310 Int. J. Services and Operations Management, Vol. 16, No.

3, 2013

The sustainable routing problem

Maurizio Faccio*, Alessandro Persona and

Giorgia Zanin
Department of Management and Engineering,
University of Padova,
Stradella San Nicola 3, 36100, Vicenza – Italy
Fax: +39-0498276816
E-mail: [email protected]
E-mail: [email protected]
E-mail: [email protected]
*Corresponding author

Abstract: Nowadays environmental, social, and political pressures to limit the

impacts linked to CO2 emissions are rapidly increasing. This paper aim at the
solution of a new class of A routing problem: the sustainable routing problem,
where the objective, unlike the classical approaches, is the CO2 emission
minimisation. The amount of CO2 emitted by a vehicle depends on different
factors. If the problem is analysed from the driver’s point of view, these factors
can be divided into ‘internal’, i.e., directly related to him/her (i.e., driving style,
acceleration, average speed, route knowledge, etc.), and into ‘external’, i.e.,
depending on the environment in which the trip is made (i.e., traffic congestion,
speed limits, vehicle type, etc.). Firstly the paper, starting from Fonseca et al.
(2011) study, proposes a CO2 emission estimation as function of the different
internal/external factors. Secondly, the authors formalise a sustainable routing
model, validating it through a case study as well as a parametric analysis. The
results suggest that, contrary to the classical vehicle routing problem, the
proposed model allows a larger reduction of CO2 emission versus the
increasing time and distance, suggesting interesting potentials in the
environmental preservation.

Keywords: routing problem; sustainability; CO2 emission; vehicle emission

factors; urban congestion.

Reference to this paper should be made as follows: Faccio, M., Persona, A. and
Zanin, G. (2013) ‘The sustainable routing problem’, Int. J. Services and
Operations Management, Vol. 16, No. 3, pp.310–336.

Biographical notes: Maurizio Faccio graduated in Engineering Management at

University of Padova with evaluation 110/110 Cum Laude, discussing a thesis
concerning: ‘Mixed-Model Assembling Line Balancing Problem’ in December
2003. On January 2005, he began his PhD in Mechatronics and Industrial
System at University of Padova. Since June 2006, he has been an Assistant
Professor at the Faculty of Engineering of University of Padova in ING-IND/17
SDS. Since 2003, he has been carrying out his own research activities at
University of Padova and the principal activities of his research are about
industrial system and service industrial system design, productive plant design,
production and logistic system management.

Copyright © 2013 Inderscience Enterprises Ltd.

 The sustainable routing problem 311

Alessandro Persona is a Full Professor of Industrial and Mechanical Plants at

the University of Padova, Italy in the disciplinary sector of ING-IND/17. The
Research Group of SSD ING-IND/17 in Padua University has been evaluated
as ‘very good-excellent’ by an international panel of referee that analysed five
years of activities: ‘This small group has excellent scientific production and
funding from local industry. Overall assessment: very good – excellent’.
Recently Alessandro Persona received an important award: Best paper award
(2008) for the paper ‘A single-vendor multi-buyer consignment stock inventory
model’ written with Battini, Faccio, Sgarbossa, received during GBID (BIRC,
Emerald, Elsevier, Science Direct Conference) in Rio De Janeiro.

Giorgia Zanin is a PhD student in Mechatronics and Industrial System at

University of Padova. She has been carrying out her own research in the
Department of Management and Technology at the University of Padua. Her
main work focuses on advanced technologies in logistic traceability and on
supply chain network design and management.

1 Introduction

The reduction of carbon dioxide emission is one of the priorities in most environmental
policies.
The transportation sector produces the largest percentage of emissions from fossil
fuel combustion by end use sector, releasing more than 1800 teragrams (Tg) of CO2
equivalents in 2008 and representing nearly one-third of the emissions from fossil fuel
combustion (United States. Environmental Protection Agency, 2010).
As the demand on the world’s resources continues to increase, cities, regions, and
states find themselves needing to foster economic growth and development while
minimising impacts on the environment (Wygonik and Goodchild, 2011).
As well known, these emissions have both direct consequences on human health (i.e.,
pollution), and indirect ones (i.e., by the depletion of the ozone). An agreement to reduce
CO2 emissions from light-duty vehicles (European Parliament, 2009) has been signed in
the European Union, limiting CO2 emissions to 130 g/km, with progressive
implementation from 2012 to 2015. Even if vehicle manufacturers have been forced to
incorporate technological improvements such as weight lightening, engine size reduction,
low-rolling-resistance tyres, improved aerodynamics and hybridisation and electrification
of vehicles, it is clear that the way of using vehicle plays an important role in the CO2
emission. As far as this last issue is concerned it is possible to say that the technical
research in the environmental sustainable vehicles design is still not enough.
It is evident that in order to go from a certain point A to a certain point B, there is a
large number of factors that influence the CO2 emission. If the problem is analysed from
the driver point of view, these factors could be divided into two sets:
• internal factors, i.e., directly depending on the driver, on the vehicle utilisation way
such as the driving style, the type of route chosen in order to reach a certain point;
the acceleration style (i.e., high/low acceleration at each stop), the average speed
(i.e., close/far to the speed limits), the street knowledge, etc.
• external factors, i.e., not depending on the vehicle utilisation way, like the vehicle
type, the speed limits, the traffic congestion, etc.
312 M. Faccio et al.

The internal and external factors are obviously connected. For example for a certain
chosen route (internal factor), a speed limit is defined for each of its parts, and in function
of the day time an expected traffic congestion (external factors) is also derived.
If the vehicle routing problem (VRP) is considered, i.e., the determination for a
certain vehicle of the specific number and sequence of customers to visit and the
determination of the way to drive according to a certain objective, it is clear how the
route solution has a large impact in the CO2 emission.
This aspect is supported by the experimental study of De Vlieger et al. (2000) who
analyse the relation among fuel consumption and emission, driving behaviour and traffic
congestion. According to their study, compared to normal driving, aggressive driving can
increase fuel consumption of 40% and emissions up to a factor 8 for the same route
covered. Moreover, they highlight how traffic conditions have a large influence on fuel
consumption and emissions. The increases in fuel consumption and emissions during rush
hours can vary from 10% to 200%. More environment-friendly route options could
require the use of ring roads and motorways even during rush hours instead of short cuts,
sometimes with a longer travelling time though. This can involve higher but sometimes
even lower emissions.
In the classical VRP are considered as typical objectives the cost minimisation, the
time minimisation or the distance minimisation. As highlighted in Figliozzi (2011)
although past and current research efforts in the vehicle routing algorithms and
scheduling are extensive (Cordeau et al., 2007), most of them have ignored the
freight-related environmental and social consequences like CO2 emission level.
The aim of the present paper is to study a new paradigm in the VRP: the sustainable
routing problem (SRP) in which the focus is just the CO2 emission minimisation. This
specific research area presents few and recent contributes (Bektaş and Laporte, 2011;
Figliozzi, 2010), although the transportation sector produces the largest percentage of
emissions from fossil fuel combustion by end use sector.
The authors, starting from the Fonseca et al. (2011) study, intend to propose a simple
model for CO2 emission increment estimation as function of the different
internal/external factors. Furthermore, the authors formalise a sustainable routing model,
validating it through a case study.
In order to demonstrate the innovation of the proposed model, a parametrical analysis
is developed to answer the following question: how can a routing model, focused on the
CO2 emission minimisation, compete with the classical routing models, which aim to
manage the short term distribution by the distance/time minimisation?
How far are these classical objectives to minimise also the CO2 emission? How far is
the proposed routing model (SRP) from the classical objectives (distance/time) that
directly impact on the companies costs? How can the internal/external factors influence
the SRP and the classical VRP? How should the modern technologies and the drivers
attitudes support these new routing perspective?
The main contributions of the present paper are, not only to introduce an innovative
point of view in the routing model optimisation, but also to understand the gap with the
classical routing approaches. Moreover, the proposed study tries to analyse the possible
situations and condition in which this new approach can be applied in the modern
distribution network and to provide some technological and organisational elements to
support it.
The remainder of the paper is organised as follows. Section 2 reports the background
and literature on the routing problem and on the CO2 emission related to the
 The sustainable routing problem 313

transportation and routing problem. Section 3 describes the CO2 estimation model and
SRP formulation, while Section 4 presents a case study. Section 5 shows the parametrical
analysis and the SRP benchmarking versus the classical routing approaches, while the
conclusions are drawn in Section 6.

