Transport Policy 50 (2016) 115125

Contents lists available at ScienceDirect

Transport Policy
journal homepage: www.elsevier.com/locate/tranpol

An effects analysis of logistics collaboration in last-mile networks

for CEP delivery services
Hyeongjun Park a, Dongjoo Park a,n, In-Jae Jeong b
a
Department of Transportation Engineering, University of Seoul, Seoul, South Korea
b
Department of Industrial Engineering, Hanyang University, Seoul, South Korea

art ic l e i nf o a b s t r a c t

Article history: Rapid increases in courier, express, and parcel (CEP) delivery demand have made environment- and
Received 16 March 2015 trafc-related issues important in metropolitan areas. This study analytically formulates CEP delivery
Received in revised form behaviors in last-mile networks to estimate the effects of logistics collaboration for apartment com-
18 March 2016
plexes. Reecting courier delivery behavior, the CEP delivery problem was divided into horizontal and
Accepted 10 May 2016
vertical routing problems. Optimization methodologies commonly utilized in the operations research
Available online 24 May 2016
area were employed for the analytical modeling of these two routing behaviors. The proposed meth-
Keywords: odologies were applied to apartment complexes in Seoul, Korea. It was found that the nancial feasibility
CEP delivery service of CEP collaboration is guaranteed when the number of households in an apartment complex exceeds
Logistics collaboration
about 900. From the nancial perspective, CEP collaboration is applicable to 9.119.4% of the apartment
Last-mile network
households in Seoul. In addition, CEP collaboration was analyzed to provide a meaningful amount of
Horizontal delivery
Vertical delivery social cost savings, implying its economic feasibility. The public sector's roles in stimulating CEP colla-
Spatial clustering boration are discussed from nancial and legislative perspectives.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction movements in apartment complexes, an efcient type of residence

With the spread of new technologies and in response to eco-
nomic and social changes, the courier, express, and parcel (CEP)
industry has experienced considerable global growth and has
prospered as a door-to-door, same- or next-day delivery service
(Kim et al., 2014). This process has increased CEP volumes, espe-
cially for home delivery (Hesse, 2002; Esser and Kurte, 2005;
Weltevreden and Rotem-Mindali, 2007). Thus, CEP actors are in-
creasingly confronted with the last-mile issue, and rms and
stores have modied their urban supply chains, increasing the
frequency of just-in-time and small, divided deliveries (Gevaers
et al., 2011; Menge and Hebes, 2011). At the same time, urban
goods distribution has become an important issue for cities (Patier
and Routhier, 2008).
Services offered in a competitive service-oriented market must
be reliable, effective, and cost-efcient. Responding to many phy-
sical and operational limitations, CEP companies have adopted
highly mobile and accessible light trucks to provide fast and ef-
cient service. This has signicantly increased commercial vehicle
movements in urban areas, which has strongly impacted urban
issues such as transport infrastructure repairs, accidents, and
pollution costs. In particular, the rise in commercial vehicle
n
Corresponding author.
for densely populated areas, has created problems such as trafc
congestion and accidents. Another issue related with last-mile CEP
service is crime. There are 20 CEP companies in Seoul, Korea, and
therefore at least 20 different service people visit apartment
complexes to provide delivery service. The apartment complexes
residents do not recognize them well, and thus many crimes have
been committed by people pretending to be service people (Kim
et al., 2014).
Delivery efciency must be improved, especially the CEP-sector
delivery service networks, which are beset by rising energy costs
and erce competition among carriers (Schwind et al., 2010). In
addition, consumers demand higher standards of ecological sus-
tainability management from the transportation business sector
(Leonardi and Baumgartner, 2004). Collaboration among CEP car-
riers is one possible solution for city logistics problems. Solutions
through CEP logistics collaboration are increasingly focused on the
optimization of last-mile delivery networks in inner city regions to
solve the problems caused by the increase in commercial vehicle
movements (Fusco et al., 2003; Taniguchi et al., 2003).
Seoul, the capital of Korea, has a quarter of the nation's popu-
lation and has the highest population density among OECD capital
cities. Its high population concentration during the 1970s and
1980s led to a serious housing shortage. The government initiated
a housing supply policy that triggered the construction of large
apartment complexes. This trend has continued, and much of

