Analyzing Energy and Mobility Impacts of Privately-owned Autonomous Vehicles

Full text


The prevailing trend toward greater automation and
connectivity requires modeling and analysis tools to explore
connectivity, automation, decision science and other future
mobility issues at multiple scales. This paper describes various
modeling efforts in order to model the mobility and energy
impact of autonomous and connected technologies; design of
scenarios under different technological, behavioral, and socio-
economic assumptions; and finally, key findings from the
scenario runs enabled by the advanced models developed. The
integrated ABM-DTA software POLARIS has been extended to
include transit, intra-household vehicle sharing, transportation
network company (TNC) operations along with updates to the
mesoscopic traffic models and value of time adjustments due to
new technologies affecting the mode, destination, and route
choice. The three scenarios are summarized as high sharing – low
automation, high sharing – high automation, and low sharing –
high automation, with VMT changes ranging from -13% to 42%.
I. INTRODUCTION
Advances in technology have brought the idea of
autonomous vehicles (AVs) very close to reality. However, the
potential effects of such disrupting technologies are still largely
unknown. This is mainly due to lack of data and the novelty of
the technology; it is still generally in the development phase at
various automobile original equipment manufacturers
(OEMs), mobility service providers, and other technology
companies. A key unknown is the ownership model that will
be adopted along with AV deployment. That is, will vehicles
continue to be largely privately owned as is currently the case
today, or will the automated, connected, electrified, and shared
(ACES) model of mobility provision predominate? The answer
to this question, and related questions regarding adoption level,
as well as impacts on usage due to the automated technology,
will have substantial impacts on the transportation system as a
whole, as observed in high-level analyses by Brown et al. [1],
Fagnant and Kockelman [2], and Wadud et al. [3]. As Wadud
et al. observed, a move to fully automated vehicles would have
high potential to negatively impact the transportation system
and energy use through faster travel, new energy-consuming
features, extending mobility to new user groups, and most
significantly, reducing the cost of travel. This is especially true
if AVs are privately owned [3], although sharing AVs could
potentially mitigate some of these negative impacts.
*Research supported by U.S. Department of Energy.
All authors are with Argonne National Laboratory, Lemont, IL 60439 USA
(630-252-5460; e-mail: jauld@anl.gov).
II. LITERATURE REVIEW
A body of research has evolved around attempts to
understand individuals’ adoption and usage pattern of AVs
using stated preference surveys for both partial [4] and full
automation [5]–[8]. The research on preference between
private ownership and shared-use vehicles is decidedly mixed,
with generally high levels of individuals reporting no desire to
use AVs, and generally higher preferences for owned AVs
versus shared AVs [6], [8]. Studies attempting to simulate the
experience of private AV ownership [9], have supported the
hypothesis that households with access to fully automated
vehicles are likely to travel substantially more. At the same
time, new shared-mobility services continue to be developed
and existing shared mobility modes continue to see user
growth, demonstrating that these are likely gaining in
acceptance. Given the uncertainty surrounding likely future
ownership preferences, it is important to consider both the
privately owned AV and shared fleet cases when exploring
potential impacts of automated vehicles. Several empirical and
simulation-based studies on potential impacts on shared and/or
private autonomous vehicle have been conducted recently. An
overview of key empirical and simulation studies to date is
shown in Table I.
In this study we apply POLARIS to study the potential
impacts of AV over a large metropolitan area. Compared to
previous literature, especially [19], here we additionally
consider: (i) effect of Transportation Network Companies; and
(ii) a behavioral model that account for a lower disutility for
trips undertaken on AV’s. The scenarios are designed to
account for the uncertainty on the AV technology and the
business models that might prevail in the future. To that end,
scenarios are designed to cover different automation levels and
different assumptions regarding TNC and personal AV costs.
III. OVERVIEW OF THE POLARIS MODELING WORKFLOW
With a recognition of the prevailing trend in mobility
toward greater automation and connectivity, a consortium of
National Laboratories under the U.S. Department of Energy
SMART Mobility initiative [10] have developed modeling and
analysis tools to explore connectivity, automation, decision
science and other future mobility issues at multiple scales. A
core focus here is on modeling and simulation to explore how
future mobility technologies interact with traveler behavior,
Exploring the mobility and energy implications of shared versus
private autonomous vehicles*
Joshua A. Auld, Felipe de Souza, Annesha Enam, Mahmoud Javanmardi, Monique Stinson, Omer
Verbas, Aymeric Rousseau
2019 IEEE Intelligent Transportation Systems Conference (ITSC)
Auckland, NZ, October 27-30, 2019
978-1-5386-7024-8/19/$31.00 ©2019 IEEE 1691
Authorized licensed use limited to: Argonne National Laboratory. Downloaded on January 24,2022 at 01:46:39 UTC from IEEE Xplore. Restrictions apply.