2 Background and literature review

The VRP is a significant issue in the supply chain management and it has been widely
studied by the researchers (Benton and Rossetti, 1992). The basic version of the problem
considers a single depot, from where all the tours start and end, and a fleet of vehicles,
which has to serve a set of customers on a given network under side constraints
(Sivakumar et al., 2012). The traditional objectives of the VRP include the minimisation
of the total distance or time travelled by all vehicles, or the minimisation of the overall
travel cost, usually a linear function of distance (Bektaş and Laporte, 2011). The most
common variant is the capacitated vehicle routing problem (CVRP) where for
each vehicle is limited the amount of load it can carry. Another common variant is the
vehicle routing problem with time windows (VRPTW) (Malairajan et al., 2012;
Anbuudayasankar et al., 2009) where it is considered a time interval, given by an earliest
arrival time and a latest arrival time; customers must be visited within this predefined
time slots. Also lean approach in the goods distribution problem is available (Faccio et
al., 2012).
Since the very beginning of the VRP, literature has become quite disjointed and
disparate. As a consequence, several versions of the problem and a wide variety of exact
and approximate algorithms have been proposed. Generally, the VRP is computationally
very hard, and cannot be solved by exact methods; therefore, the literature proposes many
heuristics (Faccio et al., 2011).
The classical heuristic can be broadly classified into three categories (Laporte and
Semet, 2002):
• constructive heuristics: these algorithms start from an empty solution and iteratively
built routes by inserting one or more customers at each iteration, until all of them are
routed
• two-phase heuristics: the problem is decomposed into two components, the
clustering, which defines a partition of the customer into subsets, corresponding to a
route, and the route construction
• improvement methods: these methods attempt to upgrade any feasible solution by
performing a sequence of edge or vertex exchanges within or between vehicle routes.
The model proposed in the present paper belongs to the category of constructive
heuristics. For a detailed classification of the literature see Eksioglu et al. (2009) and
Cordeau et al. (2007).
Due to the growing pressure to limit the Greenhouse Gas (GHG) emissions and the
awareness that the transportation field produces the largest percentage of emissions from
fossil fuel combustion, a wide range of models, mainly based on simulation, have been
proposed to predict fuel consumption and emission rates (Bektaş and Laporte, 2011). For
this reason, the Green Supply Chain is becoming an important research field in the supply
314 M. Faccio et al.

chain management, also distinguished into specific country characteristics (Kim

et al., 2011).
The amount of pollution emitted by a vehicle depends on several factors, such as
vehicle characteristics, inspection and maintenance levels, fuel characteristics etc.
Consequently, vehicle manufacturers have been forced to incorporate technological
improvements like weight lightening, engine size reduction, low-rolling-resistance tyres,
improved aerodynamics and hybridisation and electrification of vehicles (Fonseca et al.,
2011). An alternative solution technology that vehicle manufacturers have developed is
the ‘stop/start’ system, in which the internal combustion engine is automatically powered
off when the car is stopped and restarted upon driver’s demand or when needed. Fonseca
et al. (2011) studied the effect of this technology and points out that it allows A
significant reduction of CO2emissions in actual urban traffic.
Other relevant factors that influence the emission are the vehicle driving patterns
(André and Rapone, 2009), average speed (De Vlieger et al., 2000), the degree of
acceleration (Larsson and Ericsson, 2009) and also the type of street (i.e., rural or urban)
and the level of traffic congestion (Smit et al., 2008).
A wide number of CO2 emission models are developed to assess these emission
factors, like Fontaras et al. (2007), André and Rapone (2009), Larsson and Ericsson
(2009). Smit et al. (2008), proposed an exhaustive classification of the existing emission
models. The authors examined how, and to what extent, models, which are currently used
to predict emissions and fuel consumption from road traffic, include the effects of
congestion. This is because congestion can be associated with a number of interrelated
and observable effects that include increased traffic density, increased travel times on the
same route (hence lower average speeds), increased deviation from free-flow speeds
(hence increased speed fluctuation and delays) and increased queuing (hence increased
number of stops). It is not possible to take account of unexpected events that may cause
congestion, like an accident, but regular congestion related to the time of the day of travel
can be predicted from historical data. Woensel et al. (2001) point out that the use of
constant speed, like in most emission models, will lead to an underestimation of the
effective emissions. Thereby they used queuing theory to model the impact of traffic
congestion on emissions and recommended that private and public decision makers
should take into account the high impact of congestion on emissions. Maden et al. (2010)
analyse a vehicle routing and scheduling problem with time windows considering the
speed variability due to the time of travel, and the relative level of congestion. The
authors describe a heuristic algorithm to solve the problem and present a case study of a
fleet of delivery vehicles in the UK. They report up to 7% savings in CO2 emissions
when compared with a planning method using constant speed, even if the algorithm does
not plan routes that directly minimise the emission.
Although an extensive number of studies propose a CO2 emission model, few
researchers have developed routing tools that optimise emission (Wygonik and
Goodchils, 2011).
Bektaş and Laporte (2011) present the pollution-routing problem, an extension of the
classical VRP with a broader and more comprehensive objective function that accounts
not only the travel distance, but also the amount of greenhouse emissions, fuel, travel
times and their costs. The authors point out the trade-offs between various parameters
such as vehicle load, speed and total cost, and offer insights on economies of
environmental-friendly vehicle routing. Dessouky et al. (2003) present a methodology for
the joint optimisation of cost, service, and life-cycle environmental consequences in
 The sustainable routing problem 315

vehicle routing and scheduling. This research looks at a number of measures of

environmental performance and considers the life-cycle environmental impacts of each
solution anyway, it does not focus on or minimise the CO2 emissions associated with
routing.
Figliozzi (2010) formulated the emissions vehicle routing problem (EVRP) with
time-dependent travel times, hard time windows, and capacity constraints. The author
showed that a routing formulation and solution algorithm that takes into account
congestion and aims to minimise CO2 emissions can produce significant reductions in
emission levels with relatively small increases in THE distance travelled or fleet size.
Nevertheless the formulation proposed includes speed and departure time as decision
variables, which are not always possible in real condition.
As conclusion of the literature review, it is possible to highlight that there are many
contributes on the vehicles CO2 emission analysis, but few researchers have developed
routing tools that minimise the emission. For this reason the aim of this paper is to
contribute in this important research area:
• showing the effect on the amount of CO2 emitted by a vehicle due to several factors,
‘internal’, depending on the driver (i.e., driving style, acceleration, average speed,
route knowledge, etc.), and ‘external’ (i.e., traffic congestion, speed limits, vehicle
type, etc.) starting from the current state of the art in the vehicle CO2 emission
estimation models
• introducing a new sustainable routing model that gives as result sustainable routes,
i.e., a specific sequence of customers to visit and A way to drive that minimises the
CO2 emission in function of these different internal/external factors, proposing as
results sustainable routes
• highlighting how this sustainable routes are generally not equal to those obtained by
the classical routing problem solution using objectives function like distance/time
minimisation, especially for particular values of internal/external factors like traffic
congestion level or driving style.
• pointing out how, where it is not possible to have instant information on the traffic
congestion, the driver familiarity with the territory results in an important element to
support this new approach.

