or a lower cost), as compared with the coordination among more liner carriers. Hence, the coordination
stability among liner carriers is also a challenge. This paper aims to resolve these problems in the study of empty container
allocation considering the coordination among liner carriers. Furthermore, we will measure the perceived values of empty
container at different ports, under the coordination among liner carriers.
The rest of this paper is organized as follows. Notation, assumptions and problem description are described in Section 2. A
two-stage optimization method is presented in Section 3. Numerical results are given in Section 4. The conclusions are
shown in Section 5.

2. Notation, assumptions and problem description

Some mathematical notations have to be defined in order to facilitate description and formulation of the problem.
160 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169

Parameters
L Set of liner carriers
P Set of ports
S Set of surplus ports
Sk Set of surplus ports for liner carrier k 2 L
W Set of deficit ports
Wk Set of deficit ports for liner carrier k 2 L
Qk Set of OD (Origin-Destination) demands of liner carrier k 2 L
qkij Weekly number of containers of liner carrier k transported from origin port i to destination port j, i.e., the OD
demand
kij The average transportation cost for delivering one empty container from port i to port j
Decision variables
xkm
ij Weekly number of empty containers transported from surplus port i of liner carrier k 2 L to deficit port j of liner
carrier m 2 L
costj The empty container exchange cost for providing one TEU (Twenty-foot Equivalent Unit) empty container at
deficit port j 2 W

Generally, shippers will choose the proper liner carriers for delivering their containers, and they will sign a short-term or
long-term contract with their chosen liner carriers (Zheng et al., 2015). As described in Christiansen et al. (2007), it is not
unusual that some liner carriers do between 80% and 95% of their business under the long-term contracts, most of which
are negotiated once a year between liner carriers and shippers. According to the fixed container demands of different liner
carriers, the number of empty containers accumulated at the surplus ports or lacked at the deficit ports for different liner
carriers can be obtained. Generally, a practical ship route (i.e., port calling sequence by a ship) often maintains a weekly ser-
vice frequency, and the weekly container demands of different liner carriers can be used to determine the distribution of
empty containers at different ports. In order to simplify our problem, we assume that the weekly container demands of dif-
ferent liner carriers are fixed and given in advance.
Given the weekly OD demand from origin port i to destination port j of liner carrier k, qkij ð8qkij 2 Qk Þ, let nki denote the dif-
ference between the incoming and outgoing container flow for liner carrier k at port i, calculated as follows:
X X
nki ¼ qkji  qkij ; 8 k 2 L; 8i 2 P ð1Þ
j2P j2P

Clearly, nki represents the empty container supply or demand of liner carrier k at port i. When nki takes a positive value,
port i is a surplus port of liner carrier k, and port i is a deficit port of liner carrier k, when its value is negative. Generally, the
empty container allocating processes are typically time consuming and depend on the space available on vessels, both of
which are not considered in this paper. Namely, we assume: (i) travel times for delivering empty containers are neglected;
(ii) each liner carrier has sufficient space for repositioning empty containers. For simplicity, this paper mainly considers the
average transportation cost for delivering empty containers from the surplus ports to the deficit ports. Note that the empty
container repositioning is constrained by vessel spare capacities since laden containers have priority to be shipped (Dong
and Song, 2009). In practice, empty containers may be transported by a less cost-efficient liner carrier due to capacity con-
straints. In other words, the transportation cost for transporting empty containers from some certain surplus ports to some
certain deficit port may be various for different liner carriers, especially when considering the coordination among liner car-
riers. For instance, one liner carrier can buy some vessel slots from other liner carriers, leading to the different costs on trans-
porting (laden and empty) containers between some certain ports for different liner carriers. In this paper, we assume that
the average transportation cost for transporting one empty container from any surplus port to any deficit port is identical for
different liner carriers.
As mentioned before, one aim of this paper is to measure the perceived values of empty container at different ports,
which are defined as follows. For a particular surplus port, its perceived value of empty container is described by the unit
profit for delivering one empty container from this surplus port to any possible deficit port. For a particular deficit port,
its perceived value of empty container is represented by the cost paid to liner carriers for delivering one empty container
to this deficit port. This cost is termed ‘‘the empty container exchange cost” in this paper. Since empty container movements
do not generate revenue for liner carriers, we assume that the unit profits for delivering empty containers from the surplus
ports only depend on the average transportation cost for transporting empty containers and the empty container exchanges
costs at the deficit ports, independent of any other aspects (e.g., container leasing prices).
Because of the coordination among liner carriers, we assume that empty containers are repositioned in such a way that
the total transportation cost on delivering empty containers for all related liner carriers is minimized, leading to a centralized
optimization solution of empty container allocation. Then, we should determine the cost allocation (or profit sharing) among
liner carriers. For a small number of participating liner carriers, using Shapley value (Shapley, 1953) would be a meaningful
approach. Furthermore, this paper will measure the perceived values of empty container at different ports, based on the cen-
tralized optimization solution. To solve our problems, we propose a two-stage optimization method, as shown in Section 3.
 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169 161

3. Two-stage optimization method

Here, we propose a two-stage optimization method for solving our problems. In stage I, a centralized optimization solu-
tion of empty container allocation for all related liner carriers can be obtained, as shown in Section 3.1. In stage II, we aim to
determine the perceived values of empty container at different ports based on the solution obtained in stage I, as shown in
Section 3.2.

