Available online at [Link].

com
Available online at [Link]

ScienceDirect
ScienceDirect
Available online at [Link]
T ransportation Research Procedia 00 (2017) 000–000
T ransportation Research Procedia 00 (2017) 000–000
ScienceDirect [Link]/locate/procedia
[Link]/locate/procedia

Transportation Research Procedia 27 (2017) 215–221

[Link]/locate/procedia

20th
20th EURO
EURO Working
Working Group
Group on
on Transportation
Transportation Meeting,
Meeting, EWGT
EWGT 2017,
2017, 4-6
4-6 September
September 2017,
2017, Budapest,
Budapest,
Hungary
Hungary

Modeling
Modeling Transportation
Transportation Systems
Systems involving
involving Autonomous
Autonomous Vehicles:
Vehicles:
A State of the Art
A State of the Art
a b
Jaâfar
Jaâfar Berrada
Berradaa*,
*, Fabien
Fabien Leurent
Leurentb
a
a VEDECOM and LVMT, 77 rue des chantiers, Versailles, France
b VEDECOM and LVMT, 77 rue des chantiers, Versailles, France
b Paris-EstUniversity and LVMT, 6-8 Avenue Blaise Pascal, Marne-la-Vallée, France
Paris-Est University and LVMT, 6-8 Avenue Blaise Pascal, Marne-la-Vallée, France

Abstract
Abstract
Autonomous Vehicles (AVs) promise many benefits for future mobility. Several modeling studies investigated their potential impacts with
Autonomous Vehicles (AVs) promise many benefits for future mobility. Several modeling studies investigated their potential impacts with
special focus on spatial and/or socio-economic features. Spatial modeling represents (i) in detail the technical specifications of the novel mode,
special focus on spatial and/or socio-economic features. Spatial modeling represents (i) in detail the technical specifications of the novel mode,
and ii) the spatial features of the area in which the system is implemented. M ost of these models are agent -based. Socio-economic modeling
and ii) the spatial features of the area in which the system is implemented. M ost of these models are agent -based. Socio-economic modeling
addresses the conditions of market penetration and diffusion using mathematical methods with commercial or social orientation. Furthermore, it
addresses the conditions of market penetration and diffusion using mathematical methods with commercial or social orientation. Furthermore, it
investigates investment and operating costs.
investigates investment and operating costs.
This paper summarizes the main modeling works on transportation systems involving AVs that were published in the academic literature up to
This paper summarizes the main modeling works on transportation systems involving AVs that were published in the academic literature up to
end 2016. In addition, we provide some examples of applications and address their respective outreach and limitations. We present
end 2016. In addition, we provide some examples of applications and address their respective outreach and limitations. We present
recommendations for future developments.
recommendations for future developments.
This way, the paper takes part to a research project of which the ultimate goal is to build a predictive model that can be used by operators and
This way, the paper takes part to a research project of which the ultimate goal is to build a predictive model that can be used by operators and
policy-makers in order to test AVs scenarios.
policy-makers in order to test AVs scenarios.
© 2017 The Authors. Published by Elsevier B.V.
©©2017
2017The Authors.
Theunder
Authors. Published by Elsevier
Published B.V. B.V.
by scientific
Elsevier
Peer-review responsibility of the committee of the 20th EURO Working Group on Transportation M eeting.
Peer-review
Peer-reviewunder
underresponsibility of the
responsibility ofscientific committee
the scientific of the 20th
committee EURO
of the 20thWorking Group on Group
EURO Working Transportation M eeting. Meeting.
on Transportation
Keywords: Autonomous vehicles, mobility services, spatial modelling, economic modelling
Keywords: Autonomous vehicles, mobility services, spatial modelling, economic modelling

1. Introduction
1. Introduction
Since the beginning of prosperity of the automotive era, the automation of car-driving has attracted specific studies: let us quote
Since the beginning of prosperity of the automotive era, the automation of car-driving has attracted specific studies: let us quote
the car-to-car communicat ion system using radio waves in Milwaukee during the 1920s (The Milwaukee Sentinel, 1926), the
the car-to-car communicat ion system using radio waves in Milwaukee during the 1920s (The Milwaukee Sentinel, 1926), the
electro magnetic guidance of vehicles in the 1930s and 1940s, or the testing of smart highways by adding magnets to vehicles
electro magnetic guidance of vehicles in the 1930s and 1940s, or the testing of smart highways by adding magnets to vehicles
during the 1950s and 1960s (The Victoria Advocate, 1957). In 1980, Mercedes-Ben z and Bundeswehr University Munich
during the 1950s and 1960s (The Victoria Advocate, 1957). In 1980, Mercedes-Ben z and Bundeswehr University Munich
created the first autonomous car in the world, enabling to start thinking about legislation adaptation (Davidson, et al., 2015).
created the first autonomous car in the world, enabling to start thinking about legislation adaptation (Davidson, et al., 2015).
Since then, many co mpanies launched themselves in the quest for the perfect car or autonomous system, including Mercedes -
Since then, many co mpanies launched themselves in the quest for the perfect car or autonomous system, including Mercedes -
Benz, General Motors, Google, Continental Automotiv e Systems, Inc. Autoliv, Bosch, Nissan, Toyota (Google Car), Audi,
Benz, General Motors, Google, Continental Automotiv e Systems, Inc. Autoliv, Bosch, Nissan, Toyota (Google Car), Audi,
Oxford Un iversity, among others. Addition impetus was provided by the DARPA Grand Challenges I (2004), II (2005) and III
Oxford Un iversity, among others. Addition impetus was provided by the DARPA Grand Challenges I (2004), II (2005) and III
(2007). In 2016, five US states (California, Flo rida, M ichigan, Nevada, Tennessee and the District of Colu mb ia) allo wed to test
(2007). In 2016, five US states (California, Flo rida, M ichigan, Nevada, Tennessee and the District of Colu mb ia) allo wed to test
autonomous vehicles, while 16 other states are considering taking legislative bills about automated driving (Weiner, et al., 2016).
autonomous vehicles, while 16 other states are considering taking legislative bills about automated driving (Weiner, et al., 2016).
The experiments on autonomous vehicles show promising results . Such in-field experiments are main ly intended to test self -
The experiments on autonomous vehicles show promising results . Such in-field experiments are main ly intended to test self -