http://dx.doi.org/10.1016/j.tranpol.2016.05.009
0967-070X/& 2016 Elsevier Ltd. All rights reserved.
116 H. Park et al. / Transport Policy 50 (2016) 115125
Seoul's housing consists of densely populated high-rise apartment
complexes. The city's trunk line transportation strategies are di-
versied in accordance with the characteristics of each individual
CEP company, but no similar strategies are found in the city's last-
mile networks, whose urban structure is quite differentiated. This
creates several socioeconomic problems for the residential and
transportation environments.
Introducing logistics collaboration to the CEP delivery system in
last-mile networks could produce several benets, especially for
apartment complexes (Seo and Lee, 2014). For example, the R
apartment complex in Seoul, which has approximately 26 build-
ings, 2400 households, and 1000 delivery demands per day, has
recently adopted a collaborative delivery system. The new CEP
delivery agency has contracts with nine CEP delivery companies
and offers an integrated delivery service. This agency has a colla-
borative delivery center, comprising an ofce, a restroom, and fa-
cilities (e.g., cold storage, conveyor belts, pallets) installed in an
underground parking lot. The agency draws a commission on the
delivery of each order from the individual companies and pays a
fee for renting the center to the R apartment ofce. The colla-
borative delivery system increases delivery efciency in last-mile
networks, reduces delivery truck trafc, and prevents access by
criminals impersonating delivery service people. The change has
produced residential, environmental, and social benets such as
savings in vehicle operation, travel time, accident, and pollution
costs. Seoul's S apartment complex also introduced a collaborative
delivery system. However, complex ofcials treated logistics col-
laboration as a prot-making business, and the operation failed to
take root. Meanwhile, the Unied Government Building in Gwa-
cheon, where delivery service people were not allowed to enter for
security reasons, began applying logistics collaboration in Sep-
tember 2011. Although this system has reduced parking lot con-
gestion and improved delivery efciency, it has been difcult to
achieve nancial sustainability owing to the limited number of
packages it handles.
Beyond several theoretical approaches and practical cases, no
analysis of how the issues raised by logistics collaboration for CEP
service have been undertaken to determine what benets it gen-
erates. This study estimates the effect of logistics collaboration in
last-mile networks, focusing on apartment complexes. The study
formalizes both individual and collaborative CEP delivery behavior
in last-mile networks and estimates their total traveling distance
and time. The CEP delivery process in last-mile networks is di-
vided into horizontal and vertical deliveries. The capacitated ve-
hicle routing problem (CVRP) and a four-step experiment are used
to describe horizontal and vertical deliveries, respectively.
The remainder of this paper is organized as follows. After a
literature review, the mathematical formulations and solution al-
gorithms are presented. Then, the study area, sensitivity analysis,
and spatial clustering are described. Next, the numerical results of
logistics collaboration are discussed. The nal section summarizes
the conclusions, policy implications, and extensions of this study.
2. Literature review
Logistics collaboration has been thoroughly studied and discussed
by both scholars and practitioners (Caputo and Mininno, 1999;
Schmoltzi and Wallenburg, 2012). It has been referred to as a critical
factor (Naesens et al., 2009; Lindawati et al., 2014) for logistics com-
petitiveness. Among the cooperative sectors in logistics, the CEP in-
dustry remains a neglected topic, and the related literature is in its
infancy (Leitner et al., 2011). An important aspect that has not received
the attention it deserves is the reluctance of some shippers to parti-
cipate in urban consolidation centers (UCCs). Two independent sur-
veys (Regan and Golob, 2005; Holgun-Veras et al., 2008) have
estimated that carriers willingness to participate in UCCs is in the
range of 1618%. Urban staging areas (Holgun-Veras et al., 2008) could
increase load factors and reduce truck trafc. This concept is appealing
in a number of ways. First, it takes advantage of existing infrastructure,
which bypasses the need to invest in an entire building, and mini-
mizes ancillary expenses such as security and lighting. Second, it
would work well for non-perishable items, which represent a large
portion of deliveries. Third, it is scalable, as the space used could be
scaled up or down as needed. Fourth, it eliminates shippers concerns
about loss of contact time with customers and brand recognition. The
estimates suggest that anywhere between 15% and 19% of carriers
would be interested in such a concept (Holgun-Veras et al., 2008;
Holguin-Veras and Sanchez-Diaz, 2016).
Logistics collaboration is understood as a tailored business re-
lationship based on mutual trust, openness, shared risk, and shared
reward yielding a competitive advantage resulting in business per-
formance greater than rms would achieve individually (Lambert
et al., 1999). Studies on urban logistics collaboration can be categorized
into three groups (Gonzales-Feliu et al., 2014). The rst focuses on
computer modeling and simulation. Such methods use operations
research methods mainly derived from vehicle routing frameworks
(Toth and Vigo, 2002). Two main approaches are taken: a single-tier
system in which commodities are directly shipped from urban colla-
borative distribution centers (UCDCs) and two-tier systems in which
new intra-modal transfers are organized between UCDCs and the
goods nal destination (Lee and Jeong, 2008; Van Duin et al., 2008;
Crainic et al., 2009). The second group includes quantitative evaluation
and assessment methods. These studies use quantitative economics
and econometrics techniques to evaluate the feasibility of urban lo-
gistics collaboration. Once the location has been chosen, the main
strategic decisions are made (Van Duin et al., 2008; Ambrosini et al.,
2013). The European project CO3 (2014) presented a set of key per-
formance indicators (KPIs) from logistics horizontal collaboration:
(1) higher utilization rate of delivery vehicles (cube and weight ll),
(2) lower number of kilometers driven in the transport network
(especially empty ones), (3) lower number of receptions at the gate of
the Spar Retail warehouse (lower cost of labor), (4) lower inventory
levels (higher inventory rotation), (5) higher customer service level
(faster deliveries and small drops), and (6) positive impact for society
(reduced trafc). Lin. et al. (2014) examined the effectiveness of urban
delivery consolidation in terms of monetary logistics cost, energy
consumption and PM2.5 emissions with respect to a number of op-
erational (e.g., rent cost, customer demand) and policy factors (e. g.,
commercial vehicle size restriction in city centers). The third group
comprises qualitative and discussion studies that assess the suitability
and limitations of urban logistics collaboration, based on either lit-
erature surveys or case studies (Allen et al., 2012; Arvidsson and
Browne, 2013; Janjevic et al., 2013; Ville et al., 2013; Dablanc, 2014).
In seeking a sustainable logistics environment, CEP-centered
solutions have been increasingly focused on the optimization of
last-mile delivery networks in inner-city regions (Fusco et al.,
2003). A two-stage process is often formulated to simplify the CEP
network optimization problem (Zpfel and Wasner, 2002). The
most well-known approach for nding the optimal depot location
in a CEP delivery network is the location routing problem (LRP), in
which a vehicle routing problem (VRP) is integrated into the sys-
tem to consider the real pickup and delivery tour expenses needed
to visit multiple customers (Nagy and Salhi, 2007). The VRP is one
of the most thoroughly studied topics among combinatorial opti-
mization problems because of its practical relevance and difculty.
The VRP is concerned with determining the optimal route for a
eet based at one or more depots to serve a number of customers.
Many additional requirements and operational constraints are
imposed on route construction in practical VRP applications
(Sungur et al., 2008). The CVRP is a fundamental issue that con-
siders vehicle capacities in combinatorial optimization, with wide-
 H. Park et al. / Transport Policy 50 (2016) 115125
ranging applications. It forms the core of logistics planning and has
been extensively studied by the operations research community.
The last two decades have seen enormous improvements in the
ability to solve these problems through improved algorithms and
computational capabilities. Toth and Vigo (2002) provide an up-to-
date survey of problem variants, exact solution techniques, and
heuristics for the vehicle routing problem. For the stochastic VRP,
Zhong et al. (2007) consider a VRP in which customer locations
and demands are uncertain. They develop a two-stage model
using the capacity unassigned in the rst stage to adapt to the
demand uncertainty in the second stage. The rst stage creates
core areas to be serviced, and, after this demand is established,
recourse actions determine how to route in these areas, allowing
exchanges of demand nodes in the ex-zones between core
areas. The authors show that keeping those customers near the
depot unassigned is a good strategy for balancing the workload
against daily demand variations.
Nevertheless, logistics collaboration studies have not yet de-
termined what social effects can be generated from its im-
plementation, especially in last-mile networks. In the absence of
full information transparency, logistics collaboration planning will
have problems with practical introduction. This study thus pro-
poses a methodology for analyzing the changes that accompany
logistics collaboration to enable a simulation and quantitative
analysis of logistics collaboration in the last-mile networks of the
CEP industry.
3. Model formulation and solution algorithm
This section discusses the basic concepts of logistics colla-
boration in CEP delivery service and then provides mathematical
formulations and solution algorithms for its operations. The CEP
delivery problem (CDP) is divided into a CDP with horizontal
routing (CDPHR) and a CDP with vertical routing (CDPVR). The
CDPHR and CDPVR are combinatorial optimization problems
A B
C D
Fig. 1. Concept of horizontal delivery with respect to logistics collaboration.
117
seeking, respectively, the most economical ways of traveling from
building to building with a eet of vehicles and moving from
household to household within buildings.
3.1. Principle concepts
The target of logistics collaboration in CEP delivery service is
the last-mile network, the service's nal step, which is categorized
into horizontal and vertical deliveries. As shown in Fig. 1, each
company (A, B, C, and D) visits each of the buildings (1, 2, 3, and 4)
using individual trucks before collaboration. After collaboration,
only one company, the cooperative delivery company (CDC), pro-
vides an integrated delivery service using an integrated delivery
service truck. As shown in Fig. 2, before collaboration, each com-
pany's serviceperson visits the households individually using the
elevators for each line, each group of households serviced by a
specic elevator. After collaboration, the CDC's serviceperson
provides an integrated delivery service.
3.2. Mathematical formulation
CEP delivery service in last-mile networks is formalized using a
mathematical model to analyze the effect of logistics collaboration.
The CDP is dened such that the total distance traveled and time
taken for horizontal and vertical delivery, respectively, are mini-
mized. This paper assumes that the CDC's delivery mechanism,
regardless of logistics collaboration, is identical to that of each
individual CEP company.
The proposed mathematical model consists of a CDPHR and a
CDPVR. The CDPHR is regarded as a traditional CVRP, the most
basic variant of the VRP. The CVRP considers the problem of
routing a uniform eet of K vehicles at minimum cost with ca-
pacity V to serve geographically dispersed customers with de-
terministic demand. A route R is considered feasible if the total
demand of all customers on that route does not exceed the vehicle
capacity V. The CVRP is composed of a traveling salesman problem
CDC
collaboratio
n
118 H. Park et al. / Transport Policy 50 (2016) 115125

