Green Freight Distribution: A Case Study

in Greece

Grigorios D. Konstantakopoulos, Evripidis P. Kechagias, Sotiris P. Gayialis,

and Ilias P. Tatsiopoulos

Keywords Vehicle routing problem · Electric vehicles · Time windows ·

Conventional vehicles · Genetic algorithm · Greenhouse gas emissions

Introduction

Global warming has increased the concerns of nations, unions and individuals and
the main actions for addressing this crisis is through legislation made by the
competent bodies, either at the national or European level or globally. Undoubdely,
the legislation and concerns have affected the means and operations of traditional
supply chains (Chaabane et al., 2012; Zhen et al., 2019) as companies seek to find
solutions that offer the best trade-off between cost and environmental protection. In
this direction, the improvement of sustainability as a business strategy has led to
green supply chain which integrates environmental thinking (Micheli et al., 2020).
According to a research of the United States Environmental Protection Agency,
the transportation sector is the largest contributor to anthropogenic US greenhouse
gas (GHG) emissions, accounting for more than 29%, highest even from Electricity
and Industry sectors. The corresponding figure in Europe is 27% according to the
European Environment Agency. Among the sources used in this sector (Maritime,
Rail, Road, etc.), road transportation cause more than 71% of emissions. More
specifically light commercial (LC) vehicles (vans) produced around 2.5% of the
total EU emissions of CO2. On the other hand, Heavy—Duty Trucks that represent
only 4% of the on-road fleet in the EU, account for 6% of the total EU CO2
emissions.
On this premise, EU set target levels and regulations regarding GHG emissions,
in its effort to reduce them. However, these targets are hard to be detained in many
cases, since the commercial vehicles used in the distribution process are of old

G. D. Konstantakopoulos (✉) · E. P. Kechagias · S. P. Gayialis · I. P. Tatsiopoulos

School of Mechanical Engineering, Sector of Industrial Management and Operational Research,
National Technical University of Athens, Athens, Greece

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023 49

N. F. Matsatsinis et al. (eds.), Operational Research in the Era of Digital
Transformation and Business Analytics, Springer Proceedings in Business and
Economics, https://doi.org/10.1007/978-3-031-24294-6_6
50 G. D. Konstantakopoulos et al.

technology, with all that entails. More specifically, Greece has the oldest light
commercial (LC) and medium and heavy (MHC) commercial vehicles among the
countries of the European Union (EU) according to the European Automobile
Manufacturers’ Association (ACEA) “Vehicles in use” report (European Automo-
bile Manufacturers Association, 2019). The average age of LC vehicles in EU is
10.9 years, and of MHC 12.4 years, while the corresponding figures in Greece are
almost twice, and more specifically, 18.9 years and 20.9 years respectively. Based on
the same research, most EU countries have limited petrol-engine vehicles and have
invested mainly in diesel vehicles, as shown in Table 1. Greece remains relatively
behind in this case as well, since it holds the smallest percentage of diesel trucks, and
the highest in unknown fuel type, due to the lack of information since the vehicles
are almost 20 years old. The age of the vehicle and the type of the engine are the
main characteristics that determine the fuel costs, the noise and CO2 emissions,
which is of great importance nowadays. The distribution cost, probably remains the
most significant factor over the years, not only for logistics companies, in their effort
to be competitive in the market but for consumers as well, in their search for the most
value for money product. On the other hand, noise and GHG emissions are signif-
icant for inhabitants, especially in city centers, and their reduction has become the
number one priority for the EU.
Considering all the above characteristics, the transfer from the current state into
the future, which will include a complex distribution network with multiple chal-
lenges, imposes the need for investments in new technologies related to vehicles, as
well as to routing software. In the current research, in an attempt to consider both the
vehicles in use and routing software, we address the vehicle routing problem and
some of its variants that are considered important for the distribution process.
Specifically, we focus on the multiple variables and constraints that logistics com-
panies face, such as time windows, simultaneous pickups and deliveries, the costs,
driving range and emissions of vehicles, forming the Vehicle Routing Problem with
Time Windows and Simultaneous Pickups and Deliveries (VRPSPDTW) with
limited driving range. The driving range is considered mostly in electric vehicles
due to their limitation, and therefore some researchers could relate the studied
problem to the Electric VRPSPDTW. Finallly, given the fact that all specifications
related to vehicles are considered in the current research, the environmental and
economic benefits in Greece, by advancing the fleet of vehicles from conventional to
electric are estimated. To address this complex problem optimally, we develop a
genetic algorithm.
The remaining part of this paper is organized as follows: the second section
presents a literature review on the vehicle routing problem. The third section
formulates and describes the Vehicle Routing Problem with Simultaneous Pickups
and Deliveries & Time Windows (VRPSPDTW). The fourth section presents the
algorithmic approach, the fifth section the computational data, and the final section
includes the research conclusions.
Table 1 Comparison of vehicles in Greece and EU
Type of commercial vehicles Region Average age Petrol Diesel Hybrid electric Electric LPG + natural gas Other + unknown
Light Greece 18.9 41.50% 37.30% 0.00% 0.00% 0.00% 21.20%
Green Freight Distribution: A Case Study in Greece