3.1. Stage I

In stage I, we aim to determine the weekly number of empty containers delivered from surplus port i of liner carrier k 2 L
to deficit port j of liner carrier m 2 L (i.e., xkm
ij ) to obtain a centralized optimization solution of empty container allocation for
all related liner carriers. The mathematical programming model (denoted by P) to pursue this centralized optimization solu-
tion can be formulated as follows:
XXX X
min ðkij  xkm
ij Þ ð2Þ
k2L m2L i2Sk j2Wm

subject to
XX
ij ¼ ni ;
xkm k
8 i 2 Sk ; 8 k 2 L; ð3Þ
m2L j2Wm

XX
ij ¼ nj ;
xkm m
8 j 2 Wm ; 8 m 2 L; ð4Þ
k2L i2Sk

ij P 0;
xkm 8 i 2 Sk ; 8 j 2 Wm ; 8 k; m 2 L: ð5Þ

The objective function (2) minimizes the total transportation cost for delivering empty containers of all related liner car-
riers from the surplus ports to the deficit ports. Constraints (3) and (4) ensure that empty containers should be balanced after
allocating empty containers. Constraints (5) denote xkm ij as a nonnegative variable. Note that the above formulation also
accounts for the case k = m, where any single liner carrier delivers its own empty containers from its surplus ports to its def-
icit ports without considering the coordination among liner carriers.

3.2. Stage II

Based on the optimization model P ((2–5)) in stage I, liner carriers coordinate with each other to obtain a centralized opti-
mization solution for all related liner carriers. Let x  ¼ f
xkmij g denote the optimal solution of model P. In stage II, we aim to
determine the perceived values of empty container at different ports, based on the optimal solution of model P. Firstly,
we determine the empty container exchange costs ({costj}) at the deficit ports, which are paid to liner carriers for delivering
empty containers and following the optimal solution x  . The empty container exchange costs at the deficit ports can be
regarded as the parameters, which need to be properly determined in our model presented in stage II. The unit profits at
the surplus ports can be calculated with respect to the empty container exchange costs at the deficit ports.
Generally, a single liner carrier aims to maximize (or minimize) its own profit (cost). When the empty container exchange
costs at the deficit ports are determined, we assume that the cost associated with liner carrier k mainly consists of three
terms: (i) the cost for transporting empty containers from the surplus ports of liner carrier k, which is described as
P P P
j2Wm ðkij  xij Þ; (ii) the empty container exchange cost paid by liner carrier k to other liner carriers, which is for-
km
i2Sk m2L
P h  P i
mulated as j2Wk costj  nkj  i2Sk xkk ij ; (iii) the empty container exchange cost paid to liner carrier k, which is given as
P P P
j2Wm ðcost j  xij Þ. Then, the optimal decision making problem for liner carrier k to maximize its profit, can be
km
i2Sk m–k;m2L
formulated as a linear programming model (denoted by Pk), shown as follows:
XX X h i
max ðcost j  kij Þ  xkm
ij ð6Þ
i2Sk m2L j2Wm

subject to
XX
ij 6 ni ;
xkm k
8 i 2 Sk ; ð7Þ
m2L j2Wm

X
ij 6 nj ;
xkk k
8 j 2 Wk ; ð8Þ
i2Sk

ij P 0;
xkm 8 i 2 Sk ; 8 j 2 Wm ; 8 m 2 L: ð9Þ
162 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169

P h i
The objective function (6) represents the profit associated with liner carrier k plus j2Wk costj  ðnkj Þ ; which is a con-
stant when the empty container exchange costs at the deficit ports are determined. Constraints (7) and (8) ensure that empty
containers of liner carrier k originated from (or delivered to) all the surplus (or deficit) ports cannot be larger than their
surplus (or deficit). Clearly, not all empty containers of liner carrier k are used to satisfy the deficit ports of liner carrier k,
since some empty containers of liner carrier k may be delivered to the deficit ports of other liner carriers, and some empty
containers of other liner carriers may be delivered to the deficit ports of liner carrier k. From a single liner carrier point of
view, there is no necessity for liner carrier k to deliver its own empty containers to the deficit ports of other liner carriers.
As a result, liner carrier k may keep some residual empty containers at its surplus ports. In practice, this liner carrier may
lease them off in order not to keep the additional empty containers and avoid the repositioning cost. For simplicity, this
paper does not consider leasing empty containers. Actually, empty containers will be delivered if found profitable, as shown
in the following complementary slackness conditions corresponding to the primal problem Pk and the dual problem of Pk.
Hence, constraints (7) of Pk are different from constraints (3) of P. Constraints (9) denote xkm ij as a nonnegative variable.
k denote the projection of x
Let x  (the optimal solution of the model P) to the solution space of the model Pk. In the fol-
lowing, we aim to identify the empty container exchange costs at the deficit ports, in order that x k is an optimal solution of
the model Pk. According to constraints (3) and (4), we can obtain that
XX
xkm
ij ¼ ni ;
k
8 i 2 Sk ð10Þ
m2L j2Wm

X X X
xkk
ij þ
xmk
ij ¼ nj ;
k
8 j 2 Wk ð11Þ
i2Sk m–k2Li2Sm

k is a feasible solution of the model Pk.