A B

CDC

collaboration

C D

Fig. 2. Concept of vertical delivery with respect to logistics collaboration.

(TSP) and a bin-packing problem (BPP) (Sungur et al., 2008; Bra-

mel and Simchi-Levi, 2003). The CDPVR is treated as a sequential
Cv = rpqypq
simulated model. i Vh p Vv
i
q Vv
i
Indices and sets
i, j : index for individual buildings. s.t.
p, q: index for individual households.
h: index for horizontal delivery.
xij =1j Vh{0}
i Vh (1)
v: index for vertical delivery.
Vh : set of individual buildings with a delivery demand of more
than 1, including depot (i 0). xij =1i Vh{0}
j Vh (2)
Vvi : set of individual households at building i with a delivery
demand of more than 1, including sub-depot (p 0).
Parameters xi0 = K
K : total number of vehicles for horizontal delivery. i Vh (3)
bij : delivery distance from building i to building j.
M: capacity of vehicles for horizontal delivery. x 0j = K
di : demand at each building i. j Vh (4)
rpq : delivery distance from household p to household q.
L: capacity of delivery serviceperson for vertical delivery. ( )
uj ui + M 1xij dji, j Vh{ 0}, i j (5)
dp : demand at each household p.
Decision variables
di ui Mi Vh{0} (6)
Ch : total delivery distance for horizontal delivery h.
Cv : total delivery distance for vertical delivery v.
ui : ow of the vehicle after it visits building i. 1,if the vehicle goes directly from building i to j
x ij : binary variable indicating whether a vehicle goes from xij =
0,otherwise (7)
building i to building j.
up : ow of the delivery serviceperson after visiting household
p. ypq =1i, q Vvi {0}
ypq : binary variable indicating whether a serviceperson goes
p Vv
i (8)
from household p to household q
ypq =1i, p Vvi {0}
min( Ch + Cv ) q Vv (9)
i

Ch = bij xij
i Vh j Vh ( )
uq up + L 1ypq dqi, p, q Vvi { 0}, p q (10)
 H. Park et al. / Transport Policy 50 (2016) 115125 119

dp up Li, p Vvi {0} (11) Step 1 Choose an arbitrary building w, and let the cycle be Z.
Step 2 Find a building outside Z closest to a building in Z, and
1, if the service person goes directly from name it z.
Step 3 Find an edge (i, j) in Z such that bik + bkl bij is minimal.
ypq = household p to q Step 4 Construct a new cycle C by replacing (i, j) with (i, z) and

0, otherwise (12) (z, j).
Step 5 If the current cycle C contains all vertices, stop. Other-
The objective function to be minimized is the total traveling wise, go to step 2.
distance of horizontal and vertical deliveries. For horizontal de- Phase II.
livery, constraints 2, 3, 4, and 5 are vehicle routing constraints. 2 idi
j for j1,...,
Step 1 set S= M
.
Constraints 2 and 3 guarantee that only one vehicle visits each
2 idi
building, and constraints 4 and 5 guarantee that K vehicles start at Step 2 for j1 to 1 do.
M
the depot and that all vehicles return to the depot (Toth and Vigo,
2002). Constraint 6 is active when a vehicle travels directly from if x( ej) = x( bj +1) then if i =j +b1 di M , then set Sj =
b
j { x( ),,x( )}
bj ej

building i to building j. In this case, the constraint becomes

and x( bj +1) = x( bj +1+1) .
uj ui dj , implying that the vehicle is carrying a sufcient
number of items to satisfy demand at building j. If the vehicle is { x( ),,x( )} and x(
else let Sj = bj ej 1 ) = x( ej)
bj +1

not traveling directly from building i to building j, the constraint

S = { x( ),,x( )}
j
bj ej
becomes uj ui +M dj , which is not active, since uj dj and M ui
from constraint 7. Constraint 8 is a binary variable indicating
Step 3 Calculate the total traveling time for all segments, Sj , and
whether a vehicle travels directly from building i to building j. For
select the best segment with the least traveling time.
vertical delivery, constraints 9 and 10 are serviceperson-routing
Step 2 implies that if the last building in the jth segment is
constraints ensuring that only one serviceperson visits each
equivalent to the rst building of segment j 1, it is removed from
household. Constraints 11 and 12 enforce the capacity and con-
segment j 1 and merged with segment j, as long as the vehicle
nectivity of the feasible routes, and constraint 13 is a binary
capacity is not violated. If the capacity is insufcient, the building
variable indicating whether a serviceperson moves directly from
demand is removed and added to segment j1. This procedure
household p to household q (Chang and Yen, 2012).
produces a feasible vehicle delivery schedule for each segment.