EU 10.9 7.10% 91.20% 0.00% 0.30% 1.30% 0.10%

Medium and heavy Greece 20.9 0.10% 56.10% 0.00% 0.00% 0.00% 43.80%
EU 12.4 1.00% 98.30% 0.00% 0.00% 0.40% 0.20%
51
52 G. D. Konstantakopoulos et al.

Vehicle Routing and Scheduling Problem

Software can efficiently address the routing procedure and reduce operational costs
only if variants, variables and constraints that logistics companies face in their daily
distribution operations have been clearly identified and integrated into the system.
The main form of the VRP in which researchers and practitioners were based on was
proposed by Dantzig and Ramser (1959). In this initial form of the problem, the aim
is to find the optimum routing of a fleet of gasoline delivery trucks between a bulk
terminal and a large number of service stations supplied by the terminal. Conse-
quently, the CVRP was proposed, in which a fixed fleet of delivery vehicles of
uniform capacity must service known customer demands for a single commodity
from a common depot at minimum transit cost (Tarantilis et al., 2004). The CVRP
constitutes the most commonly considered variant of the VRP as it greatly simplifies
the problem.
Additionally, according to Gayialis et al. (2019), the VRP with Time Windows
(VRPTW) is among the most studied variants of the VRP. It reflects most in real-life
distribution cases faced by logistics companies, as most customers determine a time
slot that a vehicle must arrive. In most cases time windows are hard, meaning that if a
vehicle arrives before the start of the time windows the vehicle must wait till it opens,
while arriving after the end of the time windows leads to unfulfilled deliveries (Li &
Lim, 2003; Tan et al., 2006; Yu et al., 2011). The same applies in the VRP with
simultaneous pickups and deliveries (VRPSPD), as most customers in real-life have
both a delivery and a pickup demand, so that delivery items start from the depot, and
pick up items return to the depot (Hornstra et al., 2020; Konstantakopoulos et al.,
2021; Wassan & Nagy, 2014). The two variants of the VRP mentioned above are the
most studied variants, according to Konstantakopoulos et al. (2020).
Furthermore, while the distribution of goods has attracted the interest of
researchers for more than 60 years, it is only the last few years that the GHG
emissions are considered in the VRP, due to the ecological awareness of transport
sustainability, leading to green logistics. This statement is confirmed by multiple
studies on the field such as those of Konstantakopoulos et al. (2020), Elshaer and
Awad (2020) and Braekers et al. (2016). In the case of Electric VRP (EVRP), the
vehicles that are used for the distribution process, have electric batteries. This
implies a limited driving range, as well as visitations to Charging Stations (CS) in
order to recharge the batteries, when needed, for executing the routes (Erdelić et al.,
2019). The problem studied and presented in the current paper is highly related to the
work of Erdelić et al. (2019). However, the EVRP can be much more challenging
and complex if we consider that battery consumption and CO2 emissions are related
to the load and the speed of the vehicle (Lin et al., 2016). Erdelić and Carić (2019)
present a survey on the EVRP, and more specifically they focus on consumption
models, as well as on additional emerged EVRP variants, such as hybrid vehicles,
charging stations, charging functions and dynamic traffic conditions. However, we
make the assumption that consumption and CO2 emissions are fixed, irrespectively
of the load and the speed of the vehicle.
Green Freight Distribution: A Case Study in Greece 53

