ArticlePDF Available

Abstract and Figures

Efforts have been made in the last few decades to provide new urban transport alternatives. One of these is carsharing, which involves a fleet of vehicles scattered around a city for the use of a group of members. At first, part of the research effort was put into setting up real life experiments with vehicle fleets and observing the performance of major private operators. In the meantime, with the growth of this alternative and the need to better plan its deployment, researchers started to create more advanced methods to study carsharing systems’ planning issues. In this paper, we review those methods, identifying gaps and suggesting how to bridge them in the future. Based on that review we concluded that carsharing demand is difficult to model due to the fact that the availability of vehicles is intrinsically dependent on the number of trips and vice versa. Moreover, despite the existence of carsharing simulation models that offer very detailed mobility representations, no model is able to characterise accurately the supply side, thus hindering the cost-benefit assessment that is fundamental to justify investment in this transport alternative, in particular those that are being endorsed by the European Union. More complex, however, is the operation of the emerging one-way carsharing systems, where a vehicle may be dropped off at any station, which adds uncertainty as to the location where vehicles can be picked up. Several optimisation approaches have been proposed to mitigate this problem but they are always limited in scope and leave other aspects out for model simplification purposes. Some simulation models have also been developed to study the performance of this type of carsharing system, but they have not included ways of balancing the vehicle stocks.
Content may be subject to copyright.
EJTIR
Issue 13(3), 2013
pp. 201-220
ISSN: 1567-7141
www.ejtir.tbm.tudelft.nl
Carsharing systems demand estimation and
defined operations: a literature review
Diana Jorge
1
, Gonçalo Correia
2
Department of Civil Engineering, University of Coimbra
Efforts have been made in the last few decades to provide new urban transport alternatives.
One of these is carsharing, which involves a fleet of vehicles scattered around a city for the use of
a group of members. At first, part of the research effort was put into setting up real life
experiments with vehicle fleets and observing the performance of major private operators. In the
meantime, with the growth of this alternative and the need to better plan its deployment,
researchers started to create more advanced methods to study carsharing systems’ planning
issues. In this paper, we review those methods, identifying gaps and suggesting how to bridge
them in the future. Based on that review we concluded that carsharing demand is difficult to
model due to the fact that the availability of vehicles is intrinsically dependent on the number of
trips and vice versa. Moreover, despite the existence of carsharing simulation models that offer
very detailed mobility representations, no model is able to characterise accurately the supply
side, thus hindering the cost-benefit assessment that is fundamental to justify investment in this
transport alternative, in particular those that are being endorsed by the European Union. More
complex, however, is the operation of the emerging one-way carsharing systems, where a vehicle
may be dropped off at any station, which adds uncertainty as to the location where vehicles can
be picked up. Several optimisation approaches have been proposed to mitigate this problem but
they are always limited in scope and leave other aspects out for model simplification purposes.
Some simulation models have also been developed to study the performance of this type of
carsharing system, but they have not included ways of balancing the vehicle stocks.
Keywords: one-way carsharing, demand modelling, mathematical modelling, operations research,
literature review.
1. Introduction
In recent decades there have been some changes in how the use of urban transport is viewed. At
first, the increasing use of private transport in industrialised countries provided greater
accessibility. In the long-term, however, it has resulted in serious negative externalities, such as
pollution, and excessive consumption of energy and time due to congestion problems. This has
1
Rua Luís Reis Santos, 3030-788 Coimbra, Portugal, T: +351 239 797 131, F: +351 239 797 123, E:
diana.jorge@student.dec.uc.pt
2
Rua Luís Reis Santos, 3030-788 Coimbra, Portugal, T: +351 239 797 100, F: +351 239 797 123, E:
gcorreia@dec.uc.pt
EJTIR 13(3), 2013, pp.201-220 202
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
happened mainly in urban areas where demand is concentrated in peak hours (Texas Transport
Institute, 2010). Moreover, land prices and vehicle ownership costs such as fuel, parking and the
cost of purchasing and insuring the vehicle itself are increasing. ‘These last costs are sunk costs
even before a mile is driven’ (Mitchell et al., 2010), which means that they are unrecoverable,
even if the vehicle is no longer used. In addition each vehicle use is very low. In America, for
example, automobiles spend around 90% of their time parked (U.S. Department of
Transportation, 2001). Public transport could be a good alternative, but it has several
shortcomings. For instance, coverage does not allow a door-to-door service even in those
European cities that have an outstanding public transport network. Moreover, schedules are not
flexible and services lack personalisation. Providing public transport for the peak hour demand
also means that vehicles are idle for the rest of the day, which decreases its efficiency.
Since all these issues matter to society they have been handed over to policymakers, who should
act in the public interest. It is now generally agreed that strategies are needed that will minimise
these impacts while simultaneously allowing people to participate in the same activities as
before. One strategy that has been indicated to manage demand is providing a transportation
alternative in the form of carsharing which is a system that is somewhere between private and
public transport. The classical definition of carsharing states that it is a system that involves a
small to medium fleet of vehicles, available at several stations, to be used by a relatively large
group of members (Shaheen et al., 1999).
1.1 Round-trip versus one-way carsharing
Traditional carsharing system operators require users to return cars to the station where they
were picked up. These are round-trip carsharing systems, which simplify the task of the
operators because they can plan stocks based on the demand for each station. But it is less
convenient for the users. Better suited to personal needs are one-way carsharing systems. In one-
way carsharing, users can pick up a car from one station and leave it at a different one. If they
need a vehicle later on, they can pick up another one. Therefore, in theory, one-way carsharing
systems allow more trips to be captured than the alternative round-trip system, which can only
be used by a small market share for leisure, shopping and sporadic trips (Barth and Shaheen,
2002). Firnkorn and Müller (2011) concluded exactly this through a survey that shows that
market penetration of car2go, a German one-way carsharing company is about 0.37%, which is 25
times higher than the market penetration of round-trip carsharing in Germany. However, it is
relevant to note that this figure depends on their computation process, which was based on
member subscriptions and not on the number of active members.
A study based on a stated-preference survey performed in Greece (Efthymiou, 2012)
(respondents were aged between 18 and 35) also concluded that the flexibility to return the
vehicle to a station different from the one where it was picked up is a critical factor to joining a
carsharing scheme. Although stated-preference only shows how things could be and not what
would really happen if they were implemented. Costain et al. (2012) studied the behaviour of a
round-trip carsharing company in Toronto, Canada, and concluded that most trips are made for
grocery or other household shopping purposes, which supports the idea that reasons for making
trips are limited.
Despite the apparent advantages of one-way systems they do present the operational problem of
creating unbalanced vehicle stocks in the stations due to the uneven nature of the trip pattern in a
city. Nevertheless, a great effort has been made to provide these flexible systems for users in
recent years. One notable example is the car2go company (car2go, 2012), implemented first in
Germany by the Daimler carmaker and recently extended to some other European and North
American cities.
EJTIR 13(3), 2013, pp.201-220 203
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
1.2 Motivation
The complexity of managing carsharing systems, especially one-way trips, is directly linked to
the interplay effect of supply and demand. One must be able to accurately model the demand
and supply of these systems to better operate carsharing and estimate its effect on mobility
management and the accessibility that it provides in urban areas.
Carsharing has gained great momentum in the European Union as a measure to manage
transport demand, resulting in the implementation of a very significant number of private and
public carsharing initiatives. For instance the Covenant of Mayors’ initiative was created to get
European municipalities to work to reduce vehicle emissions by 20% by 2020, and one of the
measures that repeatedly appears in the plans presented by cities is the promotion of carsharing
systems. However, despite the interest shown in carsharing, there are not yet many instruments
to measure the impact of carsharing systems on the sustainability of urban mobility. Moreover, it
is often difficult to define their operational principles, especially in the fastest growing market of
one-way carsharing systems.
This paper sets out to give a comprehensive overview of the research that has been conducted on
this alternative transport system, mainly in the last few years. The focus is on the mathematical
models developed so far, especially those that address demand modelling and ways of balancing
vehicle stocks across stations in one-way carsharing systems. The objective is to create a
milestone, identifying the existing research gaps and proposing possible paths for future
development in this field. Our strategy consisted of reviewing all the research articles and
scientific reports on carsharing, especially those that use mathematical modelling to understand
the behaviour of these systems and devise better operational principles for managing carsharing.
The paper is structured as follows. The next section gives the background to carsharing systems.
This is followed by a review of the research that has produced models to characterise carsharing
demand. Then the models that have been developed to study one-way carsharing systems are
reviewed, but specifically focused on creating ways to solve the vehicle imbalance problem,
which is one of the major issues with running these systems. After that we have a tabular
summary of the studies that have been carried out so far. The paper ends with a section where
we point possible ways to plug the gaps in the literature, thus identifying potential paths of
future research in this field.
2. Carsharing: history and trends
The origins of shared-use vehicles can be traced back to 1948, when a cooperative known as
Sefage set up services in Zurich, Switzerland (Shaheen et al., 1999). These first experiments were
mainly motivated by economic reasons. Elsewhere, a series of ‘public car’ experiments were
attempted, but failed. Among the failures were a carsharing initiative known as Procotip, which
began in Montpellier, France, in 1971 and another, called Witkar, which was deployed in
Amsterdam in 1973 (Shaheen et al., 1999). However, failure breeds experience, which, coupled
with the advances in communication technology, enabled several successful programmes to be
launched in the 1980s. In these programmes we may include Mobility Carsharing in Switzerland,
and Stattauto in Germany.
