ArticlePDF Available

Evaluating the Effectiveness of Integrated Connected Automated Vehicle Applications Applied to Freeway Managed Lanes

Authors:

Abstract and Figures

The purpose of this study is to define an operational concept involving connected automated vehicle (CAV) operation on freeway managed lanes. Despite the low projected market penetration of CAVs during the next decade, the use of managed lane facilities has the potential to support the realization of increased mobility benefits by their very nature. The proposed CAV operation involves platoons of equipped vehicles governed by integrated CAV applications, including cooperative adaptive cruise control (CACC), cooperative merge, and speed harmonization. This study proposes an algorithm for integrating CAV applications. Through microscopic simulation, the study particularly examines the effectiveness of CACC, CACC plus cooperative merge, and the addition of speed harmonization under different penetration rates. Simulation results show the effectiveness of the bundled application to enhance system throughput and reduce delay, even with low CAV penetration rates. The speed harmonization shows the greatest effects on delay reduction at medium-to-high penetration rates and some benefits even at low penetration rates. The conclusions provide operational insights and guidance for traffic management centers to implement CAV-based traffic control in the future.
Content may be subject to copyright.
1
Abstract The purpose of this study is to define an operational
concept involving connected automated vehicle (CAV) operation
on freeway managed lanes. Despite the low projected market
penetration of CAVs during the next decade, the use of managed
lane facilities has the potential to support the realization of
increased mobility benefits by their very nature. The proposed
CAV operation involves platoons of equipped vehicles governed by
integrated CAV applications, including cooperative adaptive
cruise control (CACC), cooperative merge, and speed
harmonization. This study proposes an algorithm for integrating
CAV applications. Through microscopic simulation, the study
particularly examines the effectiveness of CACC, CACC plus
cooperative merge, and the addition of speed harmonization under
different penetration rates. Simulation results show the
effectiveness of the bundled application to enhance system
throughput and reduce delay, even with low CAV penetration
rates. The speed harmonization shows the greatest effects on delay
reduction at medium-to-high penetration rates and some benefits
even at low penetration rates. The conclusions provide operational
insights and guidance for traffic management centers to
implement CAV-based traffic control in the future.
Index Terms Connected Automated Vehicles (CAV), Bundled
CAV Applications, Managed Lanes, Cooperative Adaptive Cruise
Control (CACC), Cooperative Merge, Speed Harmonization
I. INTRODUCTION
managed lane is a type of highway lane that is operated
with a management scheme, such as lane use restrictions
or variable tolling, to optimize traffic flow, vehicle throughput,
or both. With rapid advancements in connected and automated
vehicle (CAV) technology, managed lanes can either benefit
directly from associated wireless communication systems and
in-vehicle innovations, or aid in their development and
deployment, thereby further adding to the original return on
investment value of these managed lanes [1-2]. Combined
vehicle connectivity and automation technologies could allow
This study is funded in part by the U.S. Department of Transportation
(Project Number: DTFH61-12-D-00020) and the University of Cincinnati
Office of Research.
Yi Guo is with the Department of Civil and Architectural Engineering and
Construction Management, University of Cincinnati, Cincinnati, OH 45221
USA (e-mail: guo2yi@mail.uc.edu)
Jiaqi Ma is with the Department of Civil and Architectural Engineering and
Construction Management, University of Cincinnati, Cincinnati, OH 45221
USA (e-mail: jiaqi.ma@uc.edu)
Edward Leslie is with the Leidos, Inc., Reston, VA 20190 USA (e-mail:
edward.m.leslie@leidos.com)
Zhitong Huang is with the Leidos, Inc., Reston, VA 20190 USA (e-mail:
zhitong.huang@leidos.com)
consistent speeds to be maintained throughout the facility [3].
This would increase the throughput and capacity of the
managed lanes and could also potentially benefit parallel
general-purpose lanes as traffic moves to the managed lane [4-
5]. Smoothed, optimized speeds would also create a reduction
in fuel consumption, harmful emissions, and highway crashes
[6-9].
Managed lanes offer several features that are favorable and,
in many respects, critical to the testing and implementation of
CAV technologies. Managed lanes are separated from general-
purpose lanes, either through barriers or markings, and provide
the opportunity for active management of traffic through access
control, vehicle eligibility restriction, pricing, or a combination
thereof. They also offer operational flexibility such that the
operating agency can proactively manage demand and capacity
on the facility by applying new strategies or modifying existing
strategies. The infrastructure and associated investment
required for several vehicle-to-infrastructure (V2I)
communication protocols are either already available or more
readily installed on several types of managed lane facilities
versus general-purpose lanes.
To date, more than three dozen connected and/or automated
vehicle applications concepts have been developed, many
through prototyping and demonstration [10-13]. In this study,
we propose the concept of the integrated CAV application,
including but not limited to speed harmonization, cooperative
adaptive cruise control (CACC), and cooperative merge. These
three applications are selected because their effectiveness has
been approved in the literature, e.g., [14-19]. In our opinion,
these selected applications are three of the most promising ones
for improving freeway system performance.
There can also be different configurations of applying the
integrated application to managed lanes. Generally, such
managed lane facilities should be positioned at the left side of
the freeways. While sharing on- and off-ramps with the regular
traffic are possible, this may create weaving bottlenecks and it
is ideal that the CAV managed lane can have dedicated ramps
to facilitate CAV cooperative operations [1, 20-21]. Note that,
in the early deployment stages when the CAV market
penetration is low, special-purpose vehicles, such as
conventional high-occupancy vehicles, may be allowed to use
the facility as well [22-23] to avoid wasting capacity on the
managed lanes. As the market penetration increases, the
eligibility requirement of the managed lane use can be
tightened, and only CAVs can be allowed at some penetration
Evaluating the Effectiveness of Integrated
Connected Automated Vehicle Applications
Applied to Freeway Managed Lanes
Yi Guo, Jiaqi Ma, Edward Leslie, and Zhitong Huang
A
2
level. This is, to a large extent, site-dependent and we do not
address it in this paper. However, it is also possible to require
these special-purpose human-driven vehicles to be equipped
with vehicle awareness devices (VAD), a device that can
broadcast real-time vehicle information at a minimum such that
other vehicles and the system managers can have full
knowledge of the real-time traffic status [24]. By transmitting
real-time data, these vehicles can serve as leaders of CAV
platoons, and this can greatly increase the likelihood of
platooning and the average length of platoons. This will be
considered in our analysis. Additionally, constructing the
managed lanes, either creating a new lane or converting existing
general-purpose lanes are also of practical significance, but
outside the scope of this paper. The focus of the paper is to
develop algorithms for the integrated application and evaluate
the effectiveness. Readers who are interested in a
comprehensive discussion on the Concept of Operations of the
CAV managed lane concept can refer to [25] authored by the
research team. We also leave discussions of more complex
scenarios, such as the existence of pure human-driven vehicles
without information communication and how to construct the
managed lanes, to future studies.