3 CO2 estimation model and SRP formulation

The SRP can be defined as follows.

For a considered geographical area in which the routing problem is set, it is possible
to define:
• a series of nodes, i.e., a series of points that can represent customers, depots, or
generally other specific locations
• a series of arcs, i.e., ways to connect two different nodes; each street can be
represented by an arc
• a graph as the union of nodes and arcs that represent the whole geographical area.
316 M. Faccio et al.

So, given the graph G = (N, A), where N = {1..n} is the set of nodes and A = {(i, j) | i, j ∈
N, i ≠ j} is the set of arcs, each vehicle emits a certain amount of CO2 when travelling
over an arc (i, j).
The proposed sustainable routing model is defined in Figure 1.

Figure 1 Sustainable routing model

A set of different factors influence the CO2 emission. They can be divided, from the
driver point of view, into internal/external factors:
• Internal factors, which depend on the driver like the driving style and the route
knowledge. These factors influence the instantaneous acceleration (i.e., high/low
acceleration at each start and stop), and consequently the route average speed.
• External factors, i.e., not depending on the driver, like streets/vehicle characteristics,
or traffic congestion level.
These factors define the inputs for the proposed model that, in the remainder of paper, are
called ‘scenario’s inputs’. These are all the variables of the problem related to the
internal/external factors, and are calculated as function of these factors as explained in the
following part of this section. The ‘scenario’s inputs’ are:
• the length of the arc (i, j), kmi,j,
• the maximum speed at which a vehicle can travel on the arc (i, j), Vmaxi,j,
• the expected level of traffic congestion in each arch, Tsi,j,
• The instantaneous acceleration, AI, related to the driving style (i.e., aggressive,
normal, etc).
The present paper suggests a CO2 estimation model starting from Fonseca et al. (2011)
study, which proposed a general mathematical expression to calculate the specific
kilometre CO2 emission:
 The sustainable routing problem 317

c1 c
evco 2 = C0 + + C2 * Av + C3 * G + + C5 * TE (1)
sv 2 t2
where
• C0, C1, C2, C3, C4 and C5 are coefficients derived by the empirical test of the model.
They depend on the vehicle characteristics.
• Sva is the average speed, [km/h].
• Av is the average acceleration, [km/h/s].
• G is the road grade, [%].
• TE is the engine oil temperature, [°C].
As assumptions in the paper the engine oil temperature, TE, is neglected, because it is
assumed that the engine is always in temperature. Also the road grade, G, is left out,
because Fonseca et al. (2011) study points out that this coefficient has lost more influence
on the CO2 emission than the others.
The basic idea of the proposed research is that using the Fonseca formulation for the
CO2 emission estimation, it is possible to calculate the CO2 increase/decrease at the
variance of the internal/external factors. In this way it will be possible to optimise the
routing according the CO2 minimisation.
It is important to highlight how the proposed model is dynamical, i.e., at the change
of the internal/external factors (i.e., traffic flow, driving style, etc.) it will be possible to
re-define the optimum sustainable routing.
The tables show the step-by-step description of the proposed sustainable routing
model, defined using the pseudo code language. Pseudo code is kind of structured
English for describing algorithms that allows to focus on the logic and on the basic idea
of the algorithm.
Table 1 Scenario’s input

Scenario’s input Description Measure unit

G Graph: G = (N, A)
N Vertex set: N = {1..n}
A Arc set: A = {(i, j) | i, j ∈ N, i ≠ j} [num.]
V Set of vertexes that must be visited: V = {S1, S2, S3} [num.]
Xi,j Binary index equal to 1 if a vehicles can travel on arc (i, j) [num.]
Kmi,j Length of the arc (i, j) [km]
Vmaxi,j Maximum speed at which a vehicle can travel on the arc (i, j) [km/h]
AI Instantaneous acceleration [km/h/s]
Tsi,j Index that evaluates the traffic flow. Being 1 for streets with [num.]
very low traffic flow and 0 for those with very high traffic
flow (Fonseca et al., 2011)
C0, C1, C2, C4 Empirical coefficients to calculate the CO2 emission [num.]
(Fonseca et al., 2011)

First from the internal/external factors influencing the CO2 emissions are derived the
scenario’s inputs (Table 1). Then A point to point SRP optimisations are considered. In
this case the set of vertex to visit is only 2, V = {A, B}.The problem aims to define the
318 M. Faccio et al.

correct series of arcs to use in order to reach a certain point B starting from a certain
point A minimising the CO2 emission.
Starting from values of the ‘scenario’s input’ a series of ‘derived inputs’ are
calculated (Table 2):

• Nsi,j, the number of estimated stop on each arc (i, j), related to the traffic flow.
According with Smit et al. (2008), congestion can be associated with a number of
interrelated and observable effects that include increased traffic density, increased
travel times on the same route (hence lower average speeds), increased deviation
from free-flow speeds (hence increased speed fluctuation and delays) and increased
queuing (hence increased number of stops).

• V’maxi,j, the maximum speed at which a vehicle can travel, also related to the traffic
flow. We assume that if the traffic flow is very high (Tsi,j ≤ 0.30) the maximum
reachable speed is equal to half of the maximum speed allowed in the considered arc.
It is in accordance with Smit et al. (2008).
From the ‘scenario’s input’ and considering the number of stop for each arc (i, j) are also
derived the kinetic values, like:

• the actual distance covered during the acceleration, Sai,j

• the acceleration and travel time, Taeffi,j and Ttoteffi,j

• the average speed and average acceleration, Svai,j and Avi,j.