transportation system design and individual vehicle
technologies. A generalized workflow for connecting multiple
simulation tools was developed for studying such problems
while a specific implementation of this workflow is shown in
Figure 1. This workflow leverages multiple simulation tools
including POLARIS [11] for regional travel modeling,
Autonomie [12] for vehicle energy simulation, UrbanSim [13]
for land use forecasting, EVI-Pro [14] for EV charging, and
studies conducted using modified driver models in AimSun for
microscopic traffic flow [15].
TABLE I KEY EXPERIMENTAL AND SIMULATION STUDIES OF AV IMPACT ON TRAVEL DEMAND
Title
Format of the
study
Application
Impact on demand
Impact on traffic flow
Impact on
ownership
M. Gucwa, 2014
[16]
Simulation
MTC Model One
 CT-RAMP ABM
 Citilabs Cube
Private
automation
Partial automation
4% - 8% increase in VMT
Input
(Modified capacity/speed
relationship)
Fixed ownership
Wadud, Zia, Don
MacKenzie, and
Paul Leiby, 2016
Computation of
Emissions using
ASIF framework-
considers
• Activity level
• Modal share
• Energy intensity
• Fuel carbon
content
Partial to full
automation
Increase in travel demand by up to 70%
 Up to 5% reduction in
congestion
 Up to 20% increase in
highway speed
Market penetration
varies from
moderate to high
M. Harb, Y. Xiao,
G. Circella, P. L.
Mokhtarian, and J.
L. Walker, 2018. [9]
Naturalistic
experiment
(Households use 60
hours of chauffer
service)
Private
automation
Full automation
 Increase in VMT by up to 85%
 Increase in long trips in evening
 Increase in ZOV trips
Not applicable
Not included in the
experiment
K. Hidaka and T.
Shiga, 2018. [17]
Simulation
 Act. generation
 Dest. choice
 Mode choice
Shared and private
ownership
Full and partial
automation
 Increase in car mode share by up to
10% in 2040 from BAU scenario
 Additional increase up to 10% with
enhanced multitasking opportunity
Not applicable
Not applicable
C. Rodier, 2018 [18]
Simulation
 MTC-ABM
 MATSim
Private and shared
automation (in
exclusive
scenarios)
Private automation
 Reduce transit 20%, walk/bike 12%
 Increase drive alone trips up to 11%
 Up to 11% increase in VMT
Shared automation
 SOV share decreases by up to 4%,
transit share decrease by up t 4%
 Up to 18% increase in VMT
Input: Increased highway
capacity
Private automation
 78% decrease in vehicle delay
 10% Increase in vehicle
volume
Shared automation
Increase travel time 20 minutes
for SOV and automated taxis; 40
minutes for transit
Assumes 100%
market penetration
of automated
vehicles (private
automation
scenario)
J. Auld, O. Verbas,
M. Javanmardi, and
A. Rousseau, 2018.
[19]
Simulation
(POLARIS)
Private
automation
Partial and full
automation
 Up to 52% increase in VMT
 Input capacity increase
 Up to 75% increase in VHT
 Up to 13% decrease in
average travel speed
Varies across
scenario as a
function of WTP
for CAV adoption
W. Zhang, S.
Guhathakurta, and
E. B. Khalil, 2018.
[20]
Simulation
 Atlanta ABM
 CUBE
Private
automation
Full automation
 Up to 13.3% increase in VMT (due to
ZOV trips, assumes fixed demand)
 Volume/capacity ratio
increases by up to 9.75%
(Due to ZOV trips)
Reduce by 12.3%
(with flexibility in
activity scheduling)
Zhao Yong and
Kockelman Kara
M., 2018. [21]
Simulation
(CAMPO travel
demand model –
with new destination
choice –TransCAD)
Private full
automation
(CAV)
and shared (SAV)
 Increase in VMT by 18% - 41%
depending on cost assumptions
Not applicable
Not applicable
X. Xu, H. S.
Mahmassani, and Y.
