Energy Policy 132 (2019) 462–473

Contents lists available at ScienceDirect

Energy Policy
journal homepage: [Link]/locate/enpol

Can autonomous vehicle reduce greenhouse gas emissions? A country-level T

evaluation
Feiqi Liua,b, Fuquan Zhaoa,b, Zongwei Liua,b, Han Haoa,b,c,∗
a
State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing, 100084, China
b
Tsinghua Automotive Strategy Research Institute, Tsinghua University, Beijing, 100084, China
c
China Automotive Energy Research Center, Tsinghua University, Beijing, 100084, China

A R T I C LE I N FO A B S T R A C T

Keywords: Autonomous vehicles are believed to be the next-generation technology for future societies. The energy and
Autonomous vehicle environmental impacts of autonomous vehicles have been realized to be important issues, but quantitative
Intelligent and connected vehicle analysis is lacking. In this study, by using China's passenger vehicle fleet as an example, we evaluate the effects of
Passenger vehicle autonomous vehicle deployment on greenhouse gas emissions in different scenarios of autonomous vehicle
Greenhouse gas
penetration rates and fuel consumption changes. A comprehensive literature review is conducted to support the
China
study. Autonomous vehicles are found to potentially affect the total greenhouse gas emissions in multiple ways,
including reducing vehicle ownership, increasing vehicle use intensity, and changing the vehicle fuel con-
sumption rate. These impacts are mostly internally offset such that the overall impact of autonomous vehicle
deployment on greenhouse gas emissions is not significant in the near-to mid-term. With a higher autonomous
vehicle penetration rate achieved, in the optimistic scenario, a net reduction in greenhouse gas emissions is
expected to be realized. In addition, the fuel economy levels of autonomous vehicles are highly uncertain and
cause major uncertainties in the simulation results. More field tests and evidence are needed to improve the
evaluation reliability.

1. Introduction (Xu and Fan, 2018). In addition, governments are enhancing their su-
pervision. Take China as an example. As of October 17th, 2018, eight
With more attention being paid to energy consumption levels and companies, including Baidu, NIO, BAIC, Daimler, [Link], Tencent,
greenhouse gas (GHG) emissions, governments worldwide have taken DiDi and Audi, had gotten driverless test licenses in Beijing, and ap-
various measures to reduce the environmental impacts of passenger proved by the local government for autonomous vehicle testing on se-
vehicle fleets (Hao et al., 2015a; Dennehy and Gallachoir, 2018; Martin lected open roads with a total distance of 123 km (Beijing innovation
et al., 2015). Light-weighting and introductions of fuel economy reg- center). Hangzhou's administration has distributed a test license to
ulations and alternative fuels are all considered effective ways to reduce Alibaba, while Shanghai's administration has issued test licenses to
the reliance on fossil fuels and eliminate GHG emissions (Das et al., SAIC, NIO, BMW China, and Momenta and a truck test license to Tu-
2016; Teixeira and Sodre, 2018; Wang et al., 2018a; Braun and Rid, simple (Shanghai Municipal).
2017). New business models have also emerged to meet people's travel There may be a dramatic disruption that is caused by autonomous
demands, such as Uber and DiDi (Hall et al., 2018; Guo et al., 2018). technology, especially if the technology becomes widely accepted. More
Driverless or autonomous vehicles (AVs) also provide a new solution for efficient road use, safer mobility, creation of a higher car value, no
mobility and are predicted to change personal vehicle ownership parking pressure, less time wasted, etc. will occur due to the AV benefits
(Lavieri et al., 2017; Bansal and Kockelman, 2017; Fagnant and (Fagnant and Kockelman, 2014; Urmson and Whittaker, 2008; Krueger
Kockelman, 2015). et al., 2016). Therefore, with more attention being paid to AVs, the
Many automakers and internet companies have joined the compe- energy, environmental and sustainability implications of AVs should
tition in the field of AV technology and have made strategic plans for also be analyzed and considered by researchers and policy makers.
future developments in related areas. Google, Ford, BMW and Daimler
have all announced their plans for future AV research and development

∗
Corresponding author. State Key Laboratory of Automotive Safety and Energy, Tsinghua University, Beijing, 100084, China.
E-mail address: hao@[Link] (H. Hao).

[Link]
Received 3 December 2018; Received in revised form 5 June 2019; Accepted 6 June 2019
0301-4215/ © 2019 Elsevier Ltd. All rights reserved.
F. Liu, et al. Energy Policy 132 (2019) 462–473

2. Literature review take AVs into consideration. Many studies have been conducted on this
topic (Yan and Crookes, 2010; Bandivadekar et al., 2008; Jochem et al.,
Limited studies have been conducted to evaluate the impacts of AVs 2015; Meyer and Wessely, 2009).
on vehicle fleet energy consumption levels or GHG emissions. All in all, the previous studies of U.S. cases mainly focused on four
Especially in China, there are few studies. Therefore, the literature re- aspects. First, some of the studies predicted the potential market of AVs
view focuses primarily on U.S. cases and contains certain discussions. from different points of view, such as customer preference, business
Chen et al. quantified the influence of introducing AVs on the U.S. light- model, etc. Second, certain studies analyzed the impacts of AV de-
duty vehicle fleet fuel consumption by integrating a vehicle stock and ployment on the vehicle fleet or traffic flow, including the vehicle stock
usage model with the national vehicle mile traveled (VMT) and speed and travel mileage, and provided a possible range of variations. Third,
database of the Federal Highway Administration (FHWA) (Chen et al., the influences of AVs on energy consumption were discussed from a
in press). Only gasoline consumption was included in the research, and gasoline perspective or an electricity perspective. Finally, changes in
electric vehicles (EVs) and fuel cell electric vehicles (FCEVs) were the individual car fuel consumption rate due to more functions in AVs
considered to be nonfuel-consuming. Additionally, aggressive assump- were also considered. Certain studies covered multiple aspects.
tions of the fuel economy change and rapid growth in the VMT were As mentioned before, China also shows great interest in the devel-
used as input data. Thus, there were wide-ranging results among the opment of AVs. Based on the planning provided by SAE-China for
different scenarios. AVs could lead to a 45% fuel consumption reduc- passenger cars, by 2019, partially AVs will be applied; by 2022, con-
tion in the optimistic scenario and a 30% increase in the pessimistic ditionally autonomous driving will be achieved and by 2025, fully
scenario. Fox-Penner et al. provided estimates and policy observations autonomous driving will be realized (SAE-China, 2016). Prior studies in
based on considering the impacts of AVs on the transportation elec- China have paid considerable attention to the development of specific
tricity use, using an expanded Kaya Identity framework with a three- technologies, such as autocalibration systems, pedestrian recognition
layer model (Fox-Penner et al., 2018). Their research emphasized the and tracking. (Zhu et al., 1808; Wang et al., 2017; Lin et al., 2017). Few
importance of EV deployment, combined with the introduction of AVs studies have analyzed the influence of AVs on energy consumption le-
and the car-sharing mode. Although the influences of AVs on the travel vels and GHG emissions.
distance, business mode, and energy efficiency were discussed, the Summarizing the research results, either only the influence of traffic
main research was still focused on EVs and electric sector dec- flow or the change of the vehicle itself was considered, or either the
arbonization. Brown et al. also used the Kaya Identity method to predict electricity consumption or fuel consumption was included. This paper
the possible effects of AVs (Brown et al., 2014). Vehicle level, fleet discusses the effects of AV deployment on the vehicle fleet energy
level, urban system level and society level effects were all discussed, consumption levels and GHG emissions in China, considering both ve-
while only the first two effects were considered in the final results. hicle and traffic level impacts and localizing certain data. The con-
Highly optimistic and highly pessimistic assumptions were input, and tribution of this study includes identifying a wide range of impacts
the final results varied from −95 to 173% compared with the baseline caused by AVs, focusing on both conventional fuel passenger cars and
scenario. The research indicated that widespread AV deployment could electric cars, and analyzing these impacts based on China. The next
lead to dramatic fuel savings when all factors were moving in the right section introduces the method used in this research, and the subsequent
direction. If not, there would also be a sharp increase in fuel con- section shows the data. The fourth section presents the results, and the
sumption. Wadud et al. also assessed the travel, energy and carbon final section summarizes the whole study.
impacts of a high market share of AVs (Wadud et al., 2016). The
findings showed that the influences on the energy and carbon emissions 3. Methods
were heavily dependent on which effects would come to dominate. The
energy intensity, travel demand effect and fuel mix change were con- The bottom-up method is widely used and applied to calculate the
sidered. Four scenarios were provided with different combinations of GHG emissions of the vehicle fleet based on the vehicle sales, survival
factors. The changes in energy consumption due to vehicle automation rates, annual travel distances, fuel economies and GHG emission in-
varied greatly, so the final outputs also showed uncertainties. In con- tensities in this article (Hao et al., 2015b, 2015c; Ou et al., 2010a; Yan
clusion, the authors emphasized the impact of the value of travel time and Crookes, 2009). Eq. (1) presents the calculation.
(VOTT) and the understanding of individual responses. Fagnant et al. i
summarized the opportunities, barriers and policy recommendations GHGi = ∑ ⎡⎢ ∑ ⎛ NAV
⎜Salesf , j × SRi − j × VMTi
NAV
× (FCRfNAV
,j × (1 − α )
for national AV development (Fagnant and Kockelman, 2014). The f
⎣ j=i−l ⎝
changes in vehicle and transportation systems were analyzed. The po-
× GIf × LHVf × ρf + PCjNAV × α × GIe ) +
tential impacts of AV deployment were analyzed. By conducting these
studies, estimates of the economic benefits from AVs in the U.S. were ∑ (SalesfAV
,j
,t
× SRi − j × VMTiAV , t × (FCRfAV
,j
,t
× (1 − α )
provided, including the travel distance increase and vehicle number t