3.3. Solution algorithms 3.3.2. CDPVR algorithm

The four-step experiment is used to solve the CDPVR. The sub-
The proposed algorithm comprises the consecutive application depot (p 0) is dened as the starting point for vertical delivery,
of algorithms for a CDPHR and a CDPVR. Applying the algorithm and the building's demand areas are aligned in ascending order by
for a CDPHR determines the traveling routes to the apartment lines and stairs, as dened by a TSP tour. This TSP tour categorizes
complexes, and a CDPVR algorithm is applied to nd the visiting
sub-tours arranged by the maximum amount of delivery, S. Sub-
sequence for individual buildings.
tour delivery is performed until all demands are delivered
sequentially:
3.3.1. CDPHR algorithm
Step 1 Initialization: Dene the sub-depot (p 0) as the starting
A simple heuristic for the CVRP, modied iterative tour parti-
point for the delivery of the tour.
tioning (MITP), was introduced by Haimovich and Kan (1985) and
Step 2 Arrangement: Array the households with demand in
later modied by Altinkemer and Gavish (1987). In the MITP, a
ascending order of the two criteria: rst the line, and then the
traveling salesperson's tour is partitioned into customer segments
oor. Consider this array as the TSP tour.
such that each is served by a single vehicle (Bramel and Simchi-
Step 3 Partitioning: Partition the path into sub-tours of delivery
Levi, 1995). The following two-phase algorithm is a route rst-
serviceperson capacity, S, sequentially.
cluster second heuristic:
Step 4 Iteration: Visit all households with demand along the
Phase I.
partitioned path until the demands are serviced.
Step 1 Relax the requirement that a building's demand cannot
be split. A building.
requesting di units can be considered as di buildings requesting
1 unit with. 4. Case study
zero distance among di buildings.
Step 2 Find a TSP tour for the depot under consideration using a 4.1. Study area and target of logistics collaboration
nearest insertion algorithm that may violate vehicle capacity.
Step 3 Starting at the depot and following the TSP tour found in Seoul is used to evaluate the effect of logistics collaboration in
Step 2 in an arbitrary orientation, partition the path such that each apartment complexes. Seoul has a population of 10,528,774 (in
segment contains no more than M . The rst segment contains only 2011). It has 3,552,453 households, comprising 1,361,168 apart-
2
ments, 1,269,741 detached houses, 632,363 row houses, and
one building. Connect the endpoint of each segment to the depot.
289,181 others. Apartment buildings in Seoul have an average of
Step 4 Repeat Step 3 such that the rst segment contains two
2 idi 14.5 oors. Since apartments represent about 38% of Seoul's re-
buildings until different solutions are generated. Note that
M sidential housing, it is an appropriate study area for considering
2 idi
is the number of segments. logistics collaboration in apartment complexes.
M
CEP delivery companies covering any specic area provide
Step 5 Let the jth segment in the best solution be
service to both households and commercial establishments in the
Sj = {x (bj )
}
, ,x(ej ) such that the. area. However, for apartment complexes, only households are as-
rst building is numbered x(bj ) and the last is numbered x(ej ). sumed to be the target of the CEP service collaboration in Seoul.
The nearest insertion algorithm used to solve the TSP in Phase I This is because the number of commercial establishments in an
is as follows: apartment complex is signicantly lower than that of households,
120 H. Park et al. / Transport Policy 50 (2016) 115125

Table 2
Cost and revenue-related assumptions for nancial feasibility analysis.Source: Korea Development Institute (2008)

Items Basic Unit (KRW) Remarks

Cost Construction cost for (KRW/month) Utilizing existing facilities

Operation cost (KRW/month) 2,800,000 Operation cost rent
Labor cost of serviceperson (KRW/person/month) 2,000,000
Vehicle operating cost (KRW/veh/month) 510,400 (98.82  distance traveled  25 days) rent fuel cost
Revenue Commission (KRW/package) 600

and therefore the effect of logistics collaboration would be mar- Table 3

ginal. In addition, in an apartment complex, commercial estab-
lishments are spatially separated from households and the crime
issue does not apply to commercial shops.
4.2. Operation-related assumptions
This study examines the CEP service collaboration of the top
four CEP companies. According to surveys on Seoul's CEP service
industry, the market shares of companies A, B, C, and D are 37%,
12%, 11%, and 11%, respectively. The CEP service per capita is
0.2022 packages per day, and the average size of households is
2.62 (Park et al., 2014). The CEP delivery service conditions are
dened assuming that four CEP companies carry out delivery ac-
cording to their market share, while only one company, the CDC,
operates through logistics collaboration. Each company, regardless
of logistics collaboration, visits buildings using small trucks for
horizontal delivery and visits households on foot for vertical de-
livery. The collaboration effect is estimated by comparing the total
vehicle-miles traveled and the total person-hours worked during
logistics collaboration.(Table 1, 2 and 3).
Since the daily use per capita and average size per household
are 0.2022 packages and 2.62 persons, respectively, this study
assumes a basic unit of demand for each household of 0.5 packa-
ges per day. The maximum truck load in horizontal delivery and
serviceperson load in vertical delivery are 300 boxes per truck and
ve boxes per serviceperson, respectively. The access time to a line
after unloading is estimated at 30 s per line and the moving time
to go up and down by elevator at three seconds per oor, without
considering acceleration or waiting time. Conveyance time to
consumers is set at 15 s per household (Park et al., 2014).
This study analyzes the effect of logistics collaboration in last-
mile networks, focusing on apartment complexes in Seoul. The
exact coordinates of each apartment complex building need to be
known. However, obtaining this information is extremely difcult,
and it would require much time to analyze. Therefore, this study
generates the location of each building using model data and
constructs a distance matrix. The shape of each apartment com-
plex is assumed to be a regular quadrangle, and the coordinates of
each building are randomly generated. For example, the
Table 1
Operation-related assumptions for logistics collaboration.
Basic units for social benet estimation.Source: The
Seoul Institute (1997); Korea Development Institute
(2008); Kim and Lee (2010)
Items Basic Unit
Travel time cost (KRW/hour) 19,521
Vehicle operating cost (KRW/km) 407
Trafc accident cost (KRW/km) 32
Environmental pollution cost (KRW/km) 67
coordinates of each building of a ve-building apartment complex
would be randomly allocated in the range (0  5, 0  5). Under
these conditions, the coordinates should be integers, and (0, 0)
represents the depot. The modeled distances must be converted
into real distances using a scale parameter. As shown in Fig. 3, this
study calculates the scale parameter using regression analysis of
61 observations to nd the relationship between the number of
buildings and the side length of the apartment complexes, ob-
taining a value of 32.411 as the scale parameter. The t-statistics
and p-value are signicant, and the R-squared value is 0.930,
which is relatively high.
Cost-related assumptions.
This study selects the targets of logistics collaboration using a
criterion based on the number of households that satises the
CDC's requirement for nancial feasibility. As obtaining relevant
data is difcult, this study utilizes the case of the R apartment
complex as a basic unit for the nancial feasibility analysis. Since
the CDC is assumed to use existing facilities such as underground
parking lots for logistics collaboration, the construction cost for
the CDC is assumed to be zero. Operation cost, consisting of of-
cials labor costs, maintenance and administration fees, and rent, is
set at 2.8 million KRW (Korean won) per month, while the labor
costs for the service people are set at 2 million KRW per month.
Vehicle operating costs for horizontal delivery comprise rent and
fuel costs. Rent, including insurance bills, costs 510,400 KRW,
based on 2013 gures, and fuel costs 98.82 KRW/km according to
the Korea Development Institute (2008). Meanwhile, the com-
mission for apartment complex delivery is set at 600 KRW per
package.

