Available online at www.sciencedirect.

com

ScienceDirect
Transportation Research Procedia 12 (2016) 213 226

The 9th International Conference on City Logistics, Tenerife, Canary Islands (Spain), 17-19 June
2015

Evaluation of cost structure and impact of parameters in locationrouting problem with time windows
Sattrawut Ponboon*, Ali Gul Qureshi, Eiichi Taniguchi
Department of Urban Management, Kyoto University, Kyotodaigaku-Katsura, Nishikyo-ku, Kyoto 615-8540, Japan

Abstract
This study presents the evaluation and application of the location-routing problem with time windows (LRPTW) in the distribution
network. The ordinary location-routing problem is considered one of the integrated logistics system tools extended from the original
facility location problem. By adding the time windows into the problem, it provides the room to improve the service quality and
customer satisfaction. To understand the impact of the main parameters, nine scenarios of the Osaka distribution network from
freight carrier were tested with different depot location, depot size, and vehicle size. It was observed that the large size depot,
located in Minato-ku, serving by large size vehicle results in the lowest overall cost. The characteristics of depot, vehicle, and
transport information were also discussed in detail.
2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
2016 The Authors. Published by Elsevier B.V.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review
under
responsibility
the organising
committee
ofInternational
the 9th International
on City Logistics.
Peer-review under
responsibility
of theoforganising
committee
of the 9th
ConferenceConference
on City Logistics
Keywords: location-routing; time windows; branch-and-price; column generation

1. Introduction
In the distribution management, determining the suitable sites of the business center is considered one of the most
essential strategies in supply chain. This decision is prepared for setting up a factory, warehouse, retail store, or in
other public services, i.e. hospital, police station, fire station, etc. Thus the facility location is required at several points
throughout the business development. Traditionally, after the facility or warehouse was set up, the transportation

* Corresponding author. Tel.: +81-75-383-3231; fax: +81-75-383-3231.

E-mail address: sattrawut@gmail.com

2352-1465 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organising committee of the 9th International Conference on City Logistics
doi:10.1016/j.trpro.2016.02.060

214

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

process has been determined and operated as a short-term basis. This method refers to the conventional location first,
routing second approach (Bruns & Klose, 1996). While the other decisions can be examined more frequently at the
short-term operational stage, the facility must be located earlier in the long term strategic planning. However, it was
proved that the combination of these problems to become a Location-Routing Problem (LRP) reduces the cost over
the long-term horizon. Solving them together early in the planning horizon provides significant cost savings, improves
productivity, and returns positive impacts for both operators and society (Rand, 1976; Madsen, 1983; Salhi & Rand,
1989).
Moreover, in the context of logistics management, much attention has been paid in improving the service quality
by increasing customer satisfaction. It was recommended by Min et al. (1998) that the future LRP studies should be
extended to consider the presence of time windows since the customers often request the service time and deadlines.
Some might argue that the delivery time is relatively small compared with the time horizon of the facility location
process. However, the technology improvement enables the growth of a new market, which is responsive to customers
behavior and their needs, especially the importance of just in time policies (Gonzalez-Feliu, et al., 2012). That is
why, the time windows are now almost essential in different businesses such as bank deliveries, postal deliveries,
industrial refuse collection, school bus routing and scheduling, or the modern e-commerce (Solomon & Desrosiers,
1988).
According to the latest comprehensive surveys (Nagy & Salhi, 2007; Drexl & Schneider, 2013; Prodhon & Prins,
2014), the LRP-related researches are growing in number. However, few of literature on location problem have been
applied to specific applications or case studies. The application-based publications are amount only less than one fifth
(Nagy & Salhi, 2007). Current et al (2002) summarized the reasons behind this phenomenon. First, the application of
location problem practices the existing model and solution technique. It does not provide a scientific approach for
research literature. Second, the real world business and industries handle this task by consultants or professional firms.
They are not interested or motivated in sharing such experiences. Third, the private sectors treat their valuable
knowledge in this area secretly. Nonetheless, it is still important to advance the modelling together with the specific
application in real world problem.
By considering the issues mentioned above, while the early LRP studies provide the framework and algorithm
development with hypothesis testing instances, the evaluation and investigation of specific applications is still being
questioned. In LRP model, two cost components, i.e. location and routing, influence the calculation process and
tradeoffs must be made between them. The ratio between these two components varies among different business. For
example, the cost ratio of pharmaceuticals business is 0.26, the paper business is 0.85, and the consumer merchandising
business is 1.56 (Srivastava, 1986). To assess the LRP under different environment and apply to the real world problem
is not only provide a broader understanding of location-routing options, but also convincing evidences of its efficiency
and practicality. This problem was motivated by the observation of LRPTW by Ponboon et al (2014) that in some
contexts, the cost saving might not be influenced by only depot location, route length, and time windows. It may be
more economical to change the configuration of the underlying parameter instead of open new facilities. Therefore, a
full understanding of that observation is examined in this study. In other words, the main purpose of this study helps
us to make decisions on;
x
x
x
x
x
x

How many and which depots to be operated?

Which customers to be assigned to which depots?
What are the sequences of the customers?
What are the truck routes?
How is the difference between opening the new depot and enlarging the old depot?
What is the effect on the size of vehicle?