change.
⎞⎤
Greenblatt et al. explored the potential importance of AVs on mo- × GIf × LHVf × ρf + PCjAV , t × α × GIe )) ⎟ ⎥ × 10−14
bility, energy use and environmental impacts with car-sharing syner- ⎠⎦ (1)
gies. By summarizing previous studies, the researchers noted that net where
decreases in energy consumption levels and GHG emissions were more
likely to occur, with at least no dramatic increases (Greenblatt and GHGi are the GHG emissions of the passenger vehicle fleet in target
Shaheen, 2015). The potential changes in energy savings would gain year i (mt CO2-eq.);
the upper hand, which was supported by another study. Gawron et al. l is the life span of the vehicle;
conducted their study from another perspective (Gawron et al., 2018). SalesfNAV
,j is the number of sales of non-AVs with fuel type f in year j
The life cycles, from cradle to grave, and GHG emissions of AVs were (unit);
assessed. Direct and indirect factors were both considered. The study SRi − j is the survival rate of the vehicle in the (i − j )th year (%);
provided a relatively comprehensive analysis of individual vehicle VMTiNAV is the annual travel distance of a non-AV in year i (km);
change. The basic case showed a 9% reduction in energy and GHG FCRfNAV,j is the fuel consumption rate of a non-AV with fuel type f
emissions in the vehicle whole life compared with those of a conven- produced in year j (L/100 km);
tional vehicle. There is no need to enumerate all of the discussions of α is the fraction of the travel distance that is powered by electricity
vehicle fleet energy consumption and GHG emission studies that did not

463
F. Liu, et al. Energy Policy 132 (2019) 462–473

(%);
GIf is the GHG emission intensity of fuel type f (g-CO2/MJ);
LHVf is the lower heating value of fuel type f (MJ/kg);
ρf is the density of fuel type f (kg/L);
PCjNAV is the power consumption of a non-AV produced in year j
(kWh/100 km);
GIe is the GHG emission intensity of electricity (g-CO2/kWh);
,t
SalesfAV
,j is the number of sales of type t AVs with fuel type f in year
j (unit);
VMTiAV , t is the annual travel distance of a type t AV in year i (km);
,t
FCRfAV,j is the fuel consumption rate of a type t AV with fuel type f
produced in year j (L/100 km); and
PCjAV , t is the power consumption of a type t AV produced in year j
(kWh/100 km).

4. Data

4.1. Sales and market share

Unlike EVs, the development of AVs is still indefinite. Uncertainties

occur when only a few brands have received road test licenses in only
certain cities and no mass production of any cars has occurred.
Although Tesla announced that their AutoPilot system can free drivers,
accidents still worry consumers. Most of the studies that focus on the
acceptance or adoption of AVs are analyzed on two sides: the demand
side and the supply side. For the demand side, the willingness to pay is
highly related to age, driving habits, car-sharing experience, gender,
etc., and plays an essential role in the consumer choice (Krueger et al.,
2016; Rödel et al., 2014; Daziano et al., 2017; Schoettle and Sivak,
2014). In addition, more travel demand might be generated because of
underserved people's travel requirements, convenience travels, lower
travel costs and lower VOTTs. However, the supply side is more de-
pendent on the development of technology than on customers. Several
studies have been performed to predict the market share in sales or in
stocks of fully or partially AVs in the future, and their results vary
Fig. 1. Fully and partially AV market shares of total new car sales (the dotted
substantially from each other (Chen et al., in press; Fox-Penner et al., lines are trend lines).
2018; Litman, 2018; Cliff, 2013; Mosquet et al., 2015; McKinsey&
Company, 2016; Laslau et al., 2014). In this article, fully AVs include
level 4 or 5 cars based on SAE international standards, and levels 1 to 3 travel demand can be satisfied with lower ownership for both house-
are collectively called partially AVs (SAE International). Fig. 1 sum- hold and public sharing uses. Private cars can be removed in two ways:
marizes the forecasts from different teams, including consulting firms sold/shed (selling existing cars) and suppressed/avoided (delaying or
and researchers. In all predictions, fully AVs will enjoy a consistent giving up plans for car purchasing). In this way, AVs would lead to a
growth, and the share of partially AVs may increase first and then de- lower ownership of private vehicles. There may be a time delay when
crease. Most of the studies are based on the U.S. case (Chen et al., in considering the impact of sold or shed vehicles. The suppression effect
press; Fox-Penner et al., 2018; Litman, 2018; Cliff, 2013). Certain stu- will show a larger impact than the shedding effects. In China, car
dies forecast the global trend (Mosquet et al., 2015; McKinsey& ownership often represents a kind of wealth or social status, and be-
Company, 2016; Laslau et al., 2014). However, few studies provide comes more noteworthy as the income grows (He et al., 2010; Cao,
predictions of China's case. 2015). Therefore, it may take a longer time for people to get rid of their
Based on existing studies, four different scenarios are provided in own cars and postpone their purchase plans. The impact of AVs on car
this research, as shown in Fig. 2. (a) Scenario 1 and (b) scenario 2 are ownership will be weakened. Fagnant et al. and Schoettle et al. studied
mainly based on the research of the Boston Consulting Group (BCG). the influence of AVs on the fleet size, as shown in Fig. 3 (Fagnant and
The forecast given by the BCG stops in 2035. Therefore, two scenarios Kockelman, 2015; Schoettle and Sivak, 2015). Their predictions were
are used. In scenario (a), partially AVs will gradually be withdrawn very optimistic and not aligned with reality. The researchers ignored
from the market, and fully AVs will finally dominate. In scenario (b), the customer choice, which would play an essential role.
after relatively slow growth, partially AVs will achieve sustained and Eqs. (2)–(5) show the relationships between the market share and
stable development when fully AVs keep pace with scenario (a). In the change in the total vehicles.
third scenario (scenario 3), there will be a rapid growth for AVs. The VSA = VSAR × (1 − δ %) (2)
market share can even reach almost 90%, and partially AVs only ac-
count for a small part. In the fourth scenario (scenario 4), no AVs will VSA
A %=
be introduced in the market, including partially and fully AVs. VSA + VSNA (3)
1
VSbefore = VSNA + VSAR = VSNA + VSA ×
4.2. Change in the vehicle stock 1 − δ% (4)

VSbefore − VSafter δ % × VSA δ % × A%

Partially AVs will not significantly change vehicle utilization, so the Δ% = = =
AVs mentioned in this part only refer to fully AVs. With the deployment VSbefore (1 − δ %) × VSNA + VSA 1 − (1 − A%) × δ %
of AVs, more of the travel demand can be met by sharing AVs. The same (5)

464
F. Liu, et al. Energy Policy 132 (2019) 462–473

Fig. 3. Change in the total vehicles as a function of the market share of AVs
(summary of research).