* Corresponding author. Tel.: +33-6-69-017-094.

* Corresponding author. Tel.: +33-6-69-017-094.
E-mail address: [Link]@[Link]
E-mail address: [Link]@[Link]
2214-241X © 2017 T he Authors. Published by Elsevier B.V.
2214-241X © 2017 T he Authors. Published by Elsevier B.V.
Peer-review under responsibility of the scientific committee of the 20th EURO Working Group on Transportation Meeting.
Peer-review under responsibility of the scientific committee of the 20th EURO Working Group on Transportation Meeting.
2352-1465 © 2017 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the scientific committee of the 20th EURO Working Group on Transportation Meeting.
10.1016/[Link].2017.12.077
216 Jaâfar Berrada et al. / Transportation Research Procedia 27 (2017) 215–221
2 Berrada et al. / Transportation Research Procedia 00 (2017) 000–000

driving technology and possibly also the attitudes, use gestures and behaviours of potential users . Yet, up to now there has been
no large scale imp lementation of AV fleet in a given territory. Prior to that, it is obviously important to deliver safe and reliable
technology and to settle a suitable regulatory framework. Even mo re important, though less obvious, is the requirement to ens ure
commercial success, i.e. the purchase of hiring of A Vs by indiv idual customers of firms, which requires in turn convincing
evidence of A V-based services attractiveness within the range of travel solutions that compete to serve mobility purposes. This is
why a number of researchers have modelled AV-based services as mobility solutions under particular territorial conditions.
This paper reviews the models developed so far, with the aim to summarize their findings and to assess their outreach and
limitat ions. As it turns out, the reviewed models fall into two broad categories depending on their main orientation that can be
geographic or socio-economic. Geographic or spatial models focus on technical conditions concerning service performance,
operations and availability in relat ion to users’ needs and alternative solutions. The choice fro m among alternative solutions
indeed leads to economic issues. Models belonging to the socio -economic category put the emphasis on the temporal conditions
of AV develop ment: this involves issues of technology readiness, legal framework, demand inclination and adoption, in relation
to the production costs of self-driving cars.
The rest of the paper is organized in four parts. Section 2 reviews models that emphasize spatial conditions: they can be further
divided according to whether they are rooted in travel demand needs and choices, or in the dynamic performance of a technical
system that links the supply and demand sides. Then, Section 3 addresses socio -economic models, fro m market penetration to
production costs passing by customership issues. Next, Section 4 reports on the evaluation of potential impacts that range from
traffic volu mes and parking demand to environ mental impacts, passing by safety. Lastly, Section 5 d iscusses the outreach and
limitations of the reviewed models and their applications, before proposing some directions for further research.

2. Spatial Models of AV-based Services

2.1. Models rooted in Travel Demand

Levin et al. (2015a) propose a four-step model div iding demand into classes by value of time and A V ownership. A Vs are
considered as private vehicles. Mode choice is between parking, repositioning, and transit based on a nested logit model. Sta tic
traffic assignment uses a generalized cost function of time, fuel, and tolls. Lev in (2015b) incorporates dynamic traffic assignment
(DTA) with endogenous departure time choices. Thus, the model considers more realistic flo w propagation and intersection
control options. In addition, it has only studied the case of full autonomous vehicles.
Results of static and dynamic assigment prove that using autonomous vehicles improves the capacity of the intersections, but
does not reduce significantly the congestion.
Auld et al. (2017) used a simulat ion model (POLA RIS) wh ich includes an activity-based model (ADA PTS) and a traffic
simu lation model. Market penetration is controlled on a reg ional scale by adjusting road capacity. Results show that capacity and
value of time affect significantly vehicle-kilometres travelled (VKT).
Kloostra et al. (2017) assumed that autonomous vehicles will change road capacity thanks to A CC technology. Then, they
modified road lin ks capacities to simu late theoretical increase in throughput enabled by AV driv ing behaviour. They
distinguished two types of road links: freeways and arterial streets. A static assignment in Emme 4 is realised. In addition, they
analysed the impacts on parking operations.