The impact of external characteristics is assessed and evaluated by various factors. The factors that influence the
distribution system include depot location, depot and vehicle capacities, their fixed costs, and time windows. To handle
the location, routing, and time windows together, the scheme of the location-routing problem with time windows
(LRPTW) with exact algorithm is used as a main tool. By using the branch-and-price algorithm developed by Ponboon
et al. (2014), this study explores the effect of various cost component factors in logistics systems. The Osaka road
network is selected as a based case of distribution network.

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

215

2. Background
2.1. Location-Routing Problem
The location problem does not require a routing determination similar to the Hamiltonian p-median problems.
Likewise, the multi-depot vehicle routing problem does not solve the facility location since it is implicitly determined.
However, in the LRP, two decisions are considered interdependently, how many and which facilities should be opened
and which routes should be built to meet customer demand. The former is based on ordinary location problem while
the latter is developed with the fundamentals of the vehicle-routing problem (VRP) seeking to serve a set of customers
with a fleet of vehicles with minimum distribution cost. The definition of location-routing problem given by Srivastava
and Benton (1990) stated as:
Given a feasible set of potential depot sites and customer sites, find the location of the
depots and the routes to customers from the depots such that the overall cost is minimized.
The overall cost is the sum of the location and distribution costs
The first theoretical study on location problem was back to 1909 when Weber identified the optimum location and
minimum cost of manufacturing plants by considering freight rates and production function (Friedric, 1929). In the
early years, the idea had been adopted mainly in military aspect. Later, Hakimi (1964) proposed the concept of center
and the median vertex of a graph to determine the optimum location of a switching center in a communication network
and the best location of a police station. Its objective sets to determine the place(s) where the total cost of satisfying
customers demand is minimize with a set of constraints, where the costs include the fixed cost to establish facility
and the distribution cost. Briefly, the facility location and vehicle routing have been carried out in the different levels.
To combine both decisions as the location-routing problem, the difficulty of computational process limited the LRP
to be presented only as the conceptual framework in the early works (Von Boventer, 1961; Maranzana, 1964; Webb,
1968; Lawrence & Pengilly, 1969; Christofides & Eilon, 1969; Higgins, 1972). Watson-Gandy and Dohrn (1973) were
the first authors who considered the vehicle routing together with depot locating using a non-linear profit function. To
evaluate the benefit of this integration, Salhi and Rand (1989) assessed the effect of ignoring routing when locating
the depots and proved that the best solution in locating stage does not guarantee the best solution in routing stage. The
early survey studies show that the LRP did not receive much attention thereafter for a long time (Balakrishman, et al.,
1987; Laporte, 1988; Laporte, 1989; Berman, et al., 1995; Min, et al., 1998). The recent surveys, however,
demonstrated a number of articles published in various forms of location-routing problem (Nagy & Salhi, 2007; Anon.,
2012; Drexl & Schneider, 2013; Prodhon & Prins, 2014).
In LRP, exact solution approaches are successful for more specific cases and provide substantial insights into
problems compared to heuristic approach. From our best knowledge, most of the LRP exact solutions were developed
based on research purposes. None of them have been tested and evaluated by the logistics application. In the early
years, a classical branch-and-bound algorithm consisted of all candidate solutions is firstly used by Larporte and
Nobert (1981). Later, the subtour elimination constraints and Gomory cuts were introduced to achieve integrality more
efficiently (Larporte, et al., 1983) while a branching procedure based on chain barring constraints was applied in
Laporte, et al. (1986) and Larporte, et al. (1989). In addition, the combination between branch-and-bound and
constrained assignment problem using a graph transformation to reformulate the location-routing problem into the
travelling salesman problem was presented in Larporte, et al. (1988). Lately, more complex algorithms have been
developed using more sophisticated integer linear programming called branch-and-price and branch-and-cut
algorithms for the LRP. The branch-and-price is the hybrid of branch-and-bound and column generation while branchand-cut uses cutting planes to tighten the relaxed problem. For the branch-and-price, Berger, et al. (2007) developed
a set-partitioning-based formulation of an uncapacitated location-routing model with distance constraints. A set of
valid inequalities was introduced to strengthen the formulation. Likewise, Akca, et al. (2009) also presented the setpartitioning-based formulation for the capacitated location and routing problem with several approaches on pricing
problem. Based on that study, Baldacci, et al. (2011) decomposed the problem into a limited set of Multi-capacitated
Depot Vehicle-Routing Problem (MDVRP) solved with dynamic programming and dual ascent methods. For the
branch-and-cut, Belengure, et al. (2011) proposed two new polynomial algorithms that strengthened by the families

