Content uploaded by Ziran Wang
Author content
All content in this area was uploaded by Ziran Wang on May 08, 2020
Content may be subject to copyright.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 21, NO. 5, MAY 2020 2029
Cooperative Eco-Driving at Signalized Intersections
in a Partially Connected and Automated
Vehicle Environment
Ziran Wang ,Student Member, IEEE, Guoyuan Wu ,Senior Member, IEEE,
and Matthew J. Barth ,Fellow, IEEE
Abstract— The emergence of connected and automated vehi-
cle (CAV) technology has the potential to bring a number
of benefits to our existing transportation systems. Specifically,
when CAVs travel along an arterial corridor with signalized
intersections, they can not only be driven automatically using
pre-designed control models but can also communicate with other
CAVs and the roadside infrastructure. In this paper, we describe a
cooperative eco-driving (CED) system targeted for signalized cor-
ridors, focusing on how the penetration rate of CAVs affects the
energy efficiency of the traffic network. In particular, we propose
a role transition protocol for CAVs to switch between a leader
and following vehicles in a string. Longitudinal control models
are developed for conventional vehicles in the network and for
different CAVs based on their roles and distances to intersections.
A microscopic traffic simulation evaluation has been conducted
using PTV VISSIM with realistic traffic data collected for the
City of Riverside, CA, USA. The effects on traffic mobility are
evaluated, and the environmental benefits are analyzed by the
U.S. Environmental Protection Agency’s MOtor Vehicle Emission
Simulator (MOVES) model. The simulation results indicate that
the energy consumption and pollutant emissions of the proposed
system decrease, as the penetration rate of CAVs increases.
Specifically, more than 7% reduction on energy consumption
and up to 59% reduction on pollutant emission can be achieved
when all vehicles in the proposed system are CAVs.
Index Terms—Connected and automated vehicles (CAVs), eco-
driving, eco-approach and departure (EAD), cooperative adaptive
cruise control (CACC), signalized intersections, mixed traffic.
I. INTRODUCTION AND BACKGROUND
MOBILITY, safety and sustainability are all critical mea-
sures of performance in the field of transportation.
In recent years, increased transportation activity continues to
have significant impacts on the above measures and raises
awareness and concerns from the general public. In terms of
traffic mobility, drivers in the U.S. spent an average of 41 hours
a year in traffic during peak hours in 2017, costing nearly
$305 billion in total, which equals to $1,445 per driver [1].
In terms of traffic safety, it is estimated that 37,461 people
died in accidents in the U.S. involving motor vehicles in 2016,
which endured a 6 percent rise from the year before [2].
Manuscript received September 16, 2018; revised January 2, 2019 and
March 22, 2019; accepted April 14, 2019. Date of publication May 1, 2019;
date of current version May 1, 2020. The Associate Editor for this paper was
M. Menendez. (Corresponding author: Ziran Wang.)
The authors are with the Bourns College of Engineering-Center for Envi-
ronmental Research and Technology (CE-CERT), University of California at
Riverside, Riverside, CA 92507 USA (e-mail: zwang050@ucr.edu).
Digital Object Identifier 10.1109/TITS.2019.2911607
And in terms of environmental sustainability, the transportation
sector was the second largest producer of GHG nationwide,
accounting for approximately 27% of total U.S. emissions
in 2013 [3].
In recent years, there has been a significant amount of
research interest on how to improve the mobility, safety and
sustainability of signalized intersections. Specifically, con-
nected and automated vehicle (CAV) technology has been
widely studied to improve the sustainability of transportation
systems, where a CAV can be driven by itself with the help of
its on-board perception sensors, and also communicate with
the driver, other vehicles on the road (through vehicle-to-
vehicle, i.e., V2V communications), roadside infrastructure
(through vehicle-to-infrastructure, i.e., V2I communications),
and the “Cloud” [4]. Such applications are often categorized
as eco-driving at signalized intersections, with specific names
such as GLOSA (Green Light Optimized Speed Advisory
(GLOSA) or Eco-Approach and Departure (EAD) [5]–[10].
On top of the advanced technology on the vehicle side, some
researchers also focus on the infrastructure side to improve
the overall energy efficiency of the traffic system. Lee et al.
proposed a cooperative vehicle intersection control system
that enables cooperation between vehicles and infrastructure
for effective intersection operations and management [11].
Guler et al. developed a CAV-based algorithm considering
platooning and signal flexibility to gain traffic mobility bene-
fits [12]. A real-time adaptive signal phase allocation algorithm
was proposed by Feng et al. using CAV data, and the results
showed a reduction of 16.33% in terms of total delay [13].
In addition to sustainability benefits, CAV technology can
also produce significant mobility and safety benefits. One
major application of the CAV technology is the cooperative
adaptive cruise control (CACC) system, which enables CAVs
to cooperate with each other to form vehicle strings. Numerous
works on the mobility and safety perspectives of CACC
systems have been conducted [14]–[21]. Researchers have
also integrated eco-driving technology with CACC systems for
traffic at signalized intersections. An Eco-CACC algorithm for
isolated signalized intersection was developed by Yang et al.,
which computes the fuel-optimum vehicle trajectory to ensure
that a CAV arrive at the intersection as soon as the last
CAV in the queue is discharged [22]. Xie et al. developed
a model of heterogeneous traffic with a mix of regular and
connected vehicles [23]. CACC and intelligent traffic signal
1524-9050 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
2030 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 21, NO. 5, MAY 2020
Fig. 1. Illustration of the proposed cooperative eco-driving system.
control technology were integrated by Malakorn et al.,where
the mobility, energy and environmental impacts of the pro-
posed system were analyzed [24]. A cluster-wise cooperative
EAD application at isolated intersections was developed by
our previous work, where CAVs are formed into different
clusters based upon CACC protocols, and conduct eco-driving
towards the intersection in a collaborative manner [25]. In this
work, we further improve the system by considering partially
CAV environment instead of a pure CAV environment, and
develop the system for signalized intersections along a corridor
instead of a single intersection. We further test our proposed
system in microscopic traffic simulations instead of numerical
simulations.