2.2. Agent-based models

Agent-based models are an effective tool for the study of innovative urban services, as agents act and react according to the
informat ion received in real t ime. On the other hand, activ ity models offer imp roved reproduction of the demand and allow a
more realistic analysis of users' mobility. Thus, agent-based models are highly used in literature to describe and analyse
operations of AV.
In 2013, Burns et al. estimate the utility of shared autonomous vehicles (SA V) for users (wait ing time) and operators (cost of
production). They consider as variables local specifications, trip length, speed, fleet size and vehicle’s cost parameters (Bu rns, et
al., 2013). The model assumes that the vehicle speed is constant and origin-destination trips are uniformly d istributed over the
study area. In addition, an application on three US cities of different sizes Ann Arbor, Babcock Ranch (Florida) and Manhattan
(New-York) confirms the economic potential of AV.
The study realized by ITF (2015) simu lates the shared mobility in the real network of Lisbon using agent -based models. Mode
choice process is based on a rule-based approach. The demand is generated based on the Lisbon Travel Survey. The user groups,
especially for new services, are not considered. A trip is generated when a user send a request. Route choice minimizes trave l
time by integrating the average speed per section per hour. Sixty stations are spread in the city and three types of vehicles ’ are
considered in the model of two-, five- and eight-passenger cars.
One of the most relevant studies is that developed by Fagnant and Kockelman (2014). They simulate SA Vs in Austin (Texas)
using an agent-based model (MATSim). SA Vs are used by 2% of the total demand. The city is composed into traffic zones. Each
traffic zone is characterized by an factor of attractiveness. All the trips are generated every 5 minutes a day using Poisson
distributions. The model is then structured by following four major steps: (1) SAV location and trip assignment, which
determines which available SA Vs are closest to waiting travellers (prioritizing those who have been waiting longest), and then
assigning available SA Vs to those trips. The assignment is done according to a First -Co me First Served (FCFS) order. A vehicle
 Jaâfar Berrada et al. / Transportation Research Procedia 27 (2017) 215–221 217
Berrada et al./ Transportation Research Procedia 00 (2017) 000–000 3

shall be assigned to a customer in an interval of 5mn; otherwise the user is put in the waiting list and is considered as a priority in
the next simulat ion. (2) SAV fleet generation, which defines the fleet size. In particu lar, the fleet size is determined by running a
SA V “seed” simulation run, in which new SA Vs are generated when any traveler has waited for 10 minutes and is still unable to
locate an availab le SA V is 10 minutes away or less. (3) SAV Movement is characterized by a vehicle speed equal in a normal
hour to 3 times the number of areas. Passengers boarding and alighting last 1mn. The calcu lation of the vehicle position is
registered every 5 minutes. (4) SAV relocation, aims to balance the vehicles distribution ahead of the demand. Four strategies are
proposed. The objective is to ensure that - for each grid cell - the difference between available vehicles and vehicles required is
minimal in absolute terms (less than a defined threshold). Zhu et al. (2017) p ropose two reactionary local repositioning strategies
and evaluate their effects in terms of empty vehicle miles travelled and level-of-service provided.
In 2015, Fagnant and Kockelman exp lored the potential of electric vehicles. They found that the average distance travelled per
day is greater than that allo wed by vehicle range (Fagnant, et al., 2015b). Chen et al. (2016) develop this model with considering
moreover charging stations. The stations are generated in order to allow vehicles to reach the user’s origin or destination. The
model is simulated for a medium-sized city, for a range of vehicles of 130km and normal charging stations (4 hour charge).
In addition, Zhang et al. (2015a; 2015b) reproduce the model of Fagnant with considering users’ incomes and dynamic
ridesharing (DRS). Vehicles and routes assignment reduce the total cost shared. In 2017, they implement a parking module in a
discrete event simulation (Zhang, et al., 2017) with various pricing strategies.
Fagnant et al. (2016) found that DRS would reduce the vehicle -kilo metres travelled by 7% and the waiting time by 25%. The
DRS as defined by Fagnant is applied only if induced extra-time for current riders and new travellers can be tolerated.
Boesch et al. (2016) use agent-based model to evaluate required A V fleet sizes to serve different levels of demand. The agent-
based model largely adopts the assumptions of the model of Fagnant (2014) with excluding the relocation policies of the AV.
The study shows that the relationship between the demand and fleet size is non-linear and the ratio increases as the demand
increases.
Levin et al. (2016) develop an event-based framework for SA V. Then, the framework was imp lemented on a cell transmission
model-based traffic simulator considering that all vehicles are SA Vs. Parking has limitless capacity. Vehicles assignment obeys
to first-come first-served policy. Algorith ms typically consist on three steps, performed iterat ively, to find a dynamic user
equilibrium assignment. Shortest paths are found for all origin-destination pairs.
Vakay il et al. (2017) propose a spatially hub-based SAV network model that analyses transfers between AVs and mass transit.
The model considers transit frequency, transfer costs and two rebalancing strategies. It proves that an integration between AV
and mass transit services leads to reduction in congestion and vehicular emissions.
Yu et al. (2017) assessed the potential of using on-demand SA V as the alternative to the low-demand buses to improve the
first/last-mile connectivity in a study area in Singapore. The agent-based model is tested for a bus -only scenario and a series of
scenarios integrating AV with various fleet sizes. Criteria are defined for each actor. For users, the out-of-vehicle time is
evaluated. For transportation services, it is the impact on road traffic. Fro m A V operators’ perspective, the profitability is
considered. Results are positive if all users accept to share the vehicle.
An application in Singapore is based on SimMobility. It is an integrated agent-based demand and supply model. It co mprises
three simulation levels: (i) a long-term level that captures land use and economic activity, with special emphasis on accessibilit y,
(ii) a mid-term level that handles agents’ activities and travel patterns, and (iii) a short -term level, that simulates movement of
agents, operational systems and decisions at a microscopic granularity . Further details could be available in (Sp ieser, et al., 2014;
Lima Azevedo, et al., 2015; Azevedo, et al., 2016). The application considers that the vehicle could anticipate the demand to
reduce wait ing times and to ensure a balance between required vehicles and ava ilable vehicles in each area. A ll existing
competing modes (taxis, trains, buses ...) are considered. The cost of service is assumed about 40% less than the regular taxi
service. The study distinguishes between internal trips, external t rips and transits. The results highlight that for 2400 veh icles and
10 stations, the waiting time is 5 minutes and the number of trips per vehicle is 16 (Azevedo, et al., 2016).