216

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

of constraints and valid inequalities Capacitated Vehicle Routing Problem Capacitated Vehicle Routing Problem.
Contardo, et al. (2013(a)) extended the work of Belengure, et al. (2011) by introducing three flow formulations for
solving capacitated LRP namely a two-index two-commodity, a three-index vehicle-, and a three-index twocommodity flow formulations. The valid inequalities were derived from each flow formulation and were computed
with separation algorithms. The extensive work of Contardo, et al. (2013(b)) proposed three stages to solve a setpartitioning problem. In addition, five new sets of valid inequalities were introduced, which successfully improved the
bounds on instances available in the literature. By adding the variant on time windows in to LRP, Ponboon et al (2014)
extended the Acka et al (2009) branch-and-price algorithm and evaluated the effect of time windows on modified
Solomon benchmarks.
2.2. Time Windows
It is most important for logistic distributions to maintain efficiency not only cut their logistics cost, but also compete
on the service qualities (Desrosiers, et al., 1995). That is why the time windows are now almost essential in different
businesses such as bank deliveries, postal deliveries, industrial refuse collection, airline, rain transportation, school
bus routing and scheduling, or the modern e-commerce (Solomon & Desrosiers, 1988). In addition, the growing
business competition and financial crisis require a readjustment in the logistics strategic decisions that now must
include consideration of the main logistics tactical issues as well. Some might argue that the delivery time is relatively
small compared with the time horizon of the facility location process. Such phenomenon discourages researchers
considering this aspect with facility location. However, the technology improvement enables the growth of a new
market, which is respondent to customers behavior and their needs, especially the importance of just in time policies
(Gonzalez-Feliu, et al., 2012).
Time window requires services to be delivered to the customer within desired period, represented by notation ai
and bi (Taniguchi, et al., 2001). If a vehicle arrives early, it will induce waiting time and probably parking costs. If a
vehicle made a delay, it diminishes customer satisfaction, even breaks the contract in the worst case. These additional
costs of early arrival and delay are called the penalty costs. In the case of the hard time window, the late arrival is
strictly prohibited setting the penalty cost equal to infinity, while the early arrival is allowed with no extra charge.
However, it is important to note that service still cannot start earlier than ai.
The impact of time windows constraints in logistics performance has been tested variously in vehicle-routing
problem, early started by the case studies (Pullen & Webb, 1967; Knight & Hofer, 1968; Madsen, 1976). Later,
Solomon (1986; 1987) was the first to examine the diversity of route construction heuristics of a two-phase sequential
algorithm for the VRPTW while the route improvement procedures have been developed by Russell (1977), Cook &
Russell (1978), and Baker & Schaffer (1986). Solving VRPTW by exact methods was firstly introduced by Desrochers
et al (1992) using column generation based on Dantzig-Wolfe decomposition (Dantzig & Wolfe, 1960). These kind of
decompositions, i.e. branch-and-price and branch-and-cut, have been widely used after that.
By adding the complexity of time window constraint into the LRP problem, a proper and effective method must be
developed to deal with a combination of facility location and routing simultaneously which is complex by its NP-hard
(Dror, 1994). Based on survey studies for the location-routing problem, the literature considering the hard time
windows solved the problem using only heuristic approaches (Min, et al., 1998; Nagy & Salhi, 2007; Gndz, 2011;
Drexl & Schneider, 2013; Prodhon & Prins, 2014). All of these methodologies decompose the problem into ordinary
location, allocation, and routing problems, and find the solution of each part repeatedly, iteratively, or simultaneously
(Toyoglu, et al., 2012). Nagy and Salhi (2007) categorize those techniques into 4 methods, namely sequential,
clustering-base, iterative, and hierarchical methods. Additional analysis is required to yield and ensure a good quality
solution, such as robustness and debility analysis (Salhi & Nagy, 1999). None of the above mentioned research
attempted the optimal solution using the exact solution approach; therefore the exact solution approach presented in
this paper is the new addition in the LRPTW related research.

217

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

3. Formulation
The description of sets, parameters, and decision variables are described below.
Sets
M
potential depots
C
customers
Km homogeneous vehicles, which belong to each depot m
Parameters
cost of setting up the depot at site m,
fm
Qm depot capacity, m M
Ok vehicle cost, kKm
di
customers demand, i C
[ai,bi] time windows, i C
q
vehicle capacity

m M

cij

operating cost between pairs of depots and customers,

cmj

vehicle fixed cost on the starting vertices,

tij

travel time on arc (i,j) and the service time at vertex i,

(i, j ) C

m M and j C

(i, j ) C

Decision Variables
1 if depot m is opened or 0 otherwise (Constraints (11))
ym
xijk 1 if arc (i,j) is used by vehicle k or 0 otherwise (Constraints (12))
sik
time when vehicle k starts to service customer i (Constraints (10))
The LRPTW can be defined in the graph G = (V, A) where V consists of set M and set C while A is the set of all
arcs (i,j), i,jV. The standard formulation for the LRPTW is shown below:

min

f m ym  Ok xmjk  cij xijk

mM

subject to

mM jAkK

xijk

iA jAkK

mM k K m jV

(1)

i C

(2)

xmjk

m M , k K m

(3)

ximk

m M , k K m

(4)

h C , k K

(5)

k K

(6)

m M

(7)

jV
iV

xihk  xhjk

iV

jV

di xijk d q

iC

jV

di xijk d Qm ym

k K m iC

jV

xmjk  U

mM j AkK

xijk sik  tij  s jk d 0

(8)

(i, j ) A , k K m

(9)

218

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

ai d sik d bi
ym ^0,1`
xijk ^0,1`

i V , k K m

(10)

m M

(11)

(i, j ) A , k K m

(12)