In this study, we focus on one representative, preferred
scenario of a single-lane managed facility and dedicated ramps
on the left side of the freeway exclusively for connected
vehicle/connected and automated vehicle (CV/CAV)
operations in an environment under various CAV market
penetration rates. In the remainder of the paper, we will first
review the development of the three key CAV applications in
the integration. Then, we detail the algorithms for vehicular
control. Last, we use a microscopic traffic simulation tool to
evaluate the effectiveness of the proposed integrated
application to enhance the operations of the managed lane
facility under different scenarios.
II. TECHNOLOGY REVIEW
The primary motivation for the development of CACC is to
reduce traffic congestion by improving highway capacity and
throughput and attenuating traffic flow disturbances [14]. The
class of CACC systems utilizing V2V communication could
potentially allow the mean following time gap to be reduced
from about 1.4 seconds when driving manually to
approximately 0.6 seconds when using CACC, resulting in an
increase in highway lane capacity [26]. Several highway traffic
simulations [14, 27] showed that autonomous ACC (i.e.,
sensor-based ACC [28]) alone, even at high market penetration
rates, had little effect on lane capacity, and recent on-the-road
experiments have shown that a stream of autonomous ACC
vehicles is string unstable, resulting in a negative impact on lane
capacity and safety. However, with the shorter following gaps
enabled by CACC systems, lane capacity could potentially be
increased from the typical 2,200 vehicles per hour to almost
4,000 vehicles per hour at 100 percent market penetration [3].
In addition to V2V-based CACC, reference [29] proposed the
concept of CACC systems utilizing I2V communication,
although it was not investigated in detail. Theoretically, the
CACC system cooperates with the infrastructure to reduce the
potential for congestion at bottleneck locations by
automatically reducing the speeds of upstream vehicle platoons
using I2V communication to set speed values, thus reducing
speed differentials and allowing the traffic flow to be
maintained at peak throughput. This is similar to the proposed
concept of the integrated application of CACC and speed
harmonization.
The concept of cooperative merge leverages V2V and V2I
communications to enable CAVs to signal other vehicles (e.g.,
via dedicated short-range communication, or DSRC) of their
intention to merge into traffic streams. Using this information,
merging vehicles may identify upcoming acceptable gaps in the
mainline and make lane changes when possible [16]. Also,
upstream managed lane vehicles may cooperate by adjusting
their speeds to create a gap for the requesting vehicle. The
trajectories of merging vehicles are then optimized. The
merging movement can then occur safely and with minimal
impact on the string's stability [30]. Reference [31] tested two
cooperative automated merging strategies for highway entry,
one using I2V communication and the other using V2V
communication in microscopic simulation. The results show
that I2V reduced travel time in the merging section when the
traffic flow was high, and the V2V case supports a significant
increase in traffic flow without increasing travel times. The
results indicate the potential advantages of using cooperative
automation to relieve the bottleneck in the merging section.
Generally speaking, speed harmonization involves gradually
lowering speeds upstream of a heavily congested area in order
to reduce the stop-and-go traffic that contributes to frustration
and crashes. To date, a related strategy known as variable speed
limits (VSL) has been applied at several locations in Europe and
a few locations in America [32], but the driver response to
suggested speed targets has not been consistent. Dynamic speed
limit adjustments are less efficient than dynamic adjustments of
recommended and/or actual speeds communicated directly into
connected and automated vehicles as the speeds are adjusted
automatically unless drivers intervene. Compared to the
segment-based speed harmonization (similar to VSL) [33],
trajectory-based speed harmonization can control and
coordinate an individual vehicle's trajectories depending on
each vehicle's location. Recent simulation studies (e.g., [34])
and field experiments [35] suggest the potential of such an
approach in enhancing traffic smoothness and therefore
improving efficiency and safety. In particular, trajectory control
can facilitate freeway merge. In this scenario, a central
controller (e.g., traffic management center) coordinates the
trajectories of upstream managed lane vehicles and merging
vehicles such that smooth and efficient merging and minimum
impact on mainline traffic can be guaranteed.
Reference [2] summarized previous studies that involve
CACC and cooperative merge. Those positive results show the
potential of integrated CAV applications and cooperative merge
cannot be isolated from other CAV operations (e.g., such as
platooning), to realize the full benefits. Additionally, the
combination of cooperative merge and speed harmonization (by
controlling and coordinating arrivals of upstream managed lane
vehicles to create gaps for merging) can further improve
3
merging area performance because the upstream traffic flow
can be smoothed to reduce the congestion at merge area.
However, those CAV applications should be integrated to
collaboratively fulfill CAV cooperative control and avoid the
potential conflicts and improve the traffic system performance.
Therefore, the integrated CAV application proposed in this
study tries to address such integration of three different
applications in the next section.
III. INTEGRATED CAV APPLICATION
A. Cooperative Adaptive Cruise Control (Platooning)
The concepts of Cooperative Adaptive Cruise Control have
been widely discussed, and several CACC implementation
methods are proposed. Among those designs, one of the major
differences is the topology of communication, which includes
decentralized communication [36], centralized communication
[37], communication with nearest vehicles [38-39], and
communication with platoon leader and nearest vehicles [40].