The importance of the above research studies, that address VRP variants and real-
life distribution cases through algorithmic approaches cannot be challenged. How-
ever, it is important that these solutions also lead to real-life applications and systems
that can be exploited by practitioners. In this direction a research methodology was
developed, aiming to find articles that propose algorithms that are also integrated
into real-life applications and systems. Specifically, Scopus database was exploited
by applying the term (“vehicle routing problem” AND (“information system” OR
“system” OR “software” OR “Application”) AND (“metaheuristic” OR “heuristic”
OR “exact”) AND (“case study”) in Title, Abstract, Keywords of journals. Among
the articles that emerged from the research, only one addresses all the above VRP
variants which is the work of Kechagias et al. (2020). The authors propose a
multiobjective algorithm that aims to minimize both emissions and costs, and that
is also integrated into a routing system. Additionally, from the above research
methodology the conclusion that only a limited number of articles develop algo-
rithms that are also itegrated into real-life applications is extracted.
Finally, the objective of the present paper is first to propose a genetic algorithm
that can optimally address the Vehicle Routing Problem with Simultaneous Pickups
and Deliveries, Time Windows (VRPSPDTW) and limited driving range, as well as
handle real-life data of customers, vehicles and refueling stations. Consequently, the
algorithm is tested in real-life distribution cases, addressed by a logistics company in
Greece, in order to calculate and compare the economic and environmental footprint,
when the fleet of vehicles is only composed of petrol, diesel and electric vehicles.
The remaining part of this paper is organized as follows: this section presents the
description of the VRPSPDTW with limited driving range. Third section describes
the studied problem and presents the mathematical formulation. Fourth section
presents the algorithmic approach, fifth section the computational data and proposed
solutions, while in final section the conclusions are analyzed.

Problem Description and Formulation

The VRPSPDTW is a variant of VRP that combines time windows and simultaneous
pickups and deliveries set by customers. More specifically, VRPSPDTW is defined
by a set of identical in terms of capacity (Q) vehicles denoted by K and by a direct
network G = {F [ N, A}, where vertex N = {0, 1, . . .n} is the set of customers,
vertex F = {n + 1, n + 2, . . .n + s} is the set of charging and gas stations (CGS), and
A = (i, j) : i ≠ j, i, j 2 F [ N denotes the set of arcs. Node 0 represents the central
depot, while N = N/{0} represents the customers. Every arc (i, j), which is a path
from node i to node j, is characterized by a distance which is indicated as dij and by a
travel time tij. Each client i 2 N is characterized by a delivery quantity qi, a pickup
quantity ri, a service time si for unloading and a time window (ei, li), where ei is the
opening and li the closing time. The depot and the CGSs are also characterized by a
service time si and a time window (ei, li) that corresponds to the working hours. The
start of service of vehicle k at node j, is denoted by akj and depends on the time of
54 G. D. Konstantakopoulos et al.

arrival as vehicles arriving before the beginning of the time window must wait until
service is possible ei. If the vehicle arrives between ei and li, the service starts
immediately, while no vehicle may arrive after li. Additionally, each vehicle k has
a specific driving range drk and a distance consumption rate f k. The decision variable
xkij is equal to 1 if vehicle k executes the route from node i to node j. Otherwise xkij is
equal to 0. The second decision variable, yki indicates the quantity of items on board
of vehicle k when departing from node i. The third decision variable is gki , which
indicates the remaining charge or gas level when arriving at customer i, according to
the vehicle type. The objective of the VRPTW is to service all customers in N, while
minimizing the distribution cost. Additionally, the fleet of vehicles is represented by
set K, while each vehicle is indexed by k 2 K. Each vehicle k is characterized by its
capacity Qk, its fixed cost fk, its variable cost vc, Finally, all the essential variables are
gathered and presented below before the mathematical model of the VRPSPDTW.

dij Distance between node i and node j

tij Transit time between node i and node j
qi Delivery demand of node i
ri Pickup demand from node i
ei Node’s i open time
li Node’s i close time
si Service time at node i
ykij Quantity of items on board of vehicle k when departing from node i
drk Driving range of vehicle k
aki Start of service of vehicle k at customer j
wki Waiting time of vehicle k at node i
gki Remaining gas or power of vehicle k when departing from node i
fk Distance consumption rate of vehicle k

X XX X XX
fk xk0j þ vk d ij xkij ð1Þ
k2K k2K j2N  k2K k2K ði, jÞ2A
X X
qi xkij ≤ Q, 8k 2 K ð2Þ
i2Ν j2Ν
X X
r i xkij ≤ Q, 8k 2 K ð3Þ
i2Ν j2N
X  
xkij yki þ r j - qj ≤ Q, 8k 2 K ð4Þ
ði, jÞ2A
XX
xkij = 1, 8i 2 N ð5Þ
k2K j2Ν
Green Freight Distribution: A Case Study in Greece 55