The LRPTW objective (1) is to find a set of open depots and a set of arcs of minimal total cost composing the routes
to meet customers demand.
Constraint (2) indicates each customer must be visited exactly once. Constraints (3) and (4) are flow conservation
constraints, meaning that if the vehicle start from a depot m, it must return to the same depot. Constraint (5) is also a
flow conservation constraint requiring the vehicle to leave the customer h if it visits the customer h. Constraints (6)
and (7) are vehicle and depot capacity constraints, respectively. Constraint (8) computes and limits the number of
vehicle U used in the solution. Constraints (9) and (10) are time windows constraints implying that if the vehicle goes
from i to j, it must serve customer i before j within its specified time windows.
4. Branch-and-Price Algorithm
In optimization problem, there are too large to consider all the variables explicitly. Since most of the variables will
be non-basic and assume a value of zero in the optimal solution, only a subset of variables needs to be considered. One
of the powerful tool to tackle such problems is column generation. It follows the idea of the Dantzig-Wolfe
decomposition by splitting original formulation into two problems: the master and the sub problems. The master
problem considers only a subset of variables from the original while the sub problem (sometimes called pricing
problem) identifies the new variables. The objective function of the sub problem considers the reduced cost of the new
variables with respect to the current dual variables. At the beginning, the master problem is solved using an initial
solution. It can be any feasible solution that meets all constraints (Perrot, 2005). In this case, we start with the depoti-depot routes. From this step, the dual prices of each constraint in the master problem are obtained. Then, the reduced
cost is calculated and utilized in the objective function of the sub problem. After solving the sub problem, the variables
(called columns in the master problem) with negative reduced cost must be identified. These variables are then added
to the master problem and resolved iteratively. The process is repeated until the sub problem solution has only nonnegative reduced costs columns. Theoretically, at that instance, the solution of the master problem is the optimal
solution. If the variables return fractional values, branching schemes are required until all variables return integer
solutions. Recently, the extended works of branch-and-price algorithm solving VRPSSW (Qureshi, et al., 2009) and
VRPSTW (Bhusiri, et al., 2014) was proved to solve the LRPTW optimally (Ponboon, et al., 2014).
The master problem
The set, parameters, and decision variables in the master problem are;
Set
P
all feasible single vehicle routes with respect to vehicle capacity that start and end at the same depot
Parameter
cp

p P
operating cost of route, p P

aip

number of times that route p serves customer i,

dp

cumulative demand,

p P

Decision Variables
ym 1 if depot m is opened or 0 otherwise (Constraints (17))
zp
1 if route p is selected or 0 otherwise (Constraints (18))

and

i C

219

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

vpk

1 if vehicle kK is assigned the route pP and 0 otherwise

The master problem can be formulated as;

f m ym  c p z p

(13)

aip z p t 1

i C

(14)

d p v pk d Qm ym

m M

min

mM

subject to

pP

pP

(15)

k K m pP

zp U

(16)

pP

ym ^0,1`

z p ^0,1`

m M

(17)

p P

(18)

The objective function of the master problem (13) is to find a set of open depots and a set of routes of minimal total
cost (starting and ending at the open depots) and meet customers demand.
Constraints (14) ensure that each customer must be visited exactly once while constraints (15) indicate the total
demand of the customers on routes originating and ending at a particular depot does not exceed its capacity. Constraints
(16) limit the number of vehicle used in the solution.
The subproblem (pricing problem)
The pricing problem attempts to generate feasible routes with negative reduced costs to be added in the master
problem. This step is called column generation. The Elementary Shortest Path Problem with Resource Constraints
(ESPPRC) is used in the pricing problem to determine the shortest path between two vertices of a graph (origin and
destination depot) and ensures that each intermediate node is visited exactly once, i.e. a feasible route (Feillet, et al.,
2004; Qureshi, et al., 2009; Bhusiri, et al., 2014). As the capacity of the vehicles is assumed to be the same for all
vehicles stationed at each depot, therefore, m different sub problems need to be solved one for each m M . The
explicit formulation of the sub problem to be considered is given as follows:

min

cijmxij

(19)

i A j A

subject to

xmj

xim

(20)

jV

(21)

iV

xih  xhj

iV

(22)

jV

di xij d q

iC

jV

(23)

xij si  tij  s j d 0

(i, j ) A

(24)

ai d si d bi

i V

(25)

(i, j ) A

(26)

xij ^0,1`

220

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

The cost shown in the objective function (19) cijm , called reduced cost is given by:

cijm

cij  S i  d i P m  K m

(27)

Here i, m, and Km are dual variables from constraints (14), (15) and (16) corresponding to the customer i and depot
m, respectively. The aim of the sub problem is to find the routes with minimal reduced cost. The column generation
process ends when the optimal objective of the sub problem is either a non-negative value or zero.
The evaluation of branch-and-price algorithm on the LRPTW instances has been tested with modified Solomon
benchmark. More details about the validation and effectiveness are presented in Ponboon et al (2014).
5. Distribution Network
The distribution data from freight carrier have been tested with the branch-and-price algorithm to evaluate the cost
structure and environmental impact on the location-routing problem. The selected logistic company provides a variety
of services including warehousing, storage and inventory management, and door to door delivery. Fresh food, medical
specimens, and large items are among the commodities to be delivered on the daily basis (Teo, et al., 2015). The
transportation information was collected by the global positioning system (GPS) during the real time operation. It
provides the customers location, delivery routes, and traveling speed. Moreover, the additional travel times on the
road networks, which have never been collected by the GPS are collected from the traffic census data from Osaka city
(MLIT, 2010). Fig. 1 shows the location of depots and selected customers in the city of Osaka. The depots are located
at Toyonaka-shi, Minato-ku, and Nishinari-ku. The travel times between depot to the customer and customer to
customer are converted and calculated the shortest path using Dijkstras algorithm (Dijkstra, 1959). It must be noted
here that some private information are specifically classified by the carrier, therefore the logistics data, including
customers demand are randomly hypothetical. The time window width (bi - ai) and service time are constantly set
equal to 30 and 10 minutes, respectively.
Three influence factors to be evaluated on the test instance are depot and vehicle configuration, i.e. location, size,
and cost. All cost components must be translated into a same time horizon, which is a daily basis. Three sizes of depots
and vehicles, i.e. small, medium, and large, make a combination of nine scenarios. For the depot, one might argue it
is quite obvious that size and cost are interdependence since the larger depot space is more costly. However, the
warehousing cost varies by their locations. We estimated the depots cost (fm) and capacity (Qm) based on the data
available from commercial real estate service firm (JLRE, 2015). For the vehicle, the capacity (q) of 1,340 kg, 2 ton,
and 3 ton trucks represent the small, medium, large vehicles. The vehicle operating cost (cij) is 14.02 yen per minute
while the vehicle fixed costs (cmj) are 5,209, 7,441, and 10,418 yen per vehicle, respectively (Qureshi, et al., 2014;
Toyota, 2015). The morning and afternoon services are assumed to be exactly the same. Table 1 and 2 show the
parameters used in our calculation.