Since this study focuses on CACC operations on one managed
lane at early stages of deployment, it adopts the single-lane
operation of the CACC operation algorithm developed by [41],
which applies the communication with nearest vehicles, and
enhances it with components that enable the speed
harmonization and cooperative merging algorithms. The
implementation also assumes the implementation of vehicle
awareness devices (VAD) on manually driven vehicles. VAD-
equipped vehicles broadcast their real-time status to
surrounding vehicles and can serve as the leader of CACC
vehicle platoons. Under this strategy, the probability of CACC-
equipped vehicles traveling in the CACC mode greatly
increases, thus offering an incentive for users to equip their
vehicles with CACC, even when the CACC market penetration
is low. Note that there may be safety risks when VAD vehicles
are platoon leaders due to the stochasticity of human-driven
behavior. This risk is partially reduced through the real-time
communication and exchanges of messages between the VAD
leader and the following CAV. While this is outside the scope
of this paper, future research may need to investigate algorithms
that particularly address the safety concerns.
We choose a maximum platoon length of 10 vehicles, as
recommended in [41]. Shorter platoon lengths would result in
more CACC platoons, which can lead to lower freeway
capacity because inter-platoon gaps are larger than the gaps
between consecutive vehicles within the platoon. On the other
hand, longer CACC platoons would lead to less versatility since
they make merging more difficult for other vehicles.
Fig. 1 provides an illustration of CACC platoons simulated
in this study. If the preceding vehicle is a conventional vehicle
and the clearance distance exceeds the detecting range of the
onboard sensors, or there is no vehicle in front of the subject
vehicle, the subject CAV will switch to the ACC speed
regulation mode to regulate the following behavior. This mode
keeps the subject vehicle cruising with target speed to reduce
unnecessary oscillations and detecting the clearance distance to
avoid the collision. The ACC controller will apply the Equation
(1) proposed by [41] in this mode.
   
(1)
where,
: acceleration recommended by the ACC controller to
the subject vehicle (m/s2)
: gain in the speed difference between the free flow speed
and the subject vehicle's current speed (  in this
study)
: free-flow speed (m/s) or reference speed commands when
speed harmonization is applied.
: current speed of the subject vehicle (m/s).
Fig. 1. Illustration for CACC Platooning Logic
Different from [41], since we assume all vehicles in the
managed lane is at least equipped with vehicle awareness
devices, the ACC gap regulation mode will not be implemented
in this study. The subject CAV may encounter two other
possible scenarios and can be either a platoon leader or
follower.
If the length of the previous CACC platoon exceeds the
maximum allowable platoon length, the subject vehicle will be
a leader of a CACC platoon and keep a constant time gap from
the preceding vehicle (1.5 seconds in this study), referred to as
an inter-platoon gap. If the time gap between the subject vehicle
and the preceding vehicle is more than 2 seconds, the subject
vehicle will switch to the speed regulation mode, which is
represented by Equation (1). Otherwise, a CACC platoon leader
gap regulation mode will be implemented [41] by using
Equation (2) and Equation (3).
        
(2)
     
(3)
where,
: time step for each update (s)
and : gains for adjusting the time gap between the
subject vehicle and preceding vehicle (  and
 ) [42]
: time gap error, which is described by the following:
          , and    
        
: is the constant time gap between the last vehicle of the
preceding CACC platoon and the subject vehicle ( 
for inter-platoon gaps in this study)
: headway between the subject vehicle and immediately
preceding vehicle
: length of the immediately preceding vehicle.
4
: speed of the leading vehicle.
If the subject vehicle can join the preceding CACC platoon,
it will be a CACC platoon follower and apply a smaller time
gap, referred to as an intra-platoon gap, to tightly follow its
preceding vehicle. A survey conducted by [27] proposed a
distribution of desired intra-platoon gaps, and we choose 0.7
seconds in this study to keep the homogeneous driving behavior
among CACC followers. If the time gap is no more than 2
seconds, this CACC platoon follower gap regulation mode uses
the same method using Equation (2) and Equation (3) except
that the following desired constant time gap   seconds
will be replaced with   seconds [41]. For time gaps
larger than 2 seconds, the subject vehicle will turn on the speed
regulation mode (i.e., Equation (1)). When the time gap is
between 1.5 seconds and 2 seconds, the subject vehicle will use
the hysteresis control rule [41], which applies the car-following
mode implemented in the previous time step.
The forward collision warning algorithm [43] developed by
the Collision Avoidance Metrics Partnership (CAMP) is
included in the CACC car following modes to determine
whether the gap between the subject vehicle and the preceding
vehicle is sufficient for safe car following. The CAMP
algorithm determines a required deceleration for the subject
vehicle:
        
    
(4)
where,
: deceleration required to avoid a rear-end collision
(in )
: deceleration of the preceding vehicle (in )
:     
 (in )
Therefore a required deceleration is obtained to avoid a
collision with the preceding vehicle. If  is less than zero,
the brake action is needed to avoid a potential collision.
Although the CACC system implementation relies on
information received from the leading vehicle in the CACC
platoon as well as from the immediately preceding vehicle, the
empirical models used in the simulation provide a simplified
description of the closed-loop vehicle-following dynamics that
are achieved relative to the immediately preceding vehicle.
Fig. 2 summarizes the steps mentioned above and
demonstrates the logic of the CACC platooning process, which
introduces many details of the algorithm implementation. For
each pre-defined step, the subject CAV will detect the
surrounding environment and communicate with the
immediately preceding and following vehicles. If there is no
vehicle in front of the subject vehicle, i.e., the gap exceeding
the pre-defined maximum range, the subject CAV will switch
to the ACC speed regulation mode. If immediately preceding
vehicle is within the pre-defined range and is a VAD vehicle,
the subject CAV will switch to CACC control logic. Because
the VAD-equipped vehicle can be a CACC platoon leader only,
the subject CAV will be the CACC platoon follower. The
specific regulation mode depends on the detected gap. If the
immediately preceding vehicle is a CAV, the subject CAV will
switch to CACC control logic and request the length of the
previous CACC platoon. If the platoon length has reached the
preset maximum value, the subject vehicle will become the
leader of a CACC platoon and maintain a constant inter-platoon
gap from the preceding vehicle (i.e., the last vehicle in the
previous platoon). Otherwise, the subject CAV will become a
CACC platoon follower and try to catch up with the front
CACC platoon using the intra-platoon gap for regulation. There
are three regulation modes, as discussed above, the
determination of which depends on the detected time gap in
real-time.