Items Assumptions Remarks

Basic unit of demand for each 0.5 package/day combining daily use and
household family size
Maximum amount of truck load 300 packages/ capacity of truck
at horizontal delivery truck
Maximum amount of service- 5 packages/ capacity of
person load at vertical delivery person serviceperson
Time for line approach 30 s/line including unloading
time from truck
Time for elevator operation 3 s/stair not considering wait
and acceleration
Time for conveyance to customer 15 s/household regardless of amount of
package Fig. 3. Relationship between the number of buildings and side length of
complexes.
 H. Park et al. / Transport Policy 50 (2016) 115125
4.3. Assumption for social benet estimation
The effects of logistics collaboration are estimated by calculat-
ing CO2 emissions and social benets based on the differences in
total traveling distance and time. The social benets in KRW are
estimated using basic unit costs provided by the Korea Develop-
ment Institute (The Seoul Institute, 1997; Korea Development In-
stitute, 2008; Kim and Lee, 2010).
4.4. Spatial clustering for grouping apartment complexes
For apartment complexes with an insufcient number of
households, the effect of logistics collaboration is inappreciable or
even negative. Since collaborative CEP delivery services are ap-
plicable only to groups of apartment complexes that are fairly
close to each other, this study applies a spatial clustering method
that groups sets of objects into classes or clusters such that the
clustered objects are highly similar (Tung et al., 2001; Malmberg
and Maskell, 2002). The chosen apartment group consists of one
hub apartment complex and other small or medium apartment
complexes. The collaborative logistics company provides CEP ser-
vices to all apartment complexes within the chosen group by using
the hub complex as a service center.
In general, Moran's I and Geary's C statistics are used to de-
termine whether a spatial clustering pattern exists in a given
geographical area (Fotheringham et al., 2002; Park et al., 2013).
However, the statistics representing the spatial dependency
among apartment complexes in Seoul are not yet known. Thus,
spatial clustering is applied to create groups of apartment com-
plexes based on the following rules: (1) a group consists of one
hub and a number of spoke complexes; (2) a hub complex should
have facilities such as a main ofce an underground parking lot,
and roads 10 m wide; and (3) an apartment complex must be
within 250 m of the hub complex to be included as a spoke.
4.5. Application procedure
This study proposes a framework for analyzing the effect of
logistics collaboration in the CEP industry's last-mile networks.
The CDP is divided into a CDPHR and CDPVR and is solved by the
procedure described below. To minimize bias during the model
application, this study repeats all procedures 10 times and yields
the average value:
1. Generate locations of apartment buildings and household IDs
using the total number of buildings and households in Seoul.
2. Calculate the amount of CEP demand for each company, in-
cluding the CDC, with basic unit and market share rates.
3. Allocate the demand for each building randomly.
4. Randomly assign the demand for each household.
5. Compute the distance matrix based on building location and
scale parameter.
6. Estimate total distance traveled using the CDPHR algorithm.
7. Estimate total time worked using the CDPVR algorithm.
8. Compare total distance traveled and total time worked for lo-
gistics collaboration.
Table 4
Scenarios for analysis.
No. of scenario Maximum amount of serviceperson Subsidies
load
Scenario 1 300packages / day None ( 0)
Scenario 2 300 packages / day None ( 0)
Scenario 3 300 packages / day 100 KRW / package
Scenario 4 200 packages / day None ( 0)
121
4.6. Scenario
This study evaluates the effects of CEP logistics collaboration in
Seoul under various conditions. That is, study scenarios are set by
assuming different maximum numbers of serviceperson-loads for
vertical delivery (300 vs. 200 packages/day), with and without
subsidy, and with and without apartment grouping. Table 4 shows
the scenarios studied. In Scenario 1, only the apartment complexes
where CEP collaboration is nancially positive without grouping
apartments are the target of collaboration. In addition, no subsidy
and 300 packages/day are assumed. In Scenario 2, apartment
grouping is added to Scenario 1. Scenario 3 is different from Sce-
nario 2 in that 100 KRW/package is supported as a subsidy from
the public sector. In Scenario 4, 200 packages/day is assumed as
the serviceperson-load for vertical delivery, and no subsidy and
apartment grouping are assumed.
Selecting appropriate target apartment complexes is very im-
portant since not all apartment complexes can gain positive effects
through logistics collaboration. This study assesses nancial fea-
sibility in terms of the number of households serviced. Financial
feasibility reects whether a project is viable given its total costs
and probable revenues. If the revenues cover the costs, the project
is viable. Figs. 46 show the nancial feasibility analysis used to
select the target apartment complexes for logistics collaboration.
In scenarios 1 and 2, nancial feasibility is satised when the
number of households in an apartment complex exceeds 1428. In
scenarios 3 and 4, nancial feasibility is guaranteed when the
numbers of households in an apartment complex exceed 935 and
2296, respectively. In the nancial analysis, revenue (commission
by package) seems to be linear, while costs seem to increase
stepwise as the number of households increases. This is because
the number of servicepeople required and vehicle operating costs
increase stepwise with the number of households.
Hub apartment complexes are assumed to be used for spatial
clustering. Since hub apartment complexes must provide facilities
for logistics collaboration such as underground parking lots and
ofces, this study denes a hub as any apartment complex con-
structed after the year 2000 with more than 1000 households.
As shown in the last column in Table 4, the numbers of
apartment households for which CEP collaboration are nancially
positive for scenarios 14 are estimated at 124,462, 237,384,
263,692, and 202,875, respectively. By grouping nearby apartment
complexes, the target apartment households for CEP collaboration
is almost doubled from 124,462 to 237,384. Note that providing
subsidies reduces the number of households in an apartment
complex required to provide nancial feasibility from 1428 (sce-
nario 2) to 935 (scenario 3), thereby increasing the total number of
households targeted by CEP service collaboration from 237,384
(scenario 2) to 263,692 (scenario 3). As expected, scenario 3 has
the largest number of target apartment households. It may be seen
that in scenario 2, 112,922 households in 171 apartment complexes
are added through spatial clustering as targets for logistics colla-
boration in addition to the 124,462 households of scenario 1, as
shown in Fig. 7. It is considered that from nancial perspective,
CEP collaboration is applicable to 9.1% (124,462/1,361,168) to 19.4%
(263,692/1,361,168) of the apartment households in Seoul.
Grouping apartment complexes by spa- Number of households where CEP collaboration is -
tial clustering nancially positive
Not applied 124,462
Applied 237,384