Clearly, x
Next, we demonstrate the use of the inverse optimization technique (Ahuja and Orlin, 2001) to determine the empty con-
tainer exchange costs at the deficit ports. The inverse optimization problems have been studied in some areas such as traffic
equilibrium (Dial, 2000; Ahuja and Orlin, 2001) and network design for liner carrier alliances (Agarwal and Ergun, 2010;
Zheng et al., 2015).
n o n o
For liner carrier k, let wk ¼ wki : wki P 0; i 2 Sk and gk ¼ gkj : gkj P 0; 8 j 2 Wk denote the dual variables associated with
constraints (7) and (8), respectively. Then the dual problem of Pk, DPk can be described as,
X X
min ðnki  wki Þ  ðnkj  gkj Þ ð12Þ
i2Sk j2Wk

subject to

wki þ gkj P costj  kij ; 8 i 2 Sk ; 8 j 2 Wk ; ð13Þ

wki P costj  kij ; 8 i 2 Sk ; 8 j 2 Wm ; 8 m–k; m 2 L; ð14Þ

wki P 0; 8 i 2 Sk ; ð15Þ

gkj P 0; 8 j 2 Wk : ð16Þ

In the above DPk, constraints (13) and (14) are the dual feasibility conditions for variables fxkk ij g and fxij g, respectively.
km

One form of the linear programming optimality conditions states that the primal solution (x  ) and the dual solution
k

((wk, gk)) are optimal for their problems if x

k is a feasible solution of Pk and (wk, gk) is a feasible solution of DPk, and together
they satisfy the complementary slackness conditions as follows:

ij > 0 (8i 2 Sk ; 8j 2 Wk ), then wi þ gj ¼ cost j  kij .

kk
(1) If x k k

km
(2) If x ij > 0 (8i 2 Sk ; 8j 2 Wm ; 8m–k; m 2 L), then wi ¼ cost j  kij .
k

P
(3) If i2Sk xij < nj (8j 2 Wk ), then gj ¼ 0.
kk k k

P P
(4) If m2L j2Wm  ij < ni (8i 2 Sk ), then wi ¼ 0.
xkm k k

According to Eq. (10), the fourth complementary slackness condition can be omitted. The inverse optimization problem
for liner carrier k aims to find a feasible dual solution and a cost vector satisfying the above complementary slackness con-
ditions. Let Ik denote the inverse optimization problem for liner carrier k. The following theorem shows the non-uniqueness
of the solution of Ik.
 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169 163

g ¼ fcostj þ cg is also a feasible cost vector, where c is an arbitrary

Theorem 1. If a cost vector cost ¼ fcostj g satisfies Ik, then cost
nonnegative number.

Proof. For the cost vector cost ¼ fcostj g, we assume that a dual solution ðwk ; gk Þ satisfies all constraints of Ik (i.e., all con-
straints of DPk and the complementary slackness conditions). Then, we can introduce another feasible dual solution
ðwk þ c; gk Þ, where wk þ c ¼ fwki þ cg and c is an arbitrary nonnegative number. For this feasible dual solution, we can find
g ¼ fcostj þ cg, since all constraints of Ik can be satisfied.h
a feasible cost vector cost

From Theorem 1, we can define the objective function of Ik as the minimization of the empty container exchange costs
associated with liner carrier k, described as follows:
XX X X XX
min ðxkm
ij  cost j Þ þ ðxmk
ij  cost j Þ ð17Þ
i2Sk m2L j2Wm m–k;m2Li2Sm j2Wk

Now, we wish to determine a common cost vector satisfying the inverse problems for all related liner carriers. Namely,
[k
INVP : I: ð18Þ
k2L
n o
Let Ak denote the set of fði; jÞj8i 2 Sk ; 8j 2 Wk g, and let Ak be the set of ði; jÞj ij > 0; 8i 2 Sk ; 8j 2 Wk . Let Bkm denote the
xkk
n o
set of fði; jÞj8i 2 Sk ; 8j 2 Wm ; 8m–k; m 2 Lg, and let Bkm represent the set of ði; jÞj ij > 0; 8i 2 Sk ; 8j 2 Wm ; 8m–k; m 2 L . Let
xkm
P

Ck be the set of fj i2Sk  ij < nj ; 8j 2 Wk g. Clearly, Ak # Ak , Bkm # Bkm and Ck # Wk . Similar to I , the optimization problem
xkk k k

INVP can be formulated as follows:

XXX X
min ðxkm
ij  cost j Þ ð19Þ
k2L i2Sk m2L j2Wm

subject to

wki þ gkj  cost j P kij ; 8 ði; jÞ 2 Ak n Ak ; 8 k 2 L; ð20Þ

wki  costj P kij ; 8 ði; jÞ 2 Bkm n Bkm ; 8 k 2 L; 8 m–k; m 2 L; ð21Þ

wki þ gkj  cost j ¼ kij ; 8 ði; jÞ 2 Ak ; 8 k 2 L; ð22Þ

wki  costj ¼ kij ; 8 ði; jÞ 2 Bkm ; 8 k 2 L; 8 m–k; m 2 L; ð23Þ

gkj ¼ 0; 8 j 2 Ck ; 8 k 2 L; ð24Þ

wki P 0; 8 i 2 Sk ; 8 k 2 L; ð25Þ

gkj P 0; 8 j 2 Wk ; 8 k 2 L; ð26Þ

costj P 0; 8 j 2 W: ð27Þ
The objective function (19) minimizes the total involved empty container exchange costs. Constraints (20) and (21) are
equivalent to constraints (13) and (14) in the model DPk. Constraints (22–24) correspond to the first three complementary
slackness conditions, respectively.
As shown in the objective function (6) of Pk, ðcostj  kij Þ can be regarded as the unit profit associated with port pair (i, j) for
transporting one empty container from surplus port i to deficit port j. According to the first two complementary slackness
conditions, we have costj  kij P 0 if the empty container flow from surplus port i to deficit port j is positive. It seems that
empty containers will be delivered if found profitable. Furthermore, we can obtain the following theorem.