VSafter is the vehicle stock after introducing AVs (unit); and

Δ% is the change rate of the total vehicle stock.

Two previous studies assumed δ% as approximately 50%, which

means that one fully AV could replace approximately two conventional
cars. This assumption is theoretically feasible, but the operating con-
ditions are not easy to realize. For example, online car-hailing services,
such as DiDi in China, provide convenient and more inexpensive ser-
vices for mobility, but people still desire to have their own cars. One
telling truth is the number of people who participate in Beijing's plate
lottery, which maintains a sustained growth. Therefore, in this study,
δ% is assumed to be only 17%, a conservative estimation.
The substitution effect of fully AVs is considered in vehicle sales, as
shown in Eq. (6). The basic assumption of vehicle sales is consistent
with the business as a usual scenario in the study of Liu et al. (2018).
1
TSafter = TSbefore × ⎡1 − MSAV , fully × ⎛ ⎞ − 1⎤
⎣ ⎝ 1 − δ% ⎠ ⎦ (6)
where

TSafter are the total vehicle sales after correction of the substitution
effect (unit);
TSbefore are the total vehicle sales before correction of the substitution
effect (unit); and
MSAV , fully is the market share of fully AVs.

4.3. Survival rate

As for traditional cars, the survival rate is a common topic (Ou et al.,
2010a; Yan and Crookes, 2009; Hao et al., 2011; Zhao et al., 2019). The
VMT mentioned next shows that the vehicle use intensity does not in-
crease significantly for AVs. Thus, the research results of Zhao et al. are
applied in this study for both conventional cars and AVs (Zhao et al.,
2019).

4.4. VMT per vehicle

Many studies have discussed the impact of AVs on the vehicle travel
distance. Positive and negative effects both exist, as shown in Table 1.
Based on Harper's study, the VMT can increase up to 14% due to
Fig. 2. Scenario assumptions of the AV market shares of total new car sales. nondrivers, the elderly and people with travel-restrictive medical con-
ditions (Harper et al., 2016a). Wadud et al. reached a similar conclusion
that underserved people can lead to an increase of 2–10% in the overall
where personal vehicle travel distance (Wadud et al., 2016). Brown's research
gave a higher result: a 40% growth due to underserved people and 50%
VSA is the vehicle stock of fully AVs (unit); growth from a higher travel demand (Brown et al., 2014). Meanwhile, a
VSAR is the non-AV stock that is replaced by fully AVs (unit); consequent change in occupancy, car sharing, would lead to a decrease
δ% is the change rate; of 12%. Childress et al. analyzed the impacts of the capacity, perceived
A% is the market share of fully AVs; travel time, parking cost and operating cost on the travel demand
VSNA is the vehicle stock of non-AVs (unit); (Childress et al., 2015). The total VMT could change from
VSbefore is the vehicle stock if no AV is introduced (unit); −35.4–19.6% among combinations of the different factors.

465
F. Liu, et al. Energy Policy 132 (2019) 462–473

Table 1
Positive and negative effects on the travel demand/distance.
Increase travel demand/distance Decrease travel demand/distance

✓ Underserved group of people ✓ Reduced private car ownership

Elderly people, young people, individuals without a driver's license, individuals with a medical ✓ Higher cost
condition, etc. Higher purchase costs lead to higher operating fees for shared cars,
✓ More travel and a mode shift VMT-based taxes, and pay-as-you-drive insurance costs
Higher car productivity, higher efficiency (less congestion, higher speeds, fewer accidents, etc.), ✓ Less hunting for parking
and lower maintenance and operating costs (lower insurance fees, higher fuel economy, no
payment to the driver, shared economy, etc.)
✓ Living farther from the destination
✓ Unoccupied travel distance

Certain researchers also evaluated the substitution effect of AVs

when no new travel demand was induced. Because of fewer cars, the
VMT would also go up. Zhang et al. performed simulations in Atlanta,
the U.S. (Zhang et al., 2018). The results showed that over 9.5% of
private vehicles can be replaced. The total VMT would increase by at
least 13%. Fagnant et al. also paid attention to the U.S. (Fagnant and
Kockelman, 2014). Considering a 10% unoccupied distance, one shared
AV could replace 11 conventional vehicles. Boschoff et al. demon-
strated that 1.1 million private cars could be replaced by 90–110
thousand autonomous taxis in Berlin, Germany (Bischoff and
Maciejewski, 2016). Spieser's research indicated that if all vehicles
operated autonomously, only 1/3 of the total number of passenger
vehicles would be needed in Singapore (Spieser et al., 2014). Fig. 5. VMT increase as a function of the market share in stocks of AVs.
Fox-Penner et al. assumed the VMT effects of AVs based on time-
lines, as shown in Fig. 4. The impact of AVs on the VMT would phase 4.5. Fuel consumption rate
linearly from 2025 or 2030 on, which is due to different scenario as-
sumptions. Although the data in this research are continuous, it is not Vehicle automation may lead to changes in the fuel economy in
reliable to adopt assumptions that rely on timelines. various ways, as shown in Fig. 6. AVs can become both a fuel economy-
The study of Fagnant et al. is applied in this research (Fagnant and additive or fuel economy-subtractive force, and certain impacts are still
Kockelman, 2015). Fagnant et al. provided a series of data about the not clear. The forces are divided into two categories in this study: the
annual benefits of AVs. When the market shares reach 10%, 50% and vehicle itself and traffic influence. The branch of the vehicle itself
90%, the VMT will be 20%, 15% and 10% higher, respectively, on consists of the changes in vehicle size, weight, shape, accessories, etc.,
account of Fagnant's research. The relationship between the market which are related to the alterations to individual cars. However, the
share and VMT increase is fitted based on these data. These three points traffic influence is based to a greater extent on the overall impact of the
follow a linear relationship. When there are no AVs, the increase in driving fleet. The following section will introduce each of the factors to
VMT should also be 0%. Thus, the VMT effect first increases with the conduct a comprehensive analysis.
market share. If a linear fitting curve is applied, there will be a turning
point, which means a sudden change that does not tie in with the facts
(the dotted line in Fig. 5). Thus, the curve is refitted, as shown in Fig. 5. 4.5.1. Eco-driving
With more AV deployment, more cars can flexibly meet people's travel Eco-driving can have fuel economy benefits in many ways, in-
needs. Therefore, to a certain extent, the unoccupied travel distance cluding moderate acceleration or deceleration, maintaining an even
will be avoided, which leads to a decline in the VMT change. The es- driving speed, eliminating excessive idling, etc. (Barkenbus, 2010).
timation of Hao et al. is regarded as the baseline for the VMT (Hao Fig. 7 shows recent research on the eco-driving effect on the fuel
et al., 2017a). economy (Barth and Boriboonsomsin, 2009; Wu et al., 2011; Li and
Gao, 2013; Sivak and Schoettle, 2012; Beusen et al., 2009; Barla et al.,
2017; Rolim et al., 2014; Zarkadoula et al., 2007). The results vary
greatly from each other, but the average value is approximately 5%.
Wadud et al. divided the impacts of eco-driving applied in AVs into

Fig. 4. VMT increase as a function of the timeline.

