Vous êtes sur la page 1sur 19

European Journal of Operational Research 237 (2014) 775793

Contents lists available at ScienceDirect

European Journal of Operational Research


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

Invited Review

A review of recent research on green road freight transportation


Emrah Demir a,, Tolga Bektas b, Gilbert Laporte c
a

School of Industrial Engineering and Operations, Planning, Accounting and Control (OPAC), Eindhoven University of Technology, Eindhoven 5600 MB, The Netherlands
School of Management and Centre for Operational Research, Management Science and Information Systems (CORMSIS), University of Southampton, Southampton,
Higheld SO17 1BJ, United Kingdom
c
Canada Research Chair in Distribution Management and Interuniversity Research Centre on Enterprise Networks, Logistics, and Transportation (CIRRELT),
HEC Montral, 3000 chemin de la Cte-Sainte-Catherine, Montral H3T 2A7, Canada
b

a r t i c l e

i n f o

Article history:
Received 25 August 2013
Accepted 23 December 2013
Available online 4 January 2014
Keywords:
Road freight transportation
Green logistics
CO2 e emissions
Operations research

a b s t r a c t
Road freight transportation is a major contributor to carbon dioxide equivalent emissions. Reducing these
emissions in transportation route planning requires an understanding of vehicle emission models and
their inclusion into the existing optimization methods. This paper provides a review of recent research
on green road freight transportation.
2014 Elsevier B.V. All rights reserved.

1. Introduction
Road freight transportation is essential for the economic development, but it is also harmful to the environment and to human
health. Until recently, the planning of freight transportation activities has mainly focused on cost minimization (see, e.g., Crainic,
2000; Forkenbrock, 1999, 2001). With an increasing worldwide
concern for the environment, logistics providers and freight carriers have started paying more attention to the negative externalities
of their operations. These include pollution, accidents, noise, resource consumption, land use deterioration, and climate change
risk (Schreyer et al., 2004).
At the local and regional levels, a signicant portion of freight
transportation is carried out by trucks, which emit a large amount
of pollutants. While transportation technologies and fuels have improved over the years, most trucks run on diesel engines, which are
major sources of emissions of nitrogen oxides (N2 O), particulate
matter (PM) and carbon dioxide (CO2 ). Repeated exposure to
N2 O-based smog and PM has been linked to a wide range of health
problems. At the global level, greenhouse gases (GHGs) signicantly contribute to global warming. In the transportation sector,
GHG emissions are dominated by CO2 emissions from burning fossil fuels. These cause atmospheric changes and climate disruptions
which are harmful to the natural and built environments and pose
health risks. Until recently, GHGs were not classied as a pollutant
Corresponding author. Tel.: +31 618132383.
E-mail addresses: E.Demir@tue.nl (E. Demir), T.Bektas@soton.ac.uk (T. Bektas),
Gilbert.Laporte@cirrelt.ca (G. Laporte).
0377-2217/$ - see front matter 2014 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.ejor.2013.12.033

in the classical sense. However, the United States Environmental


Protection Agency (EPA) has recognized in 2009 that GHGs pose
a danger to human health and welfare. GHGs absorb and emit radiations within the thermal infra-red range in the atmosphere and
signicantly raise the Earths temperature. As of October 2013,
the level of atmospheric CO2 emissions is estimated to be equal
to 393.66 parts per million and is still increasing (ESRL, 2013).
The carbon dioxide equivalent (CO2 e) measures how much global warming a given type and amount of GHG may cause, using the
functionally equivalent amount or concentration of CO2 as the reference. The selection of GHGs to include in the carbon footprint is
an important issue. Wright, Kemp, and Williams (2011) suggest
that a signicant proportion of emissions can be captured through
the measurement of CO2 and CH4 , which are the most prominent
anthropogenic GHGs. The emissions of CO2 are directly proportional to the amount of fuel consumed by a vehicle, which is in turn
dependent on a variety of vehicle, environment and trafc-related
parameters, such as vehicle speed, load and acceleration (Demir,
Bektas, & Laporte, 2011). On the other hand, CH4 emissions are a
function of many complex aspects of combustion dynamics and
of the type of emission control systems used.
Freight transportation planning has many facets, particularly
when viewed from the multiple levels of decision making.
Arguably the most famous problem at this level is the well-known
Vehicle Routing Problem (VRP), which consists of determining
least cost routes for a eet of vehicles to satisfy the demands of a
set of customers, subject to side constraints. The traditional objective in the standard VRP is to minimize the total distance traveled
by all vehicles, but this objective can be enriched through the

776

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

inclusion of terms related to fuel consumption (Bektas & Laporte,


2011; Demir, Bektas, & Laporte, 2012).
Recent developments in green road freight transportation have
heightened the importance of operations research techniques in
this area (see, e.g., Dekker, Bloemhof, & Mallidis, 2012; Dobers
et al., 2013; Lin, Choy, Ho, Chung, & Lam, 2013; Salimifard,
Shahbandarzadeh, & Raeesi, 2012; Touati & Jost, 2012). In the last
decade, the body of knowledge on the reduction of CO2 e emissions
from road transportation has grown notably. As of August 2013, we
are aware of at least 58 papers on this topic. In this study we focus
on the green logistics literature related to total energy
consumption.
The scientic contribution of this study is threefold: (i) to
analyze the factors affecting fuel consumption, (ii) to extensively
survey the available vehicle emission models, (iii) to review the
scientic literature on green road freight transportation. The
remainder of this paper is organized as follows. Section 2 reviews
the factors affecting fuel consumption. Section 3 lists the available
tools to estimate CO2 e. In Section 4, we investigate routing
problems with regard to fuel consumption. Section 5 presents an
extensive body of literature of CO2 e emissions in routing and
scheduling. Conclusions and future research directions are stated
in Section 6.

2. Factors affecting fuel consumption


The factors inuencing fuel consumption have been studied by
Ardekani, Hauer, and Jamei (1996), Bigazzi and Bertini (2009),
Demir et al. (2011) and Alwakiel (2011). A summary of these works
is provided in Fig. 1. These factors can be divided into ve
categories: vehicle, environment, trafc, driver and operations.
Most fuel consumption models concentrate on vehicle, trafc,
and environmental inuences, but do not capture driver related
issues which are relatively difcult to measure. Moreover, operations related factors are often seen as an externalities affecting fuel
consumption.
One important work by Eglese and Black (2010) studies the
emissions arising in routing and lists some of the factors affecting
fuel consumption. In contrast to the existing literature, the authors
argue that speed is more important than distance traveled when
estimating emissions. In another study, Demir et al. (2011) have
compared several emissions models, and mentioned other relevant
factors such as load weight and distribution, engine type and size,
vehicle design, and road gradient.
Road freight is mostly carried by internal combustion vehicles.
In order to move a vehicle, an engine must provide power to overcome the effects of inertia, rolling resistance, wind resistance, road

Factors affecting fuel


consumption

Vehicle

Environment

Traffic (travel)

Driver

Operations

related

related

related

related

related

(V)

(E)

(T)

(D)

(O)

Vehicle

Roadway

Speed

Driver

curb weight

gradient

(T1)

aggressiveness

mix

(V1)

(E1)

(D1)

(O1)

Vehicle shape

Pavement type

deceleration

Gear selection

Payload

(V2)

(E2)

(T2)

(D2)

(O2)

Engine size/type

Ambient

Congestion

Idle time

Empty

(V3)

temperature

(T3)

(D3)

kilometers

Fleet size and

Acceleration/

(E3)

(O3)

Altitude

Number of stops

(E4)

(O4)

Engine temperature
(V4)
Transmission
(V5)

Wind
conditions
(E5)

Fuel type/
composition

Other characteristics

(V6)

(humidity, surface
conditionsetc.)

Oil viscosity

(E6)

(V7)
Other characteristics
(maintenance, age,
accessories etc.)
(V8)

Fig. 1. Factors affecting fuel consumption.

777

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

3. Fuel consumption models


This section reviews fuel consumption (vehicle emission) models. A short description of each model is provided, along with a discussion pertaining to its development and applicability. In order to
get a full picture of these models, the reader is referred to the classication of Williams, Kemp, Coello, Turner, and Wright (2012)
which classies the models based on their simplicity and parameter uncertainty (Fig. 2). In this study, we focus on the way of
estimating fuel consumption more accurately.
Generalized activity data * emission factors
Measured activity data * emission factors
Measured activity data * assumed fuel/energy consumption * emission factors
Measured activity data * sampled fuel/energy consumption * emission factors

Specificity

 Speed: Fuel consumption depends on several factors, but speed


is the most important one because it affects inertia, rolling
resistance, air resistance and road slope. Most studies focus on
distance traveled but as Van Woensel, Creten, and Vandaele
(2001) have shown, applying an average emission value per
kilometer is inaccurate. According to Demir et al. (2011), a rule
of thumb for medium-duty vehicle (MDV) is that fuel consumption increases approximately by 0.001 liter/kilometer for every
kilometer/hour increase above 55 kilometer/hour. Decreasing
the vehicle speed from 100 kilometer/hour to 90 kilometer/
hour can reduce fuel consumption by approximately 0.02 liter
per kilometer. Demir et al. (2012) have derived an optimum
driving speed and have shown that reductions in emissions
could be achieved by varying speed over a network. It should
be noted that the optimum driving speed varies to a certain
degree between geographical areas due to speed limits and trafc density.
 Road gradient: On a slope, wheel horsepower demand increases
signicantly with vehicle weight because of road slope force. In
some regions, road gradient plays an important role and can
result higher CO2 e emissions. The study of Demir et al. (2011)
shows that fuel consumption of an MDV on a 1% road slope
may increase by up to six liters on a 100 kilometer road
segment. The reverse is also true: traveling on a negative slope
surface reduces fuel consumption. With the help of an advanced
GIS software, one can change speed on a portion of a road
segment in order to reduce fuel consumption.
 Congestion: Driving in congested areas increases fuel consumption because of lower than optimal speeds. This is supported by
Van Woensel et al. (2001) who have assessed the differences in
CO2 emissions between using constant speeds and ow dependent speeds. Based on a survey of cars on a motorway, they have
shown that ow-dependent emissions of CO2 are on average
11% higher than CO2 emissions calculated using a constant
speed. A peak increase of 40% was also observed during the congested rush hour period. In a manual of trucks engine (CAT
(2006)), traveling 15% of the total kilometers on congested
roads results approximately in an 8% increase in fuel consumption. Traveling 25% of the total kilometers on congested roads
approximately results in a 15% increase in fuel consumption.
 Driver: One of the most signicant factors affecting fuel consumption is the driver who controls vehicle speed, acceleration
rate, brake usage, shifting technique, trailer gap setting, idle
time, tire ination pressure, and more. CAT (2006) shows that
the difference in fuel consumption between the best and worst
drivers can be as much as 25% (42.256.5 liters over a 100 kilometer road segment). Idling the engine can also be seen as
driver-related consumption.
 Fleet size and mix: Selecting the right type of vehicle has a significant impact on fuel consumption. Small vehicles which have
smaller engines consume less fuel than larger vehicles. However, using a large vehicle may lead to less fuel consumption
than using two smaller vehicles. According to Demir et al.
(2011), for a certain amount of payload, the difference between
a medium and heavy-duty vehicle may be up to 14 liters of fuel
on a 100 kilometer road segment.
 Payload: Increasing the payload increases the engine demand
horsepower, which leads to a higher fuel consumption. Vehicle
payload affects the inertia force, rolling resistance and road
slope force and can be an important part of routing decisions.

Kara, Kara, and Yetis (2007) and Bektas and Laporte (2011) have
studied the effect of payload on fuel consumption. Demir et al.
(2011) show that, with an average load of one ton, fuel
consumption increases by approximately three liters over a
100 kilometer road segment. CAT (2006) shows that a
4500 kilogram reduction in payload reduces fuel consumption
by about 4.4%. Moreover, a reduction in gross weight from
36,000 kilogram to 27,000 kilogram generates an 8.8% improvement in fuel savings. In a study of DEFRA (2012), the change in
CO2 emissions is shown to be up to 25% for an articulated
heavy-duty vehicle (HDV) with a gross weight of more than
33,000 kilogram.
 Empty kilometers: Empty kilometers are kilometers driven by
empty vehicles and should be avoided as much as possible.
Reducing them always results in a lesser fuel consumption.
Most of the routing operations start or end at a depot with an
empty load. According to EC (2010), in 2010, almost a quarter
(23.9%) of all vehicle-kilometer of HDVs in the EU involved an
empty vehicle. In inbound transport, the share of empty vehicle-kilometer is usually higher than in outbound transport.
Empty kilometers cause environmental pollution and place an
unnecessary strain on the transportation infrastructures. Empty
kilometer driving is often caused by a lack of information. For
example, if a transportation company does not have the information that goods are waiting to be transported close to where
a driver has just dropped off a previous load, the driver will simply be asked to return to the depot with an empty truck.
 Green freight corridors: The concept of green corridors was
developed in recent years to minimize the environmental
related factors affecting fuel consumption. The use of green corridors enables reductions in fuel consumption and CO2 e emissions, and should improve the environmental performance of
road freight transport in the future. Green corridors can be
structured based on several environmental and trafc-related
factors. For example, CAT (2006) shows that the impact of climate is as important as vehicle speed and aerodynamics. Cold
air increases the aerodynamic drag on the vehicle. Compared
to 21 C, fuel consumption can go up by 5% if the temperature
reduces to 10 C and it can go up to 10% at 0 C. Wind and road
surface are other factors that must be considered while creating
green corridors.

Uncertainity

slope, drive train losses and accessories. Other factors which affect
engine efciency, such as airfuel-ratio, engine speed and compression ratio. Here we focus on the scientic literature related
to fuel reduction in road freight transportation by means of operations research techniques. It considers several areas separately.