Chen, 2019. [22]
Simulation
(DYNASMART
Submodules: POAV
optimizer, AV
Traffic Flow
estimator)
Private
automation
Partial and full
automation
 Increase in VMT by 20% - 24%
 Increase in VHT by 19%-23%
 Increase in transit share with no
reduction in VOT for AV
 Decrease in transit share (5.7%) with
reduction in VOT (by 50%)
Not applicable
33% reduction in
vehicle ownership
M. D. Simoni, K. M.
Kockelman, K. M.
Gurumurthy, and J.
Bischoff, 2019. [23]
Simulation (MatSim
with dynamic vehicle
routing problem
module)
Private full
automation (AV)
and shared
automation (SAV)
 Increase in VMT by 16% (AV) and
22% (SAV)
 Empty miles 0% (AV) and 6% (SAV)
 61% (AV) and 87% (SAV) increase in
travel delay
150% increase in capacity at
100% penetration
Not applicable
1692
Authorized licensed use limited to: Argonne National Laboratory. Downloaded on January 24,2022 at 01:46:39 UTC from IEEE Xplore. Restrictions apply.

Figure 1. POLARIS modeling workflow for SMART Mobility
The POLARIS agent-based transportation systems
simulator [16] forms the core of the workflow, with multiple
elements such as land use modeling, vehicle market
penetration modeling and microscopic traffic flow impacts
feeding into and informing the POLARIS model. POLARIS,
in turn, generates simulated travel episodes, network
performance characteristics, activity-demand and so on that are
used to estimate energy consumption in the Autonomie
model[12], and as feedbacks to the microscopic traffic flow
studies (new demand patterns), land use models (updated
network level of service) and fueling infrastructure models
(vehicle charging demand). The results of the workflow
ultimately feed into a metric, termed the Mobility-Energy
Productivity measure used as a basis for evaluating both energy
performance and mobility performance jointly.
POLARIS itself is a high-performance, open-source, agent-
based modeling framework that can simulate large-scale
transportation systems. It features integrated travel demand,
network flow, and a traffic assignment model, in which it can
model multiple key aspects of travel decisions (activity
planning, route choice, and tactical-level driving decisions)
simultaneously and in a continuous, fully integrated manner.
The model covers individual decision making at long-term,
mid-term, and within-day timeframes for various travel-related
decisions. The mid-term and within-day travel behavior
decisions are captured in a computational process model
representation of decision-making, which also captures the
process of individual activity episode planning and
engagement [18]. These decisions are constrained by long-term
choices regarding home/workplace choice and household
vehicle choices, and in turn, these influence activity and travel
episode planning and realization. The network model includes
a meso-scopic representation of vehicle movements based on
Newell’s kinematic wave model [19] with updates that
represent interactions with traffic control infrastructure. The
traveler agents in the model can react in real time to changing
or unexpected network conditions based on either direct
observation or information provision, using an en-route
rerouting and re-planning model.
For long-term choices, the fleet definitions within
POLARIS can either come from external market penetration
forecasts [20] coupled with baseline vehicle registration data,
or from household-level choice modeling [21]. An additional
CAV technology choice step is implemented using models
based on stated-preference survey data [4] to determine the
willingness-to-pay for various levels of CAV technology for
each household vehicle. This transportation simulation
framework connects to the Autonomie vehicle-level energy
simulation model through the SVTrip stochastic trip
reconstruction process. For more detail on the POLARIS
framework, see [16], and for an example of the use of
POLARIS in energy estimation see [11, 22, 23].
Several key features have been added to the base POLARIS
framework in order to represent future mobility technologies.
First, a household-level automated vehicle sharing
optimization model has been implemented for households that
have access to a private AV. This model optimizes the
household members’ schedules in order to minimize costs (e.g.
fuel, parking, tolls, time costs, etc.) subject to a variety of
constraints including spatio-temporal limitations, vehicle
availability, activity flexibility and others. Formulation and
further details can be found in Javanmardi et al [24].
POLARIS is extended to simulate transit (bus, rail,
commuter rail) and active modes (walking and biking) on a
multi-layered network which allows for fully intermodal
movements such as walk-to-transit, park-and-ride etc..
Travelers walk or drive to a transit stop, wait for a transit trip,
board, sit, stand, alight, transfer, get rejected to board, re-route
and so on. These movements are guided by a time-dependent
point-to-point intermodal routing algorithm [25].