3. Socio-economic models of AV development

3.1. Market penetration

Several auto makers predict the integration of fu lly self- driv ing vehicles in the market by 4 to 10 years: Audi suggests 2017, Fo rd
2020, Nissan 2020, Google 2018 and Tesla 2023 (Dav idson, et al., 2015). In 2012, the IEEE pred icts that 75% of the fleet will be
autonomous in 2040 (IEEE, 2012). According to (Lit man, 2015), this rate will be achieved in 2060 only. More recently, the
French government studied two entry scenarios: a trend -based scenario, in wh ich the deploy ment is very gradual fro m 2040; and
a breaking scenario where, by 2025, cars can be automated (Janin, et al., 2016).
In addition, surveys in Austin show that 40% (Bansal, et al., 2016) to 50% (Zmud , et al., 2016) of US respondents want to use
private A Vs for everyday use. Lavasani et al. in (2016) p roposed a market penetration model for A V by using a generalized Bass
model. Assuming that A Vs will beco me availab le in 2025, the market of new car sales may reach about 8 million annually in 10
years, and saturation may occur in 35 years assuming a 75% market size. The sensitivity analysis concluded that the market size
strongly impacts adoption rate, while the price of the technology d oes not seem influencing the diffusion process.
218 Jaâfar Berrada et al. / Transportation Research Procedia 27 (2017) 215–221
4 Berrada et al. / Transportation Research Procedia 00 (2017) 000–000

3.2. Potential customers

In recent years, various surveys investigated the general acceptance of AVs. In terms of age, some studies suggest that the
service offered by the SA V will have a strong potential to capture the elderly and those with reduced mobility (Rödel, et al.,
2014; Schoettle, et al., 2014; Schoettle, et al., 2015; Fagnant, et al., 2015a). Other studies come to the conclusion that younger
people are more open towards the introduction of A Vs (KPM G, et al., 2012; Power, 2012; Krueger, et al., 2016; Abraham, et al.,
2017). In terms of the gender, men are more likely to use AV (Piao, et al., 2005; Rödel, et al., 2014; Schoettle, et al., 2015;
Abraham, et al., 2017), but less likely to pay (Lavasani, et al., 2017). Anderson et al. (2014) suggest that non-motorized people
would rather be captured by the service, while Krueger et al. (2016) find that motorization or preferences for public transport do
not highly affect the attractiveness of A Vs. In addit ion, Krueger et al. (2016) show that current carsharing and mult iple mo de
users have a higher probability of choosing SAVs with DRS. Power et al. (2012) and Bansal et al. (2016) notice that residents of
urban areas and people with higher income are more inclined to use AV. Lavasani et al. (2017) find that willingness to pay is
affected by travel frequency, commuting distance, demand for parking and perception of AV benefits.