Then, using the kinetic values in the expression proposed in Fonseca et al. (2011),
equation (1), it is calculated the quantity of CO2 emitted in each arc (i, j), ECO2i,j.
Table 2 Derived input

Derived input
Nsi,j Number of stop on the arc (i, j) function of Tsi,j [num.]
V’maxi,j Maximum speed at which a vehicle can travel on the arc (i, j) [km/h]
Tai,j Acceleration time on the arc (i, j) [s]
Sai,j Distance covered during the acceleration on the arc (i, j) [km]
Saeffi,j Actual distance covered during the acceleration on the arc (i, j) in [km]
order to consider the number of stop
Taeffi,j Actual acceleration time on the arc (i, j) [s]
Ttoteffi,j Actual travel time of the arc (i, j) [s]
Svai,j Average speed on the arc (i, j) [km/h]
Avi,j Average acceleration on the arc (i, j) [km/h]
ECO2i,j CO2 emission on the arc (i, j) [g]

The CO2 emission estimation is derived from the estimation of the number of stops of
each arc as function of the traffic congestion level, and from the related estimation of the
kinetic variables (Table 3).
 The sustainable routing problem 319

Table 3 CO2 emissions estimation

CO2 emission estimation

for i = 1 to n Estimation of the
for j = 1 to n number of stops of
each arc (i, j)
if Tsi,j ≥ 0.9 then Nsi,j ← (0* Kmi,j) or
if 0.90 ≤ Tsi,j ≤ 0.75 then Nsi,j ← (1 * Kmi,j) or
if 0.75 ≤ Tsi,j ≤ 0.60 then Nsi,j ← (4 * Kmi,j) or
if 0.60 ≤ Tsi,j ≤ 0.45 then Nsi,j ← (8 * Kmi,j) or
if 0.45 ≤ Tsi,j ≤ 0.30 then Nsi,j ← (15 * Kmi,j) or
if 0.30 ≤ Tsi,j ≤ 0.00 then Nsi,j ← (20 * Kmi,j)
end
end
for i = 1 to n Estimation of the
for j = 1 to n kinetic variables of
each arc (i, j)
if Tsi,j ≤ 0.30
then V’maxi,j ← (Vmaxi,j / 2) It is presumed that if
else V’maxi,j ← (Vmaxi,j) the traffic flow is
very high (Tsi,j ≤
end if 0.30) the maximum
Tai,j ← (V’maxi,j/ AI); reachable speed is
equal to half of the
Taeffi,j ← (Tai,j* Nsi,j); maximum speed
Saeffi,j ← ((V’maxi,j * Tai,j) / (3600 * 2)); allowed in the
Ttoteffi,j ← ((Taeffi,j + ((Kmi,j – Saeffi,j) * 3600) / V’maxi,j); considered arc

Svai,j ← (Kmi,j *3600 / Ttoteffi,j);

Avi,j ← (AI * Taeffi,j / Ttoteffi,j);
end
end
for i = 1 to n ECO2i,j estimation
for j = 1 to n Based on the
ECO2i,j ← ((C0 + (C1 / Svai,j) + (C2 * Avi,j) + (C4 / Tsi,j)) * Kmi,j); equation proposed in
Fonseca et al. (2011)
end
end

After a first route is generated, it starts from the point A, represented by the vertex 1, and
stops on the point B, represented by the vertex N. The vertexes considered are random
generated, and an arc (i, j) is associated to the route only if Xi,j is equal to 1. Every time
that a vertex is assigned to the route the indicators of CO2 emitted, distance and time
travelled are updated. At this point, the indicators of the best route are initialised with the
first route’s data. Then the other routes, temporary routes, are generated in the same way
of the first one. The indicators of each temporary route are compared with the first route’s
indicators, and if they result better than the best route and its indicators are redefined
(Table 4) as defined in the sustainable routing model reported in Table 5.
320 M. Faccio et al.

The model generates the routes, each one starts from the vertex 1 and stops only when
all the vertexes in V are visited. It is important to highlight that is not defined the order in
which the vertexes must be visited.
Table 4 Routes indicators

Route’s indicators
count Counter [num.]
p Counter [num.]
RouteN Number of route generated for the comparison [num.]
OK1 Check, if the vertex S1 is visited OK1 is equal to 1otherwise is [num.]
equal to zero
OK2 Check, if the vertex S2 is visited OK1 is equal to 1otherwise is [num.]
equal to zero
OK3 Check, if the vertex S3 is visited OK1 is equal to 1otherwise is [num.]
equal to zero
firsteroute First route [vector]
ECO2f Indicator of CO2 emission of the first route [g]
DISTANCEf Indicator of the distance travelled in the first route [km]
TIMEf Indicator of time travelled in the first route [s]
routeminco2 Route with min CO2 emission [vector]
routemindistance Route with min distance travelled [vector]
routemintime Route with min time travelled [vector]
minco2 Indicator of CO2 emission of the route routeminco2 [g]
mindistance Indicator of distance travelled in the route routemindistance [km]
mintime Indicator of travelled time in the route routemintime [s]
temproute Temporary route [vector]
ECO2temp Indicator of CO2 emission of the temporary route [g]
DISTANCEtemp Indicator of distance travelled in the temporary route [km]
TIMEtemp Indicator of travelled time in the temporary route [s]

Table 5 Routing model

First route generation

i←1; j←1; p ←1; firsteroute(p) ← 1; Variables and indicators
OK1←0; OK2←0; OK3←0; initialisation.
ECO2f← 0; DISTANCEf← 0; TIMEf ← 0;
while OK1 = 0 or OK2 = 0 or OK3 = 0 The route starts from the vertex 1
p ← (p + 1); and the next vertex, j, is random
generated.
while Xi,j = 0 In order to generate the route it is
j ← random number between 1 and n; considered only the arc whit the
index Xi,j different to zero.
end
firsteroute(p) ← j; The route generation stops when
ECO2f ← (ECO2f+ ECO2i,j); all the vertexes S1, S2, S3 are
visited.
DISTANCEf← (DISTANCEf + Kmi,j);
TIMEf← (TIMEf + Ttoteffi,j);
 The sustainable routing problem 321

Table 5 Routing model (continued)

First route generation

i← j;
if i = S1then OK1←1 or
if i = S2 then OK2←1 or
if i = S3 then OK3←1 or
end
routeminco2 ← firsteroute; minco2 ← ECO2f; Initialisation of the indicators for
routemindistance ← firsteroute; mindistance ← DISTANCEf; the comparison of the routes
routemintime ← firsteroute; mintime← TIMEf;
Temporary route generation and comparison
count = 1; Generation of RouteN temporary
while count < RouteN routes to compare with the first.
i←1; j←1; p ←1; temproute(p) ← 1; Variables and indicators
OK1←0; OK2←0; OK3←0; initialisation.
ECO2temp← 0; DISTANCEtemp← 0; TIMEtemp← 0;
while OK1 = 0 or OK2 = 0 or OK3 = 0 Each route is created with the
p ← (p+1); same method used for the first
route
while Xi,j = 0
j← random number between 1 and n;
end
temproute(p) ← j;
ECO2temp← (ECO2temp+ ECO2i,j);
DISTANCEtemp← (DISTANCEtemp + Kmi,j);
TIMEtemp← TIMEtemp+ Ttoteffi,j);
i← j;
if i = S1then OK1←1 or
if i = S2 then OK2←1 or
if i = S3 then OK3←1 or
end
if ECO2temp < minco2 For each temporary route is
routeminco2 ← temproute; made a comparison between its
CO2 emission indicator and the
minco2 ← ECO2temp; indicator of the route with min
end CO2 emission. If the indicator of
if DISTANCEtemp < mindistance the temporary route results lower
than the other the route with min
routemindistance ← temproute;
CO2 emission and its indicator
mindistance ← DISTANCEtemp; are redefined.
end The same comparison is made
if TIMEtemp < mintime for defining the route with
minimum distance and minimum
routemintime ← temproute; travelled time.
Mintime ← TIMEtemp;
end
end
322 M. Faccio et al.

In order to compare the solution of the proposed method with the solution of the classical
method, which minimised the minimum time and distance travelled, the model generates
the route with minimum CO2 emission, the route with minimum distance travelled and
the route with minimum travelled time, and their respective indicators.