The remainder of this paper is organized as follows:
Section II illustrates the overall framework, specifications and
assumptions of our proposed system. Section III proposes
a role transition protocol for cooperative eco-driving (CED)
vehicles in the system. Section IV demonstrates different
longitudinal control models developed for different vehicle
types in the system. A simulation study of the proposed
system has been conducted using microscopic traffic simulator
PTV VISSIM, and its results are evaluated and analyzed in
Section V [26]. Section VI concludes this paper together with
further discussion on future work.
II. PROBLEM STATEMENT
The objective of this study is to develop a CED system
by CAV technology to improve the energy efficiency along
a corridor with signalized intersections. To study the effect
of penetration rate of CAVs, we define two different types
of vehicles in the system as conventional vehicles and CED
vehicles. Different role transition protocols and longitudinal
control models are proposed for different vehicles based
on their degrees of connectivity and automation. In this
study, a microscopic traffic simulation network is modeled
in VISSIM, where different vehicle longitudinal control mod-
els and their relevant logic (e.g., role transition) are integrated
into simulation network to simulate vehicles’ behavior, and
an energy/emission model is implemented to analyze the
environmental impacts of proposed methodologies.
The proposed cooperative eco-driving system can be illus-
trated as Fig. 1. Note that our study mainly focuses on
designing an integrated traffic system with proposed control
protocol, so some reasonable specifications and assumptions
Fig. 2. General framework of the proposed eco-driving system.
are made as follows to quantify the potential benefits while
modeling the system:
1) All CED vehicles in the proposed system are equipped
with appropriate on-board sensors (e.g., OBD, camera,
radar, LIDAR, etc.), and their measurements and calcu-
lations are precise without error.
2) All CED vehicles are V2V-enabled, which are equipped
with wireless communication devices such as Dedicated
Short Range Communications (DSRC) on-board units
(OBUs), to transmit vehicle information among each
other, and also V2I-enabled, which can receive MAP
and SPaT (Signal Phase and Timing) information from
intersections.
3) All intersections in the system are equipped with DSRC
roadside units (RSU) to broadcast their MAP and SPaT
message, and all signals are fixed-timing control.
4) We focus on the development of longitudinal control
strategies and application in the simulation study. The
lateral maneuvers rely on the default lane change model
of VISSIM.
The general framework of the proposed system can be
seen in Fig. 2. All models and algorithms are introduced in
section III and IV of this paper.
Compared to similar existing studies in this research area,
such as [27]–[29], our system has the following improvements:
1) Most current works assume 100% penetration rate of
CAVs in the systems, which are unrealistic for real-
world implementations. Further, we not only consider
two different types of vehicles, but also model the
interactions between different vehicle types. For exam-
ple, we study how a CED vehicle will response when
a conventional vehicle suddenly moves in front of it.
2) Different from most literatures that consider CAVs as
different individual system that conducts eco-driving
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: CED AT SIGNALIZED INTERSECTIONS IN A PARTIALLY CAV ENVIRONMENT 2031
maneuver by itself, our work proposes a cooperative
eco-driving system, where different CAVs are
categorized by different roles. Generally, leaders
conduct eco-driving maneuver with respect to signals
through V2I communication, while followers follow
leaders’ maneuvers through V2V communications.
Therefore, less conflicts will be generated among
different CAVs due to their collaborations.
3) Instead of studying the eco-approaching maneuver on
only one direction of one isolated signalized intersection,
we model a corridor with two signalized intersections
and all four directions, so both eco-approach and
eco-departure maneuvers are developed and analyzed.
The proposed algorithms allow CED vehicles to reset
their parameters once passing the current intersection,
and update parameters again while entering the
V2I communication range of the next intersection.
4) Rather than just numerical simulations, we conduct
microscopic traffic simulation based on the University
Avenue corridor in Riverside, CA, with real-world
traffic flow and signal timing data provided by the
government. Therefore, the results of implementing the
proposed CAV technology on the current transportation
system would be more realistic and convincing.
III. VEHICLE ROLE TRANSITION PROTOC OL
In the proposed system, there are generally two types of
passenger vehicles: conventional vehicles and CED vehicles.
Conventional vehicles are assumed to be driven by human
drivers with no degree of connection and automation. CED
vehicles are assumed to be CAVs with appropriate on-board
sensors to conduct automated driving, and OBUs to transmit
information among vehicles and receive information from the
infrastructure. A vehicle role transition protocol is proposed
for V2X-enabled vehicles as Algorithm 1,sinceaCEDvehicle
can transit from a leader to a follower in a vehicle string, or
vice versa.
A CED vehicle continuously checks whether there is a
preceding vehicle on the same lane, and within the V2V
communication range. If no, then it is a leader of a string.
If yes, then it compares the distance to the preceding CED
vehicle with the distance to the intersection. If the preceding
vehicle already passes the intersection, then the ego-vehicle is
a CED leader. If not, then it further checks the time-to-collision
value (if the preceding vehicle is a conventional vehicle), or
checks the estimated time-to-arrival at the intersection (if the
preceding vehicle is also a CED vehicle).
If the preceding vehicle is a conventional vehicle and
the time-to-collision value is lower than a certain thresh-
old, it means the CED vehicle has a high chance to get
a front-bumper-to-rear-bumper collision with its preceding
conventional vehicle, so the CED vehicle will be a follower
to follow its preceding conventional vehicle’s movement.