Theorem 2. For any particular surplus port i of liner carrier k, its associated unit profits for different deficit ports of other liner
carriers are identical and equal to wki , as long as these deficit ports will be served by surplus port i for delivering empty containers.

Proof. For any deficit port j ð8j 2 W n W k Þ served by surplus port i ð8i 2 Sk Þ, the empty container flow from surplus port i to
this deficit port is larger than zero. According to the second complementary slackness condition, we have wki ¼ costj  kij .h
Actually, the unit profit for delivering empty containers from surplus port i of liner carrier k to the deficit ports of liner
carrier k is also equal to wki , as shown in Section 4.2. Then, we can define wki (i.e., costj  kij ) as the unit profit at surplus port i
for delivering one empty container to any possible deficit port.
164 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169

4. Numerical experiments

Our models are efficiently solved by using CPLEX implemented in a Windows 7 environment. Numerical experiments are
performed on a 3.4 GHz Dual Core PC with 4 GB of RAM.

4.1. Data description

We perform our computational experiments on an Asia–Europe–Oceania shipping service network consisting of 46 ports
(Meng and Wang, 2011; Wang and Meng, 2011, 2012; Zheng et al., 2015), as shown in Fig. 1. In order to investigate the coor-
dination among different numbers of liner carriers, this paper considers four liner carriers: liner 1, liner 2, liner 3 and liner 4.
The OD demands for different liner carriers are derived from real data (with modification due to the commercial confiden-
tiality). In order to obtain the average transportation cost kij for shipping one TEU empty container from port i to port j, we
mainly consider the bunker cost of a 5000-TEU ship with a common sailing speed (e.g., 20 knots), since the bunker cost is a
major component of the ship operating cost. Following Brouer et al. (2014), we assume kij ¼ 0:03  Disij (i.e., 205000
3000
¼ 0:03),
where Disij is the distance between port i and port j.

4.2. Results analysis

Firstly, we investigate the benefit of liner carrier coordination. For simplicity, we show the results for two liner carriers.
In order to explore the benefit of liner carrier coordination, two cases, i.e., case I (two liner carriers do not coordinate) and
case II (two liner carriers coordinate) are mainly investigated. In case I, empty containers are not exchanged between two
liner carriers. In this case, constraints (3) and (4) of the model P can be rewritten as follows:
X
ij ¼ ni ;
xkk k
8 i 2 Sk ; 8 k 2 L ð28Þ
j2Wk

X
ij ¼ nj ;
xkk k
8 j 2 Wk ; 8 k 2 L ð29Þ
i2Sk

As shown in Table 1, it is clear that, in case I (i.e., two liner carriers do not coordinate), the total transportation costs for
two liner carriers are 1.64 million USD and 1.38 million USD, respectively. When two liner carriers coordinate (i.e., case II),

Fig. 1. Ports in an Asia–Europe–Oceania shipping service network.

 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169 165

the total transportation costs for two liner carriers become 1.18 million USD and 1.69 million USD, respectively. The overall
transportation cost for two liner carriers can be reduced by 0.15 (i.e., (1.64 + 1.38)  (1.18 + 1.69) = 0.15) million USD when
they coordinate. Table 1 also shows the total empty container exchange cost paid to liner 1 and liner 2, respectively. We can
find that liner 1 should pay 0.36 (i.e., 1.88  1.52 = 0.36) million USD to liner 2. Then, in case II, the total costs for liner 1 and
liner 2 become 1.54 million USD and 1.33 million USD, respectively. Evidently, two liner carriers can benefit from the coor-
dination with each other.
Next, we show the results for the empty container exchange cost at each deficit port and the unit profit for delivering
empty containers from each surplus port, as shown in Tables 2 and 3, respectively. In addition, the results for empty con-
tainer allocation for the surplus ports of two liner carriers are shown in Tables A1 and A2, provided in Appendix A. The
results for empty container allocation together with the empty container exchange costs can be used to verify the unit profits
at any surplus port for delivering empty containers to different deficit ports.
Table 2 shows the results for the empty container exchange costs at the deficit ports. As compared with other deficit
ports, Bremerhaven has the lowest value of empty container exchange cost (i.e., 4.56 USD), and Tokyo has the largest value
of empty container exchange cost (i.e., 339.87 USD). Moreover, the empty container exchange costs at the deficit ports in
Europe are lower than 15 USD, since many empty containers are accumulated at the surplus ports in Europe, because of
the imbalance between imports and exports for Asia-Europe trade.
According to Theorem 2, we have proved that the unit profits at any particular surplus port of a liner carrier for serving
different deficit ports of other liner carriers are identical. Here, we will further show that the unit profits for delivering empty
containers from any particular surplus port of a single liner carrier to different deficit ports of this liner carrier are also iden-
tical. As shown in Table A2, Melbourne, Chittagong, Aqabah, Port Klang, Salalah and Jeddah are the surplus ports of liner 2, all
of which only serve the deficit ports of liner 2. For instance, Brisbane and Sydney are served by Melbourne, and the
transportation distances from Melbourne to Brisbane and Sydney are 1097 nautical miles and 582 nautical miles, respec-
tively. According to Table 2, the unit profit for delivering empty containers from Melbourne to Brisbane is 260.91 USD
(i.e., 293.82  0.03  1097 = 260.91), and the unit profit for delivering empty containers from Melbourne to Sydney is also
P
260.91 USD (i.e., 278.37  0.03  582 = 260.91). It seems that we also have gkj ¼ 0 when i2Sk  ij ¼ nj (8j 2 Wk ) is satisfied,
xkk k
P
similar to the third complementary slackness condition. Actually, when i2Sk  xij ¼ nj (8j 2 Wk ) is satisfied, the empty con-
kk k