At first it was predicted that carsharing would not work in the US because ‘American cities have,
with almost no exception, become motor cities adapted to the owner-driver form of transport’
(Fishman and Wabe, 1969). So carsharing programmes only appeared later in the 1980s, under
the Mobility Enterprise programme. In contrast to early users in Europe those in the US were
motivated more by convenience than by affordability, possibly because driving is very cheap in
the US (Lane, 2005). The concept of shared vehicles only started to become popular in the US in
the 1990s. Several pilot projects were carried out to achieve a better understanding of how to
EJTIR 13(3), 2013, pp.201-220 204
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
implement and operate this kind of system. These include UCR Intellishare at the University of
California at Riverside (Barth and Todd, 2001), ZEV.NET at the University of California at Irvine,
and Carlink I and II at the Bay Area Rapid Transit station in Dublin-Pleasanton (Shaheen et al.,
2000; Shaheen and Wright, 2001). The projects provided insights on user responses to shared-use
vehicles and allowed the assessment of the possibility that these systems could be operated as a
business. Hence, a natural progression to the commercialisation of the concept in many countries
such as the United States, Japan, and Singapore, was expected (Kek et al., 2006). Currently, the
largest carsharing company in the World is Zipcar, which was founded in 2000. In May 2012, this
company had a fleet of 9,000 vehicles and 700,000 members (Wikipedia: Zipcar, 2012).
Carsharing has been observed to have a positive impact on urban mobility, mainly because each
car is used more efficiently (Litman, 2000; Schuster et al., 2005). Shared vehicles can have much
higher utilisation rates than single-user private vehicles because each vehicle spends more time
on the road and less time parked, thereby diluting the sunk costs. When cars are being used they
are not occupying parking places, so in the medium- to long-run higher vehicle utilisation rates
should also mean less land needed for parking (Mitchell et al., 2010). The use of carsharing
systems has sometimes led to a fall in car ownership rates and thus to lower car use, according to
Celsor and Millard-Ball (2007). Martin et al. (2010) conducted a stated-preference survey in North
America and concluded also that carsharing members reduced their vehicle holdings
significantly, from an average of 0.47 vehicles per household to 0.24 vehicles per household. More
recently, Schure et al. (2012) based on a survey conducted in 2010 on 13 buildings in San
Francisco concluded that the average vehicle ownership for households that use carsharing
systems is 0.47 vehicles/household compared to 1.22 vehicles/household for those that do not.
Moreover, a recent study by Sioui et al. (2010) surveyed the users of Communauto inc., a Montreal
carsharing company, and concluded that a person who does not own a vehicle and makes a high
use of the carsharing systems (more than 1.5 times per week) never reaches the car-use level of a
person who owns a vehicle: there was a 30% difference between them. This idea is reinforced by
Martin and Shaheen (2011) who found through a survey in US and Canada that the average
observed vehicle-kilometres travelled (VKT) by respondents before joining carsharing was 6468
km/year, while the average observed VKT after joining carsharing was 4729 km/year, which is a
decrease of 27% (1749 km/year). Furthermore, results of recent survey studies seem to indicate
that carsharing systems can have positive environmental effects: for instance, Martin and
Shaheen (2011) noted from the VKT estimations presented above that the greenhouse gas (GHG)
emissions of the major carsharing organisations in the US and Canada can be statistically
significantly reduced by -0.84 t GHG/year/household. While most members increase their
emissions; there are compensatingly larger reductions for other members who decrease their
emissions. Moreover, Firnkorn and Müller (2011) conducted a survey of a German carsharing
company and concluded that the CO2 emissions have decreased by between 312 to 146 kg
CO2/year per average user.
In the meantime several studies have been conducted to find out who the users of these systems
are. Most of the studies were done through user surveys and have repeatedly demonstrated
important tendencies: for instance, it has been shown that many carsharing members are frequent
public transport users and tend to live in medium to high density areas (Cervero, 2003; Shaheen
and Rodier, 2005; Burkhardt and Millard-Ball, 2006). The users tend to be in their mid 30s to mid
40s, be highly educated (Brook, 2004; Lane, 2005), belong to a household of less than average size
(Brook, 2004; Millard-Ball et al., 2005) and be environmentally aware people (Costain et al., 2012;
Efthymiou et al., 2012). Moreover, the accessibility to the stations, in terms of the distance
between home/work and the nearest station, is a critical factor to joining carsharing (Zheng et al.,
2009; Costain et al., 2012; Efthymiou et al., 2012).
EJTIR 13(3), 2013, pp.201-220 205
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
3. Demand modelling
One of the most productive streams of research on carsharing has been the study of the
characteristics of its users. Most works report on the mean value of population characteristics
using a sample of carsharing users. But other studies have used more sophisticated models to
better support their conclusions.
Stillwater et al. (2008) compared the use of carsharing vehicles over a period of 16 months with
the built environment and demographic factors for an urban US carsharing operator. They used
regression analysis to explain the average monthly hours of carsharing use and concluded that
the most significant variables were: street width, the provision of a railway service, the
percentage of drive-alone commuters, the percentage of households with one vehicle, and the
average age of the stations. The percentage of drive-alone commuters, street width, and heavy
rail availability were negatively related to carsharing, that is, the higher these factors the lower
the demand. The percentage of households with one vehicle, the average age of the stations, and
light rail only availability were positively related to carsharing use, meaning that the higher these
factors the higher the demand.
In the same year a study by Catalano et al. (2008) was published reporting on a stated-preference
survey in Palermo, Italy, with the objective of forecasting the modal split of the urban transport
demand in that city. The respondents could choose from different transportation alternatives,
which were private car, public transport, carsharing, and carpooling. A random utility model
was estimated using the survey data. The authors concluded that in a future scenario
characterised by active policies to limit private transport use the carsharing market could
increase up to 10%.
Zheng et al. (2009) studied the potential carsharing market at the University of Wisconsin-
Madison performing a stated-preference survey about transportation habits and carsharing
preferences, namely travel habits (primary mode of travel and trip purpose), attitudes on
transportation and the environment, and familiarity with carsharing, in the university
community. With the data obtained from the survey, logistic regression models were developed
to predict the willingness to join a carsharing program. This study led to the conclusion that the
status in the university (student, staff, etc.) and people’s attitudes have a great impact on the
acceptance of carsharing: students are more willing to use carsharing than other faculty
members; the same happens with people who are concerned about the environment and the cost
of owning and driving a vehicle.
More recently Lorimier and El-Geneidy (2011) studied the factors affecting vehicle usage and
availability in the carsharing stations from Communauto inc., a Montreal carsharing company.
With this data, they developed a linear regression model to explain vehicle usage, and a logistic
regression model to explain vehicle availability (binary response variable). The authors
concluded from the results provided by the two models that the size of a carsharing station has a
large impact on both variables. Larger stations offered more vehicle options and had a larger
catchment basin than smaller stations. Moreover, the seasonal impact on both availability and
usage was clear: fewer vehicles were available in the summer, which required an increase in both
the number of stations and the number of vehicles in each station. Vehicle age was also
considered as a key factor. It increased availability and decreased usage, since members tend to
prefer newer vehicles. Having child seats was another factor that corresponded to higher
availability and lower usage, probably because of the demographic characteristics of carsharing
users in the study region (Lorimier and El-Geneidy, 2011).
A study by Morency et al. (2011) analysed the carsharing transaction dataset of the same
company, Communauto inc., with a view to establishing a typology of carsharing users. This
transaction dataset included all trips of all the clients, even those that were cancelled, modified,
or not concluded. The main focus of the authors was frequency of use and distribution of
EJTIR 13(3), 2013, pp.201-220 206
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
distance travelled. With respect to the weekly distance travelled, a cluster-based classification
process resulted in two distinctive behaviours with respect to the distance travelled and trip
frequency: either urban distances throughout the week, or long distances on just one day of the
week.
Motivated by the increasing use of carsharing systems in Europe and the need to understand
their effect on urban mobility, Ciari et al. (2011) have developed an activity-based micro-
simulation model to estimate travel demand for carsharing, considering that users have different
transportation modes available, such as public transport, car, bicycle, walking and carsharing.
The authors opted for this disaggregated model because of the specificity of carsharing with
respect to the users’ characteristics, which makes it difficult to estimate travel demand using an
aggregated model, such as the classical four steps method. The authors followed the round-trip
carsharing systems’ organisation, and so the main features of the modelled system were:
carsharing is available to everybody with a driving licence (membership was not modelled);
agents can pick up, park and drop off cars only at predefined locations (stations); agents have to
drop off the car at the same station where it is picked up (round-trip); it was assumed that agents
walk to the pick-up point and from the drop-off point; and an unlimited number of cars are
available at the stations (no reservation, every agent trying to use the service will find a car)
(Ciari et al., 2011). Hence the model did not focused on the carsharing operations themselves but
on what the demand would be if a high level of this service was provided for the clients. The
model was applied to part of Zurich city centre, Switzerland. The aggregate results matched what
is happening in reality, given the specifications and the level of detail that was considered.
However, the authors stated that the model should be improved to represent the characteristics
and the way real carsharing services work, i.e. by imposing constraints on vehicle stocks.