If the preceding vehicle is also a CED vehicle and the differ-
ence between two consecutive CED vehicles’ time-to-arrival
is larger than a certain threshold (normally the total length
of an amber phase and a red phase), then the following CED
vehicle may be considered as the “string breaker” scenario
Algorithm 1: Role Transition of CED Vehicles
Input: inter-vehicle distance dgap , distance to
the intersection d1, time-to-arrival of
the ego vehicle tarr , time-to-arrival of
the preceding vehicle tarr
pre ,
time-to-collision with respect to the
preceding vehicle tcollision
Output: vehicle role
1: for
all CED vehicles
do
2: if dgap <
V2V range
then
3: if dgap >d1then
4: ego vehicle is a CED leader
5: else
6: if
preceding is a CED vehicle
then
7: if tarr −tarr
pre ≥
threshold
then
8: ego vehicle is a CED leader
9: else
10: ego vehicle is a CED follower
11: end if
12: if (tcollision )< threshold then
13: ego vehicle is a CED follower
14: else
15: ego vehicle is a CED leader
16: end if
17: else
18: end if
19: end if
20: else
21: end if
22: ego vehicle is a CED leader
23: end for
and becomes a CED leader. The method to calculate estimated
time-to-arrival can be found in Algorithm 2and Algorithm 3.
Essentially, the “string breaker” scenario happens when two
consecutive CED vehicles’ time-to-arrival fall into two differ-
ent green windows. Since we take green windows into account
when calculating the time-to-arrival values, it is possible that
the preceding one of two neighboring CED vehicles estimates
to reach the intersection at the end of the preceding green
window, but the following vehicle estimates to arrive at the
start of the following green window. Therefore, a “string
breaker” scenario is created, where the following CED vehicle
becomes a CED leader, and conducts its own EAD movement
through the intersection.
IV. LONGITUDINAL CONTROL MODEL
A. Conventional Vehicle Longitudinal Control Model
In the proposed system, conventional vehicles are assumed
to be driven by human drivers, which are not equipped with
any on-board sensor or OBU. Therefore, the car-following
model originally proposed by Rainer Wiedemann in 1974
(i.e., Wiedemann 74) is used to model the longitudinal behav-
iors of conventional vehicles [30].
The safety clearance is defined in Wiedemann 74 as
d=a·x=(b·xadd +b·xmult ·z)·√v(1)
where a·xdenotes the average standstill distance; b·xadd is the
additive part of safety distance; b·xmult is the multiplicative
part of safety distance; vis the speed of vehicle; zis a value
of range [0,1]. The tolerance of a·xlies between ±0.1m
which is normally distributed at around 0 m, with a standard
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
2032 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 21, NO. 5, MAY 2020
deviation of 0.3 m. Both b·xadd and b·xmult allow to adjust
the time requirement values.
B. CED Vehicle Longitudinal Control Model
For modeling purposes, a CED vehicle can be separated
into two different components: the CED vehicle longitudinal
control model that delivers reference values of vehicle acceler-
ation, and a vehicle powertrain model that transforms reference
acceleration values into realized throttle or brake values. In this
part, we mainly focus on the longitudinal control model, which
generates the acceleration reference demand and then feeds it
into the powertrain model.
1) CED Leader When Out of the V2I Range: When the
CED leader is running out of the V2I communication range
of the intersection, it follows its desired speed while there is no
preceding conventional vehicle in a certain range, or follows
the preceding conventional vehicle while there is one. In the
proposed system, the Intelligent Driver Model (IDM)-based
longitudinal control strategy is applied to the longitudinal
movement of the CED leader when out of V2I range of
intersections.
IDM was proposed by Treiber et al. [31] and has been
widely studied in many research work. In this study,
we develop a longitudinal control model for the CED leader
(in the case of outside the V2I communication range) based
upon IDM, which can be given as
aref =amax ·1−vego
vdes δ
−(2)
where =(
dsaf e +vego·tgap +vego·(veg o−vpr e )
2·√a2
max
dgap )2is the correlation
term with the front vehicle, which has a value when there is
a conventional vehicle gets in front of the CED leader, and
equals to 0 when no other vehicle is in front. amax is a preset
constant which denotes the maximum changing rate of speed;
vego denotes the current speed of the ego vehicle; vpre denotes
the current speed of its preceding vehicle; vdes is a preset
constant which denotes the desired speed of the ego vehicles;
dsaf e is a preset constant which denotes the minimum allowed
inter-vehicle distance; tgap is a preset constant which denotes
the desired time gap; dgap denotes the measured inter-vehicle
distance between the ego vehicle and its preceding vehicle.
The free acceleration exponent δis defined based upon IDM,
which characterizes how the acceleration of the ego vehicle
decreases with speed (e.g., δ=1 corresponds to a linear
decrease, and δ→∞leads to a constant acceleration).
2) CED Leader When in the V2I Range: Different from the
case when running out of the V2I communication range of
the intersection, the CED leader can receive MAP and SPaT
information from upcoming intersections while it is in the
V2I range, and is able to conduct EAD movement through
intersections accordingly. In the proposed system, we develop
a piecewise trigonometric-linear EAD algorithm based upon
our previous work to control the longitudinal movements of
the CED leader when in the V2I range. Similar to previous
segments, we mainly focus on the longitudinal control model
that generates the acceleration reference demand and then
feeds it into the dynamics model.