4 Case study

In this section the solution model proposed for the SRP is applied in the optimisation of
municipal distribution of fresh food product in the centre of Vicenza, a city in the north
east of Italy.
It has been assumed that the vehicle used is the same used in Fonseca et al. (2011)
consequently it is possible to keep equal the empirical coefficients. The values of the
coefficients C0, C1, C2 and C4 are respectively: 221.4, 1382.4, 46.9, 18.1.
In this applicative case is considered a graph with 20 vertexes. The representation of
the graph G is showed in Figure 2 where are indicated each one of the 20 nodes of the set
N, and where the blue links between two nodes represent the possible usable arcs.
The routing problem is related to the distribution of products to 4 customers that lie in
point 10, 15, 19 and 20 starting from point 1.

Figure 2 Map of the centre of Vicenza (see online version for colours)
Case study graph

Considering this graph the scenario’s inputs are indicated in Table 6, Table 7, Table 8
and Table 9. More in detail, Table 6 represents Xi,j the Binary index equal to 1 if a
vehicles can travel on arc (i, j), Table 7 shows kmi,j, the distance from each couple of
vertexes of the graph G and Table 8 and Table 9 indicate respectively the maximum
speed at which a vehicle can travel on the arc (i, j), Vmaxi,j, and the expected average
level of traffic at each arc, Tsi,j. These last data can be variable in function of the day time
and it is assumed constant in first part of the analysis.
 The sustainable routing problem 323

Table 6 Scenario’s input: Xi,j

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 1 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
3 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0
6 0 0 1 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0
7 0 1 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0
8 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
9 0 0 0 0 1 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0
10 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0
11 0 0 0 0 0 0 1 0 0 1 0 1 0 1 0 0 0 0 0 0
12 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 1 0 0 0 0
13 0 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0 0 0 0 0
14 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 1
15 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 1 1 0
16 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 1 0 0 0
17 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 1 0 0
18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0
19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 1
20 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0

Another important scenario’s input is the instant acceleration that depends on the driving
style. Typically rapid acceleration results in significant increase in the emission (Larsson
and Ericsson, 2009). In order to analyse this influence are considered three types of
driving behaviours: calm driving, normal driving and aggressive driving. According to
De Vlieger et al. (2000) and Larsson and Ericsson (2009) these driving styles can be
associated respectively to an instant acceleration of 2.4, 3.9 and 5.4 km/h/s. These values
have been also used in the paper.
The results of the case study are reported as follows:
• in order to make the application of the proposed routing model easier to understand,
it is first analysed the point to point routing problem where V = {1, 20} and the
problem solution aims to define the correct series of arcs to use in order to connect
the point 1 to the point 20
• to make the proposed sustainable routing model benefits clearer, the results are
compared with the classical routing objectives (distance minimisation; time
minimisation)
• the problem is solved considering different driving styles: calm driving, normal
driving and aggressive driving, changing the instant acceleration
• finally the same analysis has been developed for the routing problem related to the
distribution of fresh food products to 4 customers that lie in point 10, 15, 19 and 20
starting from point 1.
 324

Table 7

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1 0 0.7 0 0.4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 0.7 0 0.45 0 0 0 0.8 0 0 0 0 0 0 0 0 0 0 0 0 0
3 0 0.45 0 0 0 0.8 0 0 0 0 0 0 0 0 0 0 0 0 0 0
M. Faccio et al.

4 0.4 0 0.25 0 1.1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

5 0 0 0 1.1 0 0 0 0 0.15 0 0 0 0 0 0 0 0 0 0 0
Scenario’s input: kmi,j

6 0 0 0.8 0 0 0 0.25 0.5 0 0 0 0 0 0 0 0 0 0 0 0

7 0 0.8 0 0 0 0.25 0 0 0 0 0.7 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0.3 0.5 0 0 0 0 0 0 0 0 0 0 0 0 0 0
9 0 0 0 0 0.15 0 0 0.35 0 0.15 0 0 0 0 0 0 0 0 0 0
10 0 0 0 0 0 0 0 0 0.15 0 1.2 0 0 0 0 0 0.55 0 0 0
11 0 0 0 0 0 0 0.7 0 0 1.2 0 0.65 0 1.5 0 0 0 0 0 0
12 0 0 0 0 0 0 0 0 0 0 0.65 0 0.7 0 0 0.5 0 0 0 0
13 0 0 0 0 0 0 0 0 0 0 0 0.7 0 0.6 0.3 0 0 0 0 0
14 0 0 0 0 0 0 0 0 0 0 1.5 0 0.6 0 0 0 0 0 0 1.4
15 0 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0 0.9 0 0.65 1.2 0
16 0 0 0 0 0 0 0 0 0 0 0 0.5 0 0 0.9 0 0.65 0 0 0
17 0 0 0 0 0 0 0 0 0 0.55 0 0 0 0 0 0.65 0 1.6 0 0
18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.65 0 1.6 0 0.55 0
19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1.2 0 0 0.55 0 0.8
20 0 0 0 0 0 0 0 0 0 0 0 0 0 1.4 0 0 0 0 0.9 0
 Table 8

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1 0 50 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 50 0 50 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0
3 0 50 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0
4 50 0 50 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
5 0 0 0 50 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0
6 0 0 50 0 0 0 50 50 0 0 0 0 0 0 0 0 0 0 0 0
Scenario’s input: Vmaxi,j

7 0 50 0 0 0 50 0 0 0 0 70 0 0 0 0 0 0 0 0 0
8 0 0 0 0 50 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0
The sustainable routing problem

9 0 0 0 0 50 0 0 50 0 50 0 0 0 0 0 0 0 0 0 0
10 0 0 0 0 0 0 0 0 50 0 70 0 0 0 0 0 50 0 0 0
11 0 0 0 0 0 0 70 0 0 70 0 50 0 70 0 0 0 0 0 0
12 0 0 0 0 0 0 0 0 0 0 50 0 50 0 0 50 0 0 0 0
13 0 0 0 0 0 0 0 0 0 0 0 50 0 50 50 0 0 0 0 0
14 0 0 0 0 0 0 0 0 0 0 70 0 50 0 0 0 0 0 0 70
15 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 50 0 50 50 0
16 0 0 0 0 0 0 0 0 0 0 0 50 0 0 50 0 50 0 0 0
17 0 0 0 0 0 0 0 0 0 50 0 0 0 0 0 50 0 50 0 0
18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 50 0 50 0
19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 50 0 50
20 0 0 0 0 0 0 0 0 0 0 0 0 0 70 0 0 0 0 50 0
325
 326

Table 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1 0 0.6 0 0.45 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
2 0.6 0 0.3 0 0 0 0.6 0 0 0 0 0 0 0 0 0 0 0 0 0
3 0 0.3 0 0 0 0.3 0 0 0 0 0 0 0 0 0 0 0 0 0 0
M. Faccio et al.