The implemented CACC control logic is different from the
literature [41] in multiple aspects. First, in this study, since we
assume all vehicles in the managed lane is at least equipped
with VAD, the ACC gap regulation mode does not exist in this
study, which is used to regulate the gap when following a
conventional vehicle within the detection range of the onboard
sensors, and is replaced by CACC gap regulation. Second, we
allow the CACC platoon follower to exceed the speed limit but
no more than 1.1 times of the limit when it catches up with the
preceding CACC platoon to join the platoon. We allow the
maximum length to exceed the regular platoon size threshold at
the merge area. Then, we use a different forward-collision
warning algorithm. Last but not least, due to the integration
with cooperative merge and speed harmonization, the CACC
logic uses the output from other applications as the vehicle
operation input. For example, the lead vehicle of a platoon
follows the speed harmonization, while all platoon followers
strictly implement the CACC logic. The CACC logic is also
used in this study to guarantee safety (through strict gap
regulations and collision avoidance) when three applications
interact.
B. Cooperative Merge
To a large extent, algorithms for cooperative merge combine
speed harmonization and CACC control algorithms at merging
areas because CACC platoons need to operate differently at
merge points to accommodate new merging vehicles, and gaps
can be potentially created in many cases by controlling
Fig. 2. Logic of CACC Platoon
5
managed lane vehicles through I2V speed control. The
cooperative merge algorithm consists of four main steps:
detection, release, speed regulation, and gap regulation.
1) Cooperative Merge-1: Detection
As Fig. 3 (a) shows, at a merge area, a local center (e.g.,
roadside units, or RSU) collects real-time information (location,
speed, acceleration, vehicle operation mode, etc.) of all
vehicles. If there is a desired/qualified time gap between two
CACC and/or VAD vehicles, an available gap will be recorded.
These two vehicles will start to keep this gap if the following
vehicle is a CAV. If the following vehicle is VAD vehicle, the
gap may or may not be kept, and this is not controllable.
Generally, when the traffic is in the uncongested traffic regime
(i.e., the actual density is below the critical density of the local
traffic, which is also dependent on the CAV market
penetration), the qualified gap can be more likely kept.
Otherwise, the gap may be closed by human followers due to
the decelerations of the front vehicles, which run into the slow-
moving traffic at the merge area first. Prediction of human-
driver behavior may be a solution to this problem and we leave
it for future explorations.
2) Cooperative Merge-2: Release
After the gap is recorded, the on-ramp vehicle waiting at a
pre-specified "metered" location will be released when the
leader of the recorded gap arrives at the position , which
includes two components: 1) the vehicle distance traveled by
the identified gap leader when the merge vehicle is accelerating
to the speed of the leader and 2) the vehicle distance traveled
by the identified gap leader when the merge vehicle is
"platooning" (or conducting gap regulation) with the gap leader.
This virtual platooning concept will be discussed later. Here,
we are interested in calculating , the minimum distance
between the merge point  and the release point ,
which could be obtained by calculating the relative distance
 to the merge point  by using Equation (5).
        
 
  
(5)
where,
: speed of the gap leader (m/s)
: time duration for the merging vehicle to accelerate to the
gap leader's speed (m/s)
: acceleration of the on-ramp vehicle (m/s2)
: minimum gap regulation distance (m) (200 m in this
study)
3) Cooperative Merge-3: Speed Regulation
Once the on-ramp vehicle is released, a virtual leader (VL)
and a virtual follower (VF) will be assigned to the on-ramp
vehicle, as shown in Fig. 3(b). These are the two vehicles
immediately before and after the identified gap on the mainline.
The three vehicles aim to form a virtual platoon. Therefore, the
on-ramp vehicle starts to regulate the speed with the reference
speed set as the current speed of the VL when Equation (1)
is applied. When the speed of the on-ramp vehicle
approximately equals the VL ( in this study), the on-
ramp vehicle will change to the gap regulation mode with the
VL, and the VF changes to gap regulation mode with the merge
vehicle. Fig. 3(c) illustrates this regular case. However, if the
merge vehicle cannot accelerate to be approximately as fast as
the VL before the distance to the merge point becomes less than
the minimum gap regulation distance, both the VF and merge
vehicles are in gap regulation mode and follow the VL.
However, they apply different following gaps: the merge
vehicle applies an intra-platoon gap (0.7 seconds in this study),
and the VF tries to keep the identified merge gap (e.g., 1.6
seconds in this study). This is to ensure sufficiently large gaps
are available for merge. We let the VF follow VL, instead of
following the merge vehicle as in the regular cases, because the
merge vehicle is still slow (below the VL, VF, and general
traffic speed at or before the merge area) and following the slow
merge vehicle may cause the VF to significantly slow down.
Therefore, in this case, we let the VF follow the VL and
maintain the gap. Note that when the merge vehicle speed is
close to the VL, the gap regulation mode will switch to the
regular mode as shown in Fig. 3(c).
4) Cooperative Merge-4: Gap Regulation and Merge
The on-ramp vehicle will keep the relative intra-platoon gap
with the VL, and Equations (2) through (4) are applied with
 . The VF begins following the on-ramp vehicle, as
shown in Fig. 3(c) and Equations (2) through (4) are applied.
The gap for VF in the gap regulation is dependent on the vehicle
type of the VF and which vehicle the VF is following. When
the vehicle reaches the pre-specified merge point at the
acceleration lane, the vehicle can change lanes and merge into
the right lane. After this point, the cooperative merge process is
completed, and the CACC platooning logic kicks in to control
how the new vehicle will join existing platoons. Note that
because of the existence of on-ramps and off-ramps, vehicles
may join and leave the platoons. Therefore, the platoons in the
system do not always comprise ten vehicles, the preset
maximum platoon length. Instead, the merge vehicles are
usually able to join the platoon before or behind it and then
continue the operation as a part of the CACC platoon.