Recently, Morency et al. (2012) took a two-stage approach to study the behaviour of carsharing
users. In the first stage, the probability of each member being active in a given month was
studied using a binary probit model. In the second stage, the probability of an active member
using the service multiple times per month (monthly frequency of use) was determined by a
random utility-based model. These models were estimated using data from 40 months of
Communauto inc. operations. The authors concluded that the activity of members in the previous
4 months influences their behaviour in the current month, that is, the number of times that users
avail themselves of the service in the previous 4 months is directly proportional to the amount of
times they use it in the current month. Moreover, some attributes of the users, such as gender and
age, have an impact on their behaviour: males and people aged between 35 and 44 are more
inclined to favour carsharing.
In general, almost all the studies presented, except the one performed by Ciari et al. (2011), are
context specific, and local and regional characteristics make standardisation more complex.
Nevertheless, they do reveal user preferences and provide new models that can be used by other
carsharing operators to guide their system’s growth. The study by Ciari et al. (2011) is innovative
in using simulation to predict demand. However it does not include a representation of the
supply side. Furthermore, all the studies are related to the round-trip carsharing. As far as we
know, demand estimation has not so far been addressed in the literature for one-way carsharing
systems, and it is increasingly important to study it since these systems might be able to capture a
higher share of the demand and they are certainly a tendency in recent years.
4. Modelling one-way carsharing systems
Recent research has been examining one-way carsharing systems. The key issue in them is the
dynamically disproportionate distribution of vehicles across the stations, so researchers are
currently developing methods to analyse and mitigate the effects of this problem.
EJTIR 13(3), 2013, pp.201-220 207
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Comprehensive approaches have aimed at modelling the one-way system as a whole, while other
specific techniques have been proposed for balancing the systems.
4.1 A comprehensive approach to the vehicle imbalance problem
The problems of one-way carsharing systems are mainly the effect that demand has on the
supply characteristics and the effect of supply on demand. To depict this important aspect of one-
way carsharing systems we have created the following simplified causal loop diagram (Figure 1).
(a) Two station system
(b) More than two stations
Figure 1 Feedback in one-way carsharing systems
EJTIR 13(3), 2013, pp.201-220 208
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
In Figure 1-a) we can see that when the number of vehicles in station 1 is higher more demand is
satisfied and transferred to station 2 with a delay corresponding to the travel time between the
stations. If the trips that request a vehicle at station 2 have station 1 as destination a reinforcing
loop is formed, where the most important behaviour parameters are the demand pattern that
controls the vehicle flows, the delay that controls when vehicles will be available at a station, and
the capacity of each station.
But hardly any system has just 2 stations and a balanced number of trips between them, thus in
Figure 1-b) we can see that the loop may be cut open by trips from station 2 to other stations,
which will stop feeding station 1 with vehicles, thereby limiting the demand satisfied in station 1.
Papanikolaou (2011) addressed this problem in a recent paper, reporting a system dynamics
model to describe the complexity behaviour of one-way carsharing systems under non-
homogeneous demand patterns. This aggregated simulation model consisted of three stock-flow
sub-models: stations, users, and vehicles, and about 120 equations. Papanikolaou’s objective was
to develop a study tool to understand how these systems evolve over time and to explore
organisational solutions to mitigate the feedback loop effect.
After running the model using a synthetic case study area, Papanikolaou concluded that the
framework could model delays and capture the essential dynamic behaviour of the system, but
the model still needs to be validated with data from real systems. Furthermore, the aggregate
nature of system dynamics models does not take into account the strong spatial and temporal
characteristics of complex urban networks. Another limitation inherent to all continuous
forecasting models for system dynamics ‘is the simplification of behaviour due to the mixing that
occurs when aggregating resources in the stocks, that tends to drift the results as simulation time
advances sometimes overestimates performance’ (Papanikolaou, 2011). The testing of different
relocation techniques and the study of random events such as vehicle breakdowns or different
reservation schedules are highly compromised by this approach. Besides, it is harder to integrate
the effect of other transport modes in this type of model. The authors are planning to compare
this aggregate model to a micro agent-based model where the emergence of the system behaviour
comes from the decisions of the individual travellers.
4.2 The performance of one-way systems
Some authors have been trying to study the performance of these systems by considering
different scenarios and using micro models.
Li (2011) developed a discrete-event simulation model to study the performance of one-way
carsharing systems under different configurations and customer behaviours. They used different
performance measures such as average vehicle utilisation, reservation acceptance rate, full
parking time, and profit. Configurations include fleet size, number of stations, number of parking
spaces, and the distribution of vehicles and parking spaces across stations, while customer
behaviours included the time when a customer reserves a vehicle, whether or not they will cancel
the reservation, when they will pick up/return the vehicle, what the requirement for parking at
pickup/return location is, etc. A reservation is accepted if there is a vehicle available in the origin
station for the time the user needs to make the trip and a parking space available in the
destination station at the end of the trip. Round trip carsharing was also simulated to compare
the behaviour of the two systems. The author concluded that concentricity, that is high
population density in big cities where origin and destination stations are concentrated at few
locations, helps one-way carsharing to perform better, while round trip carsharing performs
better for a more dispersed population. Moreover, increasing the number of parking spaces for a
one-way system can help improve the reservation acceptance rate and vehicle utilisation. The
allocation of parking spaces among stations should correspond to the customer demand
distribution among stations, so as to achieve the best performance.
EJTIR 13(3), 2013, pp.201-220 209
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
More recently, Barrios (2012) developed an agent-based simulation model to measure and predict
the level of service offered to users of car2go, a one-way carsharing system, in Austin, Texas, and
San Diego, California. In this simulation model, cars are initially evenly distributed throughout a
schematic square city. Users come into the system at a given demand density and have random
origins and destinations, searching for the nearest car. If a user is ‘created’ in a place where there
are no cars within walking distance, they cannot travel. Then the simulation model was used to
simulate the two cities. Simulation results were compared with the reality of these systems in the
two urban areas in terms of accessibility, that is, the proportion of the operating area with an
available car within walking distance (0.33 miles). The author concluded that the model
reasonably estimates the level of service of a one-way carsharing system. Hence it can be used to
make planning level decisions.
These methods could be very useful, since they present advanced simulation tools to evaluate the
performance of one-way carsharing systems prior to their implementation, which can help
carsharing companies to plan the service. However, they do not include any relocation feature
which, as shown in the next section, should be an important leverage for the profitability of one-
way carsharing.
4.3 Solving the vehicle imbalance through relocation operations
One of the ways that has been suggested to balance vehicle stocks is the operator-based approach
that corresponds to the periodic relocation of vehicles among stations by staff members.
In 1999, Barth and Todd (1999) developed a queuing based discrete event simulation model that
included relocations and a number of input parameters that allowed different scenarios to be
evaluated, such as the vehicle-to-trip ratio and payment scheme. Some of the events were
customer arrival at a station, vehicle departure from a station, vehicle arrival at a station,
relocation start and relocation end. Three ways of deciding when relocations should be
performed were presented: ‘Static relocation’ based on immediate needs in a station; ‘Historical
predictive relocation’, which uses knowledge of expected future demand, looking 20 minutes into
the future, and ‘Exact predictive relocation’ that can be used if perfect knowledge of future
demand is available, which is impossible in the real world. The model was applied to a
community in Southern California and some measures of effectiveness were calculated, showing
that the system is most sensitive to vehicle-to-trip ratio, relocation method used, and charging
scheme employed. The authors also concluded that this system can be very competitive with
other transportation modes, particularly taxis.
Later, the operator-based approach was tested in the Honda ICVS system, which started in
March 2002. This system allowed one-way trips, with no reservation required and no return time
needing to be specified (Kek et al., 2006). Periodic communication with the backend computer
system meant that information from these vehicles prompted the system manager to relocate
vehicles between stations when needed. In May 2005 the experimental phase ended and the
programme started running as a commercial enterprise (Barth et al., 2006). The system was
cancelled in 2008 (The Business Times, 2011) because the operator could not keep up the initial
service quality as membership grew. All users expected cars to be available but this did not
happen, so dissatisfaction and complaints from members increased (The Straits Times, 2008). It
still not known if the operator based approach alleviated the company’s losses, but its ending is
an indication that this system was not profitable.
Kek et al. (2006) developed a discrete event simulation model to help multiple-station carsharing
vehicle operators implement an efficient relocation system that could minimise the allocation of
resources to vehicles, staff, and parking places, while maintaining certain levels of service. Based
on the available resources (vehicles and staff members) and the relocation policy adopted, the
model decided which relocation operations the system should implement at each time step. The
EJTIR 13(3), 2013, pp.201-220 210
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
two possible relocation policies were the shortest time, i.e. moving vehicles to or from a
neighbouring station in the shortest possible time (including staff movement, if necessary), and
inventory balancing, i.e. supplying a station which has a shortage of vehicles with a vehicle from
another station which has too many.
The proposed model was tested and validated using real data from the commercial one-way
carsharing company mentioned above. The results showed that if the company adopted an
inventory-balancing relocation technique, the system could afford a 10% reduction in car park
spaces and 25% reduction in staff strength, generating cost savings of approximately 12.8%
without lowering the level of service for users. However, this approach required an impractical
number of simulation runs to test all the combinations of different parameters. Thus in 2009, Kek
et al. introduced an optimisation module to the previous simulation tool to overcome this
limitation.
The optimiser used the current demand and station configuration with the objective of
minimising the overall cost of the relocation operations. The obtained optimised parameters were
the number of people needed to perform the relocations (staff strength), the necessary
relocations, and the resulting station status (number of vehicles at the stations at each time step).