4 0.45 0 0.3 0 0.45 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

5 0 0 0 0.45 0 0 0 0 0.3 0 0 0 0 0 0 0 0 0 0 0
Scenario’s input: Tsi,j

6 0 0 0.3 0 0 0 0.3 0.3 0 0 0 0 0 0 0 0 0 0 0 0

7 0 0.6 0 0 0 0.3 0 0 0 0 0.6 0 0 0 0 0 0 0 0 0
8 0 0 0 0 0.3 0.3 0 0 0 0 0 0 0 0 0 0 0 0 0 0
9 0 0 0 0 0.45 0 0 0.3 0 0.45 0 0 0 0 0 0 0 0 0 0
10 0 0 0 0 0 0 0 0 0.45 0 0.6 0 0 0 0 0 0.45 0 0 0
11 0 0 0 0 0 0 0.6 0 0 0.6 0 0.3 0 0.6 0 0 0 0 0 0
12 0 0 0 0 0 0 0 0 0 0 0.3 0 0.3 0 0 0.3 0 0 0 0
13 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0.3 0.3 0 0 0 0 0
14 0 0 0 0 0 0 0 0 0 0 0.6 0 0.3 0 0 0 0 0 0 0.6
15 0 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0 0.3 0 0.3 0.3 0
16 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0 0.3 0 0.3 0 0 0
17 0 0 0 0 0 0 0 0 0 0.45 0 0 0 0 0 0.3 0 0.45 0 0
18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0.45 0 0.45 0
19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0 0.45 0 0.45
20 0 0 0 0 0 0 0 0 0 0 0 0 0 0.45 0 0 0 0 0.45 0
 Table 10

Driving style Criteria CO2 (kg) Distance (km) Time (sec) CO2 (g) increment Distance (km) increment Time (s) increment ROUTE
Aggressive Min CO2 2,393.9 5.1 327.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min km 2,393.9 5.1 327.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min time 2,393.9 5.1 327.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Normal Min CO2 2,385.2 5.1 330.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
The sustainable routing problem

Min km 2,385.2 5.1 330.9 0.0% 0.0% 0.0% 1 2 7 11 14 20

Min time 2,385.2 5.1 330.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Calm Min CO2 2,375.2 5.1 334.5 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min km 2,375.2 5.1 334.5 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min time 2,375.2 5.1 334.5 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min CO2 0.0% 0.0% 0.0%
Min km 0.0% 0.0% 0.0%
Min time 0.0% 0.0% 0.0%
Point to point routing solution for values of traffic flow as reported in Table 9
327
 328

Table 11

Driving Distance CO2 (g) Distance (km) Time (s)

M. Faccio et al.

Criteria CO2 (kg) Time (sec) ROUTE

style (km) increment increment increment
Aggressive Min CO2 3,176.7 5.6 570.0 0.0% 8.8% 57.1% 1 2 7 11 12 13 14 20
Min km 3,215.6 5.1 362.7 1.2% 0.0% 0.0% 1 2 7 11 14 20
Min time 3,215.6 5.1 362.7 1.2% 0.0% 0.0% 1 2 7 11 14 20
Normal Min CO2 3,129.6 5.8 1138.9 0.0% 12.7% 208.9% 1 4 3 6 7 11 12 13 14 20
Min km 3,186.9 5.1 368.7 1.8% 0.0% 0.0% 1 2 7 11 14 20
Min time 3,186.9 5.1 368.7 1.8% 0.0% 0.0% 1 2 7 11 14 20
Calm Min CO2 3,154.2 5.1 375.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min km 3,154.2 5.1 375.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min time 3,154.2 5.1 375.9 0.0% 0.0% 0.0% 1 2 7 11 14 20
Min CO2 0.0% 7.2% 88.7%
Min km 1.0% 0.0% 0.0%
Min time 1.0% 0.0% 0.0%
point to point routing solution for values of traffic flow in Table 9 increased of 15%
 The sustainable routing problem 329

4.1 Point to points routing optimisation

The results for the point to point sustainable routing optimisation are showed in Table 10,
where are indicated the total quantity of CO2 emitted, the total distance and time
travelled, for each driving behaviour considered and for each criteria used (CO2
minimisation, distance minimisation and time minimisation). Moreover, for each driving
style is indicated the solution route given by the different criteria.
From the results reported in Table 10 it is possible to derive some important points:

• the CO2 emission increases when the driving style is more aggressive

• for the values of traffic flows as in Table 9 the optimal sustainable point to point
route (min CO2) does not change when the driving style changes

• in this condition the optimal sustainable point to point route is equal to the optimal
route obtained with the criteria of distance minimisation and time minimisation.

This low relevance of the results is consequence of the simplicity of the problem (point to
point routing).
On the other hand it is very interesting to see the influence of the traffic flow on the
results.
Table 11 reports the results considering a uniform increase of traffic flow of 15%.
When the traffic flow increases (Tsi,j decreases), the sustainable routing solution
given by the proposed model can be different from the routing solution obtained by the
distance or time minimisation.
This point is very interesting because it highlights how sustainable routing does not
means distance minimisation. From a routing point of view, as demonstrated from case
study even in the simple point to point routing optimisation, the distance minimisation
approach can be unable to really minimise the CO2 emissions that are influenced by other
important factors as showed also in Figure 1. For this reason some logistic providers are
nowadays developing important projects in managing these factors in order to achieve
results in emissions reduction, planning for example ‘Innight distribution’ when the
traffic flow indexes (Tsi,j) are greater. But this approach can be often inapplicable in
order to match customer requirements and to satisfy different kinds of constrains.
The table reports the point to point solutions for the increasing of the traffic flow of
15%, i.e., a decreasing of Tsi,j of 15% (Table 11).
Other results not reported in the paper demonstrate how at the increasing of the traffic
flow (also more than 15%) the trend in changing the optimal sustainable route versus the
other solution (time or distance minimisation) is maintained and the benefits in CO2
emission reduction typically increase, especially in case of aggressive driving style. On
the other hand according to other results, not reported in the paper, that consider a
decreasing of traffic flow (Tsi,j increases) the routing solution of the different
optimisation criteria is the same as Table 10.
Finally if traffic flow congestion indexes Tsi,j are casually generated, the optimal
sustainable route versus the other solution (time or distance minimisation) is completely
different, and the possible CO2 saving is typically very high, demonstrating how, in case
of variable traffic flow congestion it is important to:
330 M. Faccio et al.

• Obtain instant traffic flow data using traffic sensors across the streets or, otherwise,
to have familiarity with the territory in order to know the alternative possible way.

• Develop sustainable routing solutions according to specific routing models as the

proposed one.

• The point to point routing problem solution, as reported from the case study results,
demonstrates that even for simple point to point routing problem.

• The CO2 emission increases when the driving style is more aggressive and when the
traffic flow increases.

• The optimum sustainable routing, even for point to point problem, changes in
function of the traffic flow and of the driving style.

• When the driving style is calm, the routing solution is maintained the same, as
Table 9, while changing when the driving style is normal or aggressive.

• The optimal sustainable routing solution can be different from the routing solution
given by the classical routing objectives, even for simple routing problems (point to
point), especially in function of traffic flow variation. For this reason it is very
important to be able to have real time traffic flow data in order to better optimise the
sustainable routes.