Measured fuel/energy consumption * emission factors


Directly measured emissions

Fig. 2. Examples of several methods for quantifying emissions (Williams et al.,


2012).

778

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

The average fuel consumption of vehicles is provided by manufacturers, but the information provided does not make it possible
to derive CO2 e emissions. For freight trucks, these measurements
are based on engine test benches or from a standard test cycle
using a chassis dynamometer, rather than on real-life driving cycles. These emission certication tests have been shown to differ
signicantly from real driving conditions (Pelkmans & Debal,
2006).
Rakha, Ahn, and Trani (2003) mention that numerous energy
and emission models are available and differ in their modeling approaches, modeling structures, and data requirements. Here, we
categorize fuel consumption models into three main groups of
increasing levels of complexity: factor models, macroscopic models and microscopic models. Factor models include simple fuel consumption methods. Macroscopic models use average aggregate
network parameters to estimate network-wide emission rates.
Microscopic models estimate the instantaneous vehicle fuel consumption and emission rates at a more detailed level. Models can
also be grouped into the load (power)-based and the regressionbased emissions models. These denitions also refer to microand macro-emission modeling, respectively.
Emissions can be approximated through the use of an emission
factor related to one type of vehicle and a specic driving mode
(Esteves-Booth, Muneer, Kubie, & Kirby, 2002). Furthermore, the
emission factors are derived from the mean values of repeated
measurements over a particular driving cycle. This type of model
is particularly useful at a macro-scale, when the information on
trafc ows and operational modes is insufcient. An example of
a distance based factor model can be found in GHG Protocol
(2013). Another example of factor calculation can be found in
DEFRA (2012), which provides information on how to calculate
the GHG emissions. It is concerned with calculating and reporting
the direct emissions resulting from burning fuel when driving in
freight transport operations.

Table 1
Emission parameters used in MEET.
Weight class

3.5 < Weight 6 7.5


7.5 < Weight 6 16
16 < Weight 6 32
Weight > 32

110
871
765
1576

0
16.0
7.04
17.6

0
0.143
0
0

0.000375
0
0.000632
0.00117

8702
0
8334
0

0
32031
0
36067

0
0
0
0

gests the use a function the form er v K av bv cv 3


d=v e=v 2 f =v 3 , where er v is the rate of CO2 emissions
(gram/kilometer) for an unloaded goods vehicle on a road with a
zero gradient. The parameters K and a to f are predened coefcients whose values are given in Table 1.
The emission factors and functions of MEET refer to standard
testing conditions (i.e., zero road gradient, empty vehicle, etc.)
and are typically calculated as a function of the average vehicle
speed. Depending on the vehicle type, a number of corrections
may be needed to account for the effects of road gradient and
vehicle load on the emissions once a rough estimate has been produced. The following road gradient correction factor is used to take
the effect of road gradient into account: GCv A6 v 6
A5 v 5 A4 v 4 A3 v 3 A2 v 2 A1 v A0 . The coefcients A0 to A6
used to compute GCv are given in Table 2.
The following load correction factor is used to take the load
factor into account: LCv k nv pv 2 qv 3 r=v s=v 2
t=v 3 u=v ; k and n to u are coefcients whose values are presented
in Table 3.
MEET suggests estimating the total CO2 emissions (gram) as

Ev ; D er v  GCv  LCv  D:

It is noted that the parameters of MEET model were calibrated in


1999. The new engine technology and aerodynamics design of vehicles would require updates to these parameters.

3.1. Macroscopic models


This section reviews 13 macroscopic (average speed) emission
models, which are important tools in a wide-area emission assessment. These models can be used to calculate and develop a national or regional emission inventory, especially in green supply
chain management studies. Generally, emission rates are measured
for a variety of trips, each with a different average speed. Note that
the emissions of CO2 e are directly related to fuel consumption and
therefore can be easily calculated if the amount of fuel consumption is known. In order to eliminate repetition for the representation of variables, we use similar notation for all macroscopic
models. The standardized notations are the following: v is the vehicle speed (kilometer/hour) and D is the total distance (kilometer).
3.1.1. Methodology for calculating transportation emissions and
energy consumption (MEET)
A publication of the European Commission by Hickman, Hassel,
Joumard, Samaras, and Sorenson (1999) on emission factors for
road transportation (INFRAS, 1995) describes a methodology called
MEET, used for calculating transportation emissions and energy
consumption for heavy goods vehicles. MEET is based on on-road
measurements and all its parameters are extracted from real-life
experiments. It covers several vehicle technologies for different
classes of vehicles, such as vehicles weighing less than 3.5 tons,
vehicles weighing between 3.5 and 7.5 tons, vehicles weighing between 7.5 and 16 tons, vehicles weighing between 16 and 32 tons
and vehicles weighing more than 32 tons. For vehicles weighing
less than 3.5 tons, the rate of CO2 emissions per kilometer is estimated using a speed-dependent regression function of the form
er v 0:0617v 2  7:8227v 429:51. For other classes, MEET sug-

3.1.2. Network for transport and environment (NTM)


The NTM is a non-prot Swedish organization created in 1993
(NTM Road, 2010). Its aim is to establish a common way to calculate the environmental performances from different transportation
modes. The calculation pertains to the usage of natural resources
and other external effects from freight transport. The method
was mainly developed for market actors of transport services, making it possible to evaluate their own individual carbon footprint.
The NTM considers distance, load factors, type of transport mode,
positioning, empty return trips, topography and type of road
(urban, rural or motorway). The fuel consumption at the specied
load factor f lf i (liter/kilometer) for each type of road (motorway,
urban, rural) shown by i and can be calculated as

f lf i Fempty Ffull  Fempty  lf ;


where Fempty is the fuel consumption of the empty vehicle (liter/
kilometer), Ffull is the fuel consumption of the fully loaded vehicle
(liter/kilometer) and lf is the specied load factor. Moreover, Ffull
and Fempty values for different types of vehicles can be found in
NTM Road (2010). The total carbon dioxide (kg) can then be calculated as

Elf ; D1 ; D2 ; D3

3
X

f lf i  Di  eCO2 e ;

i1

where eCO2 e is the emission factor (typically 3.13 kilogram for one
liter diesel fuel) and Di is the total distance traveled for each road
type i. The NTM provides a default database if no vehicle-specic
data are available.

779

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793


Table 2
Road gradient factors for MEET.
Weight class

A6

A5

A4

A3

A2

A1

A0

Slope

Weight 6 7.5
Weight 6 7.5
7.5 < Weight 6 16
7.5 < Weight 6 16
16 < Weight 6 32
16 < Weight 6 32
Weight > 32
Weight > 32

0
0
0
0
0
0
0
0

3.01E09
1.39E10
9.78E10
6.04E11
5.25E09
8.24E11
2.04E09
1.10E09

5.73E07
5.03E08
2.01E09
2.36E08
9.93E07
2.91E08
4.35E07
2.69E07

4.13E05
4.18E06
1.91E05
7.76E06
6.74E05
2.58E06
3.69E05
2.38E05

1.13E03
1.95E05
1.63E03
6.83E04
2.06E03
5.76E05
1.69E03
9.51E04

8.13E03
3.68E03
5.91E02
1.79E02
1.96E02
4.74E03
3.16E02
2.24E02

9.14E01
9.69E01
7.70E01
6.12E01
1.45E+00
8.55E01
1.77E+00
9.16E01

[0,4]
[4,0]
[0,4]
[4,0]
[0,4]
[4,0]
[0,4]
[4,0]

Table 3
Load correction factors for MEET.
Weight class

Weight 6 7.5
7.5 < Weight 6 16
16 < Weight 6 32
Weight > 32

1.27
1.26
1.27
1.43

0.0614
0.0790
0.0882
0.121

0
0
0
0

0.00110
0.00109
0.00101
0.00125

0.00235
0
0
0

0
0
0
0

0
2.03E7
0
0

1.33
1.14
0.483
0.916

3.1.3. Computer programme to calculate emissions from road


transportation (COPERT)
COPERT is an European Economic Area funded emissions model
(Kouridis et al., 2010), which estimates vehicle emissions for a
range of vehicles by engine classication and vehicle type. It is driven by a database of emissions by vehicle class, engine technology
and speed. It estimates emissions for all major air pollutants, as
well as GHGs produced by different vehicle categories. Similar to
MEET, COPERT uses a number of regression functions to estimate
fuel consumption, which are specic to vehicles of different
weights. An example of total fuel consumption function (gram)
with different load and gradient factor is given by

Fv ; D e a expb v c expd v  D;

where a to e are the coefcient. The latest emission factor parameters of the COPERT 4 methodology have been updated for HDV on
the basis of the latest HBEFA version 3.1 (Hausberger, Rexeis,
Zallinger, & Luz, 2009). Table 4 shows a set of sample data for a
2026 ton rigid truck running on Euro-5 diesel.
3.1.4. Ecological transport information tool (ECOTRANSIT)
ECOTRANSIT was developed in 2003 with a regional scope
limited to Europe. The more recent version EcoTransIT World published in 2011 allows the estimation of environmental impacts of
worldwide transports (Knrr, Seum, Schmied, Kutzner, & Anthes,
2011), including that of freight transport for any route and any
transport mode. It covers energy consumption, GHG emissions
and air pollutants. The model can take into account upstream energy consumption. The full calculation can be performed in three
steps. The rst step is to calculate the nal energy consumption

per net ton kilometer as ECF ton kilometer ECF kilometer =CP  CU,
where ECF kilometer is the nal energy consumption per net ton kilometer, CP is the payload capacity (in ton), and CU is the capacity
utilization (%). The second step is to calculate the upstream energy
consumption per net ton kilometer as ECU ton kilometer ECF kilometer 
ECU EC , where ECU EC is the energy related upstream energy consumption. The third step is to measure total energy consumption
(megajoule) as

FD; M D  M  ECF ton kilometer ECU ton kilometer ;

where M is the mass of freight transported (ton). Energy consumption and emissions also depend on road category, gradient, and
driving pattern. The inuence of the load factor is modeled
according to the HBEFA. For example, the fuel consumption of an
empty vehicle can be one third below of that of a fully loaded
vehicle. This effect can be even higher depending on driving
characteristics and road gradient. The gradient takes into account
region-specic factors and fuel consumption also depends on the
driving pattern.
3.1.5. National atmospheric emissions inventory (NAEI)
NAEI was developed for a large range of sectors including agriculture, domestic activity, industry and transport (NAEI, 2012).
Emissions from road transportation are calculated either from a
combination of total fuel consumption data and fuel properties,
or from a combination of driving-related emission factors and road
trafc data. For each vehicle category, the fuel consumption with
average speed functions for hot exhaust in the NAEI can be calculated as (in liter/100 kilometer)
3

Fv ka bv cv 2 dv ev 4 f v g v 6 =v ;
Table 4
Set of sample data for COPERT.
Payload
(%)

Gradient
(%)

0
0
0
50
50
50
100
100
100

0
2
+2
0
2
+2
0
2
+2

530.707
546.477
1051.552
505.770
479.620
2074.874
502.941
1144.824
1882.813

0.0634
0.064
0.424
0.051
0.047
1.008
0.041
0.981
1.006

2704.528
9599.652
67.668
4762.796
7858.071
0.534
9343.090
0.400
0.422

0.512
0.766
0.084
0.609
0.677
0
0.729
0
0

157.588
61.960
0
180.436
40.246
0
195.202
0
0

where k and a to g are coefcients. Table 5 shows a setting of these


coefcients for a 2026 ton rigid truck. More information can be
found in NAEI (2012).
3.1.6. Other macroscopic models
In this section, we summarize seven other macroscopic fuel
consumption models we have identied. The reason why we
separate these seven models from those presented above is that
the sources describing them are either similar to those of earlier
models, or there does not exist sufcient information on their
methodology to provide a detailed description.

780

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

Table 5
Sample parameter sets for NAEI.
Fuel type

Pre Euro I
Euro I
Euro II
Euro III
Euro IV
Euro V
Euro VI

0.037
0.037
0.037
0.037
0.037
0.037
0.037

13690.286
13661.548
4747.573
9584.751
10297.467
10537.515
10537.515

390.990
220.208
1341.727
464.510
210.752
220.217
220.217

54.374
85.387
23.117
42.172
53.532
54.175
54.175

2.587
3.401
0.103
2.069
2.367
2.404
2.404

0.048
0.060
0.002
0.038
0.043
0.043
0.043

0
0
0
0
0
0
0

0
0
0
0
0
0
0

 MOBILE is a model developed by the Environment Protection


Agency (EPA) Ofce of Transportation and Air Quality and published in 1978. The last version MOBILE 6.2 was developed
using recent vehicle-emission testing data collected by the
EPA, CARB, and automobile manufacturers, as well as inspection
and maintenance tests conducted in various places (EPA, 2003).
The descriptive and spreadsheet outputs from MOBILE report
emission rates in grams of pollutant per vehicle-mile traveled.
In 2010, the MOBILE series of models was replaced by Motor
Vehicle Emission Simulator (MOVES) as EPAs ofcial model
for estimating emissions.
 MOVES was developed by the EPA (EPA, 2012). The purpose of
the tool is to provide an accurate estimate of emissions from
mobile sources under a wide range of user-dened conditions.
The model performs a series of calculations which were carefully developed to accurately reect vehicle operating processes, such as cold start or extended idling, and provide
estimates of bulk emissions or emission rates.
 HBEFA is an handbook of emission factors for road transport and
a database for vehicle emission factors in Europe. It was introduced in 1995 by Hausberger et al. (2009), and the current version HBEFA 3.1 was published in 2009. It provides emission
factors for all current vehicle categories of road transport, all
EU emission standards and a wide variety of trafc situations.
Besides hot emission factors, it also provides cold start and
evaporation emission factors.
 GREET is a greenhouse gases, regulated emissions, and energy
use in transportation model that was developed as a full lifecycle model by the Argonne National laboratory (Wang,
1999). It is used to evaluate energy and emission impacts of various vehicle and fuel combinations on a full fuel-cycle/vehiclecycle basis. The rst version of GREET was released in 1996.
Since then, GREET has continually been updated and expanded,
and is now called GREET2 2012. For a given vehicle and fuel system, GREET2 2012 calculates consumption of total energy and
CO2 e emissions separately.
 LEM is a lifecycle emissions model of Delucchi (2003) that estimates energy use, pollutant emissions, and CO2 e emissions
from a variety of transportation and energy life cycles. It
includes a wide range of modes of passenger and freight
transport.
 VERSIT-LD is a model that was developed to predict trafc
stream emissions for light-duty vehicles in any particular trafc situation (Smit, Smokers, & Rab, 2007). It predicts trafc
stream emissions by using accurate mean emission factors,
expressed in grams per kilometer.
 IVE is the international vehicle emissions model that was
designed to estimate emissions from motor vehicles (ISSRC,
2008). It predicts local air pollutants, toxic pollutants and GHGs
emissions. The emission prediction process of the IVE model
starts with a base emission rate, and a series of correction factors are then applied to estimate the amount of pollution from
a variety of vehicle types. These main factors refer to temperature, humidity, fuel quality and driving behavior.