Fig. 1 Ilustration of the VRPSPDTW

X
xk0j = 1, 8k 2 K ð6Þ

j2Ν
X X
xkil - xklj = 0, 8l 2 Ν , 8k 2 K ð7Þ
i2Ν j2Ν
X
xkj0 = 1, 8k 2 K ð8Þ

j2Ν
X
aki þ si þ xkij t ij ≤ lj , 8i 2 Ν, 8k 2 K ð9Þ
j2Ν
( )
X  
akj = max xkij aki þ si þ t ij , ej , 8i 2 N, 8k 2 K ð10Þ
j2N

0 ≤ gkj ≤ dr k , 8j 2 A, 8k 2 K ð11Þ

gki - f k d ij xkij ≥ 0, 8ði, jÞ 2 A, 8k 2 K ð12Þ

xkij 2 f0, 1g, 8ði, jÞ 2 A, 8k 2 K ð13Þ

yki ≥ 0, 8i 2 N, 8k 2 K ð14Þ

In the VRPSPDTW the objective is the minimization of the distribution cost and
is described by function (1). Constraints (2) and (3) state that the total quantity of
products delivered and collected in total cannot exceed the capacity of the vehicle,
while constraint (4) states that, at any stage of the distribution process, the total
quantity of products in the vehicles cannot exceed its capacity. Eq. (5) states that
each customer must be served exactly once and by one vehicle since xkij can obtain
only 1, or 0 as values. The set of constraints (6)–(8) make sure that each vehicle starts
from the depot and visits and serves the customers that belong in the route before it
finally returns to the depot. Constraint (9) indicates that for a trip from node i to node
j, no vehicle may arrive at customer j after the end of the time window, lj, while
constraint (10) that the earliest time for starting serving customer j is related to the
time of arrival, depending on whether the time of arrival is before or after the start of
the time window ei. Constraints (11) and (12) ensure that fuel/battery feasibility of
56 G. D. Konstantakopoulos et al.

each vehicle. An example of the VRPSPDTW with limited driving range is illus-
trated in Fig. 1.

Algorithmic Approach

Over the years, multiple algorithms have been developed for solving the different
variants of the VRP. Initially, exact and heuristic algorithms were implemented,
even if each category has specific disadvantages. More specifically, exact algorithms
can offer optimal solutions but cannot handle more than 100 customers in most
cases, due to the increased computational time needed. On the other hand, heuristics
can handle large-scale instances in limited computational time, but at the expense of
solution quality. Finally, metaheuristic algorithms balance both factors, as they offer
near-optimal solutions in short computational time. Therefore, a metaheuristic
algorithm, and more accurately a genetic algorithm (GA), is developed for
addressing the VRPSPDTW with limited driving range.
Genetic Algorithms (GA) belong in the class of Evolutionary algorithms, and are
inspired by the process of natural selection. Initially, a set of solutions known as
parents population Pt is constructed. The parent population Pt is of size N. In the
present paper, we apply the time-oriented nearest neighbor algorithm, as proposed in
Solomon (1987), for the construction procedure. After applying the construction
heuristic algorithm, the solutions are evaluated according to the fitness values and
sorted according to the non-dominated sorting. The sorting procedure aims to
categorize solutions into fronts (F1, F2,. . ., Fk + 1) as shown in Fig. 2. The first
front, F1, includes all non-dominated solutions which means that the solutions of this
set are not dominated by any other solution. F2 front includes solutions that are
dominated only by solutions belonging on F1 front. The same applies to the rest
fronts. Moreover, the main characteristic of solutions belonging in the same front is