In order to represent the operations of both transportation
network company (TNC) vehicles and shared-autonomous
vehicle fleets, a new SAV operator agent and SAV vehicle
agents have been developed in the POLARIS ABM. The SAV
operator receives trip requests from traveler agents and assigns
the nearest idle vehicle to pick up that traveler, and suggests
repositioning moves for vehicles that have been idled for long
periods – similar in concept to [26]. The TNC module allows
different cost structures, fleet sizes, vehicle types and with
driver ability to refuse trips and stop working at any time.
Whether an AV is part of a shared-fleet or privately owned,
its impact on traffic flow is similar, and is models as an
adjustment to the link capacity parameter in the meso-scopic
flow model based on instantaneous penetration rate on each
link [19]. The capacity adjustment are derived from simulation
studies of link performance under different CACC penetration
rates as in Lu et al. [15].
Finally, the POLARIS model has been adapted to allow all
instances of travel time costs entering choice utilities (e.g.
mode choice, destination choice, route choice…) to be
modified by a value of time adjustment factor meant to capture
the increased comfort and convenience of AV modes. The
parameters are typically varied over ranges from 0.35-1.0
rather than set by model as this is still a new area of research
(see [9], [27]) with few empirical estimates of VOT changes.
IV. SMART MOBILITY SCENARIO DESIGN
In order to explore potential differences in shared versus
private automation usage, a set of three future scenarios were
developed highlighting some of the key parameters controlling
traffic flow, travel behavior and system control under these
presumed futures as a starting point for further analysis. The
scenarios that were developed include a baseline case, a high-
sharing case with low penetration of automated vehicle
technologies (Scenario A), a high-sharing case with high
penetration of AV (i.e. auto-taxis) (Scenario B), and a low
1693
Authorized licensed use limited to: Argonne National Laboratory. Downloaded on January 24,2022 at 01:46:39 UTC from IEEE Xplore. Restrictions apply.

sharing case with high AV penetration (i.e. privately owned
AV) (Scenario C). Key parameters describing these cases in
terms of model inputs in the workflow are listed in Table II.
To be clear, this exercise was not designed as an exhaustive
exploration of all potential parameter combinations, but rather
as a set of extreme cases representative of possible future AV
scenarios to highlight key differences between shared and
private usage. Each scenario also had a low-technology
development (business as usual) and a high-technology
development (program success) variant to explore the effect of
meeting higher vehicle technology targets set by the
Department of Energy would have in each case.
TABLE II. SCENARIO INPUT PARAMETERS
Variables
Baseline
(A) High
sharing
low
automation
(B) High
tech -
mobility
(C) Low
sharing
high
Automation
Private
Ownership
98%
80%
46%
95% low tech/
90% high tech
Auto VOTT
factora
1
L3/4: 0.7-1.0
L5: 0.35-0.7
L3/4: 0.5-1.0
L5: 0.35-0.7
L3/4: 0.5-1.0
L5: 0.35-0.7
Propensity for
non-car modesb
1
0.5
1
1
Shared-use
factorc
1.3
1
1
1.3 / 1.6 (no
driver)
E-Commerce
0.08
deliveries per
person-day
0.5 deliveries
per person-
day
0.5 deliveries
per person-
day
0.2 deliveries
per person-
day
Long Haul
Freight Flows
1% CAGRd
1% CAGR
1.3% CAGR
1.3% CAGR
Vehicle
Technologye
xEV
penetration
~3%
xEV
penetration
16-25%
xEV
penetration
44-77%
xEV
penetration
from 44-77
L3/4 AV sharee
0
10% - 11%
5%-8%
5%-8%
L5 AV sharee
0
0
18% -52%
18% -52%
TNC / SAV
faref
$3.30 +
$1.25/mile +
$0.25/min.
$3.30 +
$0.95/mile +
$0.25/min.