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

Fig. 1. Location of Depots and Customers in Osaka Road Network

221

222

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

Table 1. Parameters Used in Each Scenario
Scenario
1
2
3
4
5
6
7
8
9

Depot (m)
Capacity (Qm)
Cost (fm)
22,801
8,000
14,334
21,357
34,202
12,000
21,501
32,036
68,403
24,000
43,002
64,071

Vehicle
Capacity (q)
Cost (cmj)
1,340
5,209
2,000
7,441
3,000
10,418
1,340
5,209
2,000
7,441
3,000
10,418
1,340
5,209
2,000
7,441
3,000
10,418

Table 2. Customers Demand and Time Windows

Customer (i)

Demand (di)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

730
540
700
610
680
700
570
800
620
500
680
780
500
670
800
800
580
650
540
500
720
740
640
660
550
780
630
630
780
710

Time Windows
ai
bi
70
100
30
60
120
150
60
90
100
130
80
110
40
70
70
100
30
60
50
80
70
100
40
70
50
80
20
50
80
110
120
150
10
40
10
40
30
60
90
120
40
70
30
60
50
80
80
110
30
60
120
150
120
150
110
140
70
100
80
110

Service Time
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10

6. Result and Discussion

Computational tests of the algorithm were performed on a computer with AMD Phenom II X6 1100 T 3.20 GHz
processor, and 16 GB of RAM. The algorithm was programmed in MATLAB and using Gurobi Optimizer 5.6 as the
LP solver. Table 3 to 5 show the details of solution on LRPTW performed on the test problem represented the depot,
vehicle, and transport information respectively.

223

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

Table 3. Depot Information
Scenario
1
2
3
4
5
6
7
8
9

Depot
Cost
58,492
58,492
58,492
53,537
53,537
53,537
43,002
43,002
43,002

Opened Depot
D1
D2
D3
O
O
O
O
O
O
O
O
O
X
O
O
X
O
O
X
O
O
X
O
X
X
O
X
X
O
X

Customer Demand
D1
D2
D3
6,620
5,270
7,900
7,940
3,960
7,890
7,880
5,990
5,920
7,860
11,930
7,960
11,830
8,660
11,130
19,790
19,790
19,790
-

D1
0.83
0.99
0.99
-

Depot Load
D2
0.66
0.50
0.75
0.66
0.66
0.72
0.82
0.82
0.82

D3
0.99
0.99
0.74
0.99
0.99
0.93
-

Table 4. Vehicle Information

Scenario
1
2
3
4
5
6
7
8
9

Vehicle
Cost
78,135
74,410
72,926
78,135
74,410
72,926
78,135
74,410
72,926

D1
5
4
3
-

Vehicle Used
D2
D3
Total
4
6
15
2
4
10
2
2
7
6
9
15
4
6
10
3
4
7
15
15
10
10
7
7

D1
1,324.00
1,985.00
2,626.67
-

Average Load
D2
1,317.50
1,980.00
2,995.00
1,310.00
1,990.00
2,886.67
1,319.33
1,979.00
2,827.14

D3
1,316.67
1,972.50
2,960.00
1,325.56
1,971.67
2,782.50
-

Average Truckload
D1
D2
D3
0.99
0.98
0.98
0.99
0.99
0.99
0.88
0.99
0.99
0.98
0.99
0.99
0.99
0.96
0.93
0.98
0.99
0.94
-

Table 5. Transportation Information

Scenario
1
2
3
4
5
6
7
8
9

Transport
Cost
13,223.40
9,047.89
7,623.90
14,060.09
9,706.51
7,557.83
15,082.30
10,418.26
8,300.97