Applied 263,692
Applied 202,875
122 H. Park et al. / Transport Policy 50 (2016) 115125
Fig. 4. Financial feasibility analysis for scenarios 1 and 2.
Fig. 5. Financial feasibility analysis for scenario 3.
Fig. 6. Financial feasibility analysis for scenario 4.
4.7. Results
Table 5 shows the logistics collaboration results in terms of
delivery distance and time savings. In scenario 1, the distance
traveled for horizontal delivery and the time worked for vertical
delivery decreased from 395 km/day and 1960 hours/day, respec-
tively, to 115 km/day and 1313 hours/day. In scenario 2, the re-
ductions are from 800 km/day and 3132 hours/day to 261 km/day
and 2085 hours/day. Scenario 3 shows a decline from 816 km/day
and 3675hours/day to 251 km/day and 2,397 h/day, and scenario
4 shows reductions from 708 km/day and 2712 hours/day to
234 km/day and 1813 hours/day. The sizes of savings in terms of
delivery distance and time are the same as the ranking of the
number of households for CEP service collaboration in the four
scenarios.
The savings ratios of the distance traveled for horizontal de-
livery ranged from 66.970.9%, while that of the time worked for
vertical delivery was about 34%. This discrepancy emerges because
horizontal delivery through logistics collaboration avoids visit
overlaps: before, each company would stop at each building,
causing logistical inefciencies. As few points within each building
were visited by each company before logistics collaboration, ver-
tical delivery through logistics collaboration cannot generate as
much savings as horizontal delivery can.
Table 6 shows the annual social benet produced by the dis-
tance traveled in horizontal delivery and the time worked during
vertical delivery. On average, travel time cost savings account for
more than 97% of the total benet. The annual social benets of
scenarios 14 are 4.4, 7.2, 8.7, and 6.2 billion KRW, respectively. As
expected, with the highest number of households for CEP service
collaboration, scenario 3 provided the greatest social benets. It is
considered that applying CEP service collaboration nearby apart-
ment complexes as well as to a hub apartment complex is
worthwhile because the annual social benet is increased from
4.4 to 7.2 billion KRW.
Providing a subsidy (i.e., scenario 3), however, does not sig-
nicantly increase the amount of social benet (i.e., 7.2 vs.
8.7 billion KRW per year) considering the size of the subsidy is
4.4 billion KRW. This is because even if subsidies are provided, the
number of target apartment households for CEP service colla-
boration increases marginally, as shown in Table 4. This marginal
effect of subsidy may be attributable to the fact that there are not
many apartment complexes in Seoul with between 935 and 1428
households.
4.8. Discussions and policy implications
Through the case study, it was found that from the nancial
perspective, CEP collaboration is applicable to 9.119.4% of the
apartment households in Seoul. In addition, all four scenarios re-
sulted in meaningful amounts of social cost savings. However, as
discussed earlier, only one apartment complex has been applying
CEP collaboration in Seoul, and the CEP collaboration at the
apartment complex was initiated by residents, not CEP companies,
for security reasons. This implies that from CEP companies per-
spective, CEP collaboration does not seem to be attractive enough
yet. The very low participation in CEP collaboration is probably
because (i) CEP companies worry about losing their own business
advantage or competitiveness, (ii) they do not trust each other, (iii)
they do not want to share their business information, (iv) they are
not condent about the collaboration effect, and so on.
In order to stimulate CEP collaboration in Seoul, the role of
public sector organizations such as the Seoul metropolitan gov-
ernment and the Korean government is considered essential. First,
in order to encourage CEP companies to participate in CEP colla-
boration, nancial incentives such as corporate tax reduction or
subsidy could be provided. Since 2010, the Korean government has
been selecting an environmentally friendly logistics company ev-
ery year and rewarding a number of benets. A CEP collaboration
company needs to be awarded this title as well.
Public sector policies to encourage CEP collaboration need to
target residents of apartment complexes as well. First, law/legis-
lation/rules/regulations/ordinances should be set up so that when
a new apartment complex is built, appropriate space and facilities
need to be provided in the basement parking lot. If there is no
space or facility to keep and sort parcels, then CEP collaboration is
physically impossible. Financial incentives for residents, such as
property tax discounts, might be considered as well. When an
existing apartment complex decides to start collaboration and
needs a new space or facility for it, building reform permission
should be given.
Alternatively, the public sector might launch a social enterprise
and/or cooperative association to take care of CEP collaboration.
Such a public or semi-public company could hire economically
handicapped elderly people, which is an important social-welfare
issue in Korea. A public sectordriven pilot project might be a good
strategy for motivating CEP logistics companies and apartment
 H. Park et al. / Transport Policy 50 (2016) 115125 123

Fig. 7. Target apartment complexes for spatial clustering (Scenario 2).

residents. Through such a project, the public sector could provide
detailed and comprehensive business information including a
priori implementation plans, business models (detailed business
process, legal and security issues, etc.), and posteriori evaluation
results.
5. Concluding remarks
The rapid increase in CEP delivery demand has made en-
vironment- and trafc-related issues important in metropolitan
areas such as Seoul, Korea. In the 2012 Urban Logistics Master Plan
(KOTI, 2012), the Seoul Metropolitan government proposed CEP
collaboration as a means of mitigating these social issues. How-
ever, no analytical evaluation of the effect of CEP collaboration has
been studied. This study formulated a CEP delivery model in last-
mile networks to estimate the effects of logistics collaboration for
apartment complexes. Reecting courier delivery behavior, the
CEP delivery problem was divided into horizontal and vertical
routing problems. Optimization methodologies commonly utilized
in the operations research area were employed for the analytical
modeling of these two routing behaviors.
The proposed approach was applied to apartment complexes in
Seoul, Korea. It was found that the nancial feasibility of CEP
collaboration is guaranteed when the number of households in an
apartment complex exceeds about 900, and that CEP collaboration
is applicable to 9.119.4% of the apartment households in Seoul. In
addition, CEP collaboration was analyzed to provide a meaningful
amount of social cost savings, implying its economic feasibility.
Even if CEP collaboration is nancially and economically fea-
sible, for many reasons, only one apartment complex in Seoul has
implemented CEP collaboration. In this context, this study dis-
cussed how to motivate and encourage CEP delivery companies
and apartment complexes to participate from the nancial

Table 5
Savings of delivery distance and time due to CEP service collaboration.