3.3. Production costs

Costs of AVs depend on the learning rate. Bansal et al. (2016) suggest a price of $23,950 in 2025, while Boston Consulting
Group (2015) predicts that in 2025AV will cost $9,800.
Burns et al. (2013) studied the production costs of A Vs spread in three different cities. They estimate capital costs (depreciation,
finance, reg istration and insurance) and operating costs (energy, maintenan ce and repair, and other costs) and found that SAVs
would costs to customer $0.25/km (or $0.10/km for electric and small vehicle) instead of $1/km (for taxis).
Spieser et al. (2014) analyse the impacts of a total substitution of private cars by SAVs in Singapore. Lifespan of SA Vs is
assumed over 2.5y rs, wh ich could be reduced to 1.5yrs if ridesharing is considered (Zhang, et al., 2015a). The purchase costs are
about $15,000. Thus, SA V costs on average $9,728/year instead of $11,315/year for cars. By integrating the value of time for an
average SAVs waiting time of 5,5mn, the costs gap is even more significant ($5,527/year for the AV and $18,295/year for cars).
Fagnant et al. (2016) consider a penetration rate of 1.3%. The purchase costs are about $70,000 per A V and the average lifespan
400.000km (or 7 years). The A merican Automobile Association (AAA) estimate operating costs around of $0.3/km. As a result,
the SAV costs for user about $0.625/km, which is 3 times less than the taxi fare. For operator, the rate of annual return on
investment is around 13% for a total fleet of 2118 AVs.
Considering that vehicles are electric (SA EV), operating costs vary fro m $0.41 to $0.47 per occupied mile trav elled. SA EVs are
price competitive with SAVs when gasoline reaches $4.35 to $5.70 per gallon (Chen, et al., 2016).

4. Impacts of autonomous vehicles

4.1. Impacts on mobility

Fagnant et al. (2014; 2015b) prove that one SAV would replace 9 to 11 conventional vehicles while inducing 10% more VK T.
The ITF (2015) emphasizes that the substitution of conventional cars by SAV would reduce the total fleet by 90%. In additio n,
the vehicles are more used (70% of the day against the current 5%). The ITF also shows that the deployment of A V wou ld
increase the VKT for d ifferent rates of market penetration. Sp ieser in (2014) indicates that a total replacement of cars by SA Vs in
Singapore would reduce by 2/3 the total fleet of vehicles on roads. The application of the Lev in model in Austin (2016) proves
that the replacement of cars by the private A Vs will greatly increase congestion and travel time. The anticipated relocation makes
the situation more crit ical. The DRS reduces the fleet size by 5.3% and empty VKT by 4.8% (Zhang, et al., 2015a). Boesch et al.
(2016) found that the total vehicle fleet could be significantly reduced (by 90%) if ridesharing is allowed and waiting time is over
10mn . According to Chen et al. in (2016), the use of electric vehicles would pass the ratio to only 3.7 of rep laced conventional
vehicles, with an increase in VKT fro m 7 to 14%. Gucwa (2014) used an activ ity-based model to investigate the relationships of
not shared AVs and road capacity, time value and vehicles operating costs. The study shows that VKT could increases by
between 8% and 24%. Similar studies find that VKT could increase by 3% to 30% (Ch ildress, et al., 2015) and around 20%
(Zhao, et al., 2017).

4.2. Impacts on urban parking

For a market penetration of 2%, the demand for parking is reduced by about 90% (Zhang, et al., 2015b). The ITF (2015) fou nd
that for the case of 100% A V, space savings are around 85% to 95%. Ho wever, for the case of 50% of A V, the rate of space
savings is insignificant. Fagnant et al. (2015b) suggest that the total parking demand will fall by around 8 vehicle spaces per
SA V. In addition, moving the parking fro m downtown to less dense outlying areas allows for significant savings (Lit man, 2012;
Fagnant, et al., 2015a). Zhang et al. (2017) suggest that SAV system can reduce parking land by 4.5% in Atlanta at a 5% market
penetration level.
 Jaâfar Berrada et al. / Transportation Research Procedia 27 (2017) 215–221 219
Berrada et al./ Transportation Research Procedia 00 (2017) 000–000 5

4.3. Impacts on accidents

Li et al. (2016) found that A V can save Americans $76 billion each year. The Highway Traffic Safety Ad ministration (NHTSA)
found that 94% (±2.2%) of crashes between 2005 and 2007 were caused by the driver (NHTSA, 2008) while over 40% of fatal
crashes involve driver alcohol o r drug use, driver distraction and/or fat igue (NHTSA, 2012). The French Interdepartmental
Observatory of Road Safety (ONISR, 2016) found that over 90% of crashes were human caused in France. Therefore, if
autonomous vehicles could eliminate hu man causes of accidents, the number o f crashes could ext remely decrease and mot or -
vehicle deaths could be greatly reduced.

4.4. Environmental impacts: energy consumption and pollutant emissions

The environmental impact is estimated on the basis of a Life -Cycle Analysis, which includes the vehicle operations (movement,
cold start ...), but also cars’ manufacturing and the construction of related infrastructure (parking, stations, maintenance
depots...). Since deployment of SA Vs reduces the total vehicle fleet (as cited above), using them implies significant costs
savings. The high use of A V shortens their life span to 1.5-2.5 years (Zhang, et al., 2015a; Spieser, et al., 2014), wh ich helps
improve the fleet performances. Currently, the A V technology reduces fro m 4% to 10% of energy consumed in accelerat ion a nd
deceleration (Anderson, et al., 2014). Also, the use of A V reduces by 85% (Fagnant, et al., 2015b) to 95% (Zhang, et al., 2015a)
emissions induced by cold starts. In addition, sharing vehicles could save more than 4.7% o f energy, greenhouse gas and
pollutants emissions (Zhang, et al., 2015a).
However, the use of electric vehicles would generate a significant demand for electricity during the peak charging period of the
day (53% of the fleet concurrently charg ing). Fast charging, although inducing 15% mo re in cost, is very effective at demand
spreading, with only 8% of the fleet charging during the peak charging period.