We define four different EAD scenarios for a CED leader
when it is in the V2I communication range of the intersec-
tion, which are accelerate scenario, cruise scenario, decelerate
scenario, and stop scenario. Once the CED leader receives the
SPaT information from the intersection and combines that with
the MAP information, it calculates the following variables to
decide which scenario it should be in:
tc=d1
v1
(3)
te=d1−v1·π
2α
vlim +π
2α(4)
tl=
d1−v1·π
2β
vcoast +π
2β(5)
α=min 2·amax
vlim −v1
,2·jerkmax
vlim −v1(6)
β=min 2·amax
v1−vcoast
,2·jerkmax
v1−vcoast (7)
where tc,teand tlstand for cruising time-to-arrival, earliest
time-to-arrival, and latest time-to-arrival, respectively; αand
βare variables to calculate teand tl, respectively; d1denotes
the current distance to the intersection; v1denotes the current
speed of vehicle; jerkmax is a preset constant which denotes
the maximum changing rate of acceleration or deceleration;
vlim is a preset constant which denotes the speed limit of
the current roadway; vcoast is a preset constant which denotes
the coasting speed. If the current signal phase is green,
then we can define the available green window as T=
0,tcurr _etnext _s,tnex t_e; If the current signal phase is
red, then T=tnext _s,tnext_e.Notetcur r _edenotes time-
to-the-end-of-current-green-window, tnext_sand tnex t_edenote
time to-the-start-of-next-green-window and time-to-the-end-
of-next-green-window, respectively. Then the EAD scenario
identification protocol can be demonstrated as Algorithm 2.
Algorithm 2: EAD Scenario Identification of
the CED Leader When in the V2I Range
Input: available green window T, cruising
time-to-arrival tc, earliest
time-to-arrival te, latest
time-to-arrival tl
Output: EAD scenario, estimated time-to-arrival
tarr
1: for
all CED leaders in the V2I range
do
2: if tc∈Tthen
3: cruise scenario, tarr =tc
4: else if [te,tc]∩T=∅ then
5: accelerate scenario, tarr =min[te,tc]∩T
6: else if [tc,tl]∩T=∅ then
7: decelerate scenario, tarr =min[tc,tl]∩T
8: else
9: stop scenario, tarr =tnext_s
10: end if
11: end for
Once the EAD scenario of the CED leader is identified
by Algorithm 2, the CED leader will adopt different longi-
tudinal control models with respect to different scenarios.
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: CED AT SIGNALIZED INTERSECTIONS IN A PARTIALLY CAV ENVIRONMENT 2033
If the CED leader is categorized into the cruise scenario,
it means this vehicle can travel through the intersection
during the green phase without any speed change. Therefore,
the reference longitudinal acceleration of this vehicle is zero.
If the CED leader is categorized into the accelerate
or decelerate scenario, the vehicle needs to firstly accel-
erate or decelerate to a certain speed while approaching
the intersection. This takes a half period of the trigonometric
algorithm, where the reference acceleration of the CED leader
during the first quarter period t∈0,π
2j1is
aref =vd1·j1·sin (j1t)(8)
followed by the second quarter period t∈π
2j1,π
2j1+π
2k1
aref =vd1·j1·sin k1·t+π
k1−t1 (9)
Upon finishing the first half of the trigonometric algorithm,
the CED leader maintains a constant speed and approaches the
intersection. When the CED leader departs the intersection,
it decelerates or accelerates to the target speed (if vh= vtar),
which takes the other half period of this trigonometric algo-
rithm. The reference acceleration during the third quarter
period t∈tdepart ,td epart +π
2k2is
aref =vd2·j2·sin k2·t+π
k2−tdepart (10)
followed by the last quarter period t∈
tdepart +π
2k2,tdepart +π
2j2+π
2k2
aref =vd2·j2·sin j2·t−tdepar t −π
2j2−π
2k2 (11)
In above equations, vh=d1
tarr ,vd1=vh−v1,vd2=
vh−vtar,wherevtar is a preset constant that denotes the target
speed while departing the intersection; tdepart =d2
vhdenotes
time-to-departure; d2is a preset constant that denotes the
departure distance; kiand jiare gains to control the changing
rate of acceleration or deceleration, which can be obtained by
solving the following optimization problem.
max
i=1,2ki(12a)
subject to
|ki·vdi|≤amax
|ki2·vdi|≤ jerkmax
ki≥π
2−1·vh
di
(12b)
since the vehicle dynamics should subject to the hard con-
straint of vehicle powertrain’s ability, and ride comfort of
human passengers. Once kiis solved, jican be calculated by
ji=−π
2ki−π
2ki2−4ki2·π
2−1−di
vh·ki
2π
2−1−di
vh·ki,(i=1,2)
(13)
If the CED leader is categorized into the stop scenario,
it means the vehicle cannot avoid the red phase by either accel-
erating or decelerating. The vehicle needs to firstly decelerate
all the way to full stop while approaching the intersection,
and then accelerate to the target speed while departing. The
reference acceleration of the CED leader in this scenario is
also calculated by equation (8) – (11), with
tdepart =tnext _s(14)
vh=v1
2(15)
ki=ji=vh
di·π(16)
It should be noted that while the CED leader is in the
V2I range and is conducting the EAD maneuver, it also
continuously runs Algorithm 1 to check whether potential role
transition is needed. There are chances that other vehicles cut
in front of the CED leader, so the CED leader will transform
into a CED follower and will no longer adopt the piecewise
trigonometric-linear EAD algorithm to control its longitudinal
movement. If the role of the CED leader stays unchanged,
it will still apply EAD algorithm.
3) CED Follower: When the CED follower is out of the
V2I communication range of the intersection, the distributed
consensus algorithm is proposed to control its longitudinal
movement, which is based upon the distance difference and
the speed difference between the ego vehicle and the preceding
vehicle. The reference acceleration of the ego vehicle is
calculated by the second-order consensus algorithm as
aref =β·(dgap −dref )+γ·(v pr e −vego )(17)
where βand γare damping gains; dref denotes the reference
inter-vehicle distance, which can be calculated as
dref =min(dgap,dsaf e)(18)
where dgap denotes a time gap-based inter-vehicle distance,
which is calculated by the product of ego vehicle’s current
speed and desired time gap, stated as
dgap =vego ·tgap (19)
When ego vehicle’s speed is very low (e.g., vego →0),
the reference inter-vehicle distance returned by dtime is
also very low, which might lead to front-to-rear collision.