Fig. 2 Sorting procedure

Green Freight Distribution: A Case Study in Greece 57

that while moving from one solution to another, there is always a certain amount of
sacrifice in one objective to achieve a certain amount of gain in another objective as
shown in Fig. 2.
The no-dominated procedure is important so that to identify “superior” parents
that will be combined through operators for creating even better offsprings/children.
Each pair of parents creates two new offsprings by using the crossover and mutation
procedures. The crossover and mutation operators that are applied in our case are the
same that are used in Ombuki et al. (2006), constituting key pieces of the proposed
GA. More specifically, from the P set, two parents (Pi, Pi + 1) are selected, and from
each one, a single route is also selected. The customers that belong to each route are
then removed from the opposite parent. More specifically, customers selected from
parent Pi are removed from parent Pi + 1, and reversely. Since all customers must be
served, they must also be contained in the chromosome. The insertion of a customer
is made in the position of the chromosome that minimizes the cost. After the
insertion procedure, two new chromosomes, known as offsprings, are generated.
As for the the mutation operator, swap node operator is applied. In this case, two
customers (genes) of the chromosome are randomly selected and swapped. The main
contribution of this procedure is the differentiation of chromosomes, which may help
in escaping from local optimum. Through this procedure, the set Qt is filled with N
new solutions. The merged population (Pt[Ot), composed of parents and offsprings
is also sorted according to the fittnes values so that to select the N best solutions that
will be used as parents in the next generation Pt + 1. When the maximum number of
generations has been reached, the Pareto-optimal front is given as output of the
algorithm. The Pareto-optimal front is the same as F1 front, which includes all
non-dominated solutions. In the present paper, the population size in 30, and the
maximum number of generations is 30. The general framework of the algorithm is
given in Fig. 3.
Furthermore, each solution in the specific GA is presented by an array of length
equal to the number of customers. The objects of the array, called genes, contain
integers that correlated to the customers’ ID. Additionally, the sequence of genes
determines the visitations of vehicles to customers, as well as the routes. Each time
the capacity or the time window constrain is violated, a new route starts. However, in
the case of vehicles, that have limited driving range, we also consider the need either
of recharging before the battery is discharged (electric vehicles), either of refueling
the tank (conventional vehicles). Therefore, between two consecutive genes, a
charging or a gas station (Fuel Station—FS) may intervene, but simultaneously the
station’s ID is not contained in the chromosome. Additionally, both the time and the
distance required for the vehicle to reach the FS are taken into account. In Fig. 4, a
chromosome with nine customers is presented as an example. In case the driving
range of a vehicle, which has fully charged batteries, or a full tank, is not enough for
executing a route, then the vehicle must visit a CS or a GS, according to the type of
vehicle. For example, in Route 1 of Fig. 3, after the vehicle serves customer 6, needs
to visit a FS for recharging or refueling, in order to be able to execute the route. The
possibility of a visit to a fuel station is greater in electric vehicles due to their limited
driving range.
58 G. D. Konstantakopoulos et al.

Fig. 3 GA framework

Fig. 4 Chromosome structure

Green Freight Distribution: A Case Study in Greece 59

Finally, the objective function in the present paper, which is also used in the
fitness-based procedure, is the minimization of the total distribution cost. The total
cost is composed of the fixed cost (cost per vehicle used) and the variable cost (€/
km).

Computational Data

The distribution cases that are studied in the present paper, are faced by a logistics
company in the city of Athens, Greece. The data of the customers (57 in total),
contain their addresses, their demand and return quantities (kg), as well as their time
windows. More specifically, we test five different cases, in order to extract more
reliable conclusions. Starting from the first case (Case 1), and until the last one (Case
5), the quantities of goods that each customer requests, either to be delivered or to be
returned, increase in size, and constitute a bigger part of the vehicle’s capacity.
As for the vehicles that are considered in our research, we discriminate three cases
according to the fuel type, (i) Petrol, (ii) Diesel, and (iii) Electric vehicles, as
presented in Table 2. However, in order to extract reliable conclusions from our
research, the vehicles are of the same brand (Nissan), and more specifically the same
model (NV200). Consequently, the distribution cases are tested for the three differ-
ent vehicle types. In each case, the fleet of vehicles is composed only of the same
type of vehicles. In addition, the fixed costs were calculated, taking into account
depreciation time (10 years) and annual vehicle use (250 working days per year).
The increased fixed cost of electric vehicles is at least expected since the purchase
price is higher than the other type σ of vehicles. On the other hand, the variable cost
is inversely proportional to fixed costs, since it depends on the value of fuel and the
maintenance cost of the vehicle. Furthermore, the driving range in the case of electric
vehicles is significantly lower, and leads to the need for battery recharge, which is
also time-consuming (45 min to fully recharge the empty battery). Batteries also
cause another problem, which is the reduced capacity. However, all the above
disadvantages are balanced by the zero CO2 emissions.
The computational results of our research, containing the number of vehicles
needed (number of routes) to execute the plan of routes, the total distance, CO2
emissions and the costs (variable, fixed and total), are calculated for each vehicle
type (Petrol, Diesel, and Electric), as well as for each distribution case. Tables 3, 4
and 5 contain the results of Petrol, Diesel and Electric vehicles respectively.

Table 2 Features and specifications of vehicles

Vehicle Fixed cost Variable cost (€/ Driving range Capacity CO2 emissions
type (€) km) (km) (kg) (g/km)
Petrol 8.00 € 0.0972 750 795 166
Diesel 8.82 € 0.0555 1100 728 131
Electric 11.70 € 0.0196 275 630 0
60 G. D. Konstantakopoulos et al.