In this study, two types of lane change behaviors are applied.
One is a simplified lane change behavior, which is a part of the
preferred lane change behavior for fulfilling cooperative merge
(as discussed in the "Cooperative Merge" Section). If the
identified gap can be maintained to let the on-ramp merging
vehicle merge into the mainline, the merging vehicle will
calculate the time gaps between itself and the virtual
leader/follower. If both gaps meet the critical gap requirements
(e.g., greater than 0.7 seconds in this study), then the subject
vehicle will merge into the mainline immediately. If the
identified gap cannot be maintained or is missed by the merging
vehicle, the merging vehicle will switch to the "Necessary Lane
Change" mode, which is controlled by VISSIM's lane change
model and complete the lane change manually. The subject
vehicle will try to seek a qualified gap to merge into the
mainline manually. This model allows the subject vehicle to use
a reduced safety distance during lane change to facilitate the
merge. The reduced safety distance is 10% of the original safety
distance in this study, which is calibrated by [44]. When the
merging vehicle reaches the end of the acceleration lane of the
merge area, the VISSIM's lane change model will let the vehicle
6
forcefully merge into the mainline. However, it creates a large
disturbance to the mainline traffic, similar to real-world
conditions.
Note that multiple vehicles can be released if a large enough
gap to accommodate multiple vehicles are identified. Also, the
proposed algorithm also intrinsically creates gaps for vehicles
if needed. For example, when we identify a gap of 1.6 seconds
or above on the mainline, we release the vehicle, and the
merging vehicle will form virtual platoons with the mainline
virtual leader and virtual follower. The gap of 1.6 seconds is
large enough, so the virtual follower may not need to slow down
for gap creation. But if the gap threshold value is small (say, 0.8
second), the virtual follower on the mainline will need to slow
down due to the virtual platooning rule to leave enough space
for the merging vehicle (as the middle vehicle of the virtual
platoon). This is indeed a gap creation process.
In the case studies of this paper, we also incorporated a
variation of the cooperative merge strategy, which releases and
creates gaps for on-ramp vehicles if this vehicle's waiting time
at the ramp exceeds threshold σ. The original strategy without
purposely releasing and creating gaps for on-ramp vehicles can
be regarded as  
. Note that, under this updated strategy,
once the merging vehicle is released, it will identify its virtual
leader and follower at the pre-specified gap-identification
location and start forming virtual platoons. Under this scenario,
the virtual follower will need to slow down and create gaps for
the merging vehicle.
(a) Detecting Area and Releasing Area
(b) Virtual Platoon in the Cooperative Merge
(c) Gap Regulation in Cooperative Merge
Fig. 3. Illustration for Cooperative Merge Process
C. Speed Harmonization
Speed harmonization is also referred to as I2V speed control
in this study. It aims to control each individual vehicle's
trajectory (i.e., providing each vehicle with real-time
commands) at both basic freeway segments and merge areas to
coordinate trajectories of vehicles and thus smooth traffic,
particularly at bottleneck locations. This method can be used in
conjunction with cooperative merge at merge areas to create
gaps for merging vehicles effectively. Some other studies [30,
35] also discussed this idea.
In order to achieve this ideal trajectory smoothing paradigm,
we need to be capable of detecting the speed drop and
predicting the corresponding shock wave propagation and
queue dissipation. Then we need to control the CAV to follow
a smooth trajectory so that it properly leads the vehicle queue
and enters the bottleneck right after the queue clears. In the case
of a downstream speed drop, if the congestion is moderate, the
algorithm will seek to smooth the traffic, let the queue dissipate,
and then allow the following CAVs to pass the bottleneck
smoothly at a reasonable speed. Otherwise, if traffic is too
congested and the queue is not anticipated to dissipate in a short
period of time, the CAVs will guide the upstream traffic to
avoid hitting the downstream queue at a sudden full stop by
slowing down and smoothly joining the downstream queue.
The traffic and queue status can be captured either by traffic
sensors located downstream of the traffic or by data collected
by roadside units sent from vehicles equipped with onboard
units. The algorithm records all sensor information, such that
traffic and queue status can be predicted. Then, recommended
trajectories are generated by assuming homogeneous traffic
conditions between each sensor or probe vehicle. Note that this
trajectory-based harmonization strategy is a real-time action
that is updated frequently for each vehicle (e.g., every 2
seconds). Therefore, at the beginning of each time increment,
queue characteristics are updated based on newly detected
traffic conditions, so the actions taken by CAVs may be
modified based on the updated data. Note that in this study, we
assume in the managed lane that 100 percent of vehicles are
connected vehicles (at least equipped with VADs), and
therefore real-time traffic conditions at any location can be
estimated using the data of all probe vehicles that pass the
location during a specified duration. With probe data available
for 100 percent of the vehicles, we can define "virtual detectors"
anywhere in the network and collect corresponding traffic
information. This study does not focus on how to use these data
to best estimate traffic status. In our simulation, we have full
information on the system, and therefore we extract the
information directly to generate I2V speed control commands.
For detailed information on the estimation and more advanced
trajectory construction, interested readers can refer to an earlier
publication for detailed information [34].
This study adopts a heuristic approach for speed commands
generation, building on [35]. We set many virtual loop detectors
along the speed harmonization segment (e.g., 2 kilometers
before the merge area). The traffic speed is monitored at these
points and the values are calculated as the arithmetic mean of
the speed of the past 5 minutes. Two components are considered
in the process: global and local commands. First, the speed
harmonization algorithm gradually slows the vehicle down
based on the difference in traffic conditions at the vehicle's
current location and the bottleneck location. We apply a simple
linear speed transition algorithm to obtain the global
harmonization commands, as shown in:
          
(6)
7
where,
: current speed of the subject vehicle (m/s)
: prevailing speed in the merge area (m/s)
 : the distance between the current position and the
boundary of speed harmonization area
: total length of the speed harmonization area
Second, the algorithm also has the CAVs follow local
harmonization commands, subject to the differences in the front
vehicle's distance and speed. The algorithm uses the data from
the densely deployed virtual detectors to predict the trajectory
of the front vehicle (CAV or VAD vehicle) and the trajectories
of all preceding vehicles before the merge area. When the front
vehicle is predicted to be slower at a certain speed threshold
(i.e., slow front vehicle), then local harmonization is triggered
to issue a spatiotemporal linear speed command using a method
like that shown in Equation (6), except that the distance and
speed are measured for the front vehicle at the predicted slow
moment in the future. This ensures that the CAV does not
tightly follow the front vehicle to join the congestion.