$1.65 +
$0.61/mile
$1.65 +
$0.61/mile
a. Multiplier on the in-vehicle travel time for L3/4 and L5 AVs for all choice models. Varies by
congestion level, time sensitivity of the trip and link functional class.
b. Multiplier on travel time by non-car-based modes for all choice models
c. Multiplier on in-vehicle travel time for ride-share trips
d. Compound annual growth rate from baseline freight flows
e. Range is for low technology and high technology cases, respectively
f. Baseline is a mix of TNC and taxi pricing in Chicago. A is current day TNC pricing. B and C
are SAV pricing (no driver charges + ownership cost per mile + 10% profit)
For each of the above scenarios where private ownership is
reduced, it is assumed that new shared-fleet vehicles replace 5
private vehicles (for conventional TNCs) and 10 private
vehicles for SAVs, consistent with other findings on shared
vehicle replacement rates [2], [26], and that the final share
applies to a fleet size reduced accordingly. The above scenarios
have been applied to the Chicago metropolitan area model
developed in POLARIS. Details on the Chicago baseline
model development, and a previous study on partially
automated mobility using this model, can be found in Auld et
al [19], [28].
V. RESULTS AND DISCUSSION
An overview of the key metrics relating to mobility and
energy use obtained for each scenario as well as a comparison
to the baseline is shown in Table III. The trend of each scenario
shows that increasing shared-mode usage, paired with more
efficient vehicle powertrain technologies, decreases total
energy consumption. This occurs even while serving
approximately the same total travel demand.
TABLE III. SCENARIO RESULTS
(A) High sharing -
Low automation
% 
baseline
to A-high
(B) High sharing -
High automation

baseline
to B-high
(C) Low sharing -
High automation

baseline
to C-high
Metric
Unit
Baseline
Low-tech
High-tech
Low-tech
High-tech
Low-tech
High-tech
Total trips (all types)
M trips
42.0
39.2
39.2
-7%
42.0
42.7
2%
44.8
45.0
7%
Total trips (freight)
M trips
3.8
5.2
5.2
39%
6.1
6.1
62%
5.0
5.0
33%
Total trips (auto-based)
M trips
22.6
16.7
16.7
-26%
18.4
17.9
-21%
22.0
22.3
-2%
Person-miles of travel
M miles
375.6
343.0
341.8
-9%
342.0
346.2
-8%
399.2
452.7
21%
Person-hours of travel
M hours
11.9
10.2
10.2
-14%
10.4
10.2
-14%
13.0
15.6
31%
Vehicle miles traveled
M miles
324.3
283.0
282.7
-13%
330.6
332.7
3%
368.7
459.4
42%
Vehicle hours traveled
M hours
10.3
8.9
8.9
-14%
11.4
11.1
8%
13.3
17.4
69%
Empty miles traveled
M miles
3.7
7.3
7.3
97%
16.5
23.7
543%
25.3
68.1
1744%
% auto empty miles
%
1.1%
2.6%
2.6%
126%
5.0%
7.1%
527%
6.9%
14.8%
1202%
% non-drive travel
%
46.5%
59.2%
59.2%
27%
60.1%
65.0%
40%
46.0%
43.8%
-6%
Avg. vehicle speed
mph
31.5
31.9
31.8
1%
29.1
30.0
-5%
27.6
26.5
-16%
Avg. trip speed
person-mph
31.6
33.5
33.4
6%
33.0
34.0
7%
30.7
29.1
-8%
Total fuel use
M gallons
11.3
8.0
6.6
-42%
7.1
4.8
-58%
8.0
5.9
-48%
Total electrical use
GWh
0.1
3.1
4.8
4,990%
13.7
30.5
31,975%
25.4
48.8
51,183%
Total energy
GWh
377.0
271.6
225.3
-40%
251.6
189.2
-50%
291.2
244.5
-35%
Travel efficiency
mi/KWh
1.00
1.26
1.52
52%
1.36
1.83
84%
1.37
1.85
86%
In Scenario A High-Tech, person miles traveled (PMT) is
reduced about by 9% and hours traveled is reduced by 14%,
but energy use is reduced by 40% due to increased powertrain
efficiency and the reduction in vehicle travel of 13%, for a total
travel efficiency increase of 52%. We also see that the non-
auto-drive mode share increased by 27% due to the increased
availability of TNC vehicles. Moving on to the high-
technology shared AV case (Scenario B High-Tech) has
results that are even more dramatic. The total person travel is
again approximately the same as baseline, but now vehicle
miles traveled actually increases slightly by 3%, primarily due
to the 7.1% of total miles that are unloaded. Mode share for
non-drive modes has increased to 65% due to widespread
availability of the SAVs. Travel energy efficiency has
1694
Authorized licensed use limited to: Argonne National Laboratory. Downloaded on January 24,2022 at 01:46:39 UTC from IEEE Xplore. Restrictions apply.