D1
350.71
267.23
219.55
-

Total Travel Time

D2
D3
243.21 349.25
125.59 252.54
183.30 140.94
421.08 581.78
302.84 389.49
276.49 262.59
1075.77
743.10
592.08
-

Total
943.18
645.36
543.79
1,002.86
692.33
539.08
1,075.77
743.10
592.08

Travel Time per Vehicle

D1
D2
D3
35.07 30.40
29.10
33.40 31.40
31.57
36.59 45.82
35.24
35.09
32.32
37.86
32.46
46.08
32.82
35.86
37.15
42.29
-

In Table 3, the O mark in column Opened Depot represents the depot to be opened while X mark means the
unselected depot. The capacity restriction (constraints (7)) caused all three depots (small size depots) to be opened in
scenario 1-3, two depots (medium size depots) in scenario 4-5, and one depot (large size depot) in scenario 7-9.
Customer Demand column refers to the total demand of customers which belong to each depot, while the figures in
Depot Load column are the demand/capacity ratio. It can be observed that in the small and medium size depot
scenarios (scenario 1-6), the depot 2 was dominated by other depots. While the depot load of depot 1 and 2 were
reported up to 0.83-0.99, the demand on depot 2 consumes only 0.50-0.66. The main reason is because depot 2 is
located in an industrial area in Minato-ku, which far from the city center compared to the others. It can serve only a
fractional portion to the customers. Nevertheless, the benefit of locating in that area makes depot 2 accounts the lowest
depot fixed cost (37% and 33% lower than depot 1 and 3, respectively). Therefore, considering together with vehicle
and transportation costs, it was only one depot to be selected for larger size depot scenarios (scenario 7-9).
Although the portion of depot cost varies on the scenarios, the vehicle cost remains constant as shown in Table 4
since the customers demand does not change. Fifteen small size vehicle, ten medium size vehicle, and seven large
size vehicle are used regardless of the depot size. Clearly, the numbers of vehicle used at the depot 2 (shown in column
Vehicle Used) are less than other depot because of the same reason mentioned earlier.
In general, when the time windows constraints are taken into account on routing problem, the truckload is
significantly low comparing with ordinary VRP or LRP. It happens because the time restriction does not allow the
vehicle to visit too many number of customers. Nevertheless, the truck load in our network is relatively high in all

224

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

scenarios, presented 0.94-0.99 as shown in Average Truck Load column. It can be concluded that the 30 minute
time window width (bi-ai) in this network is wide enough for each vehicle to deliver the goods in effective manner.
For the routing part shown in Table 5, if the depot size did not change, it is undoubtedly true that the travel time
reduced when the vehicle size increased. The numbers of customers that can be visited by small, medium, and large
vehicles are 2, 3, and 4.28 customers on average. One of the main benefits of using larger size vehicle over small size
vehicle is that the non-profit empty truck on return trips has been reduced dramatically, 34% of medium vehicle and
58% of large vehicle. In total, the medium and large size vehicles can save the distribution cost, i.e. vehicle plus
transport costs, 9% and 12%, respectively.
Finally, the results of the overall cost, including depot, vehicle, and transport costs has proved that the large size
depot serving by large size vehicles in scenario 9 satisfied the LRPTW in our distribution network as shown in Fig. 2.
Although the depot 2 located in Minato-ku seems not to be an appropriate location in small and medium size depot
scenario, but it proves to be the best location for the large size depot scenario.
160,000
140,000
120,000
100,000
Transport Cost
Vehicle Cost
Depot Cost

80,000
60,000
40,000
20,000
0
Small Vehicle

Medium
Vehicle
Small Depot

Large Vehicle Small Vehicle

Medium
Vehicle
Medium Depot

Large Vehicle Small Vehicle

Medium
Vehicle

Large Vehicle

Large Depot

Fig. 2. LRPTW Results on Testing Distribution Network

7. Conclusion
This study presented the evaluation of cost structure on different environmental factors of LRPTW which is rarely
examined by exact algorithm. The branch-and-price algorithm was implemented to ensure that the quality of the
solution is reliable. The distribution network was extracted from the real operation of logistics firm in Osaka while the
travel time was collected by the GPS and the traffic census data of the Osaka city. The main decision factors for
LRPTW were the location of the depot, depot size, and vehicle size. It was found that the large size depot together
with large size vehicle scenario was among the best scheme in our case study. The depot located in Minato-ku was the
best location for the large size depot scenario by the fact that it is located in the low-price area. Furthermore, the large
size vehicle of 3 ton vehicle is proved to be the best choice. Regardless of the highest vehicle fixed cost, the number
of vehicle can be reduced to only 7 vehicles, but still remain the high truck load, making the lowest delivered cost
option.
To implement the location-routing scheme into the real supply chain operation, however, it requires more
comprehensive information to achieve the goal. This tested distribution network limits only the portion of selected
customers. To handle the large network instances, the approximate approach or the combination of exact and
metaheuristic approach as a hybrid algorithm is needed to be explored in the future in order to accelerate the process
without losing much on the exact optimal solution.