 EMFAC is an emission model developed by the California Air


Resources Board. The current version of EMFAC is EMFAC2011
(CARB, 2011). EMFAC emission rates are a function of vehicle
average travel speed. Basic emission rates are derived from
emissions tests conducted under standard conditions such as
temperature, fuel, and driving cycle. Adjustments can be
applied for different temperatures, gasoline types, and
humidity.
3.2. Microscopic models
This section reviews instantaneous emission models (at time t)
for the estimation of hot-stabilized vehicle emissions. Instantaneous models are necessary to predict trafc emissions more accurately. These are based on instantaneous vehicle kinematic
variables, such as speed and acceleration, or on more aggregated
modal variables, such as time spent in each trafc mode, cruise
and acceleration.
In order to increase the readability and understandability of the
models, we describe them with a unied notation: v is the speed of
the vehicle (meter/second), T is the total journey duration (second),
M is the total weight of the vehicle (kilogram), a is the instanta2
neous acceleration (meter=second ), x is the gradient (%), g is
2
the gravitational constant (meter=second , approximately 9.81),
q is the air density (in kilogram=meter3 , typically between
1.1455 and 1.4224), A is the frontal surface area (in meter2 , typically between 3.91 and 5.88), C d is the coefcient of aerodynamic
drag (typically between 0.7 and 0.9), and C r is the coefcient of
rolling resistance (typically between 0.006 and 0.01).
3.2.1. An instantaneous fuel consumption model (IFCM)
An energy-related emissions estimation model was described
by Bowyer, Biggs, and Akelik (1985). The IFCM uses vehicle characteristics such as mass, energy, efciency parameters, drag force
and fuel consumption components associated with aerodynamic
drag and rolling resistance, and approximates the fuel consumption per second. The fuel consumption with an IFCM can be
calculated as

(
f t

a b1 Rtv b2 Ma2 v =1000 for Rt > 0


a
for Rt 6 0;

where f t is the fuel consumption per unit time (milliliter/second),


Rt is the total tractive force (kilonewton) required to move the
vehicle and calculated as the sum of drag force, inertia force and
grade force as Rt b1 b2 v 2 Ma=1000 gMx=100; 000.
Furthermore, a is the constant idle fuel rate (in milliliter/second,
typically between 0.37 and 0.56), b1 is the fuel consumption per
unit of energy (in milliliter/kilojoules, typically between 0.09 and
0.08), b2 is the fuel consumption per unit of energy-acceleration
2
(in milliliter/(kilojoulesmeter=second ), typically between 0.03
and 0.02), b1 is the rolling drag force (in kilonewton, typically
between 0.1 and 0.7), and b2 is the rolling aerodynamic force (in
kilonewton/(meter/second2), typically between 0.00003 and

781

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

0.0015). These values are extracted from Bowyer et al. (1985) and
Akelik and Besley (1996, 2003). Using IFCM, the total amount of
fuel consumption FT (milliliter) for a journey of duration T can
be calculated as

FT

f tdt:

IFCM works best at a micro-scale level and is better suited for short
trip emission estimations.
3.2.2. A four-mode elemental fuel consumption model (FMEFCM)
FMEFCM was introduced by Bowyer et al. (1985). This model includes the same parameters as IFCM but also introduces new
parameters, such as initial speed, nal speed and energy-related
parameters. The model consists of four functions, F a ta ; F d td ;
F c tc and F i ti , which correspond to fuel consumption estimations
(milliliter) for the following modes of driving, respectively: acceleration, deceleration, cruise, and idle, where t a ; td ; tc and ti are
the times spent for each mode.
 Acceleration fuel consumption (F a t a ): This function computes
the amount of fuel consumption over the acceleration phase of a
vehicle from an initial speed v i to a nal speed v f . It is dened
as

F a t a max fat a C k1 Bv 2i v 2f b1 MEk k2 b2 ME2k


0:0981b1 M xxa ; at a g;
where Ek denotes the change in kinetic energy per unit distance
during acceleration and is calculated as Ek 0:3858 104
v 2f  v 2i =xa . Furthermore, the integration coefcients are k1
p
0:616 0:000544v f  0:0171 v i
and
k2 1:376 0:00205v f
0:00538v i . When the travel distance xa and the travel time ta
are not known, they can be estimated as xa ma v i v f t a =3600,
where ma 0:467 0:00200v f  0:00210v i and ta v f  v i =
p
2:08 0:127 v f  v i  0:0182v i . In addition, C is the function
parameter (in milliliter/kilometer, typically between 21 and 100),
and B is the function parameter (in (milliliter/kilometer)/
(kilometer/hour)2, typically between 0.0055 and 0.018).
 Deceleration fuel consumption (F d td ): The amount of fuel consumption during the deceleration phase from an initial speed v i
to a nal speed v f is calculated as

F d t d max fat d kx C ky k1 Bv 2i v 2f ka b1 MEk


kx b1 ME2k 0:0981b1 Mxxd ; atd g;
where kx 0:046 100=M 0:00421v i 0:00260v f 0:05444x;
0:75
3:81
3:81
ky kx ; ka kx 2  kx
and
k1 0:621 0:000777v i 
p
0:0189 v f . If the travel distance xd and travel time t d are not
known, they can be estimated as above, although in this case the
coefcients change slightly. In addition, kx ; ky and ka are the energy
related parameters.
 Cruise fuel consumption (F c t c ): The following function can be
used to calculate the total amount of fuel consumption by a
vehicle during a cruise phase:

F c t c max ffi =v c C Bv 2c kE1 b1 MEk kE2 b2 ME2k


0:0981kG b1 M x; fi =v c gxc ;
where fi denotes the idle fuel rate (in milliliter/hour), v c is the
average cruise speed (kilometer/hour), and xc denotes the travel
distance (kilometer). The change in total positive kinetic energy
per unit distance during the cruise mode is calculated as Ek
max f0:258  0:0018v c ; 0:10g and the other parameters are
set to kE1 max f12:5=v c 0:000013v 2c ; 0:63g; kE2 3:17, and

kG 1 2:1Ek for x < 0, and 1  0:3Ek for x > 0. Furthermore,


kE1 ; kE2 and kG are the calibration parameters.
 Fuel consumption while idle (F i t i ): The following function can
be used to calculate the total amount of fuel consumption when
the vehicle is idle:

F i ti at i ;
where ti is the idle time (s), and a is the idle fuel rate (milliliter/
second).
The total fuel consumption Fta ; td ; t c ; t i (milliliter) using the
elemental model can be calculated as

Ft a ;t d ; tc ;t i

ta

F a ta dt

td

F d t d dt

tc

F c t c dt

ti

F i t i dt:
0

7
The model requires detailed data for each driving mode, such as the
total distance, speed, time and average road grade. If such data are
available, the estimation of fuel consumption can be very accurate.
3.2.3. A running speed fuel consumption model (RSFCM)
RSFCM is an aggregated form of FMEFCM and was also introduced by Bowyer et al. (1985). The model calculates fuel consumption separately during driving modes when a vehicle is running on
idle. Acceleration, deceleration and cruise modes are considered
together within a single function. The model is as follows:

Ft i ; t s max fat i fi =v r A Bv 2r kE1 b1 MEk


kE2 b2 ME2k 0:0981kG b1 M xxs ; at s g;

where Ft i ; t s is the total fuel consumption (milliliter), xs is the


total distance, v r denotes the average running speed (kilometer/
hour), t s and ti the travel and idle time, respectively. The average
speed is calculated as v r 3600xs =t s  t i . Furthermore, Ek
max f0:35  0:0025v r ;0:15g; kE1 max f0:675  1:22=v r ; 0:5g; kE2
2:78 0:0178v r . The model needs fewer data than for FMEFCM but
it is less accurate for estimating fuel consumption.
3.2.4. Average speed fuel consumption model (ASFCM)
This model is an another extension of IFCM and was also proposed by Bowyer et al. (1985). The model relates fuel consumption
per unit distance to average speed. In this model, the total fuel consumption (milliliter) can be estimated as F s D f v s D, where the
fuel consumption per unit distance is f v s fi =v s cK, and v s denotes the average travel speed (kilometer/hour), and D is the total
distance. Furthermore, K 1  K 1 1  M=1200  K 2 1  b1 =0:090
K 3 1b2 =0:045  K 4 1  b1 =0:000278MK 5 1  b2 =0:00108 is
the adjustment factor for different types of vehicles, cis the regression coefcient, and K 1  K 5 are parameters based on the analysis
of real on-road data for which a sample set is shown in Table 6.
These values are extracted from Bowyer et al. (1985).
3.2.5. Vehicle specic power (VSP)
The VSP model has proven to be very useful as a microscopic
model to estimate instantaneous vehicle fuel consumption and
emission rates (Jimenez-Palacios, 1998). It is highly correlated with
variability in second-by-second emissions of pollutants from diesel
vehicles. The VSP is a measure of the load on a vehicle and is
dened as the power per unit mass to overcome road grade, rolling
Table 6
Sample parameter sets for ASFCM.
Driving environment

K1

K2

K3

K4

K5

City center
Urban

70.6
73.8

0.893
0.72

0.79
0.867

0.21
0.134

0.421
0.406

0.109
0.280

782

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

and aerodynamic resistance, and inertial acceleration. It can be calculated using an equation given by EPA (kilowatt/ton):

F o T fidle T E

(
J
X

lMg

j1

VSPt v a1 m g sinarctanx gC r 0:5qC d Av 3 =M;


where m is the mass factor (typically between 0.1 and 0.2) accounting for the rational masses. The fuel consumption (gram) is

FT; M

VSPt  M=gdt;

where g is the efciency rate (typically between 0.85 and 0.95). The
VSP is also used as a base model to measure power demand in different emission models.
3.2.6. Emissions from trafc mode (EMIT)
EMIT was introduced by Cappiello, Chabini, Nam, and Lue
(2002). It is a statistical model with a basis in the physical system
for fuel consumption of light duty vehicles. The fuel rate (gram/
second) can be calculated as

(
f e t

afr bfr v cfr v 2 dfr v 3 ffr av if Ptract t > 0


a0fr
if Ptract t 0;

where P tract t Afr v Bfr v2 C fr v3 Mav Mg sinxhv; Afr is the


2
rolling resistance coefcient (kilowatt=meter=second ), Bfr is the
speed correction to rolling resistance coefcient (in kilowatt=
meter=second2 ), and C fr is the air drag resistance coefcient (in
3
kilowatt=meter=second ). More information on coefcients
afr ; a0fr ; bfr ; cfr ; dfr and ffr can be found in Cappiello et al. (2002). The
total fuel consumption (gram) is

F e T

fe tdt:

10

EMIT was calibrated based on the two light-duty vehicle categories. All its parameters can be found in Cappiello et al. (2002).
It provides reasonable results compared to actual measurements.
We also note that EMIT was developed and calibrated for hot-stabilized conditions with zero road grade, and without accessory
usage, and does not correspond to cold-start emissions.
3.2.7. Virginia tech microscopic (VT-Micro)
This model was developed by Ahn, Rakha, Trani, and Van Aerde
(2002) from experimentations with numerous polynomial combinations of speed and acceleration levels using chassis dynamometer
data on light duty vehicles. CO2 emissions were estimated using the
carbon balance equation in conjunction with fuel consumption
measurements. The fuel consumption rate for a cruising speed equal
to the free-ow speed and on a grade of x can be computed as

fv t exp

!
3 X
3
X
i
Li;j v f g xj ;
i1 j1

where Li;j is the model regression coefcients at speed exponent i


and acceleration exponent j. The total fuel consumption (liter) on
a link l can be calculated as
l

FDl 3600f v tDl =v f ;


where Dl is the length of link l (kilometer) and
speed on link l (kilometer/hour).