improved by 84% from baseline and energy use decreased by
50%, again primarily due to increases in powertrain technology
as travel is now less efficient due to high levels of unloaded
movements. Interestingly, although the average vehicle speed
is reduced by 5%, representing increased congestion, the
average trip speed actually increases from baseline. Hence, on
a person-basis travel is improving. Finally, the private
automation scenario (Scenario C High-Tech), shows many of
the opposite effects from B. Person miles and hours traveled
increases by 21% and 31% due to the reduction of value of
travel time in the AV. Meanwhile vehicle hours traveled has an
even greater increase of 42%, due to the above factors and
increased vehicle repositioning due to the vehicle sharing
between household members. In fact, empty vehicle travel in
this scenario now represents 15% of total VMT. Energy use has
decreased by 35% as opposed to 50% in Scenario B-high,
although a similar travel energy efficiency is still obtained.
Looking at individual scenarios in more detail, Figure 2
shows that the reduction in private vehicle ownership and high
penetration of AV technology in scenario B resulted in increase
in transit and TNC ridership. It is interesting that the increase
in transit and TNC ridership complemented each other with the
ridership growing in the city center and suburb for transit and
TNC respectively. This finding indicates that, the transit and
TNC operators can potentially maximize their ridership by
coordinating their service areas. This outcome is largely driven
by the reduction in vehicle ownership accompanying the high-
sharing automated scenario, with an assumed vehicle
replacement rate of 10 household vehicles per new SAV,
consistent with many of the previous simulation studies
referenced above (i.e. [26]).
Figure 2. Shift in transit and TNC share under scenario B high-tech
Figure 3 shows that shared automation (scenario B)
produces significantly more efficient travel compared to the
private automation scenario – with almost 50% less VMT. This
might be attributed couple of factors (i) low VOTT resulting
from high level 5 penetration results in induced travel (scenario
C); (ii) whereas reduction in private ownership allows better
repositioning of the shared fleets resulting in higher fraction of
the TNC VMT that is loaded compared to the private vehicle
VMT.
Figure 3. Loaded and unloaded miles by auto-modes
The finding that the private automation results in higher
VMT is corroborated by previous studies conducted by [9].
However, in the naturalistic experiment conducted by [9]
produced higher percentage of ZOV trips compared to the
current study resulting in higher VMT compared to the current
study. This might be due to the fact that in [9] ZOV trips
(driven by chauffeurs) conducted household errands. Whereas
in the current study ZOV trips were conducted exclusively to
reposition the vehicle without any opportunity to address
household needs. Additionally, in the current study only 52%
of household owned AV, so if ownership was extended to
100% the VMT increase of 42% could possibly have been
doubled which would match the experimental finding of 85%
increase quite closely.
The findings of the shared and private AV scenarios studied
here broadly align with previous simulation observations
regarding VMT and VHT increases as well. The 42% increase
in VMT observed in this study is substantially higher than the
16% observed in the AV scenario by [23], although in that
study AV repositioning is not simulated which accounted for a
substantial portion of the increase in this study. This could also
explain why that study found greater VMT increase in the SAV
scenario than the AV scenario, which is the opposite finding
here. Zhao and Kockelman [21] also found a VMT increase up
to 41% although significant limitations are mentioned in the
study due to the use of a 4-step travel demand model, such as
the lack of repositioning travel. Rodier [18] also found
substantially lower increase in VMT, up to 11% in AV and
18% in SAV scenarios, again opposite the findings in this
study. In this case the lower VMT can be explained by the
modest reduction in drive VOT assumed (25%), along with the
lack of vehicle repositioning trips.
VI. CONCLUSION
As a result of previous and ongoing research, the POLARIS
workflow is now able to model traveler decision making such
as activity generation, destination choice, mode choice, activity
scheduling, route choice, and vehicle sharing, as well as transit
and TNC operations within one framework. Passenger cars
1695
Authorized licensed use limited to: Argonne National Laboratory. Downloaded on January 24,2022 at 01:46:39 UTC from IEEE Xplore. Restrictions apply.