Therefore, a minimum allowed inter-vehicle distance dsaf e is
defined to ensure the reference inter-vehicle distance is always
higher than a threshold value.
When the CED follower is in the V2I range, it needs
to continuously calculate its estimated time-to-arrival based
on Algorithm 3. While running Algorithm 3, the CED follower
also runs Algorithm 1 with the updated estimated time-to-
arrival calculated from Algorithm 3. Some certain outcomes of
Algorithm 3 will trigger a role transition in Algorithm 1, which
was mentioned earlier in section III as the “string breaker”
scenario. In that case, the CED follower will transform into a
CED leader and will no longer adopt the distributed consensus
algorithm to control its longitudinal movement. If the role
of the CED follower stays unchanged, it will still apply the
distributed consensus algorithm.
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
2034 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 21, NO. 5, MAY 2020
Fig. 3. Road segments in Riverside, CA that are modeled in this simulation study.
TAB L E I
SIGNAL TIMINGDAT A O F CRANFORD AVENUE &UNIVERSITY AVENUE INTERSECTION
TAB L E I I
SIGNAL TIMING DATA OF IOWA AVENUE &UNIVERSITY AVENUE INTERSECTION
Algorithm 3: Estimated Time-to-Arrival of
the CED Follower
Input: available green window T, CED leader’s
estimated time-to-arrival tarr_l,
preceding vehicle’s estimated
time-to-arrival tarr_p, position of the
ego CED follower in the string n, length
of a red phase tred , length of an amber
phase tamber, desired time headway theadway
Output: estimated time-to-arrival tarr
1: for
all CED followers in the V2I range
do
2: tarr_temp =tarr_l+n·theadway
3: if tarr
_
temp ∈Tthen
4: tarr =tarr _temp
5: else
6: if tarr
_
p∈Tthen
7: tarr =tarr _p+tamber +tred
8: else
9: tarr =tarr _p+theadway
10: end if
11: end if
12: end for
V. S IMULATION AND DISCUSSION
We conduct a simulation study on the proposed CED system
and evaluate its system-wide impacts. The simulation network
is built based upon the six-mile University Avenue corridor
in Riverside, California. We focus on the intersections of
University & Cranford, and University & Iowa, where the
specific road segments modeled in this simulation study can
be illustrated in Fig. 3. The particular segments of roads that
lie between two ends of each black brace are built in traffic
simulation environment, where University Avenue has a length
of 1084 m, Cranford Avenue has a length of 367 m, and Iowa
Avenue has a length of 352 m. The data of signal timing and
traffic count on these two intersections are provided by the
City of Riverside. Specifically, we use the data collected during
7:00-8:00 AM on Thursday, June. 2nd, 2016 to calibrate the
inputs of our simulation network. We select this period since
we want to simulate a morning-peaktraffic network in a typical
weekday. The signal timing data are shown in TABLE I and
TABLE II, and the traffic count data are shown in TABLE III.
We adopt PTV VISSIM, a microscopic multi-modal traffic
flow simulation software, to build the simulation network
of the proposed CED system [26]. Specifically, VISSIM
Application Programmer’s Interface (API) package enables
users to integrate external applications to take influence
on the traffic simulation. In this research, we implement
the proposed role transition and longitudinal models in the
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: CED AT SIGNALIZED INTERSECTIONS IN A PARTIALLY CAV ENVIRONMENT 2035
TABLE III
TRA FFIC COUNT DATA O F UNIVERSITY &CRANFORD INTERSECTION AND UNIVERS ITY &IOWA INTERSECTION
Fig. 4. System architecture of the simulation study in VISSIM.
Fig. 5. University Avenue network built in VISSIM.
DriverModel.DLL, which can be assigned to specific vehi-
cle types in VISSIM and overwrite the standard driving
behavior. Additionally, the EmissionsModel.DLL is provided
by VISSIM to add user-defined emission models, where
we implement the U.S. Environmental Protection Agency’s
MOtor Vehicle Emission Simulator (MOVES)-based model
to perform analysis on the environmental impacts of the
system [32]. The overall architecture of this simulation study
can be illustrated as Fig. 4.
The simulation traffic network built in VISSIM can be
partially illustrated as Fig. 6 (2D mode) and Fig. 6 (3D mode).
We use light green color for conventional vehicles, and several
different colors for CED vehicles based on their scenarios
(mainly for debugging and demonstration purposes). All signal
controllers in the network are designed based on TABLE I and
TABLE II, and vehicle volumes are set by TABLE III. Para-
meters of the traffic network and vehicles in this simulation
study are listed in TABLE IV.
The microscopic traffic simulation results are shown in
TABLE V, with two baseline scenarios (1) and (2) and
ten CED scenarios (3)-(12). Specifically, we also include
EAD-Only vehicles in scenario (2), which conduct EAD
maneuvers in an ego manner. For EAD-only vehicles, only
V2I communications are enabled where they can plan their
speed trajectories based on the information received from
the infrastructure. V2V communications are not enabled for
them, which means they cannot cooperate with each other like
CED vehicles. Note baseline scenario (2) already outperforms
(1) in terms of all environmental measurements, as shown
in TABLE V.
With respect to CED scenarios, positive impacts on NOX,
HC and CO can be observed at any penetration rate of
CED vehicles in the traffic system. However, when there are
less than 70% CED vehicles in the traffic system, negative
impacts on energy and CO2can be observed compared to
these two baseline scenarios. Especially, as can be seen
from TABLE V, the worst scenario in terms of energy con-
sumption and CO2emission is with 40% CED vehicles in
the traffic network. There are basically two reasons for this
behavior:
1) The introduction of CED vehicles brings about conser-
vative driving behaviors to the traffic network: When
they know they can travel through the intersection during
the green window with current speed, they approach
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
2036 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 21, NO. 5, MAY 2020
Fig. 6. Simulation is running in VISSIM with 3D mode.