466
F. Liu, et al. Energy Policy 132 (2019) 462–473

Fig. 6. Potential fuel economy impacts of AV development and deployment.

two ways: driving practice by training drivers and optimizing the and for those who choose other modes of travel, the value is lower. In
driving cycle (Wadud et al., 2016). However, based on their opinions, addition, as mentioned before, freeway driving only accounts for
the system-wide fuel consumption might be offset by the congestion 20.74% in China, which is much lower than the 33–55% given in
generated by a higher travel demand and more convenient travel. The Wadud's article. In addition, under partial automation, drivers may not
fuel consumption might even increase. Stephens et al. also provided the exceed the existing speed limits or simply speed up slightly due to
assumption of fuel savings due to the eco-driving performance of par- safety concerns. All things considered, there will be a 0–3% fuel con-
tially and fully AVs (Stephens et al., 2016). Different scenarios were sumption increase for partially AVs and a 3–10% increase for fully AVs.
designed, including peak hours, off-peak hours, highways and city
conditions. Based on the research of Wang et al., with different city
sizes considered, the peak hours were assumed to account for 36.36% 4.5.3. Smoothing the traffic flow
and freeway travels accounted for 20.74% of the total traffic in China Various factors will make traffic smoother. The mitigation of traffic
(Wang et al., 2018b). When the results are applied to the research of congestion will be a means of effectively improving the traffic flow
Stephens et al. with regards to partially AVs, the range of fuel savings is caused by AV deployment, and intersection connectivity will be one of
0–6.8%, and for fully AVs, the range is 6.8–15.3%. The results are very the most important solutions. Li et al. reached the conclusion that
close to those of previous studies, albeit slightly higher. Because it is connected vehicles with communication at signalized intersections had
easier and more precise to apply eco-driving to AVs instead of drivers, great potentials to save fuel only under accurate control conditions, so
the estimation is considered reliable and treated as input in this re- an automatic control system would show a better performance than
search. human drivers (Li et al., 2015). Kamal et al. provided an intersection
control system for fuel economy improvement by measuring the re-
levant information of vehicles (Kamal et al., 2013). By using this
4.5.2. High speed
method, there would be a 12.98% fuel economy improvement. Wu et al.
With improved reliability levels, AVs may lead to higher travel
noted the limitations of human-machine interfaces (HMIs) and com-
speeds to provide a more convenient and faster travel experience for
pared the interfaces with partially AVs (with the use of an eco-approach
passengers. However, higher speeds will cause larger aerodynamic
and departure (EAD) application) in terms of the fuel consumption rate
losses and lead to higher fuel consumption levels correspondingly.
and CO2, CO and HC emissions (Wu et al., 2014). On average, a 5–7%
Wadud et al. proposed a very insightful point of view that the potential
reduction in fuel consumption would be reached. Hao et al. evaluated
for travel speed increase was highly related to the marginal VOTT
the impact of the EAD application in real-world testing and gave a re-
(Wadud et al., 2016). Wadud's research was based on the U.S. case and
sult of 2% fuel savings (Hao et al., 2017b). With the explosion of the
reached results where there would be an increase of 7–22% in the fuel
travel demand triggered by AVs, it is also hard to predict whether there
economy of light-duty vehicles brought on by the higher speed. How-
will be more or less congestion (Wadud et al., 2016). Yelchuru et al.
ever, China is still a developing country, so the VOTT may not be as
demonstrated that fully AVs could result in a 5% improvement in fuel
high as that in the U.S. Jiang et al. provided the VOTT in Tianjin, China,
consumption and that partially AVs could cause 1–4% in a fully co-
in 2009, and Zhao et al. calculated the VOTT in Changchun, China, in
ordinated network (Yelchuru and Waller, 2014). Thus, a moderate as-
2011 (Jiang et al., 2009; Zhao and Zhao, 2011). The VOTTs in these two
sumption is applied: 0–2% for partially AVs and 2–4% for fully AVs.
places are 15.29 RMB/h and 16.99 RMB/h for car users, respectively,

Fig. 7. The impact of eco-driving on the fuel economy.

467
F. Liu, et al. Energy Policy 132 (2019) 462–473

4.5.4. Crash avoidance [Link]. Fewer safety attributes. By introducing AVs, the weight can also
With the improvement of safety, accidents attributed to human er- be reduced without sacrificing safety. Higher safety can be achieved
rors can be prevented from occurring (Jermakian, 2011; Jeong et al., through information exchange and software control. The importance of
2017). Most studies have focused on the economic impact of crash passive safety will be compensated by an active safety system. Thus,
avoidance (Fagnant and Kockelman, 2015; Harper et al., 2016b). Thus, certain safety-related attributes can be removed, leading to a mass
the assumptions made by Stephens et al. are used in this paper reduction and fuel savings. Studies by Mackenzie et al. and Wadud et al.
(Stephens et al., 2016). According to their research, because fewer are used as references in this study, as shown in Fig. 8 (with a green
collisions will lead to less traffic congestion and lower fuel consumption background) (Wadud et al., 2016; MacKenzie, 2013; MacKenzie et al.,
levels, the operation of partially AVs can reduce fuel consumption by 2014). Drivers will still master the controls and take responsibility of
0–0.95%, and fully AVs can reduce fuel consumption by 0–1.9%. partially AVs, so nothing will be changed. While in fact, it may take a
long time for passengers to accept the fact that AVs have fewer safety
protection functions in the vehicles.
4.5.5. Platooning
Platooning will lead to a higher fuel economy improvement due to
the reduction in aerodynamic drag, especially combined with autono-
[Link]. More supporting attributes. Few studies have considered the
mous driving (Tsugawa et al., 2016; Alam et al., 2015; Deng and Ma,
impact of more supporting attributes as the result of autonomous
2015; Vahidi and Sciarretta, 2018). However, studies are mainly con-
driving. Therefore, one of the latest studies, from Gawron et al., is
centrated on heavy-duty vehicles instead of passenger vehicles. Because
applied in this part (Gawron et al., 2018). Level 4 AV sensing and
of insufficient data support, in this research, the impact of platooning
computing subsystems were discussed in their study. A large subsystem
on the passenger vehicle fuel economy is not considered. However, it
is closer to reality, so the system is used as a basic model for the fully
must be acknowledged that platooning is an important factor.
AVs in this article. An additional 55 kg mass (accounting for
approximately 4% of the total weight) will be added to the existing
4.5.6. Weight change system, which comes from cameras, sonars, radars, light detection and
With changes in the family structure and the growth of the con- ranging (LiDAR) technology, global positioning system/inertial
sumption level, Chinese consumers tend to purchase larger and higher- navigation system, dedicated short range communication system,
class vehicles (Zhao et al., 2019). Consumer behaviors put a great strain computer, wire harness and structure for both battery electric
on energy savings in the auto industry. The introduction of AVs pro- vehicles (BEVs) and internal combustion engine vehicle (ICEVs).
vides a new solution to this problem but also causes growth of the car Combining the estimations from Burns et al. and MacKenzie D., there
weight in certain aspects. Three aspects are discussed in detail in this will be a 2.6% fuel economy improvement, as shown in Fig. 8 (with a
paper, as shown and summarized in Fig. 8. blue background) (MacKenzie, 2013; Burns et al., 2012). For partially
AVs, the number of supporting attributes will decrease, and certain
parts will be changed into lower-precision and lower-cost parts. Gawron
[Link]. Rightsizing. Morrow et al. defined rightsizing as “a concept of et al. calculated a small subsystem weight of 17 kg. The value is used as
vehicles designs that are appropriate for the service they provide” (Morrow the mass change for partially AVs in this study.
et al., 2014). With shared AV deployment, more vehicles will be
provided based on occupancy need (Greenblatt and Saxena, 2015).
Wadud et al. proposed two ways to provide services depending on [Link]. Other factors related to the weight change. It was mentioned in a
demand by AVs, which both can be included as rightsizing (Wadud report from the Energy Information Administration (EIA) that
et al., 2016). One was that trips were met with current-available integrating additional comfort and entertainment features into
vehicles, but according to passenger numbers, the other was vehicles to increase the experience of passengers might also lead to a
introducing few-passenger vehicles. Thus, there would not be such a mass increase and thus a lower fuel economy. In addition, for fully AVs,
large demand for sedans. One-seat and two-seat vehicles would become user-operated driving-related equipment can be removed in the future
popular for cost-effective concern. Several studies have evaluated this (EIA, 2017). However, the former is more dependent on the customer
impact, as shown in Fig. 8 (with a yellow background) (Brown et al., choice and preference, and the latter can only be achieved with high
2014; Wadud et al., 2016; Greenblatt and Saxena, 2015). Because it is penetration and reliability of fully AVs. The influences of these two
impossible to promote all AVs as small or proper ones, we factors are not yet quantifiable. Therefore, the factors are just
conservatively assume that the penetration of demand-based AVs will mentioned, not included in this study.
be 5%. For partially AVs, there will be no change.