4.2 N points routing optimisation

The results for the N point sustainable routing optimisation are showed in Table 12,
where are indicated for each driving behaviour considered the total quantity of CO2
emitted, the total distance and time travelled, and the selected route. As reported in
Table 12, in case of traffic condition defined in Table 9, the optimal sustainable route is
the same for different condition of driving style and it is equal to the solution obtained
using as objective function the distance minimisation or the time minimisation. These
results confirm the point to point result in case of traffic condition as Table 9.
On the other hand if an increasing of traffic condition is considered, the optimal
solution of the three methods are different as reported in Table 13, in which is considered
a traffic flow increasing of 15%.
The results confirm the point to point sustainable routing results in which in case of
decrement of traffic flow the results obtained with the three criteria (Min CO2, Min
Distance, Min Time) are the same, while the results are different in case of increment of
traffic flow. If the increment of traffic flow is high (i.e., 30%) the CO2 emission reduction
using the proposed sustainable routing criteria instead of the distance-time minimisation
criteria is very high. Normally this reduction is higher than the increment of time and
distance given due to the use of the sustainable criteria instead of the other two. Table 14
reports the summary of the results considering the effect of an increment/decrement of
Tsi,j respectively of 15 and 30% compared to the values of the case zero.
 Table 12

Driving Distance CO2 (g) Distance (km) Time (s)

Criteria CO2 (kg) Time (sec) ROUTE
style (km) increment increment increment
Aggressive Min CO2 3,589.9 5.9 514.6 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
Min km 3,589.9 5.9 514.6 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
N points routing solution

Min time 3,589.9 5.9 514.6 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20

Normal Min CO2 3,564.9 5.9 522.3 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
The sustainable routing problem

Min km 3,564.9 5.9 522.3 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20

Min time 3,564.9 5.9 522.3 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
Calm Min CO2 3,536.3 5.9 531.6 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
Min km 3,536.3 5.9 531.6 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
Min time 3,536.3 5.9 531.6 0.0% 0.0% 0.0% 1 4 5 9 10 17 16 15 19 20
Min CO2 0.0% 0.0% 0.0%
Min km 0.0% 0.0% 0.0%
Min time 0.0% 0.0% 0.0%
331
 332

Table 13

Driving CO2 Distance CO2 (g) Distance (km) Time (s)

M. Faccio et al.

Criteria Time (sec) ROUTE

style (kg) (km) increment increment increment
Aggressive Min CO2 3,515.2 5.9 732.0 0.0% 0.0% 4.6% 1 4 5 9 10 17 16 15 19 20
Min km 3,515.2 5.9 732.0 0.0% 0.0% 4.6% 1 4 5 9 10 17 16 15 19 20
Min time 5,728.8 8.2 699.9 63.0% 39.0% 0.0% 1 4 5 9 10 11 14 20 14 13 15
Normal Min CO2 3,332.4 5.9 1,137.2 0.0% 0.0% 32.8% 1 4 5 9 10 17 16 15 19 20
Min km 3,332.4 5.9 1,137.2 0.0% 0.0% 32.8% 1 4 5 9 10 17 16 15 19 20
Min time 5,605.6 8.2 856.3 68.2% 39.0% 0.0% 1 4 5 9 10 11 14 20 14 13 15
Calm Min CO2 3,541.8 5.9 1,969.1 0.0% 0.0% 68.4% 1 4 5 9 10 17 16 15 19 20
Min km 3,541.8 5.9 1,969.1 0.0% 0.0% 68.4% 1 4 5 9 10 17 16 15 19 20
Min time 5,617.2 8.2 1,168.9 58.6% 39.0% 0.0% 1 4 5 9 10 11 14 20 14 13 15
Min CO2 0.0% 0.0% 35.3%
Min km 0.0% 0.0% 35.3%
Min time 63.3% 39.0% 0.0%
N points routing solution for values of traffic flow in Table 9 increased of 15%
 The sustainable routing problem 333

Table 14 N points summary results

Ts Criteria CO2 (g) increment Distance (km) increment Time (s) increment
+30% Min CO2 0.0% 0.0% 0.0%
Min Km 0.0% 0.0% 0.0%
Min Time 0.0% 0.0% 0.0%
+15% Min CO2 0.0% 0.0% 0.0%
Min Km 0.0% 0.0% 0.0%
Min Time 0.0% 0.0% 0.0%
0% Min CO2 0.0% 0.0% 0.0%
Min Km 0.0% 0.0% 0.0%
Min Time 0.0% 0.0% 0.0%
–15% Min CO2 0.0% 0.0% 35.3%
Min Km 0.0% 0.0% 35.3%
Min Time 63.3% 39.0% 0.0%
–30% Min CO2 0.0% 5.6% 133.7%
Min Km 1.6% 0.0% 159.3%
Min Time 67.5% 103.4% 0.0%

The N points routing problem solution, as reported from the case study results,
demonstrates the general results of the point to point solution, and moreover:
• The CO2 emission increases when the driving style is more aggressive, the traffic
flow grows, and in case of N points routing solution. It is possible to have an
increment of the CO2 emission saving when the number of points to reach increases
(i.e., results comparison -15% Tsi,j point to point/N points).
• The optimum sustainable routing changes in function of the traffic flow and of the
driving style, especially in case of complex problem with high number of points to
cover.
• The optimal sustainable routing solution can be different from the routing solution
given by the classical routing objectives (distance or time minimisation). On the
other hand the traffic flow is a very influencing factor in the sustainable routing
model and the acquisition of real time traffic flow data could make the proposed
model very suitable for urban routing problem.

5 Conclusions and further research

The transportation sector produces the largest percentage of emissions from fossil fuel
combustion by end use sector. On the other hand environmental, social, and political
pressures to limit the impacts associated with CO2 emissions are mounting rapidly. This
paper addresses to the solution of a new class of routing problem: the SRP, where the
objective differently from the classical approaches is the CO2 emission minimisation.
A set of different factors influence the CO2 emission. They can be divided, from the
driver point of view, into internal/external factors:
334 M. Faccio et al.

• Internal factors, which depend on the driver like driving style and the route
knowledge. These factors influence the instantaneous acceleration (i.e., high/low
acceleration at each start and stop), and consequently the route average speed.

• External factors, i.e., not depending on the driver, like streets/vehicle characteristics,
or traffic congestion level.

The basic idea of the proposed research is that using the Fonseca formulation for the CO2
estimation, it is possible to calculate the CO2 increase/decrease at the variance of the
internal/external factors. In this way it will be possible to optimise the routing according
the CO2 minimisation.
Firstly the paper, starting from the of Fonseca et al. (2011) study, proposes a CO2
emission estimation as function of the different internal/external factors. Secondly, the
authors formalise a sustainable routing model, validating it through a case study and
through a parametrical analysis comparing it versus the classical routing criteria of
distance or time minimisation.
The main obtained results are:

• Sustainable routing does not mean distance minimisation (or distance minimisation
does not mean CO2 minimisation). As demonstrated from case study even in the
simple point to point routing optimisation, the distance minimisation approach can
be unable to really minimise the CO2 emissions that are influenced by other
important factors (Figure 1).

• The CO2 emission increases when the driving style is more aggressive and when the
traffic flow increases and the optimum sustainable routing changes in function of the
traffic flow and of the driving style. These results are valid even for simple routing
problem solution (i.e., point to point) and become more evident with complex
routing problem (N points).

• The optimal sustainable routing solution can be different from the routing solution
given by the classical routing objectives, even for simple routing problems (point to
point), especially in function of traffic flow variation. In case of low traffic flow the
sustainable routing solution is in accordance with the solutions obtained with the
other methodologies, while it is very different when the traffic flow is high. In this
situation the possible CO2 saving obtainable with the proposed model is typically
very high.

• For this reason it is very important to be able to have real time traffic flow data,
using for example traffic sensors along the streets, in order to better optimise the
sustainable routes. The proposed sustainable routing model is dynamical, i.e., at the
change of the internal/external factors (i.e., traffic flow, driving style, etc) it will be
possible to re-define the optimum sustainable routing.