11

vf

is the free-ow

3.2.8. The Oguchi, Katakura, and Taniguchi (2002) fuel consumption


model (OFCM)
OFCM was developed by Oguchi et al. (2002). The function consists of ve components, namely idle consumption, friction loss,
altitude change loss, air drag loss, and acceleration loss. The total
fuel consumption F o T (milliliter) using the OFCM can be
calculated as

C d

t je

t js

tje

v dt sin hMg

t js

t je

t js

v 3 dt M m0:5v 2je  0:5v 2js

v dt

)
;

12

where fidle is the base consumption, T is the travel time, E is the fuelenergy equivalent, l is the friction coefcient, and J is the number
of interval. The parameters js and je are the start and end of interval,
and m is the internal resistance equivalent. The idle fuel consumption is the fuel used for the inertia resistance of the engine and
transmission, air conditioner, and some other electric components.
Other components estimate the energy lost by the vehicles
movement.
3.2.9. Physical emission rate estimator (PERE) model
PERE was designed by Nam and Giannelli (2005) to measure
vehicle emissions. It was developed to support the new EPA energy
and emissions inventory model, namely MOVES. PERE consists of a
series of stand-alone spreadsheets, which can be run and modied
by an user. The model performs well: in most cases the predictions
are within 10% of observed values. The fuel consumption rate (kilometer/second) can be formulated as

fpr t nfkNV Pb v =gt Pacc =gg43:2;

13

where n is the fuel air equivalence ratio (typically 1), k is the engine
friction (typically between 0.15 and 0.25), N is the engine speed
(typically between 31 and 55), V is the engine displacement volume
(typically between 5 and 7), gt is a transmission and nal drive efciency (typically between 0.4 and 0.45), Pacc is the power draw of
accessories (typically 0). The power demand (W) can be calculated
as Pb v C r v Bv 2 C d v 3 Mv a g x where A, B, C are parameters calculated as A C R0 Mg; B 0:0; C C d Ar q=2 C R2 Mg, and
C R0 and C R2 are the zero and second order in speed rolling resistance
force terms, respectively. The total fuel consumption (kilogram) can
be calculated as

F pr T

fpr tdt:

14

More information on the models parameters can be found in Nam


and Giannelli (2005).
3.2.10. Comprehensive power-based fuel consumption models (CPFMs)
Two power-based second-order polynomial models were introduced by Rakha, Ahn, Moran, Saerens, and Van den Bulck (2011).
These models were inspired from one of the base models which
can be found in the mechanical engineering literature (Wong,
2001). The rst model does not require any engine data given that
the power used by a vehicle is a function of the vehicle speed and
acceleration level. The second model requires engine data in addition to external data and therefore can be used to model eco-drive
with gear-shifting strategies. The two models can be formulated as

F 1 t a0 a1 Pt a2 Pt2 ;
F 2 t b0 xt b1 Pt b2 Pt2 ;

15

where a0 ; a1 ; a2 ; b0 ; b1 and b2 are vehicle-specic constants given for


each type of vehicle. The function Pt Rt 1:04mat=
3600gd v t measures the total power exerted by the vehicle driveline (kilowatt) at time t; Rt is the total resistance force (newton),
and gd is the driveline efciency. In addition, the total resistance
force can be calculated as Rt q=25:92C d C h Av 2 t
9:8066mC r fc1 v t c2 g 9:8066mGt. The detailed calculation
of each coefcient is shown in Rakha et al. (2011). The CPFM-1
and CPFM-2 can be easily calibrated using publicly available data,

783

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

for example using the city and highway fuel economy ratings of
EPA.
3.2.11. Passenger car and heavy duty emission model (PHEM)
The microscopic emission model PHEM was designed for simulating fuel consumption. As a result of the cooperation between
ARTEMIS (Boulter & McCrae, 2009), COST 346 (Rexeis, Hausberger,
& Riemersma, 2005) and HBEFA (Hausberger et al., 2009), the measurement program covered different engines and vehicles. PHEM
interpolates the fuel consumption from the engine maps according
to the course of engine power demand and engine speed in the
driving cycles. The actual engine power P can be calculated as
P Prr P ar P acc Pgr Pau Ptl , where P rr Mgfr 0 fr 1 v
fr 2 v 2 fr 3 v 3 fr 4 v 4 v is the rolling resistance, Par 0:5C d Aqv 3 is
the air resistance, and Pacc M mrot av is the acceleration.
Moreover, P gr Mg x0:01v is the gradient resistance, P au
P 0 Prated is the power demand for auxiliaries and Ptl Pdr =gt  P dr
is the transmission losses, where fr 0 to fr 4 are the rolling resistance
coefcients, mrot is the reduced mass for rational accelerated part,
P 0 is the power demand of the auxiliaries as ratio to the rated
power P rated , and Pdr is the power to overcome the driving
resistances. The total emission value can be calculated under the
transient conditions

Etrans t Eqs Prated F trans ;

16

where Eqs is the quasi-steady-state emission value interpolated


from steady-state emission map (in gram/hour), Prated is the rated
engine power (kilowatt) (since emission values are normalized to
the rated power), F trans t is the dynamic correction function
((g/h)/kilowatt rated power). More information can be found in
Boulter and McCrae (2009).
3.2.12. A comprehensive modal emission model (CMEM)
CMEM was developed and presented by Scora and Barth (2006),
Barth, Younglove, and Scora (2005) and Barth and Boriboonsomsin
(2008) for heavy-goods vehicles. It is based upon second-bysecond tailpipe emissions data collected from 343 light-duty
vehicles (LDVs) tested under a variety of laboratory driving cycles.

CMEM is similar to IFCM. However, to produce accurate estimations, it requires detailed vehicle specic parameters for the estimations such as the engine friction coefcient, and the vehicle
engine speed. CMEM follows, to some extent, the model of Ross
(1994) and is composed of three modules, namely engine power,
engine speed and fuel rate.
 The engine power module: The power demand function for a
vehicle is obtained from the total tractive power requirements
Ptract t (kilowatt) placed on the vehicle at the wheels:

P tract t Ma Mg sin xh 0:5C d qAv 2 MgC r


 cos xhv =1000:

17

To translate the tractive requirement into engine power requirement, the following relationship is used:

Pt Ptract t=gtf Pacc ;


where Pt is the second-by-second engine power output (kilowatt),
and gtf is the vehicle drive train efciency.
 The engine speed module: Engine speed is approximated in
terms of vehicle speed as

Nv SRL=RLg v ;
where Nv is the engine speed (in rpm), S is the engine-speed/vehicle-speed ratio in top gear Lg , R(L) is the gear ratio in gear L = 1,. . .,Lg ,
and g is the efciency parameter for diesel engines.
 The fuel rate module: The fuel rate (gram/second) is given by
the expression

fcm t nkNv V Pt=g=43:2;

18

The total fuel consumption (gram) can be calculated as

F cm T

fcm tdt:

19

CMEM can be seen as a state-of-the-art microscopic emission model


because of its ease of application.

Table 7
A detailed technical comparison of macroscopic and microscopic models.
V1
Macroscopic models
MEET
U
NTM
U
COPERT
U
ECOTRANSIT
U
NAEI
U
MOBILE
U
MOVES
U
HBEFA
U
GREET
U
LEM
U
VERSIT-LD
U
IVE
U
EMFAC
U
Microscopic models
IFCM
U
F-MEFCM
U
RSFCM
U
ASFCM
U
VSP
U
EMIT
U
VT-Micro
U
OFCM
U
PERE
U
CPFM
U
PHEM
U
CMEM
U

V2

V3

V4

V5

U
U
U

U
U

U
U
U

U
U
U
U
U
U
U
U
U
U
U

U
U
U

U
U
U
U
U

V6

U
U
U
U
U
U
U
U
U
U
U
U
U

V7

U
U
U
U
U

U
U
U
U
U

U
U
U
U
U
U
U
U
U
U
U
U

U
U
U
U

E1

U
U
U
U

U
U
U
U
U
U
U
U
U
U
U
U

E2

E3

E4

U
U
U
U

E5

T1

U
U
U
U
U
U
U
U
U
U
U

U
U
U
U
U
U

U
U
U
U
U
U

U
U
U
U
U
U

U
U
U

U
U
U

U
U

U
U

U
U
U
U
U

U
U
U
U
U

U
U
U
U
U

U
U
U
U
U

U
U
U
U
U
U
U
U

U
U
U
U
U
U
U
U

T2

U
U
U
U
U

D1

U
U

D3

O1

O2

O3

U
U
U
U
U
U
U
U
U
U
U
U
U

U
U
U
U
U
U
U
U
U
U
U
U
U

U
U

U
U
U
U
U
U
U
U
U
U
U
U

U
U
U
U
U
U
U
U
U
U
U
U

U
U
U

U
U
U

U
U
U

U
U

O4

U
U

784

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

3.3. A tabulated comparison of the models


As discussed earlier, fuel consumption depends on a number of
factors which can be grouped into ve categories: vehicle (V), environment (E), trafc (T), driver (D), and operations (O) related
factors. Using these ve categories, we present in Table 7 a
tabulated comparison of the 25 models reviewed so far.
Table 7 shows that most of the models consider vehicle load
and speed, although the way in which they incorporate these factors into the approximations, in particular for load, varies highly.
Microscopic models are similar in that they are able to reect a certain level of detail so as to consider technical and vehicle-specic
parameters such as vehicle shape (frontal area), and road conditions (e.g., gradient, surface resistance). This is not the case for
macroscopic models which mostly present mostly simpler ways
of regression-based estimations through a predened set of
parameters for a given vehicle class. It is worth mentioning that
only a few of the models listed in Table 7 consider driver-related
factors and none of the models considers congestion (factor T 3 ),
since the inclusion of such detailed factors in the estimation process is rather difcult. Furthermore, none of the models included
in Table 7 considers factors D2 and E6 .
Table 8 gives a three-point scale analysis of the models in order
to analyze several dimensions which depend on the requirements
of the users and on the performance of the models. Robustness is
dened as the ability of the model to accommodate various situations. Reliability is dened as the degree of the accuracy by which
the model performs its intended functionalities. Applicability in
optimization represents models practicability in mathematical
modeling. Responsiveness to other air pollutants measures the
capability of the model to measure other air pollutants beside fuel
consumption. The level of awareness corresponds to the popularity
of the model in the literature. The level of data requirement shows
the complexity of the parameters used to calculate vehicle emissions. The level of support and continuous improvements show

the technical support availability and the frequency of revisions


in time, respectively.
Based on our judgement and on the academic literature, all
models are investigated according to these eight dimensions.
According to Table 8, microscopic models seem more robust, reliable and more applicable in the area of optimization. Data requirement is also more important for microscopic models as was
discussed earlier. On the other hand, macroscopic emissions have
more technical support and provide continuous improvement.
Macroscopic models are also more capable of estimating other
air pollutants.

4. Fuel consumption modeling in road transportation planning


We now review emission models used at different levels of
planning for road freight transportation. The planning levels in
transportation are usually categorized based on three temporal
dimensions: strategic, tactical and operational. Green road transportation studies can also be categorized according to these three
dimensions.
The factors affecting fuel consumption should be considered
within a specic time frame. For example, green freight corridors
are long-term planning decisions. Fleet size and driver selection
are medium-term planning problems. Other routing decisions are
matters of short-term planning. Since strategic planning is not
directly related to routing and scheduling operations, it is not covered by this survey. Furthermore, tactical and operational planning
will be combined.
Freight transportation includes many facets, particularly when
viewed from the multiple levels of decision making. Arguably the
most important problem at the operational level is routing and
scheduling. This section is concerned with vehicle routing studies
which include an explicit consideration of environmental concerns,
in particular CO2 e emissions. There exists an extensive body of

Table 8
A three-point scale analysis of macroscopic and microscopic models.
Robustness

Macroscopic models
MEET
s
NTM
s
COPERT

ECOTRANSIT s
EMFAC
s
NAEI
MOBILE
s
MOVES
HBEFA

GREET
LEM
s
VERSIT-LD
s
IVE
s
Macroscopic models
IFCM
F-MEFCM
RSFCM
ASFCM
s
VSP

EMIT
s
VT-Micro
s
OFCM
s
PERE
CPFM
s
PHEM

CMEM

 High;

Medium; s Low.

Reliability

Applicability in
optimization

Responsiveness to
other air pollutants




s

s
s


s
s
s

s
s
s



s
s


s



The level of data


requirement

s


The level of
awareness

s

s
s


s





s
s
s
s


s
s

The level of
technical support

The level of
continuous
improvement

s




s




s
s
s

s
s

s
s

s



s
s
s


s
s
s












s



s
s
s

s
s
s
s
s
s
s
s
s
s

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

research on vehicle routing. In contrast, the literature on the green


routing is relatively young (Sbihi & Eglese, 2007), but the interest
in environmental issues has grown in recent years. As for freight
transportation, most of the published studies on routing are concerned with the minimization of internal (operational) costs. In
the following sections, we review several applications with different emission modeling approaches at the operational planning level of planning. In particular, we differentiate the approaches by
the use of different emission models covered in the previous
section.
4.1. Applications of IFCM
IFCM (Eq. (6)) was used by Palmer (2007) who considered CO2
emissions models in the context of the vehicle routing with time
windows (VRPTW). The objective was to develop an effective
method of constructing vehicle routes minimizing CO2 emissions.
Results indicate that there exists a potential for reducing CO2 emissions by around 5% by minimizing this objective instead of travel
time.
A second application of IFCM was given by Urquhart, Hart, and
Scott (2010), who used the IFCM to solve the VRPTW for low CO2
solutions using evolutionary algorithms. The authors looked at
the trade-offs between CO2 savings, distance and the required
number of vehicles. Their results indicate that savings of up to
10% can be achieved, depending on the problem instance and the
ranking criterion used in the evolutionary algorithm.
4.2. An application of VSP
An eco-friendly routing problem was studied by Bandeira, Carvalho, Khattak, Rouphail, and Coelho (2012) who used a microscopic
vehicle scheduling model (Eq. (9)) to assess whether eco-friendly
routes change during peak hours, based on three distinct casestudies on two continents. They authors collected more than
13,330 kilometer of data using GPS equipped light diesel and gasoline duty vehicles, in the United States and Portugal. They demonstrated empirically that most eco-friendly route during off-peak
hours is also the most eco-friendly during peak hours. Regarding
intercity routes, tests conducted during both the peak and the
off-peak hours have shown that the faster routes can reduce CO2
emissions by up to 30%. In a related research, Bandeira, Almeida,
Khattak, Rouphail, and Coelho (2013) investigated a way to generate information about emissions for drivers faced with a choice of
routes. According to their results, the choice of a route can substantially affect emission rates of the analyzed pollutants and that
smoother driving styles can also result in considerable emissions
reduction. For example, faster intercity routes are more desirable
in terms of fuel use and CO2 emissions. However, these same
routes yielded more CO and N2 O emissions.
4.3. Applications of OFCM
Using OFCM (Eq. (12)), Kono, Fushiki, Asada, and Nakano (2008)
introduced an ecological route search system which minimizes fuel
consumption by considering many factors such as trafc information, geographic information, and vehicle parameters. The authors
collected an extensive data set in areas with different geographical
features and in various trafc conditions. Their results suggest that
the fuel consumption of the eco-routes is 9% lower than that of the
time priority routes, even though the travel time of the ecological
routes is 9% longer. They argue that ecological routes could be different from time-priority routes, contrary to the conclusions of the
study by Bandeira et al. (2012).
In a similar study, Correia, Amaya, Meyer, Kumagai, and Okude
(2010) have analyzed eco-routing in France, with the use of OFCM.