Table 3 Economic and environmental impact of petrol vehicles

Fixed Variable Total
Number of Total distance CO2 emissions cost cost cost
Case routes (km) (kg) (€) (€) (€)
1 4 378.82 62.88 32.00 36.82 68.82
2 5 392.93 65.23 40.00 38.19 78.19
3 5 427.78 71.01 40.00 41.58 81.58
4 8 527.09 87.50 64.00 51.23 115.23
5 9 561.64 93.23 72.00 54.59 126.59
Total 31 2288.26 379.85 248.00 222.41 470.41

Table 4 Economic and environmental impact of diesel vehicles

Fixed Variable Total
Number of Total distance CO2 emissions cost cost cost
Case routes (Km) (kg) (€) (€) (€)
1 4 377.25 49.50 35.28 20.93 56.21
2 5 413.29 54.14 44.1 22.94 67.04
3 6 421.71 55.24 52.92 23.41 76.33
4 9 563.91 73.87 79.38 31.30 110.68
5 9 630.35 82.58 79.38 34.99 114.37
Total 33 2406.51 315.33 291.06 133.56 424.62

Table 5 Economic and environmental impact of electric vehicles

Fixed Variable Total
Number of Total distance CO2 emissions cost cost cost
Case routes (km) (kg) (€) (€) (€)
1 5 394.64 0 58.5 7.72 66.22
2 6 432.91 0 70.2 8.48 78.68
3 7 469.17 0 81.9 9.19 91.09
4 10 617.21 0 117.00 12.09 129.09
5 11 640.09 0 128.70 12.55 141.25
Total 39 2554.02 0 456.3 50.02 506.32

It can be easily observed that petrol vehicles have the highest CO2 emissions.
Diesel vehicles have also increased emissions, but they gain significant interest due
to their low distribution cost. As for electric vehicles, emissions are zero but at the
expense of high distribution cost compared to other cases of vehicles. The main
factor affecting deliveries with electric vehicles, leading to increased cost, is that in
each distribution case more vehicles are needed to execute the deliveries compared
to the other types of vehicles. That is due to the limited capacity of electric vehicles,
as batteries have increased weight and volume, which prevent them from having the
same capacity as other types of vehicles.
Moreover, in Case 1, where each customer has small demand, in terms of weight,
multiple customers can be integrated into the same route. As a result, the number of
Green Freight Distribution: A Case Study in Greece 61

Fig. 5 Correlation between Petrol

emissions and costs for each
vehicle type Diesel

Emissions (CO2)
Electric
Cost (€)

Table 6 Comparing current state with diesel and electric vehicle fleets
Current state Entirely Entirely
diesel electric Gap
Petrol Diesel Electric Mean vehicles Gap (%) vehicles (%)
Percentage 52.20 47.80 0 – 100 – 100 –
Cost 470.41 424.63 506.32 448.53 424.63 -5.98 506.32 +12.88
Emissions 379.85 315.33 0 349.00 315.33 -9.65 0 -100

electric vehicles that are utilized in deliveries are not much more than conventional
vehicles, and therefore fixed costs are on the same level. Additionally, the variable
costs remain very low in the case of electric vehicles, unlike to conventional
vehicles. On the other hand, when the items are of bigger size (Case 5), the total
distribution cost of electric vehicles is much higher, due to the increased number of
vehicles needed. In conclusion, electric vehicles could be very efficient, both in
terms of cost and CO2 emissions, in distribution of small size items such as parcels
and mail.
Surely, Greek companies should invest at least to diesel vehicles that will offer
both decreased CO2 emissions and distribution costs, if not to electric. The affect of
each vehicle type to cost and emissions is given in Fig. 5, while it is obvious that
diesel vehicles offer the best trade-off between cost and emissions.
Additionally, according to Table 1, 41.5% of LC vehicles in Greece are petrol,
37.3% diesel, while the rest is of unknown type. In Table 6 we estimate the
economical and environmental impact of replacing all petrol vehicles with diesel,
which are also compared to a fleet composed entirely of electric vehicles. The only
assumption made for the current state is that the vehicles that their type is unknown
are shared equally to petrol and diesel vehicles. Consequently, the percentages of
petrol LC vehicles is transformed to 52.2%, while the percentage of diesel LC
vehicles to 47.8%. The cost and emissions, in the “Current State” column, both for
petrol and diesel vehicles have been extracted from Tables 4 and 5. Based on the
percentage of each type, the mean cost and emissions are calculated.
Accordingly, in the “Entirely Diesel Vehicles” column, as its name idicates the
fleet of vehicles is composed entirely of diesel vehicles, leading to significant
62 G. D. Konstantakopoulos et al.