5. Conclusion: Outreach, limitations and some recommendations

This brief state of art in modelling transportation systems involving AVs is summarized in Figure 1. We observe that major
models focus on the supply operations and set-ups without detailing the demand side beyond statistical and spatial description in
the form of an origin-destination matrix of trip-flows.
On the supply side, agent-based approaches allow to assess AV fleet size required while optimizing the wait ing time and emp ty
VKT. Furthermo re, these models permit to reproduce in a realistic, detailed and robust way movements of vehicles considering
several strategies. However, s tudies considering the real network are very scarce (ITF, 2015; Adnan, et al., 2016; Anderson, et
al., 2014; A zevedo, et al., 2016). Further, urban constraints which determine the locations of stations and their capacities are n ot
considered at all, even in the case of electric vehicles. Similar to car sharing, t wo servic e configurations are possible: station-
based and free-floating. The free -floating configuration is employed for a b roader variety of uses than the station-based one
(Bereck, et al., 2016). The co mbination of fixed stations and free-floating (wh ile respecting the conditions of accessibility (Ciari,
et al., 2015)) could reduce waiting time and locating stations in low dense areas (which is also economically attractive). Using
dynamic parking cost (relevant to area’s configuration and state of congestion) could be explored as well.
Assignment strategies of vehicles to customers should be optimized as well. Indeed, almost all studies are based on a FCFS
strategy; a strategy that could be optimized using heuristic insertion or simulated annealing (Jung, et al., 2013). In addit ion, all
aforementioned models are applied on urban centres of cities. It would be interesting to explore the service potentialities in
suburban zones, freeways and around major train stations.
On the demand side, almost all of the studies estimate the A V demand based on market penetration. Studies using real inputs a re
those exploring fu ll replacement of cars by A Vs. In addition, the A Vs are not integrated in a mult imodal chain. A detailed study
of the users’ utility of SA V will enable capturing indiv idual preferences while distinguishing between (i) utility of acquiri ng
and/or maintaining and (ii) utility o f using A Vs. Th is study is essential since it enables confirming service potential and defin ing
pricing strategies. Its results are used as inputs of economic models and modal choice models. Further, taking into account d ay-
to-day traffic for the year should be explored using activity-based models, while d istinguishing between typical weekday,
weekend day, holiday or special day events .
On the other hand, to assure AVs sustainability and its dissemination, it is necessary to develop an in-depth knowledge of
production costs and demand evolutions in order to pro mote informed and rational choices. Production costs concern all
components involved in transit system operations (track and station elements, vehicles, and staff) that have to be acquired o r
hired and maintained at acceptable operating conditions. The models of new taxi apps present useful business insights. The yield
management used by UBER, LYFT, GrabTaxi… permits to smooth the demand over time.
Moreover, further studies of the acceptability of A Vs with the various stakeholders are necessary: mainly users, transport
authorities, transit operators, insurance companies and car manufacturers.
Finally, let us outline that these works would h ighly benefit fro m detailed data on production costs, as well as on commercial
revenue and individual utility of the user and environmental impacts. For an innovative service, some data cannot be observed
but must be inferred from comparison bases, and simulated by means of an ad hoc model.
220 Jaâfar Berrada et al. / Transportation Research Procedia 27 (2017) 215–221
6 Berrada et al. / Transportation Research Procedia 00 (2017) 000–000

Figure 1: Synthesis of literature review.