In phase two these optimised parameters were filtered through a series of heuristics to obtain a
set of recommended parameters. On entering the set of operating parameters into the vehicle
relocation simulator developed by Kek et al. (2006), phase three evaluated the effectiveness of
each combination using three performance indicators: zero-vehicle time (no vehicles are
available); full-station time (no parking places available), and number of relocations.
This decision support system was once again tested using operational data provided by Honda
ICVS. The performance surpassed the results of the previous simulations conducted by Kek et al.
(2006). This new three-phase decision support system indicated a 50% reduction in staff costs and
an improvement of the performance indicators. However, there is no documented effect on the
real application of the model to Honda’s system and, as mentioned, the system was cancelled.
Although mathematical programming is an interesting approach to achieving good system
configurations for running these systems, it needs a great many variables and constraints in the
mathematical programming formulation. Several simplifications were required to reduce the
problem size, such as increasing the size of the discrete time steps of the optimisation period, and
this entailed further limitations to representing accurately what happens in reality.
In 2010, Wang et al. proposed a method to forecast and relocate vehicles in carsharing systems
that consisted of three main components: microscopic traffic simulation, forecasting model, and
inventory replenishing. The model used to forecast customer demand was an aggregated model
at the station level, that is, it forecast the total number of vehicles rented out and returned over
time at each station and not the trips for each origin-destination pair. The forecast demand was
then fed into the inventory replenishing model to prepare relocations.
With respect to the inventory replenishing model, the stations with more vehicles available than
necessary (including safety stock) were defined as overstocked stations thus candidate suppliers,
while stations with fewer vehicles available than necessary were defined as understocked
stations thus candidate demanders. Once the relocation decision was made (stations and number
of vehicles involved, and when to relocate), the understocked stations were replenished from the
nearest overstocked stations in terms of the lowest trip cost at that moment of relocation, which
was determined by the microscopic traffic simulation model.
The performance of the proposed approach was tested using as case study four possible station
locations in Singapore with 12 vehicles each. Experimental results showed that the system
improved its efficiency. So the authors concluded that this approach had the potential to improve
a carsharing service. The case study was novel in that it improved the realism of travel times and
evolved from the computation of a mathematical static optimum pattern of relocation operations
EJTIR 13(3), 2013, pp.201-220 211
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
planned for the whole day to a policy proposal for real time operation in face of the predicted
demand. Even so, it is important to note that it is very small scale in terms of the number of
stations, which is smaller than most existing carsharing systems.
Cucu et al. (2010) have studied a forecasting model of one-way carsharing demand in greater
detail. The main principle was to understand and exploit customers’ preferences so as to
anticipate their needs and relocate the vehicles accordingly. To solve the problems of car
unavailability at peak periods and medium- and long-term management, and to test new station
locations, the authors considered customer preferences related to: the time of departure; the day
of the week; the weather conditions, and the traffic conditions associated with their addresses.
With this, the maximum vehicle needs for a given period of the day were computed for all
stations to distinguish priorities in terms of balancing order, other maintenance operations and to
test the implementation of new stations. This method was applied to a small city and the authors
concluded that anticipating demand would improve the vehicle availability by balancing vehicle
stocks.
Nair and Miller-Hooks (2011) continued exploring optimisation methods and proposed a
stochastic mixed-integer programming (MIP) model with the objective of generating least-cost
vehicle redistribution plans such that a proportion of all near-term demand scenarios is satisfied,
recognising the strong effect of uncertainty on carsharing planning. The model was set up for a
fixed short-term planning horizon for which demand is known probabilistically. Relocation
operations were performed throughout the day and assumed to be completed before the
beginning of the planning period considered. For relocations the operator takes into
consideration both vehicles and free parking spaces, that is, if both resources are adequate to
satisfy a p-proportion of all possible demand scenarios, no corrective actions are triggered; if not
enough vehicles are available this can be remedied by relocating from adjacent stations; and if
not enough parking spaces are available, vehicles can be relocated to free up parking spaces in
other stations. The model was also applied to Honda ICVS. Using computational experiments
and simulation, authors showed that when these relocation strategies are used, the system
operates at a reliability level that could not be achieved otherwise.
Very recently, Smith et al. (2012) studied how to minimise the number of rebalancing vehicles
travelling within a network and the number of rebalancing drivers needed to rebalance vehicle
stocks in one-way carsharing systems, considering that the number of waiting customers remains
bounded. The authors state that the “two objectives are aligned” (Smith et al., 2012), so the
optimal rebalancing strategy can be found by solving two different linear programs in a fluid
model of the system. In this system, users arrive at one of the stations (origin) and are
transported to another station (destination) by driving themselves or by being driven by an
employed driver (similar to a taxi) if the system needs to move a driver to the destination station
of the particular user. This happens because drivers can also become unbalanced among stations
and then they have to be moved to other stations without driving a rebalancing vehicle. The
results suggested that in Euclidean network topologies, the number of drivers needed is between
1/4 and 1/3 of the number of vehicles.
4.5 Using the users for system balancing
The user-based approach is a system-balancing technique that uses the clients to relocate the
vehicles through various incentive mechanisms. This is very intuitive and previous studies have
addressed this possibility by modelling its operation.
Uesugi et al. (2007) have proposed grouping or ungrouping parties of people to balance the
system. They developed a method for optimising vehicle assignment to people according to the
distribution of parked vehicles, trying to avoid vehicle imbalance. The authors proposed three
ways to assign the number of in/out vehicles between pick-up/drop-off stations, depending on
EJTIR 13(3), 2013, pp.201-220 212
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
the number of vehicles in the stations. In a normal assignment a group of people ride in one
vehicle and so they subtract one vehicle from the pick-up station and add it to the return station.
In the divided assignment, people from the group ride in vehicles and so the user subtracts
vehicles from the pick-up station and adds vehicles to the return station. In the combined
assignment groups with the same drop-off station carpool in one vehicle with the combined
group, and so the users subtract only one vehicle from the pick-up station and add only one
vehicle to the return station. The assignment is limited to the capacity of the vehicles.
In theory, a station that has excess vehicles could decrease their number by assigning a divided
ride. Equally, in a station which has few vehicles it would be possible to keep the number of
vehicles by assigning a combined trip. The effectiveness of this method was tested through
computer simulation. The authors claimed the results showed that it was effective in minimising
the imbalance problem of one-way carsharing systems. But they also stated that incentives would
have to be considered to make users behave according to the proposed model.
It should also be noted that the Intellishare research team at the University of California at
Riverside had already proposed and tried these two user-based relocation methods in 2004. They
called them trip splitting and trip joining, and they managed to reduce the number of relocations
required (Barth et al., 2004). This method was very similar to the previous one, with the
advantage of having a price incentive mechanism to encourage users to sign up to trip splitting
and trip joining. When users wanted to travel from a station with a shortage of vehicles to one
with an excess they were encouraged to share a ride in a single vehicle (trip joining), thereby
minimising the number of cars moved. Conversely, when they wanted to travel from a station
with too many vehicles to one with a shortage they were encouraged to drive separate vehicles
(trip splitting), thereby balancing the number of vehicles in the stations.
If these user-based relocation techniques were successful when applied to commercial one-way
carsharing companies it would be possible to shift the burden of relocating vehicles to the users.
But this strategy has its pitfalls. It may not be a viable option in cities where most travellers value
privacy and convenience over minor transport cost savings. Moreover, trip-joining policies make
carsharing similar to carpooling, which has severe sociological barriers associated with riding
with strangers, mainly for safety and security reasons (Chan and Shaheen, 2011; Correia and
Viegas, 2011). With respect to trip splitting, users may not be willing to be divided and this
method can only work in a market where a significant number of trips are made by groups of
people rather than by a single driver.
Mitchell et al. (2010) proposed another intuitive principle of dynamic pricing combined with
intelligent vehicles that would enable drivers to respond appropriately to pricing. Their idea
involved a pricing scheme that could be applied to several urban mobility systems, including
carsharing, to make them more efficient for users and companies alike.
This method would take advantage of trip-origin-and-destination choice elasticity combined with
price incentives for each specific trip. They considered that people would have the flexibility to
walk a block or two to find a vehicle if one is not available right outside their door. Similarly, it
might not be a problem for people to park a bit further away from their destination. And if
slightly less convenient origins and destinations result in lower-priced trips this could be an
incentive to use them, even if more convenient pick-up and return points are available (Mitchell
et al., 2010).
Febbraro et al. (2012) proposed a similar method but only taking into consideration the trip
destinations. In their system there were no stations, the city was divided into zones and the users
could park the vehicle at any parking space inside these zones. A discrete event simulation model
was developed to implement this method in which the vehicles are relocated by the users that
can opt to end their trips in proposed zones that have a shortage of vehicles or in their desired
zones. The zones were determined through a linear integer programming model aiming at
minimising the rejected reservations. If the user is happy to leave the vehicle in the proposed
EJTIR 13(3), 2013, pp.201-220 213
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
zone they will have a fare discount. The events used in the simulation were: vehicle bookings,
booking modifications, booking cancellations, vehicle pick-ups, and vehicle drop-offs.
Reservations were made throughout the day using a Poisson distribution and each had a given
probability of being modified or cancelled by users. In this model, users were required to
establish the departure time, declare their trip origin and destination, and the time at which they
will deliver the vehicle. When a vehicle is booked it is guaranteed that it will be available at the
beginning of the trip. The authors tested the model for the Restricted Traffic Zone of Turin (Italy),
using fifteen scenarios that differ according the probability of accepting the relocation proposed
by the system and with different numbers of vehicles available in the system. They concluded
that significantly fewer vehicles would be needed for the system to run efficiently using this
approach.