When the CAV is very close to the front vehicle (i.e., less
than a threshold, such as 50 meters in distance gap) and the
speed difference is less than a specified value (e.g., 5 m/s), the
CAV will be asked to slow down in a linear fashion using a
method similar to that in Equation (6), except that the distance
and speed are measured for the front vehicle at the current
moment. For a CAV approaching the front vehicle at a high
speed, this local harmonization ensures operational safety and
let the CAVs more smoothly join the queue.
D. Integrated Application Process
The three CAV applications are also integrated together
during real-time operations, referred to as integrated CAV
operations. The pseudo-code for the integrated application
process is shown below. The CACC module first determines the
CACC status and corresponding operational mode and calculate
the desired speed  for the specific CACC mode. If the
vehicle is within the speed harmonization zone or I2V speed
control is activated, the speed harmonization module calculates
the alternative speed  based on Equation (6), and the
desired speed is set to   .
When the subject CAV approaches the merge area, the
cooperative merge module is activated. This module then
releases an on-ramp vehicle when a qualified gap is identified
on the mainline upstream of the merge area. For mainline
vehicles, if the subject vehicle is a VL, the CACC module keeps
calculating and sending speed and corresponding acceleration
commands to control the VL; if the subject vehicle is a VF, the
VF follows the VL and keeps the identified gap when the on-
ramp vehicle is controlled by speed regulation mode; when the
on-ramp vehicle switches to the gap regulation mode, the VF
begins to follow the on-ramp vehicle and is controlled by gap
regulation mode. One exception is discussed above that the VF
follows the VL even the on-ramp vehicle is in gap regulation
mode (when the on-ramp vehicle speed does not match the VL).
For released on-ramp vehicles, if the on-ramp vehicle reaches
the target speed or the boundary of the gap regulation segment
(defined in the cooperative merge algorithm), it will start the
gap regulation mode; otherwise, it is controlled by the speed
regulation mode.
If a CAV needs to change lane for the cooperative merge, the
lane change behavior is still regulated by the conventional
human lane change model. However, during the lane change
process, all the longitudinal maneuvers are still governed by the
above-mentioned process, including collision avoidance. This
means that this study focuses on longitudinal control of all
automated vehicles.
IV. SIMULATION SCENARIOS AND RESULTS
The goal of the simulation is to investigate the effectiveness
of different CAV applications under the same simulation
environment. It is also of interest to understand the performance
of integrating CAV technologies together. Three analysis
scenarios are identified and investigated in this study. First, we
are interested in implementing the state-of-the-art CACC
algorithm that reflects realistic CACC operations and
evaluating its effectiveness using a separate traffic simulator
VISSIM. We want to understand the pipeline capacity of
CACC, which is the best performance of the managed lane, and
set it as a benchmark when incorporating other technologies,
such as cooperative merge and speed harmonization. Further,
CACC traffic stream performance varies under different CAV
market penetration rates, which also impacts the effectiveness
of other technologies.
Second, freeway merges are always bottlenecks that cause an
excessive delay. When considering merging traffic, the CACC
enhancement to the freeway capacity can be significantly
compromised because of the disturbance to otherwise stable,
but tightly coupled, CACC platoons. We are interested in
understanding the impact of this disturbance and how
cooperative merge can reduce such impact to a possible
minimum. In particular, we hope to understand what the
possible on-ramp volume is under different mainline CACC
mainline volumes to achieve optimal performance. This is of
particular importance when the managed lane operator hopes to
create high-performance traffic streams while accommodating
Algorithm 1 Bundled Application Algorithm
1:   __ // get the CACC status and operation mode
2:   __() // calculate the desired speed in CACC platoon
3: If Speed Harmonization is applied
4:   __
5: End
6: If Cooperative Merge is applied
7: If on-ramp vehicle released
8: If not the VL
9: If Gap Regulation
10:   _ _
11: Else
12:   _ _
13: End
14: If not the VF
15: If meet merge conditions
16: Preparing to merge into the gap
17: End
18: End
19: End
20 End
21: End
22:  =,,
23:  =_( )
24: _ ,
8
some necessary on-ramp demand.
Third, the effects of speed harmonization when it is
integrated with other CAV technologies are also of interest in
this study. There are two potential effects that speed
harmonization can have to improve traffic efficiency. One
effect is to delay or reduce the slow-down or stop-and-go
occurrences at the merging area. The other is its ability to help
create larger gaps by slowing upstream vehicles down before
the merge area and let merging vehicles to more smoothly join
the mainline traffic.
The assumed simulation network is a simple 3-lane freeway
segment with an on-ramp and an off-ramp (see Fig. 4). The
freeway mainline is 7 kilometers long. There is a 2-kilometer
'warm-up' mainline segment in the beginning. The simulated
vehicles will use this segment to reach a stable car-following
state after entering the network. This segment also allows
CACC vehicles to form stable CACC vehicle platoons in the
CACC analysis cases. There is an on-ramp 3 kilometers
downstream from the beginning of the network. An off-ramp is
located 4.65 kilometers from the network beginning. Both the
on-ramp acceleration lane and off-ramp auxiliary lane are 150
meters long. Data collection point locations: the first location
O1 is 500 meters downstream from the end of the on-ramp.
Since this study is interested in the performance of managed
lanes, in our simulation, we only study the left most managed
lane and the corresponding dedicated ramps. The bottom part
of the network is for general purpose traffic and provided for
illustration purpose.