decreases in cost and emissions. Specifically, by replacing petrol vehicles with

diesel, the CO2 emissions could decrease at least 9.65%, while the distribution
costs by 5.98%. Finally, in the last two columns of Table 6 the case in which all
vehicles are electric is studied. In this case, the cost could increase up to 12.88%
while the emissions could be eliminated, leading to a 100% decrease.
However, the emissions reduction and the cost increment concerning the case of
electric vehicles can be misleading. That is because the production of electricity in
Greece is based, in a great extend, on the use of poorer quality fuel such as lignite
and less to renewable sources. It is only the last few years that renewable sources are
utilized. Consequently, electric vehicles may have zero emissions, however the
production of electricity has increased emissions that can be minimized only by
using renewable energies. Additionally, it is clear that the acquisition cost of electric
vehicles is higher compared to petrol and diesel, as occurred with all new technol-
ogies (Hiermann et al., 2016). However, over the years, the acquisition cost is
expected to be reduced and to be more cost effective for the distribution process.

Conclusions

The type of vehicles affect in a great extend not only the distribution cost but CO2
emissions as well. In this direction multiple efforts have been made in order to
decrease both. The use of electric vehicles is one of them, that despite the high
acquisition cost offer more friendly to the environment transportation, as well as
noise reduction. Both benefits are significant for city inhabitants that are greatly
affected. Besides the numerous advantages, the main limitations of electric vehicles
are the limited driving range, the increased recharging time, as well as the decreased
capacity.
According to the presented research in distribution cases with low-volume goods
the limitations of electric vehicles do not greatly affect costs. However, as the
volume of goods increases, more electric vehicles compared to othe vehicle types
are needed, leading to an increased distribution cost. This is due to the decreased
capacity of electric vehicles. Irrespective of the cost, CO2 emissions are eliminated
by the use of electric vehicles.
Additionally, despite the advantages by using electric vehicles, in Greece the
technologies concerning the vehicles are still far behind. This is obvious by the age
and the type of vehicles used in the distribution. Moreover, the use of electric
vehicles is not sufficiently supported by the infrastructure as only a limited number
of recharging stations exist even in the capital of Greece, Athens. Considering this,
the transition to more advanced vehicles, and more specifically to electric vehicles,
may be difficult for logistics and distribution companies in Greece. However, even
the replacement of petrol vehicles with diesel would offer significant reduction both
to distribution costs and to CO2 emissions.
Concluding, integrating new technologies is not an easy task for companies
especially when the cost is high. In this direction, EU and each state should fund
Green Freight Distribution: A Case Study in Greece 63

companies to follow new technological trends. This would be benefitial both for
companies that could reduce their distribution costa and renew their equipment, as
well as for city inhabitants that could live in cleaner and green cities with less noise.