tainer exchange cost at deficit port j of liner carrier k, which is dependent on gkj , is paid to liner carrier k herself. In other
words, various values of gkj do not impact the coordination and interaction among different liner carriers. In order to
minimize the objective function of INVP (19), gkj ¼ 0 can be obtained. Furthermore, when any particular surplus port of a
single liner carrier k serves different deficit ports of different liner carriers including liner carrier k, one can find that the unit
profits at this surplus port for delivering empty containers to different deficit ports are also identical. Actually, we obtain a

Table 1
Comparison between two cases.

Total transportation cost (million USD) Total empty container exchange cost (million USD)
Liner 1 Liner 2 To liner 1 To liner 2
Case I 1.64 1.38 – –
Case II 1.18 1.69 1.52 1.88

Table 2
Empty container exchange costs at the deficit ports.

Deficit ports Empty container exchange costs (USD) Deficit ports Empty container exchange costs (USD)
Brisbane 293.82 Port Klang 258.42
Fremantle 205.2 Dalian 317.64
Adelaide 245.49 Xingang 323.64
Sydney 278.37 Qingdao 322.2
Zeebrugge 9.57 Ningbo 326.13
Le Havre 14.7 Shanghai 334.17
Bremerhaven 4.56 Karachi 207.78
Hong Kong 308.52 Manila 304.95
Cochin 210.78 Singapore 264.72
Chennai 216.42 Yantian 309.72
Nhava Sheve 204.78 Xiamen 317.13
Jakarta 258.09 Chiwan 309.72
Tokyo 339.87 Colombo 218.34
Nagoya 333.09 Kaohsiung 317.73
Yokohama 339.51 Leam Chabang 278.22
Kobe 328.92 Jebel Ali 186.45
Pusan 319.41 Thamesport 12.03
Kwangyang 320.67 Ho Chi Minh 284.1
166 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169

Table 3
The unit profits for the surplus ports.

Surplus ports The unit profits (USD) Surplus ports The unit profits (USD)
Adelaide 245.49 Kwangyang 320.67
Antwerp 7.44 Le Havre 14.7
Aqabah 118.32 Manila 304.95
Brisbane 293.82 Melbourne 260.91
Chennai 216.42 Nagoya 333.09
Chittagong 219.09 Nhava Sheve 204.78
Colombo 218.34 Port Klang 258.42
Dalian 317.64 Pusan 319.41
Fremantle 205.2 Qingdao 322.2
Hamburg 0 Rotterdam 7.59
Hong Kong 308.52 Salalah 171.78
Jakarta 258.09 Singapore 264.72
Jebel Ali 186.45 Sokhna 97.32
Jeddah 135.54 Southampton 15.15
Karachi 207.78 Sydney 278.37
Kobe 328.92 Zeebrugge 9.57

unique profit for any surplus port when delivering one empty container to any possible deficit port. In Table 3, we show the
unit profit at each surplus port for providing empty containers. Clearly, Hamburg has the lowest value of the unit profit,
which is equal to 0 USD, and Nagoya has the largest value of the unit profit (i.e., 333.09 USD). Compared with other surplus
ports, Hamburg is the farthest surplus port for delivering empty containers to the deficit ports, as shown in Fig. 1. Compared
with other deficit ports, the deficit ports in Japan (e.g., Tokyo, Yokohama, etc.) have a large value of the empty container
exchange cost, as shown in Table 2. Hence, as the nearest surplus port to Yokohama, Nagoya can obtain the largest value
of the unit profit for delivering empty containers to Yokohama.
In practice, when the empty container supply does not meet the demand, some liner carriers may lease empty containers.
As compared with the container leasing prices provided by the container leasing companies, the perceived values of empty
container (i.e., the unit profits and the empty container exchange costs) at different ports may have an impact on pricing in
the container leasing market. For any particular surplus port, liner carriers would like to reposition empty containers if the
unit profit at this port is larger than the off-leasing price. Otherwise, liner carriers would lease off their empty containers in
order to avoid the repositioning costs. For any particular deficit port, liner carriers would like to lease in empty containers if
the empty container exchange cost at this deficit port is larger than the leasing-in price. Otherwise, the coordination among
liner carriers for exchanging empty containers would be favorable.
Now, we investigate the stability of liner carrier coordination on empty container allocation by considering the coordi-
nation among different numbers of liner carriers. Table 4 shows the results of all combinations for the coordination among
four liner carriers. The second column in Table 4 shows the different coordination cases. For example, (1, 2, 3) means that
liner 1, liner 2 and liner 3 coordinate with each other. In order for easy comparison, Table 4 also shows the individual trans-
portation cost in the non-coordination case. Evidently, one can find that any two of four carriers can benefit from the coor-
dination with each other, as compared with the non-coordination case. In other words, the coordination formation is stable
when the number of liner carriers is 2. When the number of liner carriers is 3 or 4, the coordination formation among liner
carriers is also quite stable. This is because there is no motivation for some liner carriers to collude, in order to decrease their
costs. In addition, results for cost allocation by using the Shapley value are also shown in Table 4. Clearly, our results are
slightly different with the Shapley value.