References

Abraham, Hillary, e t al. 2017. Autonomous vehicles and alternatives to driving: trust, preferences, and effects of age. Transportation Research Board 96th
Annual Meeting. 2017.
Adnan, Muhammad, e t al. 2016. SimMobility: A Multi-Scale Integrated Agent-based Simulation Platform. 2016.
Anderson, James, et al. 2014. Autonomous Vehicle Technology: A Guide for Policymakers. s.l. : RAND Corporation, 2014.
Auld, Joshua, Sokolov, Vadim et Stephens, Thomas. 2017. Analysis of the impacts of CAV technologies on travel demand. Transportation Research Board
96th Annual Meeting. 2017.
Azevedo, Lima Carlos Miguel, e t al. 2016. Microsimulation of Demand and Supply of Autonomous Mobility On-Demand. Proceeding of the 95th annual
meeting of Transportation Research Board. 2016.
Bansal, Prateek, Kockelman, Kara e t Singh, Amit. 2016. Assessing public opinions of and interest in new vehicle technologies: An Austin perspective.
Transportation Research Part C: Emerging Technologies. 2016, Vol. 67, pp. 1-14.
Bereck, Henrik, Ciari, Francesco et Axhausen, Kay. 2016. Comparing Car-Sharing Schemes in Switzerland: User Groups and Usage Patterns. 2016.
Boesch, Patrick, Ciari, Francesco et Axhausen, Kay. 2016. Required Autonomous Vehicle Fleet Sizes to Serve Different Levels of Demand. Proceeding of the
95th annual meeting of Transportation Research Board. 2016.
Boston Consulting Group. 2015. Revolution in the driver's seat, The road to autonomous vehicles. [En ligne] 2015.
[Link] consumer-insight-revolution-drivers-seat-road-autonomous-vehicles/.
Burns, Lawrence, Jordan, William e t Scarborough, Bonnie. 2013. Transforming Personal Mobility. New York, United States : T he Earth Institute, Columbia
University, 2013.
Chen, Donna, Kockelman, Kara e t Hanna, Josiah. 2016. Operations of a shared, autonomous, electric vehicle fleet: implications of vehicle & charging
infrastructure decisions. Proceeding of the 95th annual meeting of Transportation Research Board. 2016.
Childress, Suzanne, et al. 2015. Using an Activity-Based Model to explore possible impacts of Automated Vehicles. Proceedings of 94th Annual Meeting of the
Transportation Research Board. 2015.
Ciari, Francesco, Weis, Claude et Balac, Milos. 2015. Evaluating the influence of carsharing stations' location on potential membership: a Swiss case study.
Euro Journal on Transportation and Logistics. 17 Mars 2015, pp. 1-25.
Davidson, Peter et Spinoulas, Anabelle. 2015. Autonomous Vehicles - What Could T his Mean For the Future Of Transport? Sydney : s.n., 2015.
Fagnant, Daniel et Kockelman, Kara. 2016. Dynamic Ride-Sharing and Optimal Fleet Sizing for a System of Shared Autonomous Vehicles. Proceeding of the
95th annual meeting of Transportation Research Board. 2016.
Fagnant, Daniel et Kockelman, Kara. 2015a. Preparing a Nation for Autonomous Vehicles: Opportunities, Barriers and Policy Recommendations.
Transportation Research Part A. 2015a, Vol. 77, pp. 167-181.
Fagnant, Daniel et Kockelman, Kara. 2014. The travel and environmental implications of shared autonomous vehicles, using agent -based model scenarios.
Transportation Research Part C: Emerging Technologies. March 2014, Vol. 40, pp. 1-13.
Fagnant, Daniel, Kockelman, Kara e t Bansal, Prateek. 2015b. Operations of a shared autonomous vehicle fleet for the Austin, Texas Market. Proceeding of
the 94th annual meeting of Transportation Research Board. 2015b.
Gucwa , Michael. 2014. Mobility and Energy Impacts of Automated Cars. s.l. : Stanford University, 2014. Dissertation.
IEEE. 2012. Look Ma, No Hands! [En ligne] 5 Septembre 2012. [Link]
 Jaâfar Berrada et al. / Transportation Research Procedia 27 (2017) 215–221 221
Berrada et al./ Transportation Research Procedia 00 (2017) 000–000 7