Fig. 8. Change in the fuel consumption rate caused by the weight.

468
F. Liu, et al. Energy Policy 132 (2019) 462–473

4.5.7. Supporting attribute energy consumption between the New European Driving Cycle (NEDC) test and real-world
A higher electricity demand will be caused by the supporting at- conditions, etc. are all included based on previous studies (Liu et al.,
tributes in AVs. According to Gawron et al., considering the density, 2018; Zhao et al., 2019).
heat value and conversion efficiency from gasoline to electricity, the
electricity demands amount to approximately 0.58 kWh/100 km and 5. Results
0.99 kWh/100 km for BEVs and 0.23 L/100 km and 0.38 L/100 km for
ICEVs, respectively (Gawron et al., 2018; Ou et al., 2010b). Fig. 10 exhibits the market share of AVs in the passenger vehicle
Data transfer will also lead to an extra burden on energy. Levinson fleet stock. Over time, the penetration of AVs reaches a relatively high
et al. estimated that the grid images required approximately 10 Mb/ level in scenario 3. By 2050, the AVs in scenarios 1 to 3 in the whole
mile, which is equivalent to 6.25 Mb/km (Levinson et al., 2007). In fleet will reach 42.1%, 52.9% and 71.9%, respectively. The shapes of
Dominguez's tests, the experimental average data flow of the map ap- the stock curves basically coincide with the sales. The fully AVs in
plication was approximately 5 Mb/km (Dominguez et al., 2015). The scenarios 1 and 2 grow with the same trend. Fig. 11 exhibits the change
energy needed for data transfer varies greatly with different technolo- in the total passenger vehicle fleet VMT based on the assumptions in
gies. Angiolini et al. presented results of 2.72–3.77 MJ/GB (Angiolini this study. Because the basic assumptions of ownership and VMT
et al., 2006). According to Abdelfattach et al., in their most energy- change in this study are not aggressive, there are no significant differ-
efficient case, only 4.5 MJ/GB was required for transferring data ences among the scenarios.
(Abdelfattah and Betz, 2013). Alignment of the data flow and energy As shown in Fig. 12, overall, the introduction of AVs will not sig-
intensity, which is approximately 2.20 MJ per 100 km of driving, is nificantly change passenger vehicle fleet GHG emissions. First, because
needed. Combined with the conversion efficiency from gasoline to the proliferation of AVs occurs relatively late, it will not have a sig-
electricity, the result is 0.61 kWh/100 km for BEVs and 0.35 L/100 km nificant influence on the peak time of the passenger vehicle fleet GHG
for ICEVs. We assume that there is no difference between partially AVs emissions.
and fully AVs in this part. There is little difference between scenarios 1 and 2, regardless of the
fuel economy upper bound or lower bound conditions. This finding
4.5.8. Aerodynamic drag means that partially AV deployment will not have a large impact on the
The exterior changes caused by the supporting components will also final results. Scenario 3, which has the highest AV penetration, ex-
increase the aerodynamic drag, which will then cause additional energy hibited both the best and worst GHG emissions. In 2045, scenario 3,
consumption. Until now, no empirical data has been available to which has a better fuel economy, even reaches a lower result than that
evaluate this impact. Taiebat et al. noted that shape alteration could be of scenario 4, which has no AV deployment. This result means that the
treated using roof racks (Taiebat et al., 2018). Chen et al. estimated that introduction of AVs will gradually become an effective way to reduce
a roof rack could increase a passenger car's fuel consumption rate by the GHG emissions of passenger vehicle fleets. Under the upper bound
0–25% (Chen and Meier, 2016). Thomas et al. reported that using a roof conditions, the GHG emissions of scenario 3 are significantly higher
rack at 80 mph could reduce the fuel economy by almost 27% (Thomas than those of all other scenarios.
et al., 2014). However, cars will not always run at such high speeds, and Under the fuel economy upper bound conditions, the curves of
the aerodynamic drag increase caused by the supporting components scenarios 1 to 3 almost overlap with the no AVs scenario. This finding
will be smaller than that caused by personal items, such as bicycles. A shows that the improvement of the fuel consumption rate compensates
study by Gawron et al. predicted the results of AVs (Gawron et al., for the increase in the vehicle driving distance. In addition, as the
2018). For partially AVs, no apparent change will occur of the exterior. growth of the VMT decreases, the fully AV begins to show its ad-
Thus, there is no new demand for energy. For fully AVs, a small roof vantages in fuel economy improvement, as mentioned before. Under
rack will cause a 0.5% increase in the fuel consumption. With the im- the upper bound conditions, all scenarios show higher results. As the
provement in design, future supporting components may be integrated growth of AVs increases in scenarios 1 and 2, the gaps between the
into the car body to reduce their impacts on the fuel consumption, but three scenarios, under the fuel economy lower bound conditions, gra-
the integration is still uncertain. dually narrow, which is partially due to the decline in the VMT growth
after the peak. The major differences are still due to the fuel con-
4.5.9. Other factors related to the fuel consumption rate sumption rate variations.
It is impossible to perform large-scale field tests of many AVs under Fig. 13 shows the sensitivity analysis of the impacts of the changes
actual road conditions at this stage. Therefore, there are certain un- in the fully AV VMTs and AV fuel consumption rates on the fleet GHG
certainties between the traffic simulations and reality. For example, as emissions. The solid lines represent the sensitivities of VMT of fully
shown in Fig. 6, it is difficult to predict whether there will be more or AVs. The dotted lines represent the sensitivities of the vehicle fuel
less congestion. The outcome depends on which of the following will be consumption. The changes in VMT and vehicle fuel consumption ex-
proactive, the surge in travel demand or traffic optimization. In addi- hibit larger long-term impact than short-term. The results also indicate
tion, as the number increases, the coordination among AVs will lead to that the GHG emissions are more sensitive to the AV fuel economy.
a more complicated problem and may cause a decrease in fuel savings. However, the gaps between the changes caused by the VMT and fuel
In this research, these factors are not considered, but the framework can economy gradually narrow. Thus, controlling the higher energy con-
output more accurate results once better data are accessible. sumption levels generated by AVs would be a key factor.