Table 4
Results for the coordination among different numbers of liner carriers.

Total cost (million USD)

Liner 1 Liner 2 Liner 3 Liner 4
Two carriers (1, 2) 1.54 1.33 – –
(1, 3) 1.63 – 1.29 –
(1, 4) 1.63 – – 1.92
(2, 3) – 1.34 1.19 –
(2, 4) – 1.31 – 1.82
(3, 4) – – 1.29 1.92
Three carriers (1, 2, 3) 1.55 1.31 1.21 –
(1, 2, 4) 1.58 1.27 – 1.85
(1, 3, 4) 1.37 – 1.57 1.80
(2, 3, 4) – 1.30 1.22 1.83
Four carriers (1, 2, 3, 4) 1.62 1.16 1.27 1.90
Non-coordination 1.64 1.38 1.80 2.50
Shapley value 1.5 1.15 1.33 1.97
 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169 167

5. Conclusions

This paper has investigated an empty container allocation problem considering the coordination among liner carriers for
exchanging empty containers. We further measure the perceived values of empty container at different ports. To solve our
problems, we propose a two-stage optimization method. Numerical implementations show the following conclusions: (i)
liner carriers can benefit from the coordination with each other; (ii) empty containers are delivered if found profitable;
(iii) the coordination formation among liner carriers is found to be stable; (iv) the perceived values of empty container at
different ports are measured. Furthermore, by comparing the perceived values of empty container at different ports with
the container leasing prices provided by the container leasing companies, we hope our work is helpful for pricing in the con-
tainer leasing market.
As shown in our assumptions, there are some limitations in our work: (i) travel times for delivering empty containers are
neglected; (ii) the weekly container demands are fixed and given in advance; (iii) we only consider the average transportation
cost when repositioning empty containers; (iv) each liner carrier is assumed to have sufficient space for repositioning empty
containers. There are some research issues we will investigate in the future. Firstly, we will extend our work by considering
the weekly dependent container demands, since the container demands vary from week to week. Secondly, we will study the
repositioning of empty containers and the routing of laden containers in a liner shipping network, by considering ship capac-
ity constraints, ship operating cost, bunker cost, inventory cost and container handling cost, etc. Thirdly, we will investigate
the impact of the container leasing prices on our results, in order to better understand the container leasing market.

Acknowledgements

We would like to thank anonymous referees for their useful comments, which significantly improve the presentation of
the paper. This research is supported by the National Basic Research Program of China (2012CB725400), the National Natural

Table A1
Empty container allocation for the surplus ports of liner 1.

Surplus ports Surplus containers Liner carriers Deficit ports with delivered empty containers
Brisbane 87 Liner 1
Liner 2 Brisbane (87)
Adelaide 192 Liner 1
Liner 2 Adelaide (192)
Melbourne 202 Liner 1
Liner 2 Brisbane (202)
Sydney 191 Liner 1
Liner 2 Sydney (191)
Chittagong 112 Liner 1
Liner 2 Ho Chi Minh (112)
Zeebrugge 107 Liner 1
Liner 2 Zeebrugge (85), Manila (22)
Sokhna 1433 Liner 1 Ningbo (386), Shanghai (857)
Liner 2 Yantian (190)
Hamburg 527 Liner 1 Bremerhaven (24), Ho Chi Minh (52)
Liner 2 Bremerhaven (148), Thamesport (303)
Nhava Sheve 96 Liner 1
Liner 2 Nhava Sheve (96)
Jakarta 296 Liner 1
Liner 2 Jakarta (155), Manila (141)
Aqabah 521 Liner 1
Liner 2 Yokohama (521)
Rotterdam 280 Liner 1 Hong Kong (59)
Liner 2 Tokyo (221)
Salalah 381 Liner 1
Liner 2 Cochin (4), Nhava Sheve (229), Karachi (148)
Karachi 121 Liner 1
Liner 2 Karachi (121)
Manila 66 Liner 1
Liner 2 Manila (66)
Jeddah 726 Liner 1
Liner 2 Cochin (726)
Singapore 146 Liner 1
Liner 2 Xiamen (146)
Colombo 126 Liner 1
Liner 2 Colombo (126)
Jebel Ali 385 Liner 1
Liner 2 Karachi (204), Jebel Ali (181)
Southampton 1685 Liner 1 Xiamen (295)
Liner 2 Ningbo (1390)
168 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169

Table A2
Empty container allocation for the surplus ports of liner 2.