The experiments are simulated in VISSIM 10.07, a
microscopic traffic simulation tool. The driver model API
(Application Programming Interface) and COM (Component
Object Model) interface are applied to model CAVs to realize
the longitudinal control functions of the CACC, speed
harmonization, and cooperative merge logic. For each
simulation run, we collected 60-minute performance after a 15-
minute warm-up period, and the free-flow speed is 104 km/h
(65 mph).
The VISSIM internal model, i.e., the Wiedemann 99 model,
is applied to simulate the car-following behavior of VAD
vehicle in the VISSIM. To simulate driving behavior
realistically, the parameters of the driver model usually need to
be calibrated, and several previous studies have calibrated
various parameter sets using real-world data [45-48]. In this
study, calibrated parameters from [49] are used. This study
suggests five parameters are adjusted from the VISSIM default,
including longitudinal oscillation (CC2), negative speed
difference (CC4), positive speed difference (CC5), safety
distance reduced factor (SDRF), and maximum lookback
distance (LBD). The adjusted driving behavior parameter
values are listed in TABLE I. All other parameters assume
default values in VISSIM, such as CC1 (inter-vehicle gap
mean) = 0.9 seconds.
This study estimates the CACC pipeline capacity under
different percentages of CAV market penetration. When data
are collected to draw fundamental diagrams under each market
penetration scenario, traffic input demand varies from 2,000
vphpl to 4,000 vphpl to generate different conditions. We
collected data of each 15 minutes (after the simulation warm-
up period) and used the four times of each 15-minute volume
data to represent a valid data point to draw the fundamental
diagram.
Note that different CAV market penetration rates are only
applied to mainline traffic. Since the VAD vehicle cannot
perform the cooperative merge in this study, the CAV market
penetration rate of on-ramp traffic is fixed at 100%. This is
corresponding to the policy that VAD vehicles can only use the
managed lane by merging from the general-purpose lane
because manual VAD vehicle merge from the dedicated ramps,
if allowed, may cause disturbance to the mainline managed lane
traffic and even affect CAV cooperative merge operations. We
leave the scenario of mixed traffic on the dedicated ramps to
future analyses. We believe the strategy of only allowing CAVs
to use the dedicated ramps will maximize the benefits.
We validated the selected CACC model [42] by comparing
simulation results and field test data. The field data were
collected from a real-world on-road experiment by using the
FHWA CARMA platform
(https://highways.dot.gov/research/research-
3000m
1500m
150m 100m 150m 100m
2500m
Managed
Lane
Fig. 4. Sketch Plot of A Simple Network
TABLE I
ADJUSTED DRIVING BEHAVIOR PARAMETERS
Parameter
Definition
Default Value
Calibrated
Value
CC2
Longitudinal oscillation: oscillation factor gains on following distance
4 m
12 m
CC4
Negative speed difference: negative speed difference during the following
process
-0.35
-0.1
CC5
Positive speed difference: positive speed difference during the following
process
0.35
0.1
SDRF
Safety distance reduced factor: reduce safety distance during lane change
0.6
0.1
LBD
Look back distance
250 m
1000 m
9
programs/operations/CARMA) and five-vehicle experimental
fleet in July 2018, representing the latest CAV hardware and
software [50]. The selected experimental location is a 17.7-
kilometer (11-mile) segment on I-95 Express Lanes. As shown
in Fig. 5, the speed, acceleration, and time headway profiles of
simulation and field data match each other well. It is expected
that the simulated acceleration is not as noisy as the field data,
but their mean and variance are similar.
Fig. 5. Model Validation
A. Analysis 1- CACC Pipeline Capacity
The density-flow rate diagrams under each market
penetration rate are presented in Fig. 6 (a)-(d). The maximum
flow is considered to be maximum capacity, which is 3,288
vphpl in this study. The results also demonstrated a significant
increase in capacity with the increase of the market penetration
of CACC vehicles. At the market penetration rates of 30%,
50%, 70% and 100%, the capacity increased by 4.1%, 20.7%,
37.8% and 42.0%, respectively. The CACC pipeline capacity
under different market penetration rates is shown in Fig. 6 (e).
We hope to mention that we have used different random
seeds for all simulation scenarios to account for the
stochasticity of the traffic systems, and all data from different
runs using different random seeds are included in the
fundamental diagrams. The arriving patterns of upstream traffic
are different from each other with different random seeds [52].
It can be seen that we can easily see the congestion dynamics
under the 0% and 30% MP scenarios. However, the congestion
dynamics are not clear when the MP is high. This is mostly
because the traffic performance (e.g., stability) and highway
capacity are much improved. In those cases, we cannot see an
obvious breakdown of traffic. While we still see some
congestion dynamics under the 70% MP, we do not see that
under 100%. It is likely because the input of the vehicles is
limited, and the traffic will not break down under normal and
high traffic input (as high as the limit of the simulation
software), meaning that the capacity may have exceeded the
limit of input of simulation. The parameter of standstill distance
(CC0) is set to 1.5 meters, and the following distance in time
(CC1) is set to 0.9 seconds for VAD vehicle. Although this
driving model is not applied for CAV, it influences the input
volume in VISSIM [52]. Since the vehicle length is about 4.8
meters, the average headway is about 1.12 seconds, therefore
lead to a limit of input volume of about 3,214 vphpl. It is also
highly likely to happen in the future when the managed lane
vehicle input is limited at the entrance of the managed lane
before which only human-driven capacity volume is possible,
meaning that the input of the managed lane at the entrance is
limited by the human-driven traffic capacity. The same
phenomena of the fundamental diagrams have been reported by
the literature [33], which reported no breakdown during high
market penetration (above 70%).
Also, Fig. 6 (e) shows that the capacity increase trend flattens
out when the MP gets higher than 70%. This is due to the
vehicle composition assumption of the simulation. In the paper,
we assume all vehicles on the managed lanes are at least VAD-
equipped, which can serve as platoon leaders. When the MP
reaches around 70%, almost all the platoons use VAD vehicles
as the leader. We also only use regular human-driven behavior
for VAD vehicles, which does not cause too much additional
disturbance. Therefore, by the time MP reaches 70%, most of
the vehicles are in relatively long platoons, and most of the
roadway capacity has been exploited. It is easy to understand
and has been shown in our simulation that, if VAD vehicles are
replaced with non-VAD vehicles, this trend line can take the
shape of an exponentially increasing curve because higher MPs
indicate higher probabilities of forming platoons.