785

The authors proposed an analysis method using geographical features for the needs of eco-route search. According to their results,
the fuel savings afforded by eco-routes could be between 1% and
36% as compared to fastest routes.
4.4. An application of CPFM
CPFM (Eq. (15)) was used by Minett, Salomons, Daamen, Van
Arem, and Kuijpers (2011) who studied eco-routing to reduce fuel
consumption considering motorways, local and provincial routes.
The authors developed a tool in which historical link speed data
were used as a basis for replicating vehicle speed proles, thus enabling the calculation of fuel costs per link. These fuel costs and
speed proles were validated by eld test data. According to the
results, based on eld tests, provincial route would offer an average
time saving of 10.41 (25%) minutes over the local route, and an
average fuel saving of 1.25 liter over the motorway route (45%).
4.5. Applications of CMEM
A rst application of the CMEM within the vehicle routing is
that of Bektas and Laporte (2011). The authors have introduced
the pollution-routing problem (PRP), an extension of the classical
vehicle routing problem with time windows (VRPTW) which consists of routing a number of vehicles to serve a set of customers,
and determining their speed on each route segment, so as to minimize a function comprising fuel, emission and driver costs. In estimating pollution, the authors consider factors such as speed, load,
and time windows, using the emissions function Eq. (17). They assume that in a vehicle trip all parameters will remain constant on a
given arc, but load and speed may change from one arc to another.
Their model approximates the total amount of energy consumed
on the arc, which directly translates into fuel consumption and further into GHG emissions. Computational results reported by the
authors suggest that by using the proposed approach, energy savings can be up to 10% when time windows are in place, and up to
4% when the demand variation is high.
The second application of the CMEM to the PRP is due to Demir
et al. (2012), who computed the total fuel consumption per second
via Eq. (18). The authors proposed a heuristic to solve the PRP. The
algorithm iterates between the solution of the VRPTW and of a
speed optimization problem. The former problem was solved
through adaptive large neighborhood search (ALNS) (Ropke &
Pisinger, 2006) and the latter was solved through a polynomial
time speed optimization procedure (SOP) as will be discussed later.
The authors introduced new removal and insertion operators
aimed at improving solution quality in terms of CO2 emissions.
In order to analyze the fuel consumption more accurately, the
authors generated benchmark instances by randomly selecting cities from the United Kingdom and therefore using real geographical
distances. All instances can be downloaded from http://
www.apollo.management.soton.ac.uk/prplib.htm. According to
their results, the reduction on CO2 emissions were around 10%.
In a related research, Demir, Bektas, and Laporte (2013) investigated the trade-offs between fuel consumption and driving time
via CMEM. They showed that trucking companies need not compromise greatly in terms of driving time in order to achieve a
signicant reduction in fuel consumption and CO2 emissions. The
converse of this insight also holds, i.e., considerable reductions in
driving time are achievable if one is willing to increase fuel
consumption only slightly.
Using CMEM, Ramos, Gomes, and Barbosa-Pvoa (2012) investigated the service areas and routes that minimize the CO2 emissions of a transportation system with multiple products and
depots. The authors proposed a decomposition solution approach
and applied it to a case study in order to reshape the current

786

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

system and create more environmental-friendly routes. According


to their results, by comparing the CO2 emissions of the current system with the proposed one, a decrease of 23% can be achieved if
the company reshapes both its service areas and vehicle routes.
The total potential savings can be up to 20% if the company keeps
its current service areas.
One last application of CMEM reviewed here is by Franceschetti,
Honhon, Van Woensel, Bektas, and Laporte (2013) who investigated fuel consumption, CO2 emissions and driver costs under trafc congestion. The authors identied the conditions under which it
is optimal to wait idly at certain locations in order to avoid congestion and to reduce the cost of emissions. Their results suggest that
a 20% reduction in the cost of fuel, CO2 emissions and driver cost is
achievable based on the instances generated by Demir et al. (2012).
4.6. Applications of MEET
The MEET model (Eq. (1)) has been widely used by several
researchers. One of the rst applications of MEET is due to Kim,
Janic, and Van Wee (2009) who investigated the relationships between freight transport costs and CO2 emissions in intermodal
and truck only freight networks, as a part of multi-objective study.
CO2 emissions were estimated as a function of cruising speed and
distance traveled, and average values were used for other factors,
such as cold-start emissions and ambient temperature.
The second application is due to Figliozzi (2010) who studied the
minimization of emissions and fuel consumption. The author introduced the emissions vehicle routing problem (EVRP), an extension
of the time-dependent vehicle routing problem (TDVRP). The
author described a formulation and an algorithm for the problem.
The objective of the EVRP is the minimization of emission cost,
which is proportional to the amount of GHG emitted, which in turn
is a function of travel speed and distance traveled. In the proposed
algorithm, a partial EVRP is rst solved to minimize the number of
vehicles by means of a TDVRP algorithm, following which and emissions are optimized subject to a eet size constraint. The departure
times are also optimized using the proposed algorithm for any pair
of customers. The author worked with three trafc conditions:
uncongested, somewhat congested, and congested trafc conditions. Results presented on the Solomon (1987) instances suggested
that uncongested travel speeds tend to reduce emissions on average; however, this is not always the case, and the opposite trend
is sometimes observed. The author suggests that a 20% reduction
is possible by optimizing departure times. In another study
(Figliozzi, 2011), the same author analyzed CO2 emissions for different levels of congestion and time-denitive customer demands.
Travel time data from an extensive archive of freeway sensors,
time-dependent vehicle routing algorithms, and instances with different types of binding constraints were used to study the impacts
of congestion on commercial vehicle emissions.
Pan, Ballot, and Fontane (2010) have used MEET to analyze the
effect of pooling supply chain networks on the reduction of transport-related CO2 emissions. The authors have computed CO2 emissions for two transport modes with real data. Their results show
that reductions of 14% in CO2 emissions for road transport only
and of 52% for joint road and rail transport can be achieved.
Another application of MEET is due to Jabali, Van Woensel, and
de Kok (2012) who have investigated travel times and CO2 emissions in the context of the TDVRP by analyzing the effect of limiting
vehicle speed. The authors also addressed the issue of congestion,
where the vehicle is forced to drive slower and therefore emits
more CO2 . The authors provided a formulation of the problem in
which the costs of driving, fuel and CO2 are included. Using tabu
search algorithm, they achieved a reduction of 11.2% in CO2 emissions on average, which required increasing the travel time by

14.8% under a 90 kilometer/hour speed prole. Under a speed limit


of 80 kilometer/hour, the authors report that an increase of 4.7% in
travel time yields a reduction of 3.7% in CO2 emissions.
Omidvar and Tavakkoli-Moghaddam (2012) introduced a vehicle routing model for alternative fuel vehicles (AFV) (e.g., hybrid,
electric and fuel cell vehicles) in order to minimize fuel consumption and CO2 emissions. They have tested their model on the
Solomon benchmark instances. Their results suggest that the
importance of CO2 emissions costs for these vehicle is around 3%
of the total costs.
Another application of MEET is due by Saberi and Verbas (2012)
who looked at emissions minimization in the context of the TDVRP.
The authors have developed a continuous approximation model to
analyze strategic planning of one-to-many distribution systems
and have evaluated the effects of emissions costs. Their results suggest that a 50% reduction in emissions is possible.
4.7. Applications of NAEI
The NAEI (Eq. (5)) model was applied by Maden, Eglese, and
Black (2010) to the vehicle routing and scheduling problem with
an objective of minimizing the total travel time under congestion.
The authors take into account regular congestion due to volume of
trafc, and long-term road works, which can be predicted from historical data. Their results suggest that the proposed approach may
yield CO2 emissions savings of up to 7%.
Urquhart, Scott, and Hart (2010) have sought to identify traveling salesman tours with low CO2 emissions. The authors examined
two different fuel emission models: a power based instantaneous
fuel consumption model (IFCM) and a simpler spreadsheet based
model (NAEI). The authors have applied their algorithm to the latter model. Computational results on six randomly generated instances having between 10 and 30 delivery points suggest that
only a small improvement can be achieved by using the fuel emission model because of the inadequacy of the emission model.
4.8. An application of MOVES
MOVES model was used by Wygonik and Goodchild (2011) who
introduced the emissions minimization VRP with time windows
(EVRPTW). The authors have modeled a specic case study involving a real eet with specic operational characteristics and have
analyzed different external policies and internal operational rules.
Their results show a stable relationship between monetary cost
and kilograms of CO2 emissions: each kilogram of CO2 results in
an increase in cost by 3.5 dollars.
4.9. Applications of COPERT
COPERT (Eq. (3)) emission models have been widely used in different studies. The rst study was conducted by Tavares, Zsigraiova, Semiao, and da Graa Carvalho (2008), who looked at the
optimization of routing networks for waste transportation. Their
model takes into account both the road angle and the vehicle load.
Their ndings indicate that optimizing fuel consumption can yield
savings of up to 52% in fuel when compared with minimizing distance. Another paper of Tavares, Zsigraiova, Semiao, and Carvalho
(2009) is concerned with the optimization of municipal solid waste
collection routes to minimize fuel consumption using 3D GIS modeling. The results of this study suggest that the proposed approach
reduces traveled distance by 29% and fuel consumption by 16%.
Scott, Urquhart, and Hart (2010) have studied the effect of
topology and payload in the context of CO2 optimized vehicle routing. They worked on two different data sets: delivery of groceries
to households and delivery of paper from a central warehouse.

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

They also worked on several traveling salesman instances randomly chosen from problem data sets, and assigned average
speeds for each road category. Their results suggest that the effect
of gradient and payload are highly dependent on the instances considered. The difference in CO2 emissions between the instances
was found to be less than 2.1% for the solutions obtained with
COPERT.
Using COPERT models, Jovicic et al. (2010) have investigated the
energy efciency to estimate the potential for reduction of fuel
consumption and thus CO2 emissions through the optimization
of solid waste collection. The authors show that in the City of Kragujevac in Serbia, savings of up to 20% can be achieved in costs and
the associated emissions.
The PhD thesis of Qian (2012) studies fuel emission optimization in VRP with time-varying speeds. The author aims to generate
routes and schedules for a eet of heavy goods vehicles so as to
minimize the emissions in a road network where travel speeds
are time-dependent. Computational results obtained by the algorithm on a London case study suggest that 67% savings in fuel
consumption may be achieved using the proposed approach.

Fig. 3. Number of papers per year until August 2013.

787

4.10. An application of IVE


Another eco-routing study was carried out by Kang, Ma, Ma,
and Huang (2011) who used the IVE model to minimize fuel consumption and vehicle emissions. The authors sought to determine
the vehicle specic power, which is derived from their vehicle
routing and scheduling model. Their eco-routing approach may
yield CO2 emissions savings of up to 18% compared a time-minimizing objective and by 8% compared to a distance-minimizing
objective.
4.11. Other applications
EMVRP: The energy-minimizing vehicle routing problem was
introduced and formulated by Kara et al. (2007). The objective of
the EMVRP is to minimize a weighted load function as a way of
estimating fuel consumption. This function is based on a physics
rule stating that on a at surface work equals force times distance.
The integer linear programming model proposed for the EMVRP is
based on that of the capacitated VRP. Since the model minimizes
the total work done on the road, the authors argue that this leads
to minimizing the total energy requirements, at least in terms of
total fuel consumption. They study the differences between distance-minimizing and energy-minimizing solutions on benchmark
capacitated VRP instances from the literature and nd that energy
usage increases as total distance decreases. They conclude that
there is a considerable difference between energy-minimizing
and distance-minimizing solutions, and that the cost of routes
minimizing total distance may be up to 13% less than those minimizing energy.
FCVRP: Xiao, Zhao, Kaku, and Xu (2012) studied the fuel consumption rate in the context of the capacitated VRP. To estimate
fuel consumption, the authors use a regression model based on statistical data, proposed by the Ministry of Land, Infrastructure,
Transport and Tourism of Japan. They present a mathematical
model and apply a simulated annealing algorithm to solve the
problem.
EVRP-VC: The emission minimization vehicle routing problem
with vehicle classes were introduced by Kopfer, Schnberger, and
Kopfer (in press). The model minimizes fuel consumption instead
of driving distance by offering the possibility of using several types
of vehicles. The authors compared their model to that of the

Fig. 4. The analysis of publications with respect to the factors used. Note: The number refers to the publication in Tables 9 and 10.

788

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

DSOP: The departure time and speed optimization procedure is


an extension of SOP and was studied by Franceschetti et al. (2013)
to calculate the departure times from depot and customer sites
while optimizing speeds on each arc of a given route.