TAB L E I V
PARAMETERS OF THE SIMULATIONTR AFFI C NETWORK AND VEHICLES
TAB L E V
SIMULATION RESULTS OF ENERGY CONSUMPTION AND POLLUTANT EMISSIONS
the intersection by cruising instead of accelerating.
When they depart the intersection after a full stop
during the red window, they conduct an eco-departure
maneuver with low acceleration process to save energy.
Such conservative driving behaviors surely impede the
movements of conventional vehicles, which always try
to travel through the intersection as fast as possible.
A very good example of this can be observed in Fig. 6,
specifically at the right intersection (Iowa Avenue &
University Avenue). As can been seen from TABLE III,
the volume of eastbound left-turn vehicles at that inter-
section during that one hour is 138, however, there is
only an 11-second green window during a 100-second
signal cycle on that direction (derived from TABLE II).
Therefore, the relatively slow departure rate of CED
vehicles introduces a long queue along the upstream of
eastbound University Avenue, so many through vehicles
have to make unnecessary speed changes/full stops and
consume more energy.
2) Since the proposed cooperative eco-driving methodology
focus on collaborations among different CED vehicles,
when the penetration rate of CED vehicles in the traffic
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
WANG et al.: CED AT SIGNALIZED INTERSECTIONS IN A PARTIALLY CAV ENVIRONMENT 2037
network is lower than some certain threshold, it is
difficult for CED vehicles to connect with each other and
conduct collaborative maneuvers, and therefore brings
about negative impacts on energy consumption. Based
on the simulation results, 40% is the threshold when
CED vehicles are not enough in the traffic network to
conduct cooperative eco-driving maneuver, so the energy
consumption comes out as the worst case scenario.
When the penetration rate of CED vehicles is higher
than 40%, it starts to compensate the negative impacts
on conventional vehicles, so the energy consumption
becomes better.
As can be seen from TABLE V, when there are enough CED
vehicles in the traffic system, positive environmental impacts
are shown. Specifically, compared to baseline scenario (1),
3.9% reduction on energy consumption can be observed when
there are 80% CED vehicles. When all vehicles are CED
vehicles, 7.1% and 2.5% reduction on energy consumption
can be observed compared to baseline scenario (1) and (2),
respectively. These results indicate that, only when the penetra-
tion rate of CED vehicles in the traffic system is high enough,
the proposed CED system can work efficiently with more CED
followers, instead of conventional vehicles, follow the move-
ments of CED leaders and conduct eco-driving in a cooperative
manner. It shall also be noted that CACC maneuver introduces
energy savings for string followers since the reduction of
their aerodynamic drag, but this factor is not included in the
MOVES-based model. Therefore, if we consider aerodynamic
drag of vehicle in the energy consumption model, the proposed
system will get further reductions of energy consumption and
pollutant emissions.
VI. CONCLUSIONS AND FUTURE WORK
In this study, we have developed a CED system in a partially
CAV environment, aiming to reduce energy consumption and
pollutant emissions along a corridor with signalized inter-
sections. Role transition protocol and longitudinal control
models have been developed for different vehicles of the
system. A microscopic traffic simulation has been conducted
in PTV VISSIM with realistic traffic inputs, such as signal
timing data and traffic count data. Simulations have been run
with different penetration rates of CED vehicles, and their
effects on traffic mobility have been evaluated. The MOVES
model has been integrated to evaluate the environmental
effects of the proposed CED system, where more than 7%
reduction on energy consumption and up to 59% reduction
on pollutant emission have been shown at 100% penetration
rate of CED vehicles, respectively, when compared to the
baseline.
Based on the simulation results and our analysis in
Section V, there are several future work can be conducted to
improve the proposed cooperative eco-driving system, espe-
cially to compensate the negative impacts when the penetration
rate of CED vehicles is relatively low:
1) A joint vehicle-infrastructure eco-driving system can be
developed based on the current version, where traffic
signals can be designed to cooperate with CAVs under
different traffic volumes and congestion levels to max-
imize the energy efficiency of the whole system. For
example, the green window at University & Iowa inter-
section on eastbound left-turn direction can be rationally
increased, given the relatively large traffic volumes on
that direction.
2) The eco-departure algorithm of this cooperative
eco-driving system can be further improved with the
consideration of upstream queue length. If upstream
vehicles are already waiting in a long queue, the
CED vehicle may adopt a relatively quicker departure
algorithm to allow more upstream vehicles to be dis-
charged during the same green window.
3) Queue prediction algorithm can also be integrated to
the eco-approach algorithm, so the estimated time-to-
arrival value can be calculated more precisely, and the
EAD scenario of the CED leader can be set more
accurately. In such a manner, the dynamically chang-
ing queue length at the downstream intersection will
less likely to introduce frequent speed changes of the
approaching CED leader, and hence further reduce
energy consumption.
ACKNOWLEDGEMENTS
The authors sincerely thank the City of Riverside for
providing the traffic and signal data on University Avenue,
Riverside.
REFERENCES
[1] INRIX. Los Angeles Tops INRIX Global Congestion Ranking.
Accessed: Feb. 9, 2018. [Online]. Available: http://inrix.com/press-
releases/scorecard-2017/
[2] National Safety Council. NSC Motor Vehicle Fatality Estimates.
Accessed: Feb. 9, 2018. [Online]. Available: http://www.nsc.
org/NewsDocuments/2017/12-month-estimates.pdf
[3] U.S. Department of Transportation. Bureau of Transportation Statistics.