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

225

References
Akca, Z., Berger, R., & Ralphs, T. (2009). A Branch-and-Price Algorithm for Combined Location and Routing Problems under Capacity
Restrictions. In J. Chinneck, B. Kristjansson, M. Saltzman, J. W. Chinneck, B. Kristjansson, & M. J. Saltzman (Eds.), Operations Research
and Cyber-Infrastructure (Vol. 47, pp. 309-330). Springer US. doi:10.1007/978-0-387-88843-9_16
Baker, E., & Schaffer, J. (1986). Computational experience with branch exchange heuristics for vehicle routing problems with time window
constraints. American Journal of Mathematical and Management Sciences, 6, 261-300.
Balakrishman, A., Ward, J. E., & Wong, R. T. (1987). Integrated facility location and vehicle routing models: Recent work and future prospects.
American Journal of Mathematical and Management Sciences, 7, 35-61.
Baldacci, R., Mingozzi, A., & Rolfer Calvo, R. (2011). An Exact Method for the Capacitated Location-Routing Problem. Operations Research,
59(5), 1284-1296.
Belengure, J., Benavent, E., & Prins, C. (2011). A Branch-and-Cut Method for the Capacitated Location-Routing Problem. Computers &
Operations Research, 38, 931-941.
Berger, R., Coullard, C., & Daskin, M. (2007). Location-Routing Problems with Distance Constraints. Transportation Science, 41, 29-43.
Berman, O., Jaillet, P., & Simchi-Levi, D. (1995). Location-routing problems with uncertainty. In Z. Drezner, Facility Location: A Survey of
Applications and Methods (pp. 427-452). New York: Springer.
Bhusiri, N., Qureshi, A. G., & Taniguchi, E. (2014). The trade-off between fixed vehicle costs and time-dependent arrival penalties in a routing
problem. Transportation Research Part E: Logistics and Transportation Review, 62, 1-22.
Bruns, A., & Klose, A. (1996). A Locate First - Route Second Heuristic for a Combined Location-Routeing Problem. In U. Zimmermann, U.
Derigs, W. Gaul, R. Mhring, & K. Schuster (Eds.), Operations Research Proceedings 1996 (pp. 49-54). Berlin: Springer.
Christofides, N., & Eilon, S. (1969). An algorithm for the vehicle dispatching problem. Operational Research Quarterly, 20(3), 309-318.
Contardo, C., Cordeau, J. F., & Gendron, B. (2013(a)). A Computational Comparison of FLow Formulations for the Capacitated Location-Routing
Problem. Discrete Optimization, 10(4), 263-295.
Contardo, C., Cordeau, J. F., & Gendron, B. (2013(b)). An Exact Algorithm based on Cut-and-Column Generation for the Capacitated LocationRouting Problem. INFORMS Journal on Computing, 26(1), 88-102.
Cook, T., & Russell, R. (1978). A simulation and statistical analysis of stochastic vehicle routing with timing constraints. Decision sciences, 9(4),
673-687.
Cost optimisation in freight distribution with cross-docking: N-echelon location routing problem. (2012). PROMET-Traffic&Transportation, 24(2),
143-149.
Current, J., Daskin, M., & Schilling, D. (2002). Discrete network location models. In Z. Drezner, & H. Hamacher (Eds.), Facility Location:
Applications and Theory (pp. 81-108). Berlin: Springer.
Dantzig, G. B., & Wolfe, P. (1960). Decomposition principle for linear programs. Operations Research, 8(1), 101-111.
Desrochers, M., Desrosiers, J., & Solomon, M. (1992). A new optimization algorithm for the vehicle routing problem with time windows.
Operations research, 40(2), 342-354.
Desrosiers, J., Dumas, Y., Solomon, M., & Soumis, F. (1995). Time constrained routing and scheduling. In M. Ball, T. Magnanti, C. Monma, & G.
Nemhauser (Eds.), Handbooks in operations research and management science (pp. 35-139). North-holland.
Dijkstra, E. W. (1959). A note on two problems in connexion with graphs. Numerische mathematik, 1(1), 269-271.
Drexl, M., & Schneider, M. (2013). A survey of location-routing problems. Mainz: Johannes Gutenberg University.
Dror, M. (1994). Note on the complexity of the shortest path models for column generation in VRPTW. Operations Research, 42(5), 977-978.
Feillet, D., Dejax, P., Gendreau, M., & Gueguen, C. (2004). An exact algorithm for the elementary shortest path problem with resource constraints:
Application to some vehicle routing problems. Network, 44(3), 216-229.
Friedric, C. J. (1929). Alfred Weber's Theory of the Location of Industries. Chicago: University of Chicago Press.
Gonzalez-Feliu, J., Ambrosini, C., & Routhier, J. L. (2012). New trends on urban goods movement: Modelling and simulation of e-commerce
distribution. European Transport, 50, 23.
Gndz, H. I. (2011). The Single-Stage Location-Routing Problem with Time Windows. In J. Bose, H. Hu, C. Jahn, X. Shi, R. Stahlbock, S. VoB,
J. W. Bse, H. Hu, C. Jahn, X. Shi, R. Stahlbock, & S. Vo (Eds.), Computational Logistics (pp. 44-58). Hamburg: Springer Berlin Heidelberg.
Hakimi, S. L. (1964). Optimum Locations of Switching Centers and the Absolute Centers and Medians of a Graph. Operations Research, 12(3),
450-459.
Higgins, J. C. (1972). On the merits of simple models in distribution planning. International Journal of Physical Distribution, 2(2), 144-148.
JLRE. (2015). Warehouse for rent/Logistics real estate. Retrieved 1 25, 2015, from http://lw.e-sohko.com/
Knight, K., & Hofer, J. (1968). Vehicle scheduling with timed and connected calls: A case study. OR, 19(3), 299-310.
Laporte, G. (1988). Location-routing problems. In B. L. Golden, & A. A. Assad, Vehicle Routing: Methods and Studies (pp. 163-198). Amsterdam:
North-Holland.
Laporte, G. (1989). A survey of algorithms for location-routing problems. Investigacion Operativa, 1, 93-123.
Laporte, G., Nobert, Y., & Arpin, D. (1986). An Exact Algorithm for Solving a Capacitated Location-Routing Problem. Annals of Operations
Research, 6, 293-310.
Laporte, G., Nobert, Y., & Taillefer, S. (1988). Solving a family of multi-depot vehicle routing and location-routing problems. Transportation
Science, 22(3), 161-172.
Larporte, G., & Nobert, Y. (1981). An Exact Algorithm for Minimizing Routing and Operating Costs in Depot Location. European Journal of
Operational Research, 6, 224-226.
Larporte, G., Louveaux, F., & Mercure, H. (1989). Models and Exact Solutions for a Class of Stochastic Location-Routing Problems. European
Journal of Operational Research, 39, 71-78.
Larporte, G., Nobert, Y., & Pelletier, P. (1983). Hamiltonian Location Problems. European Journal of Operational Research, 12, 82 - 89.
Lawrence, R. M., & Pengilly, P. J. (1969). The number and location of depots required for handling products for distribution to retail stores in
South-East England. Operational Research Quarterly, 20(1), 23-32.