5. Descriptive statistics
In this section, we presents several statistics of the growing
body of literature on green road freight transportation. The section
is organized based on three indicators, namely the type of publication, the factors taken into account in estimating fuel consumption,
and the reported savings.
Fig. 5. Savings in terms of CO2 e.

traditional VRP. They replaced the distance minimizing objective


function with an afne and piecewise linear fuel consumption
function which considers payload. According to their results, a signicant amount of reduction is possible through the use of a heterogeneous eet of vehicles.
SOP: The speed optimization procedure (SOP) is a polynomial
time exact algorithm and used by Demir et al. (2012) in solving
the PRP. Given a vehicle route, the SOP consists of nding the optimal speed on each arc of the route between successive nodes so as
to minimize an objective function comprising fuel consumption
costs and driver wages. The objective of SOP is non-linear due to
the function used to estimate the fuel consumption of a vehicle.

5.1. Type of publication


We rst analyze the number of manuscripts and categorize
them according to publication type and year: journal articles (JA),
conference proceedings (CP), technical reports (TR), and PhD theses
(PT). The number of publications of each type is shown in Fig. 3.
As can be seen from Fig. 3, a total of 58 publications are associated with the reduction of fuel consumption in vehicle routing and
scheduling. There is an indication of the increase of publications
after 2009. The number of JA; CP; TR and PT publications are 33,
16, 6 and 3, respectively. The increase in the number of publications is approximately 430% from 2009 to 2012.

Table 9
Literature on green road freight transportation-1.
Reference

Publication type

Problem

Mathematical model

Solution Approach

Case study

Emission model

CO2 e savings

1. Ando and Taniguchi (2006)


2. Ericsson et al. (2006)
3. Tavares et al. (2008)
4. Apaydin and Gonullu (2008)
5. Tavares et al. (2009)
6. Chalkias and Lasaridi (2009)
7. Maden et al. (2010)
8. Kuo (2010)
9. Figliozzi (2010)
10. Jovicic et al. (2010)
11. Pan et al. (2010)
12. Kuo and Wang (2011)
13. Figliozzi (2011)
14. Bektas and Laporte (2011)
15. Suzuki (2011)
16. Ubeda et al. (2011)
17. Faulin et al. (2011)
18. Xiao et al. (2012)
19. Jabali et al. (2012)
20. Demir et al. (2012)
21. Treitl et al. (2012)
22. Aranda et al. (2012)
23. Saberi and Verbas (2012)
24. Li (2012)
25. Paksoy and zceylan (in press)
26. Bandeira et al. (2013)
27. Kwon et al. (2013)
28. Franceschetti et al. (2013)
29. Oberscheider et al. (2013)
30. Pradenas et al. (2013)
31. Gaur et al. (2013)
32. Kopfer et al. (in press)
33. Demir et al. (2013)

JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA
JA

VRPTW-P
Eco
Eco
Eco
Eco
Eco
VRPTW
TDVRP
EVRP
Eco
SCD
VRP
TDVRP
PRP
TSPTW
VRPB
CVRP
FCVRP
ETDVRP
PRP
IRP
Eco
EVRP
VRPTW
SCD
Eco
HVRP
TDPRP
MDVRPPDTW
VRPBTW
CumVRP
EVRP
BiPRP

o,
s
s
s
s
s
h
h
o,
s
o
h
o,
o
o
o
h
o,
o,
o,
o
h
h
o,
o
s
o,
o
o,
o
o,
o,
h

U
U
U
U
U
U
U

Factor
Factor
COPERT
Factor
COPERT
Factor
NAEI
Factor
MEET
COPERT
MEET
Factor
MEET
CMEM
Factor
Factor
Factor
Factor
MEET
CMEM
Factor
Factor
MEET
Factor
Factor
VSP
Factor
CMEM
Factor
CMEM
Factor
Factor
CMEM

II
II
IV
IV
III
I
II
IV
IV
IV
III
II
I
II
II
IV

U
U
U
U
U
U
U
U
U
U
U

U
U
U
U
U
U
U
U

h
U
U
h

U
U
h
h
h
U
U
h
U
h
h
h
h

I
I
II
III

IV
IV
III
IV
IV
IV
IV
II

VRPTW-P: Vehicle Routing and Scheduling Problems with Time Windows - Probabilistic; Eco: Eco-routing; VRPTW: Vehicle Routing Problem with Time Windows; TDVRP:
Time-Dependent Vehicle Routing Problem; EVRP: Emissions Vehicle Routing Problem; SCD: Supply Chain Design; VRP: Vehicle Routing Problem; PRP: Pollution-Routing
Problem; TSPTW: Traveling Salesman Problem with Time Windows; VRPB: Vehicle Routing Problem with Backhauling; CVRP: Capacitated Vehicle Routing Problem; FCVRP:
Fuel Consumption rate considered Capacitated Vehicle Routing Problem; ETDVRP: Emissions-based Time-Dependent Vehicle Routing Problem; IRP: Inventory Routing
Problem; HVRP: Heterogenous Vehicle Routing Problem; TDPRP: Time-Dependent Pollution-Routing Problem; MDVRPPDTW: Multi-Depot Vehicle Routing Problem with
Pickup and Delivery and Time Windows; VRPBTW: Vehicle Routing Problem with Backhauling and Time Windows; CumVRP: Cumulative Vehicle Routing; BiPRP: Bi-objective
Pollution-Routing Problem; o: Optimal algorithm; h: Heuristic algorithm; s: Software application.

789

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

5.2. Factors used in publications to estimate fuel consumption

and 10, most of the papers are supported by case studies, which
indicates the interest of grounding the research in real-life applications. The most popular models are the COPERT at a macroscopic
level and CMEM at a microscopic level. Solution approaches remain mostly heuristic, given the increasing complexity of the problems when emission models are integrated.

In this section, we analyze publications based on the factors


used in their research. Fig. 4 illustrates the factors used in 58 publications reviewed in this paper.
This gure is quite revealing in several ways. First, we see that
most of the studies consider speed and load in their formulations.
As discussed in Demir et al. (2012), speed has a signicant effect in
fuel consumption, and optimal speed could lead better reduction in
terms of CO2 . Second, congestion is another important factor but
only very few studies have looked at this factor. Third, driver
behavior and road gradient are mostly studied in eco-routing studies with the help of GIS softwares. Finally, eet size and mix are
attracting increasing attention due to the benets of selecting the
right type of vehicle. As discussed in Section 2, the total vehicle
mass is directly related to the fuel consumption and any extra
weight will lead more consumption.

6. Conclusions and future research directions


Over the years, the minimization of the total traveled distance
has been accepted as the most important objective in the eld of
vehicle routing and freight transportation. However, the interaction of operational research with mechanical and trafc engineering shows that there exist factors which are critical to explain fuel
consumption. This triggered the birth of the green road transportation studies in operational research. In recent years, the number,
quality and the availability of the available models have increased
considerably. Our research leads to the following ve important
conclusions:

5.3. CO2 e emissions savings

 Most studies in the eld of green road freight transportation


have focused on a limited number of factors, mainly vehicle
load and speed. In particular, vehicle speed is shown to be quite
crucial and that it is important to travel at a speed that leads to
minimum fuel consumption for a given routing plan.
 Several studies suggest that, for a same load, light duty vehicles
should be preferred over medium duty (MD) and heavy duty
(HD) vehicles. MD should be also preferred to HD vehicles if
possible.
 A positive road gradient leads to an increase in fuel consumption and it should be taken into account in route planning in
future applications. Current GIS software can provide information of the road gradient.

In this section, we analyze the CO2 e savings reported in the


literature based on various applications. These are grouped into
four categories: Group I: 05.99%, Group II: 610.99%, Group III:
1115.99%, and Group IV: higher than 16%. We provide a graph
of savings in Fig. 5. It is apparent from this gure that the actual
savings are heavily dependent on the vehicle emission model used,
but signicant reductions in CO2 e emissions are possible. We note,
however, that the savings are related to the particular factors
considered in the solution methodology.
Tables 9 and 10 contain a summary of the publications
reviewed in this paper. Since the topic has received attention from
researchers of various backgrounds, this categorization can be very
useful to the research community. As can be seen from Tables 9

Table 10
Literature on green road freight transportation-2.
Reference

Publication
type

Problem

34. Christie and Satir (2006)


35. Kara et al. (2007)
36. Palmer (2007)
37. Kono et al. (2008)
38. Yong and Xiaofeng (2009)
39. Urquhart, Hart et al. (2010)
40. Urquhart, Scott et al. (2010)
41. Scott et al. (2010)
42. Correia et al. (2010)
43. Hao (2010)
44. Pitera et al. (2011)
45. Goodchild (2011)
46. Kang et al. (2011)
47. Minett et al. (2011)
48. Wygonik and Goodchild (2011)
49. Qian (2012)
50. Demir (2012)
51. Jemai et al. (2012)
52. Chen and Chen (2012)
53. Bandeira et al. (2012)
54. Omidvar and Tavakkoli-Moghaddam
(2012)
55. Huang et al. (2012)
56. Ramos et al. (2012)
57. Rao and Jin (2012)
58. Peiying et al. (2013)

CP
CP
PT
TR
CP
CP
TR
CP
TR
CP
CP
TR
CP
CP
CP
PT
PT
CP
TR
CP
CP

CVRSO
EMVRP
CVRP
Eco
VRPFC
TSP
VRPTW
Eco
Eco
VRP
TDVRP
TDVRPPDTW
Eco
Eco
EVRPTW
CVRP
VRPTW
GVRP
VRP
Eco
TDVRP

CP
TR
CP
CP

VRPSPD
MDVRP
VRPMEC
VRPPD

Mathematical
model

U
U
U

U
U
U
U

Solution
Approach
s
o
h
s
h
h
h
s
s
o
h
h
s
s
s
h
o, h
h
s
s
h
o
o, h
o, h
h

Case
study

U
U
U
U
U
U
U
U
U
U
U

Emission
model

CO2 e
savings

Factor
Factor
IFCM
OFCM
Factor
IFCM
NAEI
COPERT
OFCM
Factor
Factor
Factor
IVE
CPFM
MOVES
COPERT
CMEM
Factor
Factor
VSP
MEET

IV
III
I
II

Factor
CMEM
CMEM
CMEM

II
I
II
IV
IV
IV
IV
I
II
II
I
IV
III
I
I
IV
I

CVRSO: Computerized Vehicle Routing and Scheduling Optimization; EMVRP: Energy minimizing Vehicle Routing Problem; VRPFC: Vehicle Routing Problem with Fuel
Consumption; TSP: Traveling Salesman Problem; TDVRPPDTW: Time-dependent Vehicle Routing Problem with Pickup and Delivery and Time Windows; EVRPTW: Emissions
minimization Vehicle Routing Problem with Time Windows; GVRP: Green Vehicle Routing Problem; VRPSPD: Vehicle Routing Problem Simultaneous Pickup and Delivery;
MDVRP: Multi-Depot Vehicle Routing Problem; VRPMEC: Vehicle Routing Problem Minimizing the Energy Consumption; VRPPD: Vehicle Routing Problem with Pickup and
Delivery.

790

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

 Further studies should focus on the selection of appropriate


emission tools and standardize them. Our ndings show that
the COPERT and CMEM models serve different purposes but
greatly depend on the availability of data for accurate
calculations.
 Besides CO2 e emissions, other trafc externalities could be
examined at the local and regional levels, such as other pollutants, noise, accidents, and environmental damage.
Considering current studies on green road freight transportation, the following ve areas were identied as further research
directions.
 The current GIS software uses very simple regression models,
based solely on the distance traveled, to estimate fuel consumption. The microscopic models could be integrated within such
software to take fuel consumption into account. The integration
of the algorithms reviewed in this paper and GIS software could
save both fuel and time.
 The interactive selection of solutions generated via multi-objective optimization should allow for exibility, productivity and
support for the route planners of freight companies. Multiobjective optimization may come to play a vital role in green
road transportation studies, particularly as the different objectives (e.g., distance, time, emissions) are incommensurable.
 Most studies only focus on the routing aspect of green logistics.
Other problems that can be linked to routing may offer former
reductions in emissions. For example, facility location is concerned with physically locating a set of facilities (depots) so as
to maximize the prot generated by providing service to a set
of customers. The relocation of a depot may lead to reductions
in CO2 e emissions. Some attempts in this direction have already
appeared in the literature (see, e.g., Harris, Mumford, & Naim,
2011).

 The selection of the right vehicles from an available set is


another promising area to minimize CO2 emissions. The effect
of the characteristics of a vehicle on CO2 e emissions is reviewed
in Section 2 but this is an area in need of more effort for comprehensive investigations. The eet size and mix vehicle routing
problem usually consists of determining the type and the number of vehicles of each type in order to minimize acquisition and
routing costs. To our knowledge, there exist only very few studies carried out within the perspective of fuel consumption and
emissions.
 Driver working hours should be considered in the green logistics studies because of law requirements, as well as the
acknowledged health hazards arising from intensive workload
of some routing plans. Although some of the available studies
could conceivably be modied to reect these type of real-life
requirements, this issue still requires further attention.

Acknowledgements
The authors gratefully acknowledge funding provided by the
European Unions Seventh Framework Programme (FP7/20072013) under Grant Agreement 318275 (GET Service), by the Technology University of Eindhoven, by the University of Southampton
and by the Canadian Natural Sciences and Engineering Research
Council under Grant 39692-10. Thanks are due to the Professor
Roman Slowinski and to the three referees for their valuable
comments.

Appendix A
We summarize the common notation used in describing the
emission functions in Table 11.