4.5.10. Summary of the fuel consumption rate impacts 6. Conclusion

Fig. 9 summarizes the final results regarding the percentage change
in fuel economy compared with conventional passenger cars. The cur- As more companies join the AV technology competition, AVs are
rent fuel consumption rates of ICEVs and BEVs are used as baselines to considered a new way to change people's lives and mobility choices.
evaluate the impacts. The upper bounds show the best energy saving Meanwhile, the energy consumption levels and GHG emissions of pas-
scenario, and the lower bounds represent the worst scenario. senger vehicles have drawn much attention. AVs will bring changes in
vehicle ownership, travel distance, fuel economy, etc. Though AVs will
4.6. Other relevant data provide a convenient mobility choice for people, as one fully AV may
replace several cars, it will still take a long time to eliminate car
Basic fuel consumption rates, discharge/charge efficiencies, share of ownership, especially in China. Therefore, a conservative prediction is
electricity-powered plug-in hybrid electric vehicles (PHEVs), difference assumed in this article. The assumptions of the changes in vehicle sales

469
F. Liu, et al. Energy Policy 132 (2019) 462–473

Fig. 9. The summary of impacts on the fuel economy.

Fig. 10. Market share of AVs and fully AVs in stocks.

Fig. 11. Total VMTs of the passenger vehicle fleets in the different scenarios.

470
F. Liu, et al. Energy Policy 132 (2019) 462–473

Fig. 12. GHG emissions of the whole passenger vehicle fleet.

and travel distances are not aggressive. The travel distance per vehicle As more AVs are applied on roads, the government should pay more
and fuel economy will play important roles in the final results. attention to this new fleet. Based on the sensitivity analysis, the pas-
As for the fuel economy, there are numerous influencing factors at senger vehicle fleet GHG emissions are more sensitive to the fuel
both the vehicle and fleet levels, even from an urban planning per- economy. In the short term, due to the low deployment, AVs will not
spective. Due to the lack of a test environment, it is almost impossible to have significant impacts on passenger vehicle fleet GHG emissions.
evaluate the impacts accurately under the present circumstances. In Thus, when there are only a few AVs, policy makers should encourage
addition, few studies have tested the fuel consumption rates of AVs. technology development more actively to improve the fuel economy.
Thus, significant uncertainties exist in assessing the impact of AVs on Meanwhile, policy makers should also focus on the data collection of
energy demand and GHG emissions. In this study, upper bounds and AV fuel consumption rates. The stages of technology and actual data are
lower bounds are provided to make the results more reliable. A more supposed to be used as references for enacting or updating policies and
precise output may come with test or real-world input data. regulations. Neither policy makers nor researchers should overlook the
Based on the existing research, passenger vehicle fleet GHG emis- possible emissions of AVs. More studies should be focused on this topic.
sions are calculated in China with AV deployment. By aligning the In addition, the energy consumption of the vehicle manufacturing
changes caused by AVs and passenger vehicle fleet GHG emissions, the process will also be affected by more supporting components and fewer
results indicate that the introduction of AVs does not lead to GHG safety protection attributes. Evaluating this impact will be an inter-
emission reductions before 2050. With higher fuel consumption rates, esting topic for future studies. The substitution effects of AVs should
AVs may even lead to more emissions in most cases. Although the in- also be recorded with more practical data. Moreover, the uncertainties
crease is not large in absolute terms, in terms of the relative magnitude, of ownership and VMT still exist. Therefore, it is essential that further
the increase can even reach 14.1%. With a better fuel economy, after surveys are conducted.
2045, a larger share of AVs in the passenger vehicle fleet can begin to
show the advantages of AVs in GHG emission reduction. A long-term Acknowledgement
plan for AVs may have a better result. Passenger vehicle fleets with
different deployments of partially AVs will not show significantly dif- This study is sponsored by the National Natural Science Foundation
ferent results. of China (71774100, 71690241, U1764265), Young Elite Scientists

Fig. 13. Sensitivity analysis of the changes in the fully AV VMTs, AV fuel consumption rates and GHG emissions.