(conventional or autonomous), TNC vehicles (conventional or
autonomous), transit vehicles, as well as walking and biking
along with all possible intermodal connections are simulated in
a single network. The output of this software feeds into SVTrip
and Autonomie to calculate energy consumption. This highly
integrated modeling capability enables us to model the
mobility and energy implications of future technologies not
only from the perspective of traffic or operational efficiency
but also from the perspective of behavioral changes such as
reductions in value of travel time, and the circular effect of
supply and demand on each other. This workflow has been
applied to a set of hypothetical future scenarios regarding
shared and/or automated vehicles to study possible impacts on
mobility and energy. We find that high-sharing scenarios tend
to lead to more efficient outcomes, with lower overall
congestion, empty travel miles and energy use compared to
scenarios with high private AV ownership.
For future work we will perform a sensitivity analysis on
the scenarios to identify the key drivers of the results and the
relationship among them.
ACKNOWLEDGMENT
This report and the work described were sponsored by the
U.S. Department of Energy Vehicle Technologies Office
under the Systems and Modeling for Accelerated Research in
Transportation Mobility Laboratory Consortium, an initiative
of the Energy Efficient Mobility Systems Program. David
Anderson, a Department of Energy Office of Energy
Efficiency and Renewable Energy manager, played an
important role in establishing the project concept, advancing
implementation, and providing ongoing guidance.
REFERENCES
[1] A. Brown, J. Gonder, and B. Repac, “An Analysis of Possible Energy
Impacts of Automated Vehicle,” in Road Vehicle Automation, G.
Meyer and S. Beiker, Eds. Cham: Springer International Publishing,
2014, pp. 137–153.
[2] D. J. Fagnant and K. M. Kockelman, “The travel and environmental
implications of shared autonomous vehicles, using agent-based model
scenarios,” Transp. Res. Part C Emerg. Technol., vol. 40, pp. 1–13,
Mar. 2014.
[3] Z. Wadud, D. MacKenzie, and P. Leiby, “Help or hindrance? The
travel, energy and carbon impacts of highly automated vehicles,”
Transp. Res. Part Policy Pract., vol. 86, pp. 1–18, Apr. 2016.
[4] P. Bansal and K. M. Kockelman, “Forecasting Americans’ long-term
adoption of connected and autonomous vehicle technologies,” Transp.
Res. Part Policy Pract., vol. 95, pp. 49–63, Jan. 2017.
[5] M. W. Levin and S. D. Boyles, “Effects of Autonomous Vehicle
Ownership on Trip, Mode, and Route Choice,” Transp. Res. Rec. J.
Transp. Res. Board, vol. 2493, pp. 29–38, Jan. 2015.
[6] C. J. Haboucha, R. Ishaq, and Y. Shiftan, “User preferences regarding
autonomous vehicles,” Transp. Res. Part C Emerg. Technol., vol. 78,
pp. 37–49, May 2017.
[7] F. Nazari, M. Noruzoliaee, and A. Mohammadian, “Shared Mobility
Versus Private Car Ownership: a Multivariate Analysis of Public
Interest in Autonomous Vehicles,” 2018, p. 7p.
[8] P. S. Lavieri, V. M. Garikapati, C. R. Bhat, R. M. Pendyala, S.
Astroza, and F. F. Dias, “Modeling Individual Preferences for
Ownership and Sharing of Autonomous Vehicle Technologies,”
Transp. Res. Rec. J. Transp. Res. Board, no. 2665, p. pp 1–10, 2017.
[9] M. Harb, Y. Xiao, G. Circella, P. L. Mokhtarian, and J. L. Walker,
“Projecting travelers into a world of self-driving vehicles: estimating
travel behavior implications via a naturalistic experiment,”
Transportation, vol. 45, no. 6, pp. 1671–1685, Nov. 2018.
[10] R. Sarkar and J. Ward, “DOE SMART Mobility: Systems and
Modeling for Accelerated Research in Transportation,” in Road
Vehicle Automation 3, G. Meyer and S. Beiker, Eds. Cham: Springer
International Publishing, 2016, pp. 39–52.
[11] J. Auld, M. Hope, H. Ley, V. Sokolov, B. Xu, and K. Zhang,
“POLARIS: Agent-based modeling framework development and
implementation for integrated travel demand and network and
operations simulations,” Transp. Res. Part C Emerg. Technol.
[12] Argonne National Laboratory, “Autonomie,” Autonomie. [Online].
Available: www.autonomie.net.
[13] P. Waddell, A. Borning, M. Noth, N. Freier, M. Becke, and G.
Ulfarsson, “Microsimulation of Urban Development and Location
Choices: Design and Implementation of UrbanSim,” Netw. Spat.