Despite the apparent advantages of this option, the authors also recognised that not all trip
origins and destinations are elastic to price: for instance, a commuting trip may be more
constrained since it is a mandatory trip, usually with very rigid schedules. Other limitations are
related to its practical application, since it relies on a very efficient real-time use of information
and communication technology that enables people to be aware of price changes. In addition,
there must be a willingness to access this information allied to a significant trip-and-station
choice elasticity to price. This is yet to be fully tested in real carsharing companies operating on a
one-way basis. The price instability may also be a disincentive to using carsharing, which would
have a contrary effect.
4.4 Full control over where and how to supply vehicles
Other ways of balancing one-way carsharing systems through controlling the supply have been
devised. Several authors have proposed trip selection for vehicle allocation in order to achieve a
more favourable balance of vehicle stocks, that is, only the trips that help to balance the system
should be served. Fan et al. (2008) formulated a mathematical programming model for vehicle
fleet management to maximise the profits of one-way carsharing operators. In their model the
carsharing operator decides which vehicle reservations should be accepted or refused and how
many vehicles should be relocated or held to maximise profit. Thus, if any request was regarded
as unprofitable or the system could not accommodate it, it would be declined.
A multistage stochastic linear integer model was formulated that could account for demand
variations. A five-day sample network with four carsharing locations was used to test the model
and some results were obtained which indicated profit improvement. Limitations of computation
time and solver capability, however, meant that the model was not applied to a real network and
several unreal conditions were assumed.
A more recent study has addressed the effect of the location of carsharing stations on capturing a
trip pattern more favourable to a balanced distribution of the vehicles in the network, thus
transferring the system imbalance to the clients by decreasing their possibility of accessing this
system. Correia and Antunes (2012) developed three mathematical programming models to
balance vehicle stocks through a convenient choice of the location, number and size of stations.
The objective was to maximise the profit from operating a one-way carsharing system,
considering all the revenues (price paid by clients) and costs involved (vehicle depreciation,
vehicle maintenance and parking space maintenance).
The first scheme (model), which was similar to that tested by Fan et al. (2008), assumed that the
carsharing organisation has total control over trip selection, based on a list of requests made by
the clients. The second assumed that all trips requested by the clients would be accepted when
they occurred between any pair of stations in the solution. And the third was a hybrid scheme in
which there was no obligation to satisfy all trips between stations, but trips could only be rejected
if there were no vehicles available at the pick-up station.
EJTIR 13(3), 2013, pp.201-220 214
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Each model was applied to the case study city of Lisbon, Portugal, and results showed that the
scheme yielding the highest profits was the one where the carsharing operator had full control
over trip selection. This was expected since it is the scheme offering the most freedom to
maximise profit. The authors concluded that the imbalance situation would lead to severe
financial loss in a scenario where all demand should be satisfied, even if the client is charged a
very high price. They also found that financial losses could be reduced by making appropriate
choices of the stations’ configuration (number, location, and size), but profits could only be
achieved with full control over trip selection.
The problem with these approaches is once again the limitations of computation time and solver
capability, which could not accurately represent the reality of carsharing systems because some
simplifications were necessary, such as 10-minute time steps. It was not possible to consider the
choice of station location, trip selection schemes and vehicle relocation operations all in the same
formulation, which hindered an integrated view of these systems planning. For instance the
planning of station locations is intuitively dependent on the existence or non-existence of
relocation operations which could mitigate the effects of an uneven trip pattern and so allow the
supply to expand, as some of the previous research shows.
5. Summary
In Table 1 we present a summary of the studies where carsharing has been modelled. For each
study we indicate the topic addressed, the modelling approach used and the type of carsharing.
The references are in chronological order.
Table 1 Summary of the studies presented
Authors
Year
Modelling Approach
Type of
carsharing
Bonsall and Kirby
1979
Microsimulation
Round-trip
Bonsall
1982
Microsimulation
Round-trip
Arnaldi, Cozot,
Donikian and
Parent
1996
Simulation
Round-trip
Barth and Todd
1999
Queuing-based discrete-
event simulation
One-way
Barth and Todd
2001
Trip joining
One-way
Barth, Todd and
Xue
2004
Simulation
One-way
Kek, Cheu and
Chor
2006
Discrete-event simulation
One-way
Uesugi, Mukai
and Watanabe
2007
Simulation
One-way
Stillwater,
Mokhtarian and
Shaheen
2008
Regression analysis
Round-trip
Catalano, Lo
Casto and
Migliore
2008
Random utility model
Not -defined
Fan, Machemehl
and Lownes
2008
Optimisation
One-way
Zheng et al.
2009
Regression analysis
Not defined
EJTIR 13(3), 2013, pp.201-220 215
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Kek, Cheu, Meng
and Fung
2009
Optimisation and Discrete-
event simulation
One-way
Wang, Chang and
Lee
2010
Microsimulation and
inventory replenishing
model
One-way
Cucu, Ion, Ducq
and Boussier
2010
Optimisation
One-way
Febbraro, Sacco
and Saeednia
2010
Discrete-event simulation
and Optimization
One-way
Lorimier and El-
Geneidy
2011
Regression analysis
Round-trip
Morency,
Trépanier and
Agard
2011
Cluster analysis
Round-trip
Ciari, Schuessler
and Axhausen
2011
Activity-based simulation
Round-trip
Papanikolaou
2011
System Dynamics
One-way
Li
2011
Discrete-event simulation
One-
way/Round-
trip
Nair and Miller-
Hooks
2011
Optimisation
One-way
Morency, Habib,
Grasset and Islam
2012
Random utility model
Round-trip
Barrios
2012
Agent-based simulation
model
One-way
Smith, Pavone,
Schwager,
Frazzoli and Rus
2012
Optimisation
One-way
Correia and
Antunes
2012
Optimisation
One-way
From Table 1 we can see that most of studies related to carsharing modelling were done after
2000. Moreover, despite the considerable number of studies related to demand modelling for
round-trip carsharing, there is a clear predominance of studies about balancing vehicle stocks
across stations in one-way carsharing, mainly through relocation operations performed by the
company or the users. With respect to the modelling approaches used for studying demand,
regression analysis is the most popular technique, while issues related to one-way systems tend
to be studied by means of optimisation and different types of simulation.
6. Suggestions for bridging the gaps found
Carsharing has gained great momentum in the last two decades as an alternative to private
vehicle ownership, especially for urban trips. Despite this growth, there are still many questions
about its true position among the other modes of transport and the markets that it should serve.
Moreover, these systems operation standards are hard to define. Modelling is increasingly being
used to address these issues. Previous research has emphasised demand estimation using
techniques ranging from regression analysis to more complex tools such as agent-based
simulation models. In general, the statistical methods have helped to improve our understanding
of the systems and they have found structural relationships between service characteristics and
demand patterns. However, most of them were too context specific and therefore difficult to
apply to other realities (Stillwater et al., 2008; Lorimier and El-Geneidy, 2011; Morency et al.,
2011; Morency et al., 2012). Moreover, they tend to neglect the supply side and the organisational
configurations needed to offer the service, which is directly connected to its economic viability.
This is a key aspect for any possible candidate operator of a carsharing service.
EJTIR 13(3), 2013, pp.201-220 216
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Advanced demand studies were carried out, like the one by Ciari et al. (2011), who have set up
an agent-based model to represent in detail the type of users who are likely to have carsharing as
part of their mode choice set, recognising the fact that carsharing is not likely to be used for
commuter trips or by certain traveller groups. However, this study did not consider the supply
side, so it is not possible to understand the equilibrium between supply and demand in this
system.
The articles reviewed on demand estimation have generally ignored the one-way option, which is
understandable if we take into account its youth when compared to the round-trip mode. But
more importantly, they tended to disregard the integration of carsharing with traditional
transport modes. Hence it is our belief that a significant effort must be made to develop more
general and realistic models to estimate demand, that is, models that can accurately represent the
characteristics of carsharing, be valid for different contexts, and apply to one-way carsharing.
One of the important questions that still remain to be answered is if carsharing has a greater
effect on reducing the use of private vehicles or if, on the contrary, it reduces the number of
public transport users. This is a paramount question for policymakers who may be deciding
whether or not to endorse carsharing. Currently there are generous funds available for energy
and emissions reduction associated with the transport sector, particularly in the European Union,
and carsharing systems have been considered candidate recipients for those funds.
We have concluded that most of the literature on the modelling of carsharing systems is
concerned specifically with one-way carsharing systems. Operations researchers are devoting
increasing attention to its particular problems which we have seen to be mostly linked to the
natural imbalance of vehicle stocks caused by the uneven pattern of trips during the course of the
day. We have concluded that this is a complex problem with a feedback loop that results from the
interaction of demand and supply. In classic transport systems, such as bus and underground
services, the directional capacity is offered to clients irrespective of the existing demand;
however, in one-way carsharing, demand can completely change the system’s supply in ways
that are hard to predict.
Researchers have used complex system modelling tools to address this problem, such as system
dynamics to try to translate their behaviour and extract better operating principles
(Papanikolaou, 2011). However, these tools bring many limitations especially in what concerns to
their adherence to reality due to their aggregate nature.