Table 11
Nomenclature for functions mentioned in the paper.
Notation

Description

er v
Ev ; D
fi lf
Elf ; D
Fv ; D
FD; M
Fv
f t
FT
F a ta
F d t b
F c tc
F i td
Fta ; td ; tc ; t i
Fti ; t s
F s D
VSPt
FT; M
fe t
F e T
fv t
FDl
F o T
fpr t
F pr T
F 1 T
F 2 T
Etrans t
fcm t
F cm T

The rate of CO2 emissions for speed v (gram/kilometer)


The total amount of CO2 emissions for v and distance D (gram)
The rate of fuel consumption for each type of road i at the specied load factor lf (liter/kilometer)
The total amount of CO2 emissions for the specied load factor lf and distance D (kilogram)
The total amount of fuel consumption for speed v and distance D (gram)
The total amount of fuel consumption for distance D and load M (megajoule)
The total amount of fuel consumption for speed v (liter/100 kilometer)
The rate of fuel consumption per unit time (milliliter/second)
The total amount of fuel consumption for a journey of duration T (milliliter)
The total amount of fuel consumption during acceleration (milliliter)
The total amount of fuel consumption during deceleration (milliliter)
The total amount of cruise fuel consumption (milliliter)
The total amount of fuel consumption while idle (milliliter)
The total amount of fuel consumption including all modes (milliliter)
The total amount of fuel consumption for idle time ti and traveling time t s (milliliter)
The total amount of fuel consumption for distance D (milliliter)
Vehicle specic power per unit ton (kilowatt/ton)
The total amount of fuel consumption for a journey of duration T and load M (gram)
The rate of fuel consumption per unit time (gram/second)
The total amount fuel consumption for a journey of duration T (gram)
The rate of fuel consumption per unit time (liter/second)
The total amount of fuel consumption for distance of link l (liter)
The total amount of fuel consumption for a journey of duration T (milliliter)
The rate of fuel consumption per unit time (kilogram/second)
The total amount of fuel consumption for a journey of duration T (kilogram)
The total amount of fuel consumption for a journey of duration T (liter)
The total amount of fuel consumption for a journey of duration T (liter)
The total amount of CO2 emissions under the transient conditions per unit time (gram)
The rate of fuel consumption per unit time (gram/second)
The total amount of fuel consumption for a journey of duration T (gram)

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

References
Ahn, K., Rakha, H., Trani, A., & Van Aerde, M. (2002). Estimating vehicle fuel
consumption and emissions based on instantaneous speed and acceleration
levels. Journal of Transportation Engineering, 128(2), 182190.
Akelik, R., & Besley, M. (1990). SIDRA 5 User Guide. Technical report, ARRB
Transport Research Ltd., Vermont South, Australia.
Akelik, R, & Besley, M. (2003). Operating cost, fuel consumption, and emission
models in aaSIDRA and aaMOTION. In 25th Conference of Australian Institutes
of Transport Research (CAITR 2003), University of South Australia, Adelaide,
Australia.
Alwakiel, H.N. (2011). Leveraging Weigh-In-Motion (WIM) data to estimate linkbased heavy-vehicle emissions. PhD thesis. Portland State University, USA.
Ando, N., & Taniguchi, E. (2006). Travel time reliability in vehicle routing and
scheduling with time windows. Networks and Spatial Economics, 6(3), 293311.
Apaydin, O., & Gonullu, M. T. (2008). Emission control with route optimization in
solid waste collection process: A case study. Sadhana, 33(2), 7182.
Aranda, U. A., Ferreira, G., Bribin, Z. I., & Vsquez, Z. D. (2012). Study of the
environmental performance of end-of-life tyre recycling through a simplied
mathematical approach. Thermal Science, 16(3), 889899.
Ardekani, S., Hauer, E., & Jamei, B. (1996). Trafc impact models. In Trafc ow
theory (pp. 17). Washington, DC, USA.
Bandeira, J., Almeida, T. G., Khattak, A. J., Rouphail, N. M., & Coelho, M. C. (2013).
Generating emissions information for route selection: Experimental monitoring
and routes characterization. Journal of Intelligent Transportation Systems, 17(1),
317.
Bandeira, J., Carvalho, D. O., Khattak, A. J., Rouphail, N. M., & Coelho, M. C. (2012). A
comparative empirical analysis of eco-friendly routes during peak and off-peak
hours. In Transportation research board 91st annual meeting. Washington, USA.
Barth, M., & Boriboonsomsin, K. (2008). Real-world CO2 impacts of trafc
congestion. Transportation Research Record: Journal of the Transportation
Research Board, 2058(1), 163171.
Barth, M., Younglove, T., & Scora, G. (2005). Development of a heavy-duty diesel
modal emissions and fuel consumption model. Technical report. UC Berkeley:
California Partners for Advanced Transit and Highways (PATH), California, USA.
<http://www.path.berkeley.edu/PATH/Publications/PDF/PRR/2005/PRR-200501%.pdf> (01.12.13).
Bektas, T., & Laporte, G. (2011). The pollution-routing problem. Transportation
Research Part B: Methodological, 45(8), 12321250.
Bigazzi, A. Y., & Bertini, R. L. (2009). Adding green performance metrics to a
transportation data archive. Transportation Research Record: Journal of the
Transportation Research Board, 2121(1), 3040.
Boulter, P. G., & McCrae, I. (2009). ARTEMIS: Assessment and reliability of transport
emission models and inventory systems: Final report. Technical report,
Berkshire, United Kingdom. <http://www.inrets.fr/ur/lte/publi-autresactions/
chesresultats/chear%temis/artemis.html> (01.12.13).
Bowyer, D. P., Biggs, D. C., & Akelik, R. (1985). Guide to fuel consumption analysis
for urban trafc management. Australian Road Research Board Transport
Research, 32.
Cappiello, A., Chabini, I., Nam, E. K., Lue, A., & Abou Zeid M. (2002). A statistical
model of vehicle emissions and fuel consumption. In Proceedings the IEEE 5th
international conference intelligent transportation systems, 2002 (pp. 801
809). Singapore.
CARB, (2011). EMFAC2011 technical documentation. Technical report, California,
USA, 2011. <http://www.arb.ca.gov/msei/emfac2011-documentation-nal.pdf>
(01.12.13).
CAT, (2006). Tractor-trailer performance guide. Technical report, 2006. <http://
www.cat.com/cda/les/2222280/> (01.12.13).
Chalkias, C., & Lasaridi, K. (2009). A GIS based model for the optimisation of
municipal solid waste collection: The case study of Nikea, Athens, Greece.
WSEAS Transactions on Environment on and Development, 1, 1115.
Chen, B. S., & Chen. Y. H. (2012). The carbon emission reduction strategies: An
example of mobile phone logistics. Technical report, Taipei City, Taiwan, 2012.
<http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2176576> (01.12.13).
Christie, J. S., & Satir, S. (2006). Saving our energy sources and meeting Kyoto
emission reduction targets while minimizing costs with application of vehicle
logistics optimization. In 2006 Annual conference and exhibition of the
Transportation Association of Canada: Transportation without boundaries.
Charlottetown, Prince Edward Island.
Correia, A., Amaya, S., Meyer, S., Kumagai, M., & Okude, M. (2010). Eco Routing for
European Market. In 17th ITS World Congress. Busan, Korea.
Crainic, T. G. (2000). Service network design in freight transportation. European
Journal of Operational Research, 122(2), 272288.
DEFRA, (2012). 2012 Guidelines to DEFRA/ DECCs GHG conversion factors for
company reporting: Methodology paper for emission factors. Technical report,
London, United Kingdom. <http://https://www.gov.uk/government/uploads/
system/uploads/attachment_data/%le/69568/pb13792-emission-factormethodology-paper-120706.pdf> (01.12.13).
Dekker, R., Bloemhof, J., & Mallidis, I. (2012). Operations Research for green
logisticsAn overview of aspects, issues, contributions and challenges. European
Journal of Operational Research, 219(3), 671679.
Delucchi, M. (2003). A lifecycle emissions model (LEM): Lifecycle emissions from
transportation fuels, motor vehicles, transportation modes, electricity use,
heating and cooking fuels, and materials. Technical report, California, USA.
<http://www.escholarship.org/uc/item/9vr8s1bb> (01.12.13).

791

Demir, E. (2012). Models and algorithms for the pollution-routing problem and its
variations. PhD thesis, University of Southampton, United Kingdom.
Demir, E., Bektas, T., & Laporte, G. (2011). A comparative analysis of several vehicle
emission models for road freight transportation. Transportation Research Part D:
Transport and Environment, 6(5), 347357.
Demir, E., Bektas, T., & Laporte, G. (2012). An adaptive large neighborhood search
heuristic for the pollution-routing problem. European Journal of Operational
Research, 223(2), 346359.
Demir, E., Bektas, T., & Laporte, G. (2013). The bi-objective pollution-routing
problem. European Journal of Operational Research, 232(3), 464478.
Dobers, K., Klukas, A., Lammers, W., Laux, M., Mauer, G., & Schneider, M. (2013).
Green logistics: Optimisation approaches for resource-efcient logistics
services. In Efciency and logistics (pp. 149161).
EC, (2010). Road freight transport vademecum. Technical report. <http://
ec.europa.eu/transport/modes/road/doc/2010-road-freight-vademecu%m.pdf>
(01.12.13).
Eglese, R. W., & Black, D. (2010). Optimizing the routing of vehicles. In A. McKinnon,
S. Cullinane, M. Browne, & W. Whiteing, (Eds.), Green Logistics: Improving the
environmental sustainability of logistics (pp. 215228). Great Britain/
KoganPage.
EPA, (2003). Users guide to MOBILE6.1 and MOBILE6.2: Mobile source emission
factor model. Technical report. United States Environmental Protection
Agency,
USA.
<http://www.epa.gov/oms/models/mobile6/420r03010.pdf>
(01.12.13).
EPA, (2012). Motor vehicle emission simulator (MOVES): Users guide for
MOVES2010b. Technical report, United States Environmental Protection
Agency,
USA.
<http://www.epa.gov/oms/models/moves/documents/
420b12001b.pdf> (01.12.13).
Ericsson, E., Larsson, H., & Brundell-Freij, K. (2006). Optimizing route choice for
lowest fuel consumptionPotential effects of a new driver support tool.
Transportation Research Part C: Emerging Technologies, 14(6), 369383.
ESRL, 2013. NOAA In Situ Carbon Dioxide (CO2 ) Measurements. Technical report,
National Oceanic and Atmospheric Administration. <http://www.esrl.noaa.gov/
gmd/obop/mlo/programs/esrl/co2/co2.html> (01.12.13).
Esteves-Booth, A., Muneer, T., Kubie, J., & Kirby, H. (2002). A review of vehicular
emission models and driving cycles. Journal of Mechanical Engineering Science,
216(8), 777797.
Faulin, J., Juan, A., Lera, F., & Grasman, S. (2011). Solving the capacitated vehicle
routing problem with environmental criteria based on real estimations in road
transportation: A case study. Procedia-Social and Behavioral Sciences, 20,
323334.
Figliozzi, M. A. (2010). Vehicle routing problem for emissions minimization.
Transportation Research Record: Journal of the Transportation Research Board,
2197(1), 17.
Figliozzi, M. A. (2011). The impacts of congestion on time-denitive urban freight
distribution networks CO2 emission levels: Results from a case study in
Portland, Oregon. Transportation Research Part C: Emerging Technologies, 19,
766778.
Forkenbrock, D. J. (1999). External costs of intercity truck freight transportation.
Transportation Research Part A: Policy and Practice, 33(78), 505526.
Forkenbrock, D. J. (2001). Comparison of external costs of rail and truck freight
transportation. Transportation Research Part A: Policy and Practice, 35(4),
321337.
Franceschetti, A., Honhon, D., Van Woensel, T., Bektas, T., & Laporte, G. (2013). The
time-dependent pollution routing problem. Transportation Research Part B:
Methodological, 56, 265293.
Gaur, D. R., Mudgal, A., & Singh, R. R. (2013). Routing vehicles to minimize fuel
consumption. Operations Research Letters, 41(6), 576580.
GHG Protocol, (2013). GHG emissions from transport or mobile sources. Technical
report, Greenhouse Gas Protocol, USA. <http://www.ghgprotocol.org/
calculation-tools/all-tools> (01.12.13).
Goodchild, A. (2011). Cost, emissions, and customer service trade-off analysis in
pickup and delivery systems. Technical report, Civil & Environmental
Engineering, University of Washington, USA.
Hao, X. (2010). A fuel consumption objective of VRP and the genetic algorithm. In
IEEE Conference (pp. 14). Henan, China.
Harris, I., Mumford, C. L., & Naim, M. M. (2011). An evolutionary bi-objective
approach to the capacitated facility location problem with cost and CO2
emissions. In Proceedings of the 13th annual conference on Genetic and
evolutionary computation (pp. 697704). NY, USA: ACM.
Hausberger, S., Rexeis, M., Zallinger, M., & Luz, R. (2009). Emission factors from the
Model PHEM for the HBEFA: Version 3. Technical report. TU Graz, Austria.
Hickman, J., Hassel, D., Joumard, R., Samaras, Z., & Sorenson, S. (1999). MEETMethodology for calculating transport emissions and energy consumption.
European Commission, DG VII. Technical report. <http://www.transportresearch.info/Upload/Documents/200310/meet.pdf> (01.12.13).
Huang, Y., Shi, C., Zhao, L., & Van Woensel, T. (2012). A study on carbon reduction in
the vehicle routing problem with simultaneous pickups and deliveries. In
Service Operations and Logistics, and Informatics (SOLI), 2012 IEEE
international conference (pp. 302307).
INFRAS, (1995). Workbook on emission factors for road transport: explanatory
notes. Technical report, Bern, Switzerland, 1995. <http://www.inrets.fr/ur/lte/
cost319/M22.pdf> (01.12.13).
ISSRC, 2008. IVE model users manual version 2.0. Technical report, International
Sustainable Systems Research Center. <http://www.issrc.org/ive/downloads/
manuals/UsersManual.pdf> (01.12.13).