Accessed: Jun. 17, 2017. [Online]. Available: https://www.rita.dot.
gov/bts/sites/rita.dot.gov.bts/files/publications/transportation_statistics_
annual_report/2015/chapter7.html
[4] Center for Advanced Automotive Technology. Connected and Auto-
mated Vehicles. Accessed: Feb. 12, 2018. [Online]. Available:
http://autocaat.org/Technologies/Automated_and_Connected_Vehicles/
[5] K. Katsaros, R. Kernchen, M. Dianati, and D. Rieck, “Performance study
of a green light optimized speed advisory (GLOSA) application using
an integrated cooperative ITS simulation platform,” in Proc. IEEE 7th
IWCMC, Jul. 2011, pp. 918–923.
[6] G. De Nunzio, C. C. de Wit, P. Moulin, and D. Di Domenico, “Eco-
driving in urban traffic networks using traffic signal information,”
in Proc. IEEE Conf. Decision Control, Florence, Italy, Dec. 2013,
pp. 892–898.
[7] Y. Saboohi and H. Farzaneh, “Model for developing an eco-driving
strategy of a passenger vehicle based on the least fuel consumption,”
Appl. Energy, vol. 86, no. 10, pp. 1925–1932, 2009.
[8] H. Xia, K. Boriboonsomsin, and M. Barth, “Dynamic ECO-
driving for signalized arterial corridors and its indirect network-wide
energy/emissions benefits,” J. Intell. Transp. Syst., vol. 17, no. 1,
pp. 31–41, 2013.
[9] O. Altan, G. Wu, M. J. Barth, K. Boriboonsomsin, and J. Stark,
“GlidePath: Eco-friendly automated approach and departure at signalized
intersections,” IEEE Trans. Intell. Vehicles, vol. 2, no. 4, pp. 266–277,
Nov. 2017.
[10] X. Xiang, K. Zhou, W. B. Zhang, W. Qin, and Q. Mao, “A closed-
loop speed advisory model with driver’s behavior adaptability for eco-
driving,” IEEE Trans. Intell. Transp. Syst., vol. 16, no. 6, pp. 3313–3324,
Dec. 2015.
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.
2038 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 21, NO. 5, MAY 2020
[11] J. Lee and B. Park, “Development and evaluation of a cooperative
vehicle intersection control algorithm under the connected vehicles
environment,” IEEE Trans. Intell. Transp. Syst., vol. 13, no. 3, pp. 81–90,
Mar. 2012.
[12] S. I. Guler, M. Menendez, and L. Meier, “Using connected vehicle
technology to improve the efficiency of intersections,” Transp. Res. C,
Emerg. Technol., vol. 46, pp. 121–131, Sep. 2014.
[13] Y. Feng, K. L. Head, S. Khoshmagham, and M. Zamanipour,
“A real-time adaptive signal control in a connected vehicle environment,”
Transp. Res. C, Emerg. Technol., vol. 55, pp. 460–473, Jun. 2015.
[14] B. van Arem, C. J. G. van Driel, and R. Visser, “The impact of
cooperative adaptive cruise control on traffic-flow characteristics,” IEEE
Trans. Intell. Transp. Syst., vol. 7, no. 4, pp. 429–436, Dec. 2006.
[15] V. Milanés, S. E. Shladover, J. Spring, C. Nowakowski, H. Kawazoe,
and M. Nakamura, “Cooperative adaptive cruise control in real traffic
situations,” IEEE Trans. Intell. Transp. Syst., vol. 15, no. 1, pp. 296–305,
Feb. 2014.
[16] J. Ploeg, B. T. M. Scheepers, E. van Nunen, N. van de Wouw, and
H. Nijmeijer, “Design and experimental evaluation of cooperative adap-
tive cruise control,” in Proc. 14th Int. IEEE Conf. Intell. Transp. Syst.,
Oct. 2011, pp. 260–265.
[17] Z. Wang, G. Wu, and M. J. Barth, “Developing a consensus-based coop-
erative adaptive cruise control (CACC) system for heterogeneous vehi-
cles with predecessor following topology,” J. Adv. Transp., vol. 2017,
Aug. 2017, Art. no. 1023654. doi: 10.1155/2017/1023654.
[18] Z. Wang, G. Wu, P. Hao, K. Boriboonsomsin, and M. Barth, “Developing
a platoon-wide eco-cooperative adaptive cruise control (CACC) system,”
in Proc. IEEE Intell. Vehicles Symp., Jun. 2017, pp. 1256–1261.
[19] Z. Wang, G. Wu, and M. J. Barth, “A review on cooperative adaptive
cruise control (CACC) systems: Architectures, controls, and applica-
tions,” in Proc. IEEE 21st Int. Conf. Intell. Transp. Syst., Nov. 2018,
pp. 2884–2891.
[20] S. Maiti, S. Winter, and L. Kulik, “A conceptualization of vehicle
platoons and platoon operations,” Transp. Res. C, Emerg. Technol.,
vol. 80, pp. 1–19, Jul. 2017.
[21] Y. Meng, L. Li, F.-Y. Wang, K. Li, and Z. Li, “Analysis of cooperative
driving strategies for nonsignalized intersections,” IEEE Trans. Veh.
Technol., vol. 67, no. 4, pp. 2900–2911, Apr. 2018.
[22] H. Yang, H. Rakha, and M. V. Ala, “Eco-cooperative adaptive cruise
control at signalized intersections considering queue effects,” IEEE
Trans. Intell. Transp. Syst., vol. 18, no. 6, pp. 1575–1585, Jun. 2017.
[23] D.-F. Xie, X.-M. Zhao, and Z. He, “Heterogeneous traffic mixing regular
and connected vehicles: Modeling and stabilization,” IEEE Trans. Intell.