Surplus ports Surplus containers Liner carriers Deficit ports with delivered empty containers
Fremantle 1296 Liner 1 Fremantle (19)
Liner 2 Adelaide (475), Sydney (436), Jakarta (366)
Melbourne 821 Liner 1
Liner 2 Brisbane (547), Sydney (274)
Chittagong 51 Liner 1
Liner 2 Manila (25), Ho Chi Minh (26)
Antwerp 923 Liner 1 Cochin (18), Kaohsiung (224), Leam Chabang (160), Ho Chi Minh (134)
Liner 2 Manila (59), Yantian (76), Colombo (132), Kaohsiung (120)
Sokhna 1796 Liner 1 Port Klang (41), Ningbo (431), Chiwan (398)
Liner 2 Singapore (543), Leam Chabang (383)
Le Havre 800 Liner 1 Le Havre (5), Chiwan (626)
Liner 2 Xiamen (169)
Hamburg 880 Liner 1 Thamesport (4)
Liner 2 Bremerhaven (876)
Hong Kong 56 Liner 1 Chiwan (56)
Liner 2
Chennai 551 Liner 1 Chennai (22), Yokohama (20), Port Klang (221)
Liner 2 Chiwan (288)
Nagoya 114 Liner 1 Nagoya (48), Yokohama (46)
Liner 2 Yokohama (20)
Kobe 246 Liner 1 Yokohama (41), Kobe (162)
Liner 2 Tokyo (43)
Aqabah 284 Liner 1
Liner 2 Yokohama (284)
Pusan 290 Liner 1 Pusan (224)
Liner 2 Shanghai (66)
Kwangyang 856 Liner 1 Kwangyang (249)
Liner 2 Shanghai (607)
Port Klang 174 Liner 1
Liner 2 Singapore (174)
Rotterdam 594 Liner 1 Hong Kong (25), Xiamen (569)
Liner 2
Dalian 1531 Liner 1 Dalian (59), Xingang (95), Shanghai (812)
Liner 2 Xingang (565)
Qingdao 1431 Liner 1 Qingdao (201), Shanghai (90)
Liner 2 Shanghai (1140)
Salalah 361 Liner 1
Liner 2 Cochin (361)
Jeddah 275 Liner 1
Liner 2 Nhava Sheve (275)
Southampton 1435 Liner 1 Hong Kong (419), Tokyo (5), Yantian (583)
Liner 2 Karachi (428)

Science Foundation of China (71501021, 61304179), the Fundamental Research Funds for the Central Universities
(3132015069), and Research Fund for the Doctoral Program of Higher Education of China (20130009120001).

Appendix A

The results for empty container allocation for the surplus ports of two liner carriers are shown in Tables A1 and A2,
respectively.

References

Agarwal, R., Ergun, O., 2010. Network design and allocation mechanisms for carrier alliances in liner shipping. Oper. Res. 58 (6), 1726–1742.
Ahuja, R.K., Orlin, J.B., 2001. Inverse optimization. Oper. Res. 49 (5), 771–783.
Bell, M.G.H., Liu, X., Angeloudis, P., Fonzone, A., Hosseinloo, S.H., 2011. A frequency-based maritime container assignment model. Transp. Res. Part B 45 (8),
1152–1161.
Bell, M.G.H., Liu, X., Rioult, J., Angeloudis, P., 2013. A cost-based maritime container assignment model. Transp. Res. Part B 58, 58–70.
Bonney, J., 2004. Changing the landscape. J. Commer. (April 26)
Braekers, K., Janssens, G.K., Caris, A., 2011. Challenges in managing empty container movements at multiple planning levels. Transp. Rev. 31 (6), 681–708.
Brouer, B.D., Alvarez, J.F., Plum, C.E.M., Pisinger, D., Sigurd, M.M., 2014. A base integer programming model and benchmark suite for liner-shipping network
design. Transp. Sci. 48 (2), 281–312.
Brouer, B.D., Pisinger, D., Spoorendonk, S., 2011. Liner shipping cargo allocation with repositioning of empty containers. INFOR 49 (2), 109–124.
Cheung, R.K., Chen, C.Y., 1998. A two-stage stochastic network model and solution methods for the dynamic empty container allocation problem. Transp.
Sci. 32 (2), 142–162.
Christiansen, M., Fagerholt, K., Nygreen, B., Ronen, D. 2007. Maritime Transportation. Barnhart, C., Laporte, G. (Eds.), Handbook in Operations Research &
Management Science, vol. 14.
 J. Zheng et al. / Transportation Research Part E 83 (2015) 158–169 169