Realising the need to have a micro-scale approach to the problems of one-way carsharing
systems, several researchers have been developing simulation models to study their performance
(Li, 2011; Barrios, 2012) and achieve higher profits, but these have not been able to handle the
balancing of vehicle stocks. Other researchers have focused on individual techniques to balance
those stocks. The objective has been to accomplish an optimum configuration of the systems
towards some objective that has varied from the general profit maximisation of the operator and
the specific minimisation of relocation costs. Researchers have mostly used simulation and
mathematical programming optimisation for this.
Despite the positive results of demonstrating that there are operating principles that can be used
to improve these systems’ performance, there have been limitations. While simulation requires an
impractical number of runs to test all the combinations of the different operational parameters
(Kek et al., 2006), optimisation requires a large number of variables to integrate several decisions
into the same problem in a real case study context (Correia and Antunes, 2012). Several authors
tackled this by simplifying the formulations to a level where they start to be unrealistic and the
efficiency improvements gained from them may be hard to transfer to real carsharing ventures.
Most of these models have worked with station-based systems, but stationless systems are
currently emerging, where the vehicle may be dropped off at any parking space. If this trend
develops most of the models devised for studying one-way carsharing systems will be overtaken
by a reality that is not compatible with a fixed set of concentrated demand points.
EJTIR 13(3), 2013, pp.201-220 217
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Researchers should test the techniques on real case scenarios and use more advanced simulation
models to address different aspects of the business, from the most strategic to the operational. We
believe that it will be difficult to find optimum solutions for the operational configuration of
these systems, so research should head towards developing more detailed simulation models that
integrate other modes of transport and also consider operational issues. This will be at the cost of
model size and limited control over the experiments, but the lack of a realistic functioning of the
systems is restricting the view of the big picture on how to run carsharing schemes and find their
true effect on mobility. A detailed and accurate computation of the effect of carsharing systems
should make it possible to bridge a major gap in the literature: there is no measure of the balance
of cost and benefits of these systems, largely because of the uncertainty about the effect that they
have on users of public and private transport. A new trend of peer-to-peer systems has appeared
in recent years. It has changed the cost of the systems for the operator, who now does not need to
buy a whole vehicle fleet. This could greatly change the cost/benefit balance and reduce the risk
of managing the systems.
Researchers must continue to closely watch the big commercial round-trip carsharing ventures
such as Zipcar and Cambio Car to observe the management of their operations and the behaviour
of their clients, and see how any new services provided evolve. The idea is to keep research
abreast of the latest tendencies in this market, which may become more relevant in a context of
financial crisis where falling household budgets could boost the use of cheaper transport
alternatives. It will be especially interesting to watch the first companies that are adopting one-
way trips as their core business, such as car2go.
Acknowledgements
The research upon which this article is based was carried out within the framework of the SCUSSE
project (Smart Combination of passenger transport modes and services in Urban areas for maximum
System Sustainability and Efficiency), MIT Portugal Program. The authors wish to thank the Fundação
para a Ciência e a Tecnologia for financing this project.
References
Arnald, B., Cozot, R., Donikian, S. and Parent, M. (1996). Simulation Models for French Praxitele
Project, Transportation Research Record: Journal of the Transportation Research Board, 1521, 118-125.
Barrios, J. (2012). On the Performance of Flexible Carsharing A Simulation-Based Approach.
http://iceusa.org/GS1%20J%20A%20Barrios_On%20the%20Performance%20of%20Flexible%20Carsh
aring.pdf. 10 February, 2012.
Barth, M., Han, J. and Todd, M. (2001). Performance evaluation of a multi-station shared vehicle
system. Paper presented at the IEEE Intelligent Transportation Systems Conference, Aug. 2001, Oakland,
CA.
Barth, M. and Shaheen, S. (2002) Shared-Use Vehicle Systems: Framework for Classifying Carsharing,
Station Cars, and Combined Approaches, Transportation Research Record: Journal of the Transportation
Research Board, 1791, 105-112.
Barth, M., Shaheen, S., Fukuda, T. and Fukuda, A. (2006). Carsharing and Station Cars in Asia:
Overview of Japan and Singapore, Transportation Research Record: Journal of the Transportation Research
Board, 1986, 106-115.
Barth, M. and Todd, M. (1999). Simulation model performance analysis of a multiple shared vehicle
system, Transportation Research Part C, 7(4), 237-259.
Barth, M. and Todd, M. (2001). User Behavior Evaluation of an Intelligent Shared Electric Vehicle
System, Transportation Research Record: Journal of the Transportation Research Board, 1760, 145-152.
EJTIR 13(3), 2013, pp.201-220 218
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Barth, M., Todd, M. and Xue, L. (2004). User-Based Vehicle Relocation Techniques for Multiple-Station
Shared-Use Vehicle Systems. Paper presented at the Transportation Research Board 83rd Annual Meeting,
Jan. 2004, Washington DC.
Bonsall, P. (1982). Microsimulation: Its application to carsharing, Transportation Research Part A:
General, 16(5-6), 421-429.
Bonsall, P. and Kirby, H. (1979). Microsimulation of Organised Car Sharing Model Predictions and
Policy Implications. Institute of Transport Studies, University of Leeds, Working Paper 114.
Brook, D. (2004). Carsharing Start Up Issues and New Operational Models. Paper presented at the
Transportation Research Board 83rd Annual Meeting, Jan. 2004, Washington DC.
Burkhardt, J. and Millard-Ball, A. (2006). Who Is Attracted to Carsharing?, Transportation Research
Record: Journal of the Transportation Research Board, 1986, 98-105.
car2go (2012.) http://www.car2go.com/. Accessed 15 May, 2012.
Catalano, M., Lo Casto, B. and Migliore, M. (2008). Car sharing demand estimation and urban
transport demand modelling using stated preference techniques, European Transport, 40, 33-50.
Celsor, C. and Millard-Ball, A. (2007). Where Does Carsharing Work?: Using Geographic Information
Systems to Assess Market Potential, Transportation Research Record: Journal of the Transportation Research
Board, 1992, 61-69.
Cervero, R. (2003). City CarShare: First-Year Travel Demand Impacts, Transportation Research Record:
Journal of the Transportation Research Board, 1839, 159-166.
Chan, N. and Shaheen, S. (2011). Ridesharing in North America: Past, Present, and Future., Transport
Reviews, 32(1), 93-112.
Ciari, F., Schuessler, N. and Axhausen, K. (2011). Estimation of Carsharing Demand Using an Activity-
Based Microsimulation Approach: Model Discussion and Some Results, International Journal of
Sustainable Transportation (In press).
Correia, G. and Antunes, A. (2012). Optimization approach to depot location and trip selection in one-
way carsharing systems, Transportation Research Part E: Logistics and Transportation Review, 48(1), 233-
247.
Correia, G. and Viegas, J. (2011). Carpooling and carpool clubs: Clarifying concepts and assessing
value enhancement possibilities through a stated preference web survey in Lisbon, Portugal,
Transportation Research Part A: Policy and Practice, 45(2), 81-90.
Costain, C., Ardron, C. and Habib, K. (2012). Synopsis of users behaviour of a carsharing program: a
case study in Toronto. Paper presented at the Transportation Research Board 91st Annual Meeting, Jan.
2012, Washington DC.
Cucu, T., Ion, L., Ducq, Y. and Boussier, J. (2010). Management of a public transportation service:
carsharing service, La Rochelle, France.
Efthymiou, D., Antoniou, C. and Waddell, P. (2012). Which factors affect the willingness to join vehicle
sharing systems? Evidence from young Greek drivers. Paper presented at the Transportation Research
Board 91st Annual Meeting, Jan. 2012, Washington DC.
Fan, W., Machemehl, R. and Lownes, N. (2008). Carsharing: Dynamic Decision-Making Problem for
Vehicle Allocation, Transportation Research Record: Journal of the Transportation Research Board, 2063, 97-
104.
Febbraro, A., Sacco, N. and Saeednia, M. (2012). .‘One-way carsharing: solving the relocation problem.
Paper presented at the Transportation Research Board 91st Annual Meeting, Jan. 2012, Washington DC.
Firnkorn, J. and Müller, M. (2011). What will be the environmental effects of new free-floating
carsharing systems? The case of car2go in Ulm, Ecological Economics, 70(8), 1519-1528.
EJTIR 13(3), 2013, pp.201-220 219
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Fishman, L. and Wabe, S. (1969). Restructuring the Form of Car Ownership: A Proposed Solution to
the Problem of the Motor Car in the United Kingdom, Transportation Research, 3(4), 429-442.
Kek, A., Cheu, R. and Chor, M. (2006). Relocation Simulation Model for Multiple-Station Shared-Use
Vehicle Systems, Transportation Research Record: Journal of the Transportation Research Board, 1986, 81-
88.
Kek, A., Cheu, R., Meng, Q. and Fung, C. (2009). A decision support system for vehicle relocation
operations in carsharing systems’, Transportation Research Part E: Logistics and Transportation Review,
45(1), 149-158.
Lane, C. (2005). PhillyCarShare: First-Year Social and Mobility Impacts of Car Sharing in Philadelphia,
Pennsylvania. Transportation Research Record: Journal of the Transportation Research Board, 1927, 158-166.
Li, L. (2011). Design and Analysis of a Car Sharing System Offering One-way Journeys. Master Thesis.
University of Wisconsin-Milwaukee.
Litman, T. (2000). Evaluating Carsharing Benefits, Transportation Research Record: Journal of the
Transportation Research Board, 1702, 31-35.
Lorimier, A. and El-Geneidy, A. (2011). Understanding the Factors Affecting Vehicle Usage and
Availability in Carsharing Networks: A Case Study of Communauto Carsharing System From
Montréal, Canada, International Journal of Sustainable Transportation (In press).