471
F. Liu, et al. Energy Policy 132 (2019) 462–473

Sponsorship Program by CAST (YESS20160140), State Key Laboratory Greenblatt, J.B., Shaheen, S., 2015. Automated vehicles, on-demand mobility, and en-
of Automotive Safety and Energy (ZZ2019-023), Tsinghua-Rio Tinto vironmental impacts. Current Sustainable Renewable Energy Reports 2, 74–81.
Guo, Y., Xin, F., Barnes, S.J., Li, X., 2018. Opportunities or threats: the rise of Online
Research Center for Resources Energy and Sustainable Development. Collaborative Consumption (OCC) and its impact on new car sales. Electron. Commer.
Res. Appl. 29, 133–141.
References Hall, J.D., Palsson, C., Price, J., 2018. Is Uber a substitute ot complement for public
transit? J. Urban Econ. 108, 36–50.
Hao, H., Wang, H.W., Ouyang, M.G., Cheng, F., 2011. Vehicle survival patterns in China.
Abdelfattah, M.S., Betz, V., 2013. The Power of Communication: Energy-Efficient NOCS Sci. China Technol. Sci. 54 (3), 625–629.
for FPGAS[C]//International Conference on Field Programmable Logic & Hao, H., Liu, Z., Zhao, F., Liu, W., Hang, W., 2015a. Scenario analysis of energy con-
Applications. IEEE. sumption and greenhouse gas emissions from China's passenger vehicles. Energy 91,
Alam, A., Martensson, J., Johansson, K.H., 2015. Experimental evaluation of decen- 151–159.
tralized cooperative cruise control for heavy-duty vehicle platooning. Contr. Eng. Hao, H., Liu, Z., Zhao, F., Li, W., Hang, W., 2015b. Scenario analysis of energy con-
Pract. 38, 11–25. sumption and greenhouse gas emissions from China's passenger vehicles. Energy 91,
Angiolini, F., Meloni, P., Carta, S., Benini, L., Raffo, L., 2006. Contrasting a NoC and a 151–159.
Traditional Interconnect Fabric with Layout awareness[C]//Conference on Design. Hao, H., Geng, Y., Li, W., Guo, B., 2015c. Energy consumption and GHG emissions from
European Design and Automation Association. China's freight transport sector: scenarios through 2050. Energy Policy 85, 94–101.
Bandivadekar, A., Cheah, L., Evans, C., Groode, T., Heywood, J., Kasseris, E., Kromer, M., Hao, H., Cheng, X., Liu, Z., Zhao, F., 2017a. Electric vehicles for greenhouse gas reduction
Weiss, M., 2008. Reducing the fuel use and greenhouse gas emissions of the US ve- in China: a cost-effectiveness analysis. Transport. Res. Part D 56, 68–84.
hicle fleet. Energy Policy 36 (7), 2754–2760. Hao, P., Wu, G., Boriboonsomsin, K., Barth, M., 2017b. J. Eco-approach and departure
Bansal, P., Kockelman, K.M., 2017. Forecasting Americans' long-term adoption of con- (EAD) application for actuated signals in real-world traffic. IEEE Trans. Intell. Transp.
nected and autonomous vehicle technologies. Transport. Res. Pol. Pract. 95, 49–63. Syst. 99, 1–11.
Barkenbus, J.N., 2010. Eco-driving: an overlooked climate change initiative. Energy Harper, C.D., Hendrickson, C.T., Mangones, S., Samaras, C., 2016a. Estimating potential
Policy 38 (2), 762–769. increases in travel with autonomous vehicles for the non-driving, elderly and people
Barla, P., Gilbert-Gonthier, M., Lopez Castro, M.A., Miranda-Moreno, L., 2017. Eco- with travel-restrictive medical conditions. Transport. Res. C Emerg. Technol. 72, 1–9.
driving training and fuel consumption: impact, heterogeneity and sustainability. Harper, C.D., Hendrickson, C.T., Samaras, C., 2016b. Cost and benefit estimates of par-
Energy Econ. 62, 187–194. tially-automated vehicle collision avoidance technologies. Accid. Anal. Prev. 95 (Pt
Barth, M., Boriboonsomsin, K., 2009. Energy and emissions impacts of a freeway-based A), 104–115.
dynamic eco-driving system. Transport. Res. Transport Environ. 14 (6), 400–410. He, Y., Zou, D., Jin, L., 2010. Exploiting the goldmine: a lifestyle analysis of affluent
Beijing innovation center for mobility intelligent. [Link] Chinese consumers. J. Consum. Mark. 27 (7), 615–628.
Beusen, B., Broekx, S., Denys, T., Beckx, C., Degraeuwe, B., Gijsbers, M., et al., 2009. Jeong, E., Oh, C., Lee, S., 2017. Is vehicle automation enough to prevent crashes? Role of
Using on-board logging devices to study the longer-term impact of an eco-driving traffic operations in automated driving environments for traffic safety. Accid. Anal.
course. Transport. Res. Transport Environ. 14 (7), 514–520. Prev. 104, 115–124.
Bischoff, J., Maciejewski, M., 2016. Simulation of city-wide replacement of private cars Jermakian, J.S., 2011. Crash avoidance potential of four passenger vehicle technologies.
with autonomous taxis in Berlin. Procedia Computer Science 83, 237–244. Accid. Anal. Prev. 43 (3), 732–740.
Braun, A., Rid, W., 2017. Energy consumption of an electric and an internal combustion Jiang, Y., Li, T., Bona, R.F.D., 2009. Value of travel time: an SP survey in Tianjin. Urban
passenger car. A comparative case study from real world data on the Erfurt circuit in Transport of China 7 (2), 68–73.
Germany. Transportation Research Procedia 27, 468–475. Jochem, P., Babrowski, S., Fichtner, W., 2015. Assessing CO2 emissions of electric ve-
Brown, A., Gonder, J., Repac, B., 2014. An analysis of possible energy impacts of hicles in Germany in 2030. Transport. Res. Pol. Pract. 78, 68–83.
Automated vehicles. In: Meyer, G., Beiker, S. (Eds.), Road Vehicle Automation, Kamal, M.A.S., Mukai, M., Murata, J., Kawabe, T., 2013. Model predictive control of
Lecture Notes in Mobility. Springer International Publishing, pp. 137–153. vehicles on urban roads for improved fuel economy. IEEE Trans. Control Syst.
Burns, L.D., Jordan, W.C., Scarborough, B.A., 2012. Transforming Personal Mobility[R]. Technol. 21 (3), 831–841.
The Earth Institute, Columbia University. [Link] Krueger, R., Rashidi, T.H., Rose, J.M., 2016. Preferences for shared autonomous vehicles.
mobility/files/2012/12/[Link]. Transport. Res. C Emerg. Technol. 69, 343–355.
Cao, J., 2015. The demand for private cars and status seeking. Statistics & Information Laslau, C., Holman, M., Saenko, M., See, K., Zhang, Z., 2014. Set Autopilot for Profits:
Forum 30 (10), 72–78. Capitalizing on the $87 Billion Self-Driving Car opportunity[R]. Lux Research.
Chen, Y., Meier, A., 2016. Fuel consumption impacts of auto roof racks. Energy Policy 92, Lavieri, P.S., Garikapati, V.M., Bhat, C.R., Pendyala, R.M., Astroza, S., Dia, F.F., 2017.
325–333. Modeling individual preferences for ownership and sharing of autonomous vehicle
Chen Y, Gonder J, Young S, Wood E. Quantifying Autonomous Vehicles National Fuel technologies. Transportation Research Record: Journal of the Transportation
Consumption Impacts: A Data-Rich approach[J]. Transportation Research Part A: Research Board 2665, 1–10.
article in press. Levinson, J., Montemerlo, M., Thrun, S., 2007. In: Map-Based Precision Vehicle
Childress, S., Nichols, B., Charlton, B., Coe, S., 2015. Using an activity-based model to Localization in Urban Environments[C]//Robotics: Science & Systems Iii, June.
explore the potential impacts of automated vehicles. Transportation Research Board Georgia Institute of Technology, Atlanta, Georgia, Usa DBLP.
Annual Meeting 2493, 99–106. Li, J.M., Gao, Z., 2013. Exploring the impact of speed synchronization through connected
Cliff, D., 2013. Half of new vehicles shipping in north America to have driverless, robotic vehicle technology on fleet-level fuel economy. SAE International Journal of
capabilities by 2032. Microw. J. 56 (10), 59. Passenger Cars - Electronic and Electrical Systems 6 (1), 213–221.
Das, S., Graziano, D., Upadhyayula, V.K.K., Eric, Masanet, Riddle, M., Cresko, J., 2016. Li, S., Xu, S., Huang, X., Cheng, B., Peng, H., 2015. Eco-Departure of connected vehicles
Vehicle lightweighting energy use impacts in U.S. light-duty vehicle fleet. Sustainable with V2X communication at signalized intersections. IEEE Trans. Veh. Technol. 64
Materials & Technologies 8, 5–13. (12), 5439–5449.
Daziano, R.A., Sarrias, M., Leard, B., 2017. Are consumers willing to pay to let cars drive Lin, P., Liu, J., Jin, P., Ran, B., 2017. Autonomous vehicle-intersection coordination
for them? Analyzing response to autonomous vehicles. Transport. Res. C Emerg. method in a connected vehicle environment. IEEE Intelligent Transportation Systems
Technol. 78, 150–164. Magazine 9 (4), 37–47.
Deng, Q., Ma, X., 2015. A Simulation Platform for Autonomous Heavy-Duty Vehicle Litman, T., 2018. Autonomous Vehicle Implementation Predictions, Implications for
Platooning in Mixed Traffic[C]//Transportation Research Board Meeting. Transport Planning. Victoria Transport Policy Institute.
Dennehy, E.R., Gallachoir, B., 2018. Ex-posr decomposition analysis of passenger car Liu, F., Zhao, F., Liu, Z., Hao, H., 2018. The impact of fuel cell vehicle deployment on road
energy demand and associated CO2 emissions. Transport. Res. Part D 59, 400–416. transport greenhouse gas emissions: the China case. Int. J. Hydrogen Energy 43 (50),
Dominguez, S., Khomutenko, B., Garcia, G., Martinet, P., 2015. An Optimization 22604–22621.
Technique for Positioning Multiple Maps for Self-Driving Car's Autonomous MacKenzie, D., 2013. Fuel Economy Regulations and Efficiency Technology
Navigation[C]//IEEE International Conference on Intelligent Transportation Improvements in U.S. Cars since 1975. Doctoral dissertation. Massachusetts Institute
Systems. IEEE. of Technology.
EIA, 2017. Study of the Potential Energy Consumption Impacts of Connected and MacKenzie, D., Zoepf, S., Heywood, J., 2014. Determinants of US passenger car weight.
Automated Vehicles. Energy Information Administration. Int. J. Veh. Des. 65 (1), 73–93.
Fagnant, D.J., Kockelman, K.M., 2014. The travel and environmental implications of Martin, N., Bishop, J., Choudhary, R., Boies, A.M., 2015. Can UK passenger vehicles be
shared autonomous vehicles, using agent-based model scenario. Transport. Res. C designed to meet 2020 emissions targets? A novel methodology to forecast fuel
Emerg. Technol. 40, 1–13. consumption with uncertainty analysis. Appl. Energy 157, 929–939.
Fagnant, D.J., Kockelman, K., 2015. Preparing a nation for autonomous vehicles: op- McKinsey&Company, 2016. Automotive Revolution-Perspective towards 2030[R].
portunities, barriers and policy recommendations. Transport. Res. Part A 77, Meyer, I., Wessely, S., 2009. Fuel efficiency of the Austrian passenger vehicle fleet—-
167–181. analysis of trends in the technological profile and related impacts on CO2 emission.
Fox-Penner, P., Gorman, W., Hatch, J., 2018. Long-term U.S transportation electricity use Energy Policy 37 (10), 3779–3789.
considering the effect of autonomous-vehicles: estimates & policy observations. Morrow, W.R., Greenblatt, J.B., Sturges, A., Saxena, S., Gopal, A., Millstein, D., et al.,
Energy Policy 122, 203–213. 2014. Key Factors Influencing Autonomous Vehicles' Energy and Environmental
Gawron, J.H., Keoleian, G.A., De Kleine, R.D., Wallington, T.J., Kim, H.C., 2018. Life Outcome[M]. Road Vehicle Automation. Springer International Publishing, pp.
cycle assessment of connected and automated vehicles: sensing and computing sub- 127–135.
system and vehicle level effects. Environ. Sci. Technol. 52, 3249–3256. Mosquet, X., Dauner, T., Lang, N., Rurmann, M., Mei-pochtler, A., Agrawal, R., Schmieg,
Greenblatt, J.B., Saxena, S., 2015. Autonomous taxis could greatly reduce greenhouse-gas F., 2015. Revolution in the Driver's Seat: the Road to Autonomous vehicles[R].
emissions of US light-duty vehicles. Nat. Clim. Change 5, 860–865. Boston Consulting Group.