Econ., vol. 3, no. 1, pp. 43–67, Jan. 2003.
[14] E. W. Wood, C. L. Rames, and M. Muratori, “New EVSE Analytical
Tools/Models: Electric Vehicle Infrastructure Projection Tool (EVI-
Pro),” National Renewable Energy Lab. (NREL), Golden, CO (United
States), NREL/PR-5400-70831, Jan. 2018.
[15] X.-Y. Lu, J. Lee, D. Chen, J. Bared, D. Dailey, and S. E. Shladover,
“Freeway Micro-simulation Calibration: Case Study Using Aimsun
and VISSIM with Detailed Field Data,” in Proceedings of the 93rd
Transportation Research Board Annual Meeting, 2014.
[16] M. Gucwa, “Mobility and energy impacts of automated cars,” in
Proceedings of the Automated Vehicles Symposium, San Francisco,
2014.
[17] K. Hidaka and T. Shiga, “Forecasting Travel Demand for New
Mobility Services Employing Autonomous Vehicles,” Transp. Res.
Procedia, vol. 34, pp. 139–146, Jan. 2018.
[18] C. Rodier, “Automated Vehicle Scenarios: Simulation of System-
Level Travel Effects Using Agent-Based Demand and Supply Models
in the San Francisco Bay Area,” National Center for Sustainable
Transportation, 2018.
[19] J. Auld, O. Verbas, M. Javanmardi, and A. Rousseau, “Impact of
Privately-Owned Level 4 CAV Technologies on Travel Demand and
Energy,” Procedia Comput. Sci., vol. 130, pp. 914–919, 2018.
[20] W. Zhang, S. Guhathakurta, and E. B. Khalil, “The impact of private
autonomous vehicles on vehicle ownership and unoccupied VMT
generation,” Transp. Res. Part C Emerg. Technol., vol. 90, pp. 156–
165, May 2018.
[21] Zhao Yong and Kockelman Kara M., “Anticipating the Regional
Impacts of Connected and Automated Vehicle Travel in Austin,
Texas,” J. Urban Plan. Dev., vol. 144, no. 4, p. 04018032, Dec. 2018.
[22] X. Xu, H. S. Mahmassani, and Y. Chen, “Impact of Autonomous
Vehicles on Household Activity and Travel Scheduling: An Integrated
Dynamic Network Modeling Approach,” presented at the
Transportation Research Board 98th Annual MeetingTransportation
Research Board, 2019.
[23] M. D. Simoni, K. M. Kockelman, K. M. Gurumurthy, and J. Bischoff,
“Congestion pricing in a world of self-driving vehicles: An analysis of
different strategies in alternative future scenarios,” Transp. Res. Part
C Emerg. Technol., vol. 98, pp. 167–185, Jan. 2019.
[24] M. Javanmardi, J. Auld, and O. Verbas, “Analyzing Intra-household
Fully Autonomous Vehicle Sharing,” presented at the Transportation
Research Board 97th Annual MeetingTransportation Research Board,
2018.
[25] Ö. Verbas, J. Auld, H. Ley, R. Weimer, and S. Driscoll, “Time-
Dependent Intermodal A* Algorithm: Methodology and
Implementation on a Large-Scale Network,” Transp. Res. Rec., vol.
2672, no. 47, pp. 219–230, Dec. 2018.
[26] J. Bischoff and M. Maciejewski, “Simulation of City-wide
Replacement of Private Cars with Autonomous Taxis in Berlin,”
Procedia Comput. Sci., vol. 83, pp. 237–244, Jan. 2016.
[27] S. Hess, A. Daly, T. Dekker, M. O. Cabral, and R. Batley, “A
framework for capturing heterogeneity, heteroskedasticity, non-
linearity, reference dependence and design artefacts in value of time
research,” Transp. Res. Part B Methodol., vol. 96, pp. 126–149, Feb.
2017.
[28] J. Auld, M. Hope, H. Ley, V. Sokolov, B. Xu, and K. Zhang,
“POLARIS: Agent-based modeling framework development and
implementation for integrated travel demand and network and operations
simulations,” Transp. Res. Part C Emerg. Technol., vol. 64, pp. 101–116,
Mar. 2016.
1696
Authorized licensed use limited to: Argonne National Laboratory. Downloaded on January 24,2022 at 01:46:39 UTC from IEEE Xplore. Restrictions apply.