ITF. 2015. Urban Mobility System Upgrade: How shared self-driving cars could change city traffic. Paris : OECD, 2015.
Janin, Lionel, Nemri, Mehdi e t Raynard, Christine. 2016. La voiture sans chauffeur, bientôt une réalité. Paris : France Stratégie, 2016.
Jung, Jaeyoung, Jayakrishman, R. e t Park, Ji Young. 2013. Design and modeling of real-time shared-taxi dispatch algorithms. 2013.
Kloostra, Bradley e t Roorda, Mattew J. 2017. Fully Autonomous Vehicles: Analyzing transportation network performance and operating scenarios in the
Geater T oronto Area. Transportation Research Board 96th Annual Meeting. 2017.
KPMG e t CAR. 2012. Self-Driving Cars: The Next Revolution. Ann Arbor : s.n., 2012.
Krueger, Rico, Rashidi, Taha et Rose, John. 2016. Adoption of Shared Autonomous Vehicles - A Hybrid Choice Modeling Approach based on a Stated
Choice Survey. Transportation Reearch Board 95th Annual Meeting. 2016.
Lavasani, Mohammad, e t al. 2017. Investigating the Willingness to Pay for Autonomous Vehicles and the Likelihood of Residential Relocation. Transportation
Research Board 96th Annual Meeting. 2017.
Lavasani, Mohammad, Jin, Xia e t Du, Yiman. 2016. Market Penetration Model for Autonomous Vehicles Based on Previous T echnology Adoption
Experiences. Proceeding of the 95th annual meeting of Transportation Research Board. 2016.
Levin, Michael W. et Boyles, Stephen D. 2015a. Effects of autonomous vehicle ownership on trips, mode and route choice. Proceeding of the 94th annual
meeting of the Transportation research board. 2015a.
Levin, Michael William. 2015b. Integrating Autonomous Vehicle Behavior into Planning Models. Austin : T he University of Texas, 2015b. Thesis.
Levin, Michael, et al. 2016. A general framework for modeling shared autonomous vehicles. Proceeding of the 95th annual meeting of Transportation Research
Board. 2016.
Li , Tianxin et Kockelman, Kara. 2016. Valuing the Safety Benefits of Connected and Automated Vehicle T echnologies. Transportation Research Board TRB
95th Annual Meeting. 2016.
Lima Azevedo, Carlos, et al. 2015. Microsimulation of Demand and Supply Autonomous Mobility On-Demand. Proceeding of the 94th annual meeting of
Transportation Research Board. 2015.
Litman, Todd. 2015. Autonomous Vehicle Implementation Predictions: Implications for Transport Planning. s.l. : Victoria Transport Policy Institute, 2015.
Litman, Todd. 2012. Parking Management: Strategies, Evaluation and Planning. Victoria : Victoria T ransport Policy Institute, 2012.
NHTSA. 2012. Fatal Analysis Reporting System. Washington D.C : U.S. Department of T ransportation, 2012.
NHTSA. 2008. National Motor Vehicle Crash Causation Survey. s.l. : U.S. Department of Transportation, 2008.
O NISR. 2016. Bilan définitif de l’accidentalité routière 2015. 2016.
Piao, J. e t McDonald, M. 2005. An assessment of user acceptance of intelligent speed adaptation systems. Proc. Intelligent Transportation Systems. 2005.
Power, J.D. 2012. Vehicle Owners Show Willingness to Spend on Automotive Infotainment Features. Westlake Village : s.n., 2012. Technical Report.
Rödel, Christina, et al. 2014. Towards Autonomous Cars: The Effect of Autonomy Levels on Acceptance and User Experience. Proceedings of the 6th
International Conference on Automotive User Interfaces and Interactive Vehicular Applications. AutomotiveUI ’14, 2014, Vol. 11.
Schoettle, Brandon et Sivak, Michael. 2014. A survey of public opinion about autonomous and self-driving vehicles in the U.S., the U.K., and Australia. s.l. :
T he University of Michigan T ransportation Research Institute, 2014. Technical Report.
Schoettle, Brandon et Sivak, Michael. 2015. Motorists’ preferences for different levels of vehicle automation. s.l. : T he University of Michigan Transportation
Research Institute, 2015. Technical Report.
Spieser, Kevin, et al. 2014. Toward a Systematic Approach to the Design and Evaluation of Automated Mobility-on-Demand Systems: A Case Study in
Singapore. [éd.] Springer. Road Vehicle Automation. Septembre 2014.
The Milwaukee Sentinel. 1926. Phantom Auto' will tour city. [En ligne] 8 Decembre 1926. [Citation : 23 Juillet 2013.]
[Link]
The Victoria Advocate. 1957. Power companies build for your new electric living. [En ligne] 24 Mars 1957.
[Link]
Vakayil, Akhil e t Samaranayake, Samitha. 2017. Innovative Applications and Evaluations of Ridesharing and Ridesourcing Services. Transportation
Research Board TRB 96th Annual Meeting. 2017.
Weiner, Gabriel et Walker, Smith. 2016. Automated Driving: Legislative and Regulatory Action. [En ligne] 2016.
[Link]/wiki/[Link]/Automated_Driving:_Legislative_and_Regulatory_Action.
Yu, Shen, Zhang, Hongmou et Jinhua, Zhao. 2017. Embedding Autonomous Vehicle Sharing in a Public T ransit System: Example of Last-Mile Problem.
Transportation Research Board TRB 96th Annual Meeting. 2017.
Zhang, Wenwen et Guhathakurta, Subhrajit. 2017. Parking Spaces in the Age of Shared Autonomous Vehicles: How Much Parking Will We Need and
Where? Transportation Research Board TRB 96th Annual Meeting. 2017.
Zhang, Wenwen, e t al. 2015b. Exploring the Impact of Shared Autonomous Vehicles on Urban Parking Demand: An Agent-Based Simulation Approach.
Cambridge, Massachusetts, United States : s.n., 2015b.
Zhang, Wenwen, e t al. 2015a. The performance and benefits of a shared autonomous vehicles based dynamic ridesharing system: an agent. Washington DC,
United States : s.n., 2015a.
Zhao, Yong e t Kockelman, Kara. 2017. Anticipating T he Regional Impacts of Connected and Automated Vehicle T ravel in Austin, Texas. Proceedings of the
96th annual meeting of Transportation Research Board. 2017.
Zhu, Shirley et Kornhauser, Alain. 2017. The interplay between fleet size, level-of-service and empty vehicle repositioning strategies in large-scale, shared-
ride autonomous taxi mobility-on-demand scenarios. Transportation Research Board 96th Annual Meeting. 2017.
Zmud , Johanna, Sener, Ipek et Wagner, Jason. 2016. Consumer Acceptance and Travel Behavior - Impacts of Automated Vehicles. s.l. : T exas A&M
T ransportation Institute, 2016. Technical Report.