226

Sattrawut Ponboon et al. / Transportation Research Procedia 12 (2016) 213 226

Madsen, O. (1976). Optimal scheduling of trucks - A routing problem with tight due times for delivery. In H. Strobel, R. Genser, & M. Etschmaier
(Eds.), Optimization Applied to Transportation Systems (pp. 210-223). Laxenburgh: International Institute for Applied System Analysis.
Madsen, O. B. (1983). Methods for solving combined two level location-routing problems of realistic dimensions. European Journal of Operational
Research, 12(3), 295-301.
Maranzana, F. (1964). On the location of supply points to minimize transport costs. Operational Research Quarterly, 15, 261-270.
Min, H., Jayaraman, V., & Srivastara, R. (1998). Combined location-routing problem: A synthesis and future research directions. European Journal
of Operational Research, 108(1), 1-15.
MLIT. (2010). Ministry of Land, Infrastructure, Transport, and Tourism: 2010 nationwide road and street traffic situation investigation. Retrieved
1 20, 2012, from http://www.mlit.go.jp/road/census/h22-1/
Nagy, G., & Salhi, S. (2007). Location-Routing: Issues, Models, and Methods. European Journal of Operational Research, 177, 649-672.
Perrot, N. (2005). Integer Programming Column Generation Strategies for the Cutting Stock Problem and its Variants. University of Bordeaux.
PhD Thesis, University Bordeaux 1, France.
Ponboon, S., Qureshi, A. G., & Taniguchi, E. (2014). Exact Solution for Location-Routing Problem with Time Windows using Branch-and-Price
Algorithm. Transportation Research Part E: Logistics and Transportation Review, (Submitted).
Prodhon, C., & Prins, C. (2014). A survey of recent research on location-routing problems. European Journal of Operational Research, 238(1), 117. Retrieved from http://dx.doi.org/10.1016/j.ejor.2014.01.005
Pullen, H., & Webb, M. (1967). A computer application to a transport scheduling problem. The Computer Journal, 10(1), 10-13.
Qureshi, A., Taniguchi, E., & Yamada, T. (2009). An Exact Solution Approach for Vehicle Routing and Scheduling Problem with Soft Time
Windows. Transportation Research Part E: Logistics and Transportation Review, 45(6), 960-977.
Qureshi, A., Taniguchi, E., Thompson, R., & Teo, J. (2014). Application of exact route optimization for the evaluation of a city logistics truck ban
scheme. International Journal of Urban Sciences, 18(2), 117-132.
Rand, G. (1976). Methodological Choices in Depot Location Studies. Operational Research Quarterly, 27, 241-249.
Russell, R. (1977). Technical NoteAn Effective Heuristic for the M-Tour Traveling Salesman Problem with Some Side Conditions. Operations
Research, 25(3), 517-524.
Salhi, S., & Nagy, G. (1999). Consistency and Robustness in Location-Routing. Studies in Locational Analysis, 13, 3-19.
Salhi, S., & Rand, G. (1989). The Effect of Ignoring Routes when Locating Depots. European Journal of Operation Research, 39, 150-156.
Solomon, M. (1986). On the worstcase performance of some heuristics for the vehicle routing and scheduling problem with time window
constraints. Networks, 16(2), 161-174.
Solomon, M. (1987). Algorithms for the vehicle routing and scheduling problems with time window constraints. Operations research, 35(2), 254265.
Solomon, M. M., & Desrosiers, J. (1988). Survey Paper - Time Window Constrained Routing and Scheduling Problems. Transportation Science,
22(1), 1-13.
Srivastava, R. (1986). Algorithms for solving the location-routing problem. Columbus, Ohio: PhD dissertation, The Ohio State University.
Srivastava, R., & Benton, W. C. (1990). The location-routing problem: Consideration in physical distribution system design. Computers and
Operations Research, 6, 427-435.
Taniguchi, E., Thompson, R. G., Yamada, T., & Duin, R. V. (2001). City Logistics. Oxford: Elsevier Science Ltd.
Teo, J., Taniguchi, E., Qureshi, A., Mai, V., & Uchiyama, N. (2015). Towards a safer and healthier urbanization by improving land use footprint
of last-mile freight delivery. Washington D.C.: 94th Transportation Research Board Annual Meeting Compendium of Papers.
Toyoglu, H., Karasan, O., & Kara, B. (2012). A New Formulation Approach for Location-Routing Problems. Networks and Spatial Economics,
12(4), 635-659.
Toyota. (2015). Rate and Car Model List. Retrieved 1 25, 2015, from https://rent.toyota.co.jp/en/rental/main09.html
Von Boventer, E. (1961). The relationship between transportation costs and location rent in transportation problems. Journal of Regional Science,
3, 27-40.
Watson-Gandy, C., & Dohrn, P. (1973). Depot location with van salesmen - A practical approach. Omega, 1(3), 321-329.
Webb, M. (1968). Cost functions in the location of depots for multi-delivery journeys. Operational Research Quarterly, 19(3), 311-320.