Transp. Syst., to be published.
[24] K. J. Malakorn and B. Park, “Assessment of mobility, energy, and
environment impacts of intellidrive-based cooperative adaptive cruise
control and intelligent traffic signal control,” in Proc. IEEE Int. Symp.
Sustain. Syst. Technol., May 2010, pp. 1–6.
[25] Z. Wang, G. Wu, P. Hao, and M. J. Barth, “Cluster-wise cooperative
eco-approach and departure application for connected and automated
vehicles along signalized arterials,” IEEE Trans. Intell. Vehicles,vol.3,
no. 4, pp. 404–413, Dec. 2018.
[26] PTV Group. PTV Vissim. Accessed: Mar. 1, 2018. [Online]. Available:
http://vision-traffic.ptvgroup.com/en-us/products/ptv-vissim/
[27] H. Jiang, J. Hu, S. An, M. Wang, and B. Park, “Eco approaching at
an isolated signalized intersection under partially connected and auto-
mated vehicles environment,” Transp. Res. C, Emerg. Techn., vol. 79,
pp. 290–307, Jun. 2017.
[28] B. Asadi and A. Vahidi, “Predictive cruise control: Utilizing upcoming
traffic signal information for improving fuel economy and reducing trip
time,” IEEE Trans. Control Syst. Technol., vol. 19, no. 3, pp. 707–714,
May 2011.
[29] M. A. S. Kamal, S. Taguchi, and T. Yoshimura, “Intersection vehicle
cooperative eco-driving in the context of partially connected vehicle
environment,” in Proc. IEEE 18th Int. Conf. Intell. Transp. Syst.,
Sep. 2015, pp. 1261–1266.
[30] R. Wiedemann, “Simulation des strassenverkehrsflusses,” Inst. Transp.
Sci., Univ. Karlsruhe, Karlsruhe, Germany, 1974. [Online]. Available:
https://trid.trb.org/view/59623
[31] M. Treiber, A. Hennecke, and D. Helbing, “Congested traffic states
in empirical observations and microscopic simulations,” Phys. Rev. E,
Stat. Phys. Plasmas Fluids Relat. Interdiscip. Top., vol. 62, no. 2,
pp. 1805–1824, 2000.
[32] MOVES2014a User Guide, U.S. Environ. Protection Agency,
Washington, DC, USA, Nov. 2015.
Ziran Wang (S’16) received the B.E. degree in
mechanical engineering and automation from the
Beijing University of Posts and Telecommunica-
tions in 2015. He is currently pursuing the Ph.D.
degree in mechanical engineering with the College
of Engineering–Center for Environmental Research
and Technology, University of California at River-
side, Riverside. He was a Research Intern with the
Toyota Infotechnology Center, Mountain View, CA,
USA, in 2018. He is also a Research Assistant with
the College of Engineering–Center for Environmen-
tal Research and Technology, University of California at Riverside. His
research focuses on connected and automated vehicle technology, including
V2X, ADAS, motion planning and control. He holds memberships in
various societies, including the IEEE, the Society of Automotive Engineers
(SAE), the Transportation Research Board (TRB), the International Chinese
Transportation Professional Association (ICTPA), and the Chinese Overseas
Transportation Association (COTA). He received the National Center for
Sustainable Transportation (NCST) Dissertation Award in 2018.
Guoyuan Wu (M’09–SM’15) received the Ph.D.
degree in mechanical engineering from the Univer-
sity of California at Berkeley, Berkeley, in 2010.
He is currently an Associate Research Engineer with
the Transportation Systems Research (TSR) Group,
Bourns College of Engineering–Center for Envi-
ronmental Research and Technology (CE–CERT),
and an Associate Adjunct Professor in electrical
and computer engineering with the University of
California at Riverside. His research focuses on
the development and evaluation of sustainable and
intelligent transportation system (SITS) technologies, including connected
and automated transportation systems (CATS), optimization and control of
vehicles, and traffic modeling and simulation. He is a Board Member of the
Chinese Institute of Engineers–Southern California Chapter (CIE-SOCAL)
and a member of the Chinese Overseas Transportation Association (COTA).
He is an Associate Editor of the SAE International Journal of Connected and
Automated Vehicles and also a member of the Vehicle-Highway Automation
Committee (AHB30) of the Transportation Research Board (TRB).
Matthew J. Barth (M’90–SM’00–F’14) received
the B.S. degree in electrical engineering/computer
science from the University of Colorado in 1984 and
the M.S. and Ph.D. degrees in electrical and com-
puter engineering from the University of California
at Santa Barbara, Santa Barbara, in 1985 and 1990,
respectively.
He is currently the Yeager Families Professor with
the College of Engineering, University of California
at Riverside. He is a part of the intelligent systems
faculty in electrical and computer engineering and
is also serving as the Director of the Center for Environmental Research
and Technology (CE-CERT). His research focuses on applying engineering
system concepts and automation technology to transportation systems, and,
in particular, how it relates to energy and air quality issues. His current
research interests include ITS and the environment, transportation/emissions
modeling, vehicle activity analysis, advanced navigation techniques, electric
vehicle technology, and advanced sensing and control. He is active with
the U.S. Transportation Research Board serving in a variety of roles in
several committees, including the Committee on ITS and the Committee on
Transportation Air Quality. He has also been active in the IEEE Intelligent
Transportation System Society for many years, serving as a Senior Editor
for both the Transactions of ITS and the Transactions on Intelligent Vehicles.
He served as the IEEE ITSS President for 2014 and 2015 and is currently the
IEEE ITSS Vice President for Finance.
Authorized licensed use limited to: Univ of Calif Riverside. Downloaded on May 08,2020 at 00:37:09 UTC from IEEE Xplore. Restrictions apply.