Martin, E. and Shaheen, S. (2011). Greenhouse gas emission impacts of carsharing in North America,
IEEE Transactions on Intelligent Transportation Systems, 12(4), 1074-1086.
Martin, E., Shaheen, S. and Lidicker, J. (2010). Impact of Carsharing on Household Vehicle Holdings
Results from North American Shared-Use Vehicle Survey, Transportation Research Record: Journal of the
Transportation Research Board, 2143, 150-158.
Millard-Ball, A., Murray, G., Schure, J., Fox, C. and Burkhardt J. (2005). TCRP report 108: Car-Sharing:
Where and How It Succeeds (Washington DC: Transportation Research Board of the National
Academies).
Mitchell, W., Borroni-Bird, C. and Burns, L. (2010). Reinventing the Automobile: personal urban mobility
for the 21st century, MIT Press, Cambridge, Massachusetts.
Morency, C., Habib, K., Grasset, V. and Islam, M. (2012). Understanding members’ carsharing
(activity) persistency by using econometric model, Journal of advanced Transportation, 46, 26-38.
Morency, C., Trépanier, M. and Agard, B. (2011). Typology of carsharing members. Paper presented at
the Transportation Research Board 90th Annual Meeting, Jan. 2011, Washington DC.
Nair, R. and Miller-Hooks, E. (2011). Fleet management for vehicle sharing operations, Transportation
Science, 45(4), 105-112.
Papanikolaou, D. (2011). A New System Dynamics Framework for Modelling Behaviour of Vehicle
Sharing Systems, Proceedings of the 2011 Symposium on Simulation for Architecture and Urban
Design, Society for Computer Simulation International, Boston, Massachusetts, 126-133.
Schure, J., Napolitan, F. and Hutchinson, R. (2012). Cumulative impacts of carsharing and unbundled
parking on vehicle ownership & mode choice. Paper presented at the Transportation Research Board 91st
Annual Meeting, Jan. 2012, Washington DC.
Schuster, T., Byrne, J., Corbett, J. and Schreuder, Y. (2005). Assessing the potential extent of carsharing
- A new method and its applications, Transportation Research Record: Journal of the Transportation
Research Board, 1927, 174-181.
Shaheen, S. and Rodier, C. (2005). Travel Effects of a Suburban Commuter Carsharing Service:
CarLink Case Study, Transportation Research Record: Journal of the Transportation Research Board, 1927,
182-188.
EJTIR 13(3), 2013, pp.201-220 220
Jorge and Correia
Carsharing systems demand estimation and defined operations: a literature review
Shaheen, S., Sperling, D. and Wagner, C. (1999). A Short History of Carsharing in the 90's, Journal of
World Transport Policy and Practice, 5(3), 16-37.
Shaheen, S. and Wright, J. (2001). Carlink II: Research Approach and Early Findings, California PATH
Research Record, Institute of Transportation Studies, University of California, Berkeley.
Shaheen, S., Wright, J., Dick, D. and Novick, L. (2000). Carlink: A Smart Carsharing System Field Test
Report, California PATH Research Record, Institute of Transportation Studies, University of California,
Berkeley.
Sioui, L., Morency, C., Trépanier, M., Viviani, M. and Robert, B. (2010). How Carsharing Affects the
Travel Behavior of Households: A Case Study of Montréal, Canada, International Journal of Sustainable
Transportation (In press).
Smith, S., Pavone, M., Schwager, M., Frazzoli, E. and Rus, D. (2013). Rebalancing the Rebalancers:
Optimally Routing Vehicles and Drivers in Mobility-on-Demand Systems. Paper presented at The 2013
American Control Conference, Jun. 2013, Washington, DC.
Stillwater, T., Mokhtarian, P. and Shaheen, S. (2008). Carsharing and the Built Environment: A GIS-
Based Study of One U.S. Operator, Institute of Transportation Studies, University of California.
Texas Transport Institute (2010). The 2010 Urban Mobility Report.
The Business Times (2011). History of car sharing in Singapore. 2 February, 2011.
The Straits Times (2008). End of the Road for Honda Car Sharing Scheme. 29 February, 2008.
Uesugi, K., Mukai, N. and Watanabe, T. (2007). Optimization of Vehicle Assignment for Car Sharing
System, Knowledge-Based Intelligent Information and Engineering Systems, 4693, 1105-1111.
U.S. Department of Transportation (2001). National Household Travel Survey: Summary of Travel Trends.
Wang, H., Cheu, R. and Lee, D.-H. (2010). Dynamic Relocating Vehicle Resources Using a Microscopic
Traffic Simulation Model for Carsharing Services, Computational Science and Optimization (CSO),
2010 Third International Joint Conference, Huangshan, Anhui, 108-111.
Wikipedia: Zipcar (2012). http://en.wikipedia.org/wiki/Zipcar Accessed 1 June, 2012.
Zheng, J., Scott, M., Rodriguez, M., Sierzchula, D., Guo, J. and Adams, T. (2009). Carsharing in a
University Community Assessing Potential Demand and Distinct Market Characteristics,
Transportation Research Record: Journal of the Transportation Research Board, 2110, 18-26.
... These preferences and thus the demand vary during the day (uneven nature of the travel pattern). This is the so-called "tide phenomenon", which represents the oscillation of demand intensity throughout the day (spatio-temporal demand asymmetries) (Côme 2014;Jorge and Correia 2013). Besides, the current number of available vehicles in an location affects the number of future rentals in that location. ...
... operator-based relocation of bikes is handled with trucks (Dötterl et al. 2017). Although operator-based relocation is one of the main cost drivers (Jorge and Correia 2013) and (time) inefficient (vehicle cannot be used while it is relocated by staff) (Schiffer et al. 2021), almost all VSS providers use it to obtain reduced imbalance. ...
Article
Full-text available
Vehicle sharing systems have become increasingly popular. However, one-way vehicle sharing system providers face a major challenge. The uneven distribution of vehicles across locations caused by the uneven nature of the demand patterns poses a problem, since there are accumulations of vehicles where the demand is low. This challenge can be solved with an appropriate pricing approach that creates incentives for user-based relocation by considering supply-side network effects. While the literature mostly focuses on trip-based pricing, we were inspired by the majority of car sharing providers who use origin-based minute pricing that differentiates based on the origins of rentals, such as Share Now. Therefore, we develop two different and practicable solution approaches to determine spatially and temporally differentiated origin-based minute prices that take into account supply-side network effects. The first solution approach does not differentiate between rentals and demand and calculates continuous prices for every period and location. The second solution approach determines the vehicle distribution for each period and then calculates the optimal prices for each period backwards. Extensive computational experiments show that our solution approaches anticipate supply-side network effects and thus generate a near-optimal profit in less computational time compared to more complex benchmarks from the literature. In a sensitivity analysis we additionally show that the results are robust against stochasticity of demand and that the solution approaches perform well for different price sets.
... Carsharing is a form of shared mobility that provides easy access to on-demand car use without the burden of car ownership responsibilities, the need to process paperwork such as for car rental services, or even the need to return the vehicle to the pickup points in most of the cases (Liao and Correia 2022). Carsharing services operate on different schemes when it comes to pickup and drop-off arrangements, with three main schemes: round-trip systems, where users pick up and return vehicles to the same locations, and more flexible options like one-way trips or free-floating systems, where vehicles can be dropped off at any designated point within a specified area (Amirnazmiafshar and Diana 2022;Jorge and Correia 2013). ...
Article
Full-text available
Carsharing services have a significant potential for improving urban mobility by increasing the independence and freedom of travel and reducing traffic externalities. Although carsharing has been used for over a decade, several aspects need further investigation, such as the impact of user’s psychological factors on service use, as well as the factors impacting users’ choices between different carsharing operators, in particular their preferences for different payment schemes, and their perceptions of the operators’ application rating. Accordingly, four hybrid choice models (HCM) were estimated to investigate factors impacting (i) the knowledge about carsharing services, (ii) carsharing adoption, (iii) the shift from other modes to carsharing, (iv) the choice between carsharing operators with different payment schemes, using a large survey sample (N = 1044 responses 9469 SP observation) from Munich, Germany. The models showed the significance of sociodemographics, such as income level, education level, household size, employment status, ownership of a bike, access to a car, the availability of a driving license, and public transport subscription-based tickets on the carsharing use directly and indirectly, and four psychological factors encompassing different personality traits (i.e., adventurous), travel behavior, and attitudes were found to be significant in the various models; the latter covered service-related attitudes (perceived carsharing app importance) and travel behavior attitudes or profiles (frequent public transport user and frequent shared micromobility user). This research raises questions regarding the inequitable use of carsharing, the impacts of mobile applications on using the service, and the potential of integrating carsharing in mobility as a Service platforms to increase the potential for multimodality.
... The return of the vehicle can take place at the same station (two-way, round trip), at another station (one-way trip), or at the address indicated. In the case of car sharing, it is frequently possible to leave the vehicle anywhere (freefloating car sharing) [6]. ...
Article
Optimization of business processes and policies in a car rental company is an ongoing topic. Especially that each of these companies operates in slightly different local conditions and external environments. Testing new optimization algorithms is best done with simulation methods. A comprehensive rental simulation model can be built, for example, using the SimEvents library of the Matlab/Simulink environment. The paper focuses on the problem of preparing a sequence of customer requests in the short-term car rental system, necessary to carry out simulations in the SimeEvents environment. It was assumed that these data may come from the real world, or they may also be artificial data. A method of generating artificial sequences of customer requests and the structure of input data necessary to carry out this process have been proposed. The use of machine learning to build models that transform real data in such a way that it can be randomized was also tested.