Christiansen, M., Fagerholt, K., Nygreen, B., Ronen, D., 2013. Ship routing and scheduling in the new millennium. Eur. J. Oper. Res. 228, 467–483.
Christiansen, M., Fagerholt, K., Ronen, D., 2004. Ship routing and scheduling: status and perspectives. Transp. Sci. 38 (1), 1–18.
Crainic, T.G., Gendreau, M., Dejax, P., 1993. Dynamic and stochastic models for the allocation of empty containers. Oper. Res. 41 (1), 102–126.
Di Francesco, M., Crainic, T.G., Zuddas, P., 2009. The effect of multi-scenario policies on empty container repositioning. Transp. Res. Part E 45, 758–770.
Di Francesco, M., Lai, M., Zuddas, P., 2013. Maritime repositioning of empty containers under uncertain port disruptions. Comput. Ind. Eng. 64 (3), 827–837.
Dial, R., 2000. Minimal revenue congestion pricing Part II: An efficient algorithm for the general case. Transp. Res. 34, 645–665.
Dong, J.X., Song, D.P., 2009. Container fleet sizing and empty repositioning in liner shipping systems. Transp. Res. Part E 45 (6), 860–877.
Drewry, 2011. Container Market – Annual Review and Forecast. Drewry Shipping Consultants, London
Erera, A.L., Morales, J.C., Savelsbergh, M., 2009. Robust optimization for empty repositioning problems. Oper. Res. 57 (2), 468–483.
Feng, C.M., Chang, C.H., 2008. Empty container reposition planning for intra-Asia liner shipping. Marit. Policy Manage. 35 (5), 469–489.
Jula, H., Chassiakos, A., Ioannou, P., 2006. Port dynamic empty container reuse. Transp. Res. Part E 42, 43–60.
Lam, S.W., Lee, L.H., Tang, L.C., 2007. An approximate dynamic programming approach for the empty container allocation problem. Transp. Res. Part C 15 (4),
265–277.
Le, D.H. 2003. The logistics of empty cargo containers in the Southern California Region. Final Report (Long Beach, CA: METRANS Transportation Center).
Li, J.A., Leung, S.C.H., Wu, Y., Liu, K., 2007. Allocation of empty containers between multi-ports. Eur. J. Oper. Res. 182 (1), 400–412.
Li, J.A., Liu, K., Leung, S.C.H., Lai, K.K., 2004. Empty container management in a port with long-run average criterion. Math. Comput. Model. 40, 85–100.
Long, Y., Chew, E.P., Lee, L.H., 2015. Sample average approximation under non-i.i.d. sampling for stochastic empty container repositioning problem. OR
Spectrum 37 (2), 389–405.
Long, Y., Lee, L.H., Chew, E.P., 2012. The sample average approximation method for empty container repositioning with uncertainties. Eur. J. Oper. Res. 222
(1), 65–75.
Meng, Q., Wang, S., 2011. Liner shipping service network design with empty container repositioning. Transp. Res. Part E 47 (5), 695–708.
Meng, Q., Wang, S., Andersson, H., Thun, K., 2014. Containership routing and scheduling in liner shipping: overview and future research directions. Transp.
Sci. 48 (2), 265–280.
Moon, I., Do Ngoc, A.D., Hur, Y.S., 2010. Positioning empty containers among multiple ports with leasing and purchasing consideration. OR Spectrum 32,
765–786.
Moon, I., Do Ngoc, A.D., Konings, R., 2013. Foldable and standard containers in empty container repositioning. Transp. Res. Part E 49, 107–124.
Shapley, L.S., 1953. A value for n-person games. In: Kuhn, H.E., Tucker, A.W. (Eds.), Contributions to the Theory of Games II, Annals of Mathematical Studies,
vol. 28. Princeton University Press, Princeton, NJ, pp. 307–317.
Shintani, K., Imai, A., Nishimura, E., Papadimitriou, S., 2007. The container shipping network design problem with empty container repositioning. Transp.
Res. Part E 43 (1), 39–59.
Shintani, K., Konings, R., Imai, A., 2010. The impact of foldable containers on container fleet management costs in hinterland transport. Transp. Res. Part E 46,
750–763.
Song, D.P., Carter, J., 2009. Empty container repositioning in liner shipping. Marit. Policy Manage. 36 (4), 291–307.
Song, D.P., Dong, J.X., 2011. Effectiveness of an empty container repositioning policy with flexible destination ports. Transp. Policy 18 (1), 92–101.
Song, D.P., Dong, J.X., 2012. Cargo routing and empty container repositioning in multiple shipping service routes. Transp. Res. Part B 46 (10), 1556–1575.
Song, D.P., Dong, J.X., 2013. Long-haul liner service route design with ship deployment and empty container repositioning. Transp. Res. Part B 55, 188–211.
Song, D.P., Dong, J.X., 2015. Empty container repositioning. Handbook of Ocean Container Transport Logistics. Springer International Publishing, pp.
163–208.
Song, D.P., Zhang, J., Carter, J., Field, T., Marshall, J., Polak, J., Schumacher, K., Woods, J., 2005. On cost-efficiency of the global container shipping network.
Marit. Policy Manage. 32, 1–16.
Theofanis, S., Boile, M., 2009. Empty marine container logistics: facts, issues and management strategies. GeoJournal 74 (1), 51–65.
Tran, N.K., Haasis, H.-D., 2015. Literature survey of network optimization in container liner shipping. Flex. Serv. Manuf. J. 27, 139–179.
Wang, S., Liu, Z., Bell, M.G.H., 2015. Profit-based maritime container assignment models for liner shipping networks. Transp. Res. Part B 72, 59–76.
Wang, S., Meng, Q., 2011. Schedule design and container routing in liner shipping. Transp. Res. Rec. 2222 (1), 25–33.
Wang, S., Meng, Q., 2012. Liner ship route schedule design with sea contingency time and port time uncertainty. Transp. Res. Part B 46 (5), 615–633.
Zheng, J.F., Gao, Z.Y., Yang, D., Sun, Z., 2015. Network design and capacity exchange for liner alliances with fixed and variable container demands. Transp. Sci.
Articles Adv., 1–14