References

Braekers, K., Ramaekers, K., & Van Nieuwenhuyse, I. (2016). The vehicle routing problem: State
of the art classification and review. Computers & Industrial Engineering, 99, 300–313. https://
doi.org/10.1016/j.cie.2015.12.007
Chaabane, A., Ramudhin, A., & Paquet, M. (2012). Design of sustainable supply chains under the
emission trading scheme. International Journal of Production Economics, 135(1), 37–49.
https://doi.org/10.1016/j.ijpe.2010.10.025
Dantzig, G. B., & Ramser, J. H. (1959). The truck dispatching problem. Management Science, 6(1),
80–91. https://doi.org/10.1287/mnsc.6.1.80
Elshaer, R., & Awad, H. (2020). A taxonomic review of metaheuristic algorithms for solving the
vehicle routing problem and its variants. Computers & Industrial Engineering, 140, 106242.
https://doi.org/10.1016/j.cie.2019.106242
Erdelić, T., & Carić, T. (2019). A survey on the electric vehicle routing problem: Variants and
solution approaches. Journal of Advanced Transportation, 2019, 5075671. https://doi.org/10.
1155/2019/5075671
Erdelić, T., Carić, T., Erdelić, M., & Tišljarić, L. (2019). Electric vehicle routing problem with
single or multiple recharges. Transportation Research Procedia, 40, 217–224. https://doi.org/
10.1016/j.trpro.2019.07.033
European Automobile Manufacturers Association. (2019). Vehicles in use—Europe 2019. https://
www.acea.be/publications/article/report-vehicles-in-use-europe-2019
Gayialis, S. P., Konstantakopoulos, G. D., & Tatsiopoulos, I. P. (2019). Vehicle routing problem for
urban freight transportation: A review of the recent literature. In A. Sifaleras & K. Petridis
(Eds.), Operational research in the digital era—ICT challenges (pp. 89–104). Springer. https://
doi.org/10.1007/978-3-319-95666-4_7
Hiermann, G., Puchinger, J., Ropke, S., & Hartl, R. F. (2016). The electric fleet size and mix vehicle
routing problem with time windows and recharging stations. European Journal of Operational
Research, 252(3), 995–1018. https://doi.org/10.1016/j.ejor.2016.01.038
Hornstra, R. P., Silva, A., Roodbergen, K. J., & Coelho, L. C. (2020). The vehicle routing problem
with simultaneous pickup and delivery and handling costs. Computers & Operations Research,
115, 104858. https://doi.org/10.1016/j.cor.2019.104858
Kechagias, E. P., Gayialis, S. P., Konstantakopoulos, G. D., & Papadopoulos, G. A. (2020). An
application of an urban freight transportation system for reduced environmental emissions.
Systems, 8(4), 49. https://doi.org/10.3390/systems8040049
Konstantakopoulos, G. D., Gayialis, S. P., & Kechagias, E. P. (2020). Vehicle routing problem and
related algorithms for logistics distribution: A literature review and classification. Operational
Research International Journal, 22(3), 2033–2062. https://doi.org/10.1007/s12351-020-
00600-7
Konstantakopoulos, G. D., Gayialis, S. P., Kechagias, E. P., Papadopoulos, G. A., & Tatsiopoulos,
I. P. (2021). Delivering and picking goods under time window restrictions: An effective
evolutionary algorithm. Journal of Intelligent & Fuzzy Systems, Preprint, 40(3), 5323–5336.
https://doi.org/10.3233/JIFS-202129
Li, H., & Lim, A. (2003). Local search with annealing-like restarts to solve the VRPTW. European
Journal of Operational Research, 150(1), 115–127. https://doi.org/10.1016/S0377-2217(02)
00486-1
64 G. D. Konstantakopoulos et al.

Lin, J., Zhou, W., & Wolfson, O. (2016). Electric vehicle routing problem. Transportation
Research Procedia, 12, 508–521. https://doi.org/10.1016/j.trpro.2016.02.007
Micheli, G. J. L., Cagno, E., Mustillo, G., & Trianni, A. (2020). Green supply chain management
drivers, practices and performance: A comprehensive study on the moderators. Journal of
Cleaner Production, 259, 121024. https://doi.org/10.1016/j.jclepro.2020.121024
Ombuki, B., Ross, B. J., & Hanshar, F. (2006). Multi-objective genetic algorithms for vehicle
routing problem with time windows. Applied Intelligence, 24(1), 17–30. https://doi.org/10.
1007/s10489-006-6926-z
Solomon, M. M. (1987). Algorithms for the vehicle routing and scheduling problems with time
window constraints. Operations Research, 35(2), 254–265. https://doi.org/10.1287/opre.35.
2.254
Tan, K. C., Chew, Y. H., & Lee, L. H. (2006). A hybrid multiobjective evolutionary algorithm for
solving vehicle routing problem. Computational Optimization and Applications, 34(1),
115–151. https://doi.org/10.1007/s10589-005-3070-3
Tarantilis, C. D., Diakoulaki, D., & Kiranoudis, C. T. (2004). Combination of geographical
information system and efficient routing algorithms for real life distribution operations.
European Journal of Operational Research, 152(2), 437–453. https://doi.org/10.1016/S0377-
2217(03)00035-3
Wassan, N. A., & Nagy, G. (2014). Vehicle routing problem with deliveries and pickups: Modelling
issues and meta-heuristics solution approaches. International Journal of Transportation, 2(1),
95–110. https://doi.org/10.14257/ijt.2014.2.1.06
Yu, B., Yang, Z. Z., & Yao, B. Z. (2011). A hybrid algorithm for vehicle routing problem with time
windows. Expert Systems with Applications, 38(1), 435–441. https://doi.org/10.1016/j.eswa.
2010.06.082
Zhen, L., Huang, L., & Wang, W. (2019). Green and sustainable closed-loop supply chain network
design under uncertainty. Journal of Cleaner Production, 227, 1195–1209. https://doi.org/10.
1016/j.jclepro.2019.04.098