• In case it is not possible to have instant information on the traffic congestion, the
driver familiarity with the territory results in an important element to support this
new approach, giving to the driver the knowledge of the street and the level of traffic
congestion on different day time, that allow him/her to choose carefully a different
route.
 The sustainable routing problem 335

Two possible further researches will be developed. Actually, the authors are developing a
case study in which the proposed sustainable routing model will be integrated with a real
time traffic level data acquisition system. In this way it will be possible to analyse, as
function of the instant traffic level change, the possible benefits of the use of the
proposed approach in the CO2 emissions reduction versus the classical other routing
criteria and to propose to companies and generally speaking to all the interested subject a
sustainable routing prototype to implement in their business. The second research will
integrate the sustainable routing model with the fixed routing model with driver learning,
proposed in Battini et al. (2011), in witch is showed the important effect of the driver
familiarity with the territory and the customers. This familiarity allows, not only to
reduce the service time at each customers, by the drivers learning, but results also an
important element to support the new approach proposed in this paper in those cases
where it is not possible to have instants information on the traffic congestion, and the
driver must change the routes-based only on his/her knowledge of the streets

References
Anbuudayasankar, S.P., Ganesh, K., Lee, T-R. and Mohandas, K. (2009) ‘COG: a composite
genetic algorithm with local search methods to solve a mixed vehicle routing problem with
backhauls’, Int. J. of Services and Operations Management, Vol. 5, No. 5, pp.617–636.
André, M. and Rapone, M. (2009) ‘Analysis and modeling of the pollutant emissions from
European cars regarding the driving characteristics and test cycles’, Atmospheric Environment,
Vol. 43, No. 5, pp.986–995.
Battini, D., Faccio, M., Persona, A. and Zanin, G. (2011) ‘Routing strategy in a distribution
network: fixed route with driver learning versus variable daily optimized route’,
PROCEEDING XVI Summer School ‘Francesco Turco’ Impianti Industriali Meccanici Abano
Terme, 14–16 September, Padova, Italy.
Bektaş, T. and Laporte, G. (2011) ‘The pollution-routing problem’, Transportation Research Part
B, Vol. 45, No. 8, pp.1232–1250.
Benton, W.C. and Rossetti, M.D. (1992) ‘The vehicle scheduling problem with intermittent
customer demands’, Computers & OR, Vol. 19, No. 6, pp.521–531.
Cordeau, J-F., Laporte, G., Savelsbergh, M.W.P. and Vigo, D. (2007) ‘Vehicle routing’, in
Barnhart, C. and Laporte, G. (Eds.): Transportation, Handbooks in Operations Research and
Management Science, Chapter 6, Vol. 14, pp.367–428, Elsevier, Amsterdam, The
Netherlands.
De Vlieger, I., De Keukeleere, D. and Kretzschmar, J.G. (2000) ‘Environmental effects of driving
behaviour and congestion related to passenger cars’, Atmospheric Environment, Vol. 34,
No. 27, pp.4649–4655.
Dessouky, M., Rahimi, M. and Weidner, M. (2003) ‘Jointly optimizing cost, service, and
environmental performance in demand-responsive transit scheduling’, Transportation
Research: Part D, Vol. 8, No. 6, pp.433–465.
Eksioglu, B., Vural, A.V. and Reisman, A. (2009) ‘The vehicle routing problem: a taxonomic
review’, Computers & Industrial Engineering, Vol. 57, No. 4, pp.1472–1483.
European Parliament (2009) ‘Setting emission performance standards for new passenger cars as
part of the community’s integrated approach to reduce co2 emissions from light-duty vehicles.
Regulation (EC) No. 443/2009 of the European Parliament and of the Council of 23 April
2009’, Official Journal of the European Union, 5, June, p.L140.
Faccio, M., Ferrari, E., Persona, A. and Vecchiato, P. (2012) ‘Lean distribution principles to food
logistics: a products category approach’, International Journal of Operational Research,
Vol. 16, No. 2, p.2013.
336 M. Faccio et al.

Faccio, M., Persona, A., Sgarbossa, F. and Zanin, G. (2011) ‘Multi-stage supply network design in
case of reverse flows: a closed-loop approach’, International Journal of Operational
Research, Vol. 12, No. 2, pp.157–191.
Figliozzi, M.A. (2010) ‘Vehicle routing problem for emissions minimization’, Transportation
Research Record, Vol. 2197, No. 1, pp.1–7.
Figliozzi, M.A. (2011) ‘The impacts of congestion on time-definitive urban freight distribution
networks CO2 emission levels: results from a case study in Portland, Oregon’, Transportation
Research Part C, Vol. 19, No. 5, pp.766–778.
Fonseca, N., Casanova, J. and Valdés, M. (2011) ‘Influence of the stop/start system on CO2
emissions of a diesel vehicle in urban traffic’, Transportation Research Part D, Vol. 16,
No. 2, pp.194–200.
Fontaras, G., Kouridis, H., Samaras, Z., Elst, D. and Gense, R. (2007) ‘Use of a vehicle-modelling
tool for predicting CO2 emissions in the framework of European regulations for light goods
vehicles’, Atmospheric Environment, Vol. 41, No. 14, pp.3009–3021.
Kim, J.H., Youn, S. and Roh, J.J. (2011) ‘Green supply chain management orientation and firm
performance: evidence from South Korea’, Int. J. of Services and Operations Management,
Vol. 8, No. 3, pp.283–304.
Laporte, G. and Semet, F. (2002) ‘Classical heuristics for the capacitated vehicle routing problem,
in Toth, P. and Vigo, D. (Eds.): The Vehicle Routing Problem, Monographs on Discrete
Mathematics and Applications, pp.109–128, SIAM, Philadelphia.
Larsson, H. and Ericsson, E. (2009) ‘The effects of an acceleration tool in vehicles for reduced fuel
consumption and emissions’, Transportation Research Part D, Vol. 14, No. 2, pp.141–146.
Maden, W., Eglese, R.W. and Black, D. (2010) ‘Vehicle routing and scheduling with time varying
data: a case study’, Journal of the Operational Research Society, Vol. 61, No. 3, pp.515–522.
Malairajan, R.A., Ganesh, K., Muhos, M. and Iskanius, P. (2012) ‘CLING: heuristic to solve
integrated resource allocation and routing problem with time window’, Int. J. of Services and
Operations Management, Vol. 13, No. 2, pp.247–266.
Sivakumar, P., Ganesh, K., Ducq, Y. and Anbuudayasankar, S.P. (2012) ‘Class of sustainable
supply chain routing problems – framework and comprehensive review’, Int. J. of Services
and Operations Management, Vol. 12, No. 2, pp.188–220.
Smit, R., Brown, A.L. and Chan, Y.C. (2008) ‘Do air pollution emissions and fuel consumption
models for roadways include the effects of congestion in the roadway traffic flow?’,
Environmental Modelling & Software, Vol. 23, Nos. 10–11, pp.1262–1270.
United States. Environmental Protection Agency (2012) Inventory of U.S. Greenhouse Gas
Emissions and Sinks: 1990–2010 [online]
http://www.epa.gov/climatechange/ghgemissions/usinventoryreport.html.
Woensel, T.V., Creten, R. and Vandaele, N. (2001) ‘Managing the environmental externalities of
traffic logistics: the issue of emissions’, Production and Operations Management, Vol. 10,
No. 2, pp.207–223.
Wygonik, E. and Goodchild, A. (2011) ‘Evaluating CO2 emissions, cost, and service quality
trade-offs in an urban delivery system case study’, IATSS Research, Vol. 35, No. 1, pp.7–15.