472
F. Liu, et al. Energy Policy 132 (2019) 462–473

Ou, X., Zhang, X., Chang, S., 2010a. Scenario analysis on alternative fuel/vehicle for energy savings. IEEE Transactions on Intelligent Vehicles 1 (1), 68–77.
China's future road transport: life-cycle energy demand and GHG emissions. Energy Urmson, C., Whittaker, W., 2008. Self-driving cars and the urban challenge. IEEE Intell.
Policy 38 (8), 3943–3956. Syst. 23 (2), 66–68.
Ou, X., Yan, X., Zhang, X., 2010b. Using coal for transportation in China: life cycle GHG of Vahidi, A., Sciarretta, A., 2018. Energy saving potentials of connected and automated
coal-based fuel and electric vehicle, and policy implications. International Journal of vehicles. Transport. Res. Part C 95, 822–843.
Greenhouse Gas Control 4 (5), 878–887. Wadud, Z., MacKenzie, D., Leiby, P., 2016. Help or hindrance? The travel, energy and
Rödel, C., Stadler, S., Meschtscherjakov, A., Tscheligi, M., 2014. Towards Autonomous carbon impacts of highly automated vehicles. Transport. Res. Part A 86, 1–18.
Cars: the Effect of Autonomy Levels on Acceptance and Uer experience[C]// Wang, H., Wang, B., Liu, B., Meng, X., Yang, G., 2017. Pedestrian recognition and tracking
International Conference on Automotive User Interfaces & Interactive Vehicular using 3D LiDAR for autonomous vehicle. Robot. Autonom. Syst. 88, 71–78.
Applications. ACM. Wang, S., Zhao, F., Liu, Z., Hao, H., 2018a. Impacts of a super credit policy on electric
Rolim, C.C., Baptista, P.C., Duarte, G.O., Farias, T.L., 2014. Impacts of on-board devices vehicle penetration and compliance with China's Corporate Average Fuel
and training on light duty vehicle driving behavior. Procedia - Social and Behavioral Consumption regulation. Energy 155, 746–762.
Sciences 111, 711–720. Wang, Q., Huo, H., He, K., Yao, Z., Zhang, Q., 2018b. Characterization of vehicle driving
SAE International. SAE J3016 (R) Taxonomy and Definitions for Terms Related to Driving patterns and development of driving cycles in Chinese cities. Transport. Res. Part D
Automation Systems for On-Road Motor Vehicles. 13, 289–297.
SAE-China, 2016. Technology Roadmap for Energy Saving and New Energy vehicles[M]. Wu, C., Zhao, G., Ou, B., 2011. A fuel economy optimization system with applications in
China Machine Press, pp. 208. vehicles with human drivers and autonomous vehicles. Transport. Res. Part D 16 (7),
Schoettle, B., Sivak, M., 2014. A Survey of Public Opinion about Autonomous and Self- 515–524.
Driving Vehicles in the U.S., the U.K., and Australia[R]. The University of Michigan Wu, G., Boriboonsomsin, K., Xia, H., Barth, M., 2014. Supplementary benefits from partial
Transportation Research Institute. vehicle automation in an ecoapproach and departure application at signalized in-
Schoettle, B., Sivak, M., 2015. Potential Impact of Self-Driving Vehicles on Household tersections. Transportation Research Record: Journal of the Transportation Research
Vehicle Demand and usage[R]. The University of Michigan, Transportation Research Board 2424, 66–75.
Institute. Xu, X., Fan, C.K., 2018. Autonomous vehicles, risk perceptions and insurance demand: an
Shanghai Municipal Commission of Economy and Information [Link] individual survey in China. Transportation Research Part A: Policy and Practice ar-
gydt/[Link]. ticle (in press).
Sivak, M., Schoettle, B., 2012. Eco-driving: strategic, tactical, and operational decisions of Yan, X., Crookes, R.J., 2009. Reduction potentials of energy demand and GHG emissions
the driver that influence vehicle fuel economy. Transport Pol. 22 (none), 96–99. in China's road transport sector. Energy Policy 37 (2), 658–668.
Spieser, K., Treleaven, K.B., Zhang, R., Frazzoli, E., Morton, D., Pavone, M., 2014. Toward Yan, X., Crookes, R.J., 2010. Energy demand and emissions from road transportation
a Systematic Approach to the Design and Evaluation of Automated Mobility-On- vehicles in China. Prog. Energy Combust. Sci. 36 (6), 651–676.
Demand Systems: A Case Study in Singapore[M]//Road Vehicle Automation. Yelchuru, B., Waller, T., 2014. Preliminary Eco-Traffic Signal Timing Modeling Results.
Springer International Publishing. [Link]
Stephens, T.S., Gonder, J., Chen, Y., Lin, Z., Liu, C., Gohlke, D., 2016. Estimated Bounds Zarkadoula, M., Zoidis, G., Tritopoulou, E., 2007. Training urban bus drivers to promote
and Important Factors for Fuel Use and Consumer Costs of Connected and Automated smart driving: a note on a Greek eco-driving pilot program. Transport. Res. Transport
Vehicles. National Renewable Energy Laboratory. [Link] Environ. 12 (6), 449–451.
fy17osti/[Link]. Zhang, W., Guhathakurta, S., Khalil, E., 2018. The impact of private autonomous vehicles
Taiebat, M., Brown, A.L., Stafford, H.R., Qu, S., Xu, M., 2018. A review on energy, en- on vehicle ownership and unoccupied VMT generation. Transport. Res. C Emerg.
vironment, and sustainability implications of connected and automated vehicles. Technol. 90, 156–165.
Environ. Sci. Technol. 52, 11449–11465. Zhao, S., Zhao, B., 2011. Value of travel time of urban resident under multifactor influ-
Teixeira, A.C.R., Sodre, J.R., 2018. Impacts of replacement of engine powered vehicles by ence. Journal of Jilin University (Engineering and Technology Edition 41 (1), 46–50.
electric vehicles on energy consumption and CO2 emissions. Transport. Res. Part D Zhao, F., Liu, F., Liu, Z., Hao, H., 2019. The correlated impacts of fuel consumption
59, 375–384. improvements and vehicle electrification on vehicle greenhouse gas emissions in
Thomas, J., Huff, S., West, B., 2014. Fuel economy and emissions effects of low tire China. J. Clean. Prod. 207, 702–716.
pressure, open windows, roof top and hitch-mounted cargo, and trailer. Exteriors 7 Zhu F, Ma L, Xu X, Guo D, Cui X, Kong Q. Baidu apollo auto-calibration system – an
(2), 862–872. industry-level data driven and learning based vehicle longitude dynamic calibrating
Tsugawa, S., Jeschke, S., Shladover, S.E., 2016. A review of truck platooning projects for algorithm[J]. arXir:1808.10134 ([[Link]]).

473