... The latter refers to the short-term decision problem faced by the SEV operation enterprises, that is, the problems faced by the enterprises in daily service management (Ciari et al., For SEV operators, the main challenge is to rebalance the system before a decrease in customer satisfaction and a shortage of available vehicles. This is also the core of SEV relocation (Jorge and Correia, 2013). Currently, most SEV operation systems have low vehicle turnover and uneven supplies and demands of vehicles between stations, which results in high operation costs and low customer satisfaction. ...
Article
The carsharing service has experienced significant growth over the past few years in China, yet few studies have scrutinized the multi-city variations of this service. Using carsharing data from 61 cities in China, we analyzed the usage and efficiency of each city and investigated the impact of system and urban factors on the service performance. The study reveals vast differences in carsharing supply and demand across Chinese cities. Our results show the parking station density and the parking lot to vehicle ratio of the carsharing system are positively related to the usage. Urban factors such as public transportation availability, educational attainment levels, and vehicle restriction policies, are found to have significantly positive associations with the carsharing usage. However, no urban factors demonstrate significant associations with the efficiency measured by vehicle utilization rate. Moreover, the presence of other competing carsharing services within a city exhibits a positive impact on the performance of carsharing systems. This study also examined nonlinear effects of the factors. It provides valuable insights into the management of carsharing services in China, which can inform policy-making and operational strategies for sustainable development of carsharing.
Article
Full-text available
In the past 10 to 15 years, carsharing has attracted significant attention as a green transportation alternative that can reduce car dependence and promote smarter, healthier mode choice in urban areas. Despite rapid growth and aggressive funding, carsharing remains a niche product with quantifiable but relatively small benefits. A new type of carsharing seeks to expand its reach, and perhaps its benefits, by eliminating some of the inconveniences of traditional services, such as having to book in advance and having to return the car to the same location. Flexible carsharing, pioneered by Daimler's car2go program in late 2008, also introduces a logistical problem not present in traditional carsharing: the spatial distribution of vehicles tends to be irregular because of the randomness of demand. Vehicle redistribution, which can be done periodically (e.g., at the end of the day) or continuously, can be used to ameliorate this problem. By using an agent-based model of a flexible carsharing system, this research explores the trade-offs between fleet size and hired vehicle redistributors, with the objective of maximizing the demand level that can be satisfactorily served. Production functions for representative city sizes and demand densities are presented. Finally, the results of the simulation are compared with actual data from car2go's Austin, Texas, operation. The comparison shows that the simulation results are reasonably close to the measured performance of one-way carsharing in Austin.
Article
The Praxitele project was charged with designing a new kind of transportation in an urban environment, which consisted of a fleet of electric public cars. These public cars are capable of autonomous motion on certain displacements between stations. The realization of such a project requires experimentation regarding the behaviors of autonomous vehicles in the urban environment. Because of the danger connected with these kinds of experiments at a real site, it was necessary to design a virtual urban environment in which simulations could be done. To perform an authentic simulation of a real environment composed of a large set of vehicles (some of which are autonomous and others of which are controlled by the user or by some specific control law), different models need to be implemented: geometric modeling of the environment, mechanical simulation, motion control models, driver models, sensor models, and visualization algorithms. To implement these different models into a unique system, a new simulator system was designed. This simulator takes into account real-time synchronization and communication between cooperative processes implementing the models mentioned earlier. First, the aims and goals of the Praxitele project are presented. The motion control algorithm for automatic platooning of autonomous vehicles is then briefly presented. The focus is on the simulation of a virtual urban environment that includes Praxitele vehicles. The implementation of all of these models is described. Finally, results of a simulation of cooperative driving of the Praxitele vehicles in a virtual urban environment are given.
Article
Members of carsharing organizations reduce both the number of vehicles owned and vehicle miles traveled (VMT). Given these benefits at the individual level, carsharing may interest policy makers as another tool to address the negative environmental, economic, and social consequences of automobile dependence. However, the aggregate effects of carsharing must be estimated before sound policy decisions can be made. This paper describes a Monte Carlo simulation of the economic decision to own or share a vehicle on the basis of major cost components and past vehicle use. The simulation estimates the percentage of vehicles that would be cheaper to share than own. In Baltimore, Maryland, this result ranged from 4.2% under a traditional neighborhood carsharing model to 14.8% in a commuter-based carsharing model. Sensitivity analyses identified travel time and VMT as the most important economic factors, which likely incorporate other factors such as transit access and environmental attitudes. Because travel behavior, not ownership cost, drives the economic carsharing decision, the model hypothesizes that there will be increasing marginal societal benefits from policies that promote carsharing. The model can be applied to any geographic area and can be used to assess carsharing impacts of various policies that change the economics of owning or driving an auto. These results indicate that carsharing can become prevalent enough to be considered an important policy tool.
Article
Carsharing provides members access to a fleet of shared-use vehicles in a network of locations on a short-term, as-needed basis. It allows individuals to gain the benefits of private vehicle use without the costs and responsibilities of ownership. The dynamic vehicle allocation problem is addressed in a carsharing context, that is, as a decision-making problem for vehicle fleet management in both time and space to maximize profits for the carsharing service operator. A multistage stochastic linear integer model with recourse is formulated that can account for system uncertainties such as carsharing demand variation. A stochastic optimization method based on Monte Carlo sampling is proposed to solve the carsharing dynamic vehicle allocation problem. Preliminary results are discussed and related insights are presented on the basis of a five-stage experimental network pilot study.
Article
The provision of sufficient parking for new development in urban areas often is a subject of debate for developers, city agencies, businesses, and the general public, given the differing priorities of each group. The challenge is to create walkable, transit-supportive neighborhoods with minimized parking and traffic impacts that also ensure the marketability of new development. To address that goal, several cities have enacted policies and programs in recent years to encourage carsharing and unbundled parking. Although the independent effects of carsharing and unbundled parking have been studied, research is lacking on the joint effects on parking demand and vehicle use. This paper discusses the findings of a study conducted in 2010 that surveyed residents of 13 buildings in downtown San Francisco, California, to determine any differences between those residents of buildings that provided both carsharing and unbundled parking and those who lived in buildings that offered neither. The differences were examined in terms of vehicle ownership, commute patterns, and carshare vehicle use. San Francisco was selected because of its relatively large number of residential developments that provided on-site carshare vehicles and unbundled parking. These facilities were provided, in part, in response to recent, local ordinances that required new residential and commercial developments to unbundle parking and to host carshare vehicles. The findings of this study suggested that the presence of both carsharing and unbundled parking in residential developments had a significant impact on vehicle ownership and that carshare members had significantly lower levels of vehicle ownership and drive-alone rates than nonmembers did.
Article
Astochastic, mixed-integer program (MIP) involving joint chance constraints is developed that generates least-cost vehicle redistribution plans for shared-vehicle systems such that a proportion of all near-term demand scenarios are met. The model aims to correct short-term demand asymmetry in shared-vehicle systems, where flow from one station to another is seldom equal to the flow in the opposing direction. The model accounts for demand stochasticity and generates partial redistribution plans in circumstances when demand outstrips supply. This stochastic MIP has a nonconvex feasible region. A novel divide-and-conquer algorithm for generating p-efficient points, used to transform the problem into a set of disjunctive, convex MIPs and handle dual-bounded chance constraints, is proposed. Assuming independence of random demand across stations, a faster cone-generation method is also presented. In a real-world application for a system in Singapore, the potential of redistribution as a fleet management strategy and the value of accounting for inherent stochasticities are demonstrated.
Article
In this paper we study rebalancing strategies for a mobility-on-demand urban transportation system blending customer-driven vehicles with a taxi service. In our system, a customer arrives at one of many designated stations and is transported to any other designated station, either by driving themselves, or by being driven by an employed driver. The system allows for one-way trips, so that customers do not have to return to their origin. When some origins and destinations are more popular than others, vehicles will become unbalanced, accumulating at some stations and becoming depleted at others. This problem is addressed by employing rebalancing drivers to drive vehicles from the popular destinations to the unpopular destinations. However, with this approach the rebalancing drivers themselves become unbalanced, and we need to "rebalance the rebalancers" by letting them travel back to the popular destinations with a customer. Accordingly, in this paper we study how to optimally route the rebalancing vehicles and drivers so that stability (in terms of boundedness of the number of waiting customers) is ensured while minimizing the number of rebalancing vehicles traveling in the network and the number of rebalancing drivers needed; surprisingly, these two objectives are aligned, and one can find the optimal rebalancing strategy by solving two decoupled linear programs. Leveraging our analysis, we determine the minimum number of drivers and minimum number of vehicles needed to ensure stability in the system. Interestingly, our simulations suggest that, in Euclidean network topologies, one would need between 1/3 and 1/4 as many drivers as vehicles.
Article
In this paper, we present an optimization approach to depot location in one-way carsharing systems where vehicle stock imbalance issues are addressed under three trip selection schemes. The approach is based on mixed-integer programming models whose objective is to maximize the profits of a carsharing organization considering all the revenues and costs involved. The practical usefulness of the approach is illustrated with a case study involving the municipality of Lisbon, Portugal. The results we have obtained from this study provided a clear insight into the impact of depot location and trip selection schemes on the profitability of such systems.