Scenario Horizontal delivery (km/day) Vertical delivery (hour/day)

Distance w/o colla- Distance with colla- Savings (a- Saving ratio Time w/o collabora- Time with collabora- Savings ((c- Saving ratio
boration (a) boration (b) b) (%) tion (c) tion (d) d)) (%)

Scenario 1 395 115 280 70.9 1960 1313 647 33.0

Scenario 2 800 261 539 67.4 3132 2085 1,047 33.4
Scenario 3 816 254 562 68.9 3675 2397 1,278 34.8
Scenario 4 708 234 474 66.9 2712 1813 899 33.2
124 H. Park et al. / Transport Policy 50 (2016) 115125
Table 6
Annual social benet of CEP service collaboration for apartment complexes.
Classication
Subsidies (KRW/year)
Social benet Travel time cost saving 4,349,744,995
(KRW/year) Vehicle operating cost saving
Trafc accident cost saving
Environmental pollution cost saving
Total cost saving 4,396,539,277
incentive and legislative points of view. The role of the public
sector is considered to be essential, and could include launching a
social enterprise and/or cooperative association to take care of CEP
collaboration.
A number of issues remain to be addressed in future studies.
First, this study did not consider several factors relevant to the
introduction of logistics collaboration in last-mile networks, such
as existing facilities for logistics collaboration (e. g., depots, sub-
depots), delivery strategies (e. g., handcarts or electronic vehicles
for horizontal delivery, retirees or the handicapped for vertical
delivery). Therefore, various strategies for optimal delivery must
be further investigated. Second, this study generated building lo-
cations and allocated demand using model data rather than real
data. The model requires validation using real-world data. Third,
more realistic vertical delivery time needs to be surveyed and
reected when estimating the effect of CEP collaboration. Fourth,
this study dened social benet as savings in travel time cost,
vehicle operating cost, trafc accident cost, and pollution cost.
Were logistics collaboration introduced throughout the city,
crimes committed by those impersonating servicepeople would be
prevented, and the landscapes of apartment complexes would be
improved. The effects of these aspects need to be quantied as
well. Finally, the effect of CEP collaboration in commercial and
ofce areas needs to be analyzed.
Acknowledgments
This research was supported by the Basic Science Research
Program through the National Research Foundation of Korea (NRF)
funded by the Ministry of Science, ICT and Future Planning
(2014R1A2A2A01006906).
References
Allen, J., Browne, M., Woodburn, A., Leonardi, J., 2012. The role of urban con-
solidation centres in sustainable freight transport. Transp. Rev. 32 (4), 473490.
Altinkemer, K., Gavish, B., 1987. Heuristics for unequal weight delivery problems
with a xed error guarantee. Oper. Res. Lett. 6 (4), 149158.
Ambrosini, C., Gonzales-Feliu, J., Toilier, F., 2013. A design methodology for sce-
nario-analysis in urban freight modeling. Eur. Transp. 54 (7), 121.
Arvidsson, N., Browne, M., 2013. A review of the success and failure of tram systems
to crry urban freight: The implications for a low emission intermodal solution
using electric vehicles on trams. Eur. Transp. 54 (5), 118.
Bramel, J., Simchi-Levi, D., 1995. A location based heuristic for general routing
problems. Oper. Res. 43 (4), 649660.
Bramel, J., Simchi-Levi, D., 2003. The Logic of Logistics: Theory, Algorithms, and
Applications for Logistics Management. Springer.
Caputo, M., Mininno, V., 1999. Internal, vertical and horizontal logistics integration
in italian grocery distribution. Int. J. Phys. Distrib. Logist. Manag. 26 (9), 6491.
Chang, T., Yen, H., 2012. City-courier routing and scheduling problems. Eur. J. Oper.
Res. 223 (2), 489498.
Crainic, T.G., Ricciardi, N., Storchi, G., 2009. Models for evaluating and planning city
logistic systems. Transp. Sci. 43 (4), 432454.
Dablanc, L., 2014. Logistics sprawl and urban freight planning issues in a major
gateway city. Sustain. Urban Logist. : Concepts Methods Inf. Syst., 4969.
Esser K. and Kurte J., 2005. B2C e-commerce: impact on transport in urban areas.
Recent Advances in City Logistics: Proceedings of the 4th International Con-
ference on City Logistics. 437448.
Scenario 1 Scenario 2 Scenario 3 Scenario 4
None ( 0) None ( 0) 4,350,918,000 None ( 0)
7,092,455769 8,594,486,737 6,099,714,629
37,638,879 72,427,860 75,445,688 63,736,517
2,959,322 5,694,574 5,931,848 5,011,225
6,196,081 11,923,014 12,419,806 10,492,252
7,182,501,217 8,688,284,079 6,178,954,623
Fotheringham, A.S., Brundson, C., Charlton, M., 2002. Geographically Weighted
Regression: The Analysis of Spatially Varying Relationships. John Wiley And
Sons Ltd., West Sussex, England.
Fusco, G., Tatarelli, L., and Valentini, M., 2003. Last-Mile, a Procedure to Set-Up an
optimized Delivery Scheme. Logistics Systems for Sustainable Cities. Proceed-
ings the 3rd International Conference on City Logistics, 147161.
Gevaers, R., Van de Voorde, E., Vanelslander, T., 2011. Characteristics and typology
of last-mile logistics from an innovation perspective in an urban context. City
Distrib. Urban Freight Transp.: Mult. Perspect., 5671.
Gonzalez-Feliu, J., Grau, J., Beziat, A., 2014. A location-based accessibility analysis to
estimate the suitability of urban consolidation facilities. Int. J. Urban Sci. 18 (2),
166185.
Haimovich, M., Kan, R., 1985. Bounds and heuristics for capacitated routing pro-
blems. Math. Oper. Res. 10 (4), 527542.
Hesse, M., 2002. Shipping news: the implications of electronic commerce for lo-
gistics and freight transport. Resour. Conversat. Recycl. 36 (3), 211240.
Holgun-Veras, J., Silas, M., Polimeni, J., 2008. An investigation into the attitudinal