(a) Market Penetration = 0 percent
(b) Market Penetration = 30 percent
10
(c) Market Penetration = 70 percent
(d) Market Penetration = 100 percent
(e) Pipeline Capacity under Different Market Penetration
Fig. 6. CACC Pipeline Capacity Analysis
B. Analysis 2 CACC and Cooperative Merge
Vehicles entering via on-ramps bring disturbance to the
mainline managed lane. Their impacts mainly depend on the
mainline managed lane volume. Lighter traffic volume on the
managed lane can provide more and longer gaps that will be
available for vehicles from the on-ramp to merge into. This
section aims to compare managed lane performance with and
without cooperative merge.
Several sensitive analyses have been conducted to find out
the optimal critical gap and releasing strategies. First, we
compared two strategies that release either one on-ramp vehicle
for any identified gaps and multiple vehicles depending on the
sizes of the qualified gaps. The critical gap for one vehicle is
set as 1.6 seconds. As shown in Fig. 7, releasing and operating
more than one merging vehicle can increase traffic performance
for 100% and 30% market penetration cases. For the 70% case,
the result is worse than releasing only one vehicle. This is
mainly because, in this scenario, most CACC platoons are
formed with a VAD leader and CAV followers. Then the
qualified gaps are usually identified between two CACC
platoons (i.e., the inter-platoon gap). Thus, most virtual
followers are VAD vehicles. They cannot maintain the
identified gaps, especially under the congested situation. Even
though the same situation can also happen in the 30% case, the
mainline density is relatively lower than the 70% case, which
could tolerate some disturbance.
We also tested the updated releasing strategy that releases
and creates gaps for on-ramp vehicles if this vehicle's waiting
time at the ramp exceeds threshold . We selected four different
levels of σ, which are 4, 8, 12, and 15 seconds, representing the
minimum guaranteed ramp metering rates of 900, 450, 300, and
240 veh/h, respectively. The strategy that only releases vehicle
when big enough gaps are identified can be regarded as σ=∞.
The results show that the performances of σ=12 seconds are
better than others in both 100% and 30% cases. The scenario of
σ=12 seconds can be seen as a balance point where more on-
ramp vehicles can be released with only moderate disturbance
to the mainline traffic. That can help reduce both the number of
waiting vehicles and total waiting times so that the average
delay of the entire system decreases and the throughput
increases. For the case of 70% market penetration, as we
discussed in the last paragraph, the frequent release on-ramp
vehicles cause significant disturbance to the mainline traffic
because many of the identified gaps may disappear when the
follower does not create gaps. In this case, the case of σ=∞
performs the best because fewer vehicles are released, and they
are only released when there are potential gaps.
Note that in the case studies, the merge algorithm only allows
merge vehicles to join gaps over a threshold. This threshold is
set at 1.6 seconds in this study, which is slightly higher than the
inter-platoon gap. If the threshold is too large, the roadway is
not fully utilized. As shown in Fig. 7, two smaller thresholds,
0.9 and 1.2 seconds, are tested. It can be found, as expected,
that if the threshold is too small, the virtual follower needs to
slow down frequently during congestion to create gaps for the
merging vehicle due to the virtual platooning process of the
cooperative merge, therefore causing significant disturbance to
the local traffic. In this paper, this value has been selected
through many simulation runs under different scenarios (e.g.,
congestion level) to ensure the merging maneuver does not
cause significant disturbance to the local traffic and the
corresponding throughput and delay are best on average across
the trail runs.
TABLE II shows selected simulation results with combined
CACC and cooperative merge under different CAV market
penetrations. The managed lane input (ML) volume is around
85% of the pipeline capacity, such that additional vehicles are
allowed to enter the segment through on-ramps. The merge
input volume is generated by actual simulation runs with
cooperative merge algorithms, and it is partly dependent on the
available merging gaps. We use the same merge volume for the
corresponding non-cooperative merge of the same MP
scenarios to compare results with the cooperative merge.
As we see, the system improvement due to the cooperative
merge is dramatic when the market penetration is at 100%. The
throughput improvement is 12.77% (from 2,850 vphpl to 3,214
vphpl) and the delay reduction is 80.88% (163.82 seconds to
11
31.32 seconds). For the 70% MP scenarios, we release a single
vehicle based on release gaps, instead of multiple-vehicle
platoon release, because this strategy performs better as shown
and discussed above. The throughput improvement at 70% is
larger than in other cases, but the average delay is slightly
worse. But with lower penetration rates such as 30%, the benefit
of cooperative merge still maintains at a good level. It is mainly
because the low penetration rates mean additional gaps that
merging vehicles can use to merge into the traffic since the
desired gaps between manual vehicles are usually larger than
the critical gap (i.e., 1.6 sec) used in this study. But a lot of these
gaps cannot be maintained because of the human-driven virtual
followers. Therefore, the net benefits of cooperative merge
under lower MP rates are smaller than under high MP rates,
though by a small margin.
In terms of reduction from pipeline capacity, it is intuitive
that the 100% case is better than 70% because the 100% MP
allows full cooperation and, therefore, less disturbance from the
merging traffic to the mainline. However, under lower
penetration, the reduction in capacity is not obvious. This is
because 1) the pipeline capacity is not high, and many gaps are
available for merging, and 2) more CAVs entering the network
from ramp actually increases the percentage of cooperation
more than other high-MP cases.
C. Analysis 3 Effects of Speed Harmonization
For all the scenarios in Analysis 2, this section evaluates
them again with the addition of speed harmonization. Speed
harmonization effects are also analyzed with and without a
cooperative merge.
TABLE III shows the simulation results with integrated
CACC, cooperative merge, and speed harmonization under
different CAV market penetrations. We use the same method
for generating the managed lane and on-ramp traffic volumes.
Speed harmonization is