792

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793

Jabali, O., Van Woensel, T., & de Kok, A. G. (2012). Analysis of travel times and CO2
emissions in time-dependent vehicle routing. Production and Operations
Management, 21(6), 10601074.
Jemai, J., Zekri, M., & Mellouli, K. (2012). An NSGA-II algorithm for the green vehicle
routing problem. In Evolutionary computation in combinatorial optimization:
12th European Conference, Malaga, Spain (Vol. 7245, pp. 3748).
Jimenez-Palacios, J. L. (1998). Understanding and quantifying motor vehicle
emissions with vehicle specic power and TILDAS remote sensing. PhD thesis,
USA: Massachusetts Institute of Technology.
Jovicic, N. M., Bokovic, G. B., Vujic, G. V., Jovicic, G. R., Despotovic, M. Z.,
Milovanovic, D. M., et al. (2010). Route optimization to increase energy
efciency and reduce fuel consumption of communal vehicles. Thermal
Science, 14, 6778.
Kang, J., Ma, T., Ma, F., & Huang, J. (2011). Link-based emission model for eco
routing. In ITS Telecommunications (ITST), 11th international conference (pp.
207212). Saint Petersburg, Russia.
Kara, I., Kara, B. Y., & Yetis, M.K. (2007). Energy minimizing vehicle routing problem.
In A. Dress, Y. Xu, & B. Zhu, (Eds.), Combinatorial optimization and applications,
Lecture notes in computer science Vol. 4616. Berlin/Heidelberg.
Kim, N. S., Janic, M., & Van Wee, B. (2009). Trade-off between carbon dioxide emissions
and logistics costs based on multiobjective optimization. Transportation Research
Record: Journal of the Transportation Research Board, 2139(1), 107116.
Knrr, W., Seum, S., Schmied, M., Kutzner, F., & Anthes, R. (2011). Ecological
transport information tool for worldwide transports. Technical report. <http://
www.ecotransit.org/download/ecotransit_background_report.pdf> (01.12.13).
Kono, T., Fushiki, T., Asada, K., & Nakano, K. (2008). Fuel consumption analysis and
prediction model for eco route search. In 15th world congress on intelligent
transport systems and ITS Americas 2008 annual meeting.
Kopfer, H. W., Schnberger, J., & Kopfer, H. (2013). Reducing greenhouse gas
emissions of a heterogeneous vehicle eet. Flexible Services and Manufacturing
(in press).
Kouridis, C., Gkatzoias, D., Kioutsioukis, I., Ntziachristos, L., Pastorello, C., & Dilara,
P. (2010). Uncertainty estimates and guidance for road transport emission
calculations. Technical report, European Commission Joint Research Centre
Institute
for
Environment
and
Sustainability.
<http://
publications.jrc.ec.europa.eu/repository/handle/111111111/14202> (01.12.13).
Kuo, Y. (2010). Using simulated annealing to minimize fuel consumption for the
time-dependent vehicle routing problem. Computers & Industrial Engineering,
59(1), 157165.
Kuo, Y., & Wang, C. C. (2011). Optimizing the VRP by minimizing fuel consumption.
Management of Environmental Quality, 22(4), 440450.
Kwon, Y.-J., Choi, Y.-J., & Lee, D.-H. (2013). Heterogeneous xed eet vehicle routing
considering carbon emission. Transportation Research Part D: Transport and
Environment, 23, 8189.
Li, J. (2012). Vehicle routing problem with time windows for reducing fuel
consumption. Journal of Computers, 7(12), 30203027.
Lin, C., Choy, K. L., Ho, G. T. S., Chung, S. H., & Lam, H. Y. (2013). Survey of green
vehicle routing problem: Past and future trends. Expert Systems with
Applications, 41(4), 11181138.
Maden, W., Eglese, R. W., & Black, D. (2010). Vehicle routing and scheduling with
time-varying data: A case study. Journal of the Operational Research Society,
61(3), 515522.
Minett, C. F., Salomons, A. M., Daamen, W., Van Arem, B., & Kuijpers, S. (2011). Ecorouting: Comparing the fuel consumption of different routes between an origin
and destination using eld test speed proles and synthetic speed proles. In
Integrated and Sustainable Transportation System (FISTS), 2011 IEEE Forum
(pp. 3239). Vienna, Austria.
NAEI, (2012). Emission factors. Technical report, National Atmospheric Emissions
Inventory, United Kingdom. <http://naei.defra.gov.uk/data/emission-factors>
(01.12.13).
Nam, E. K., & Giannelli, R. A. (2005). Fuel consumption modeling of conventional
and advanced technology vehicles in the physical emission rate estimator
(PERE). Technical report, Washington, USA. <http://www.epa.gov/oms/models/
ngm/420p05001.pdf> (01.12.13).
NTM Road, (2010). Environmental data for international cargo transport. Technical
report.
Oberscheider, M., Zazgornik, J., Henriksen, C. B., Gronalt, M., & Hirsch, P. (2013).
Minimizing driving times and greenhouse gas emissions in timber transport
with a near-exact solution approach. Scandinavian Journal of Forest Research,
28(5), 493506.
Oguchi, T., Katakura, M., & Taniguchi, M. (2002). Carbondioxide emission model in
actual urban road vehicular trafc conditions. Journal of the Japan Society of Civil
Engineers (695), 125136.
Omidvar, A., & Tavakkoli-Moghaddam. R. (2012). Sustainable vehicle routing:
Strategies for congestion management and refueling scheduling. In Energy
Conference and Exhibition (ENERGYCON), 2012 IEEE international (pp. 1089
1094). Florence, Italy.
Paksoy, T., & zceylan, E. (2013). Environmentally conscious optimization of
supply chain networks. Journal of the Operational Research Society (in press).
Palmer, A. (2007). The development of an integrated routing and carbon dioxide
emissions model for goods vehicles. PhD thesis. Bedford, United Kingdom:
Craneld University.
Pan, S., Ballot, E., & Fontane, F. (2010). The reduction of greenhouse gas emissions
from freight transport by pooling supply chains. International Journal of
Production Economics, 143(1), 8694.

Peiying, Y., Jiafu, T., & Yang, Y. U. (2013). Based on low carbon emissions cost model
and algorithm for vehicle routing and scheduling in picking up and delivering
customers to airport service. In 25th Chinese Control and Decision Conference
(CCDC). Proceedings. The IEEE international conference (pp. 16931697).
Guiyang, China.
Pelkmans, L., & Debal, P. (2006). Comparison of on-road emissions with emissions
measured on chassis dynamometer test cycles. Transportation Research Part D:
Transport and Environment, 11(4), 233241.
Pitera, K., Sandoval, F., & Goodchild, A. (2011). Evaluation of emissions reduction in
urban pickup systems. Transportation Research Record: Journal of the
Transportation Research Board, 2224(1), 816.
Pradenas, L., Oportus, B., & Parada, V. (2013). Mitigation of greenhouse gas
emissions in vehicle routing problems with backhauling. Expert Systems with
Applications, 40, 29852991.
Qian, J. (2012). Fuel emission optimization in vehicle routing problems with
time-varying speeds. PhD thesis. Lancaster, United Kingdom: Lancaster
University.
Rakha, H., Ahn, K., & Trani, A. (2003). Comparison of MOBILE5a, MOBILE6, VTMICRO, and CMEM models for estimating hot-stabilized light-duty gasoline
vehicle emissions. Canadian Journal of Civil Engineering, 30(6), 10101021.
Rakha, H., Ahn, K., Moran, K., Saerens, B., & Van den Bulck, E. (2011). Simple
comprehensive fuel consumption and CO2 emission model based on
instantaneous vehicle power. In 90th transportation research board annual
meeting. Washington, USA.
Ramos, T. R. P., Gomes, M. I., & Barbosa-Pvoa, A. P. (2012) Minimizing CO2
emissions in a recyclable waste collection system with multiple depots. In
EUROMA/POMS joint conference. Amsterdam, The Netherlands.
Rao, W., & Jin, C. (2012). A model of vehicle routing problem minimizing energy
consumption in urban environment. In 2012 Asian conference of management
science and Applications. Proceedings (pp. 2129).
Rexeis, M., Hausberger, S., & Riemersma, I. (2005). Emissions and fuel consumption
from Heavy Duty Vehicles. Technical report, COST 346, TU Graz, Austria.
<http://www.transport-research.info/Upload/Documents/200906/
20090619_171%904_26922_II_COST346_WGA_FinalReport.pdf> (01.12.13).
Ropke, S., & Pisinger, D. (2006). An adaptive large neighborhood search heuristic for
the pickup and delivery problem with time windows. Transportation Science,
40(4), 455472.
Ross, M. (1994). Automobile fuel consumption and emissions: Effects of vehicle and
driving characteristics. Annual Review of Energy and the Environment, 19(1),
75112.
_ . (2012). Continuous approximation model for the vehicle
Saberi, M., & Verbas, I.
routing problem for emissions minimization at the strategic level. Journal of
Transportation Engineering, 138(11), 13681376.
Salimifard, K., Shahbandarzadeh, H., & Raeesi, R. (2012). Green transportation and
the role of operation research. In International Conference on Trafc and
Transportation Engineering (ICTTE 2012). Belgrade, Serbia.
Sbihi, A., & Eglese, R.W. (2007). The relationship between vehicle routing &
scheduling and green logistics-A literature survey. Technical report, Lancaster
University Management School, United Kingdom. <http://halshs.archivesouvertes.fr/hal-00644133/> (01.12.13).
Schreyer, C., Schneider, C., Maibach, M., Rothengatter, W., Doll, C., & Schmedding, D.
(2004). External costs of transport: Update study. Technical report, INFRAS.
<http://ecocalc-test.ecotransit.
org/CE_Delft_4215_External_Costs_of_Tran%
sport_in_Europe_def.pdf> (01.12.13).
Scora, G. & Barth, M. (2006). Comprehensive modal emission model (CMEM),
version 3.01: Users guide. Technical report, University of California, USA.
<http://www.cert.ucr.edu/cmem/docs/CMEM_User_Guide_v3.01d.pdf>
(01.12.13).
Scott, C., Urquhart, N., & Hart, E. (2010). Inuence of topology and payload on CO2
optimised vehicle routing. Applications of Evolutionary Computation, 6025,
141150.
Smit, R., Smokers, R., & Rab, E. (2007). A new modelling approach for road trafc
emissions: VERSIT+. Transportation Research Part D: Transport and Environment,
12(6), 414422.
Solomon, M. M. (1987). Algorithms for the vehicle routing and scheduling problems
with time window constraints. Operations Research, 35(2), 254265.
Suzuki, Y. (2011). A new truck-routing approach for reducing fuel consumption and
pollutants emission. Transportation Research Part D: Transport and Environment,
16(1), 7377.
Tavares, G., Zsigraiova, Z., Semiao, V., & Carvalho, M. G. (2009). Optimisation of
MSW collection routes for minimum fuel consumption using 3D GIS modelling.
Waste Management, 29(3), 11761185.
Tavares, G., Zsigraiova, Z., Semiao, V., & da Graa Carvalho, M. (2008). A case study of
fuel savings through optimisation of MSW transportation routes. Management
of Environmental Quality, 19(4), 444454.
Touati, N., & Jost, V. (2012). On green routing and scheduling problem. Technical
report, 2012. <http://arxiv.org/pdf/1203.1604.pdf> (01.12.13).
Treitl, S., Nolz, PC., & Jammernegg, W. (2012). Incorporating environmental aspects
in an inventory routing problem. A case study from the petrochemical industry.
Flexible Services and Manufacturing Journal, 127.
Ubeda, S., Arcelus, F. J., & Faulin, J. (2011). Green logistics at Eroski: A case study.
International Journal of Production Economics, 131(1), 4451.
Urquhart, N., Hart, E., & Scott, C. (2010). Building low CO2 solutions to the vehicle
routing problem with Time Windows using an evolutionary algorithm. In
Evolutionary Computation (CEC), 2010 IEEE congress (pp. 16). Barcelona, Spain.

E. Demir et al. / European Journal of Operational Research 237 (2014) 775793


Urquhart, N., Scott, C., & Hart, E. (2010). Using an evolutionary algorithm to discover
low CO2 tours within a travelling salesman problem. Applications of Evolutionary
Computation, 6025, 421430.
Van Woensel, T., Creten, R., & Vandaele, N. (2001). Managing the environmental
externalities of trafc logistics: The issue of emissions. Production and
Operations Management, 10(2), 207223.
Wang, M. Q. (1999). GREET 1.5-transportation fuel-cycle model-Vol. 1:
Methodology, development, use, and results. Technical report. Illinois, USA:
Argonne National Laboratory.
Williams, I., Kemp, S., Coello, J., Turner, DA., & Wright, L. A. (2012). A beginners
guide to carbon footprinting. Carbon, 3(1), 5567.
Wong, J. Y. (2001). Theory of ground vehicles. New York: Wiley-Interscience.
Wright, L. A., Kemp, S., & Williams, I. (2011). Carbon footprinting: towards a
universally accepted denition. Carbon, 2(1), 6172.

793

Wygonik, E., & Goodchild, A. (2011). Using a GIS-based emissions minimization


vehicle routing problem with time windows (EVRPTW) model to evaluate CO2
emissions and cost trade-offs in a case study of an urban delivery system. In
Transportation research board 90th annual meeting (number 111109),
Washington, DC, USA.
Xiao, Y., Zhao, Q., Kaku, I., & Xu, Y. (2012). Development of a fuel consumption
optimization model for the capacitated vehicle routing problem. Computers &
Operations Research, 39(7), 14191431.
Yong, P., & Xiaofeng, W. (2009). Research on a vehicle routing schedule to reduce
fuel consumption. In Measuring technology and mechatronics automation (Vol.
3, pp. 825827). Shanghai, China.