factors determining participation in cooperative multi-carrier delivery in-
itiatives. In: Taniguchi, E., Thomson, R. (Eds.), Innovations in City Logistics IV.
Nova Science Publishers, pp. 5568.
Holgun-Veras, J., Snchez-Daz, I., 2016. Freight Demand Management and the
Potential of Receiver-Led Consolidation programs. Transport. Res. Part A: Pol.
Pract. 84, 109130.
Janjevic, M., Kaminsky, P., Ball Ndiaye, A., 2013. Downscaling the consolidation of
goodsstate of the art and transferability of micro-consolidation initiatives. Eur.
Transp. 54 (4), 123.
Kim, D., Lee, J., 2010. Application of Neural Network Model to Vehicle Emissions.
Int. J. Urban Sci. 14 (3), 264275.
Kim, S., Park, D., Kim, S., and Park, H., 2014. Modeling Courier Vehicle's Travel Be-
havior: A Case of Seoul, Korea. Transportation Research Board 93rd Annual
Meeting. CD-ROM. Transportation Research Board of the National Academies,
Washington, DC.
Korea Development Institute, 2008. A Study on Standard Guidelines for pre-feasi-
bility Study on Road and Railway projects.
Korea Transport Institute (KOTI), 2012. A study on mater plan of the logistics in
Seoul, Goyang.
Lambert, D., Emmelhainz, M., Gardner, J., 1999. Building Successful Logistics Part-
nership. J. Bus. Logist. 20 (1), 165181.
Lee, D., Jeong, I., 2008. Regression approximation for a partially centralized in-
ventory system considering transportation costs. Comput. Ind. Eng. 56 (4),
11691176.
Leitner, R., Meizer, F., Prochazka, M., Sihn, W., 2011. Structural concepts for hor-
izontal cooperation to increase efciency in logistics. CIRP J. Manuf. Sci. Tech-
nol. 4 (3), 332337.
Leonardi, J., Baumgartner, M., 2004. CO2 efciency in road freight transportation:
Status quo, measures and potential. Transp. Res. Part D: Transp. Environ. 9 (6),
451464.
Lin J., Chen Q. and Kawamura K., Sustainability SI: logistics cost and environmental
impact analyses of urban delivery consolidation strategies, Net Spat Econ. on-
line rst April 2014.
Lindawati, van Schagen, J., Goh, M., de Souza, R., 2014. Collaboration in urban lo-
gistics: motivations and barriers. Int. J. Urban Sci. 18 (2), 278290.
Malmberg, A., Maskell, P., 2002. The elusive concept of localization economies:
towards a knowledge-based theory of spatial clustering. Environ. Plan. A. 34,
429450.
Menge, J., Hebes, P., 2011. Optimization of urban deliveries: evaluating a courier,
express and parcel services pilot project in Berlin, City Distribution and Urban
Freight Transport: Multiples Perspectives. Edward Elgar Publishing., North-
ampton, pp. 201216.
Naesens, K., Gelders, L., Pintelon, L., 2009. A swift response framework for mea-
suring the strategic t for a horizontal collaborative initiative. Int. J. Prod. Econ.
121, 550561.
Nagy, G., Salhi, S., 2007. Location-routing. issues, models, and methods. Eur. J. Oper.
Res. 177, 649672.
Park, H., Park, D., Kim, C., Kim, H., Park, M., 2013. A comparative study on sampling
strategies for truck destination choice model: case of Seoul Metropolitan Area.
Can. J. Civil. Eng. 40, 1926.
Park, H., Park, D., Shin, S., Lee, C., and Nam, D., 2014. A Study on Estimation of
Logistics Collaboration in Parcel Delivery Service: focused on apartment com-
plex in Seoul. Korean Society of Transportation 69th Conference, CD-ROM.
Korea Society of Transportation.
Patier, D. and Routhier, J.L., 2008. How to improve the capture of urban goods
 H. Park et al. / Transport Policy 50 (2016) 115125
movement data. 8th International conference on survey methods in transport,
Annecy, France.
Regan, A.C., Golob, T.F., 2005. Trucking industry demand for urban shared use
freight terminals. Transportation 32 (1), 2336.
Schmoltzi, C., Wallengburg, C., 2012. Operational governance in horizontal co-
operations of logistics service providers: performance effects and the moder-
ating role of cooperation complexity. J. Supply Chain Manag. 48 (2), 5374.
Schwind, M., Kunkel, M., 2010. Collaborative optimization of last mile networks for
courier, express and parcel delivery services. MKWI Inf. Transp. Verk.,
19611973.
Seo, S., Lee, J., 2014. Improving the parcel service performance in the Seoul me-
tropolitan area: Differentiated approaches by urban land use pattern. Transport
Research Arena 2014, Paris, France.
Sungur, I., Ordez, F., Dessouky, M., 2008. A robust optimization approach for the
capacitated vehicle routing problem with demand uncertainty. IIE Trans. 40 (5),
509523.
Taniguchi, E., Thompson, R.G., Yamada, T., 2003. Predicting the effects of city lo-
gistics schemes. Transp. Rev. 23 (4), 489515.
125
The Seoul Institute, 1997. A Study on the Effective Management of Urban Goods
Movement.
Toth, P., Vigo, D., 2002. Models, relaxations and exact approaches for the capaci-
tated vehicle routing problem. Discret. Appl. Math. 123, 487512.
Tung, A., Hou, J., and Han, J., 2001. Spatial clustering in the presence of obstacles.
Proceeding of International Conference on Data Engineering. 359367.
Van Duin, R., Quak, H., Munuzuri, J., 2008. Revival of the cost benet analysis for
evaluating the city distribution center concept. Innov. City Logist., 97114.
Ville, S., Gonzalez-Feliu, J., Dablanc, L., 2013. The limits of public policy intervention
in urban logistics: Lessons from Vicenza. Eur. Plan. Stud. 21 (10), 15281541.
Weltevreden, J., Rotem-Mindali, O., 2007. Mobility effects of B2C and C2C e-com-
merce in the Netherlands: a quantitative assessment. J. Transp. Geogr. 17 (2),
8392.
Zpfel, G., Wasner, M., 2002. Planning and optimization of hub-and-spoke trans-
portation networks of cooperative third-party logistics providers. Int. J. Prod.
Econ. 78 (2), 207220.
Zhong, H., Hall, R.W., Dessouky, M.M., 2007. Territory planning and vehicle dis-
patching. Transp. Sci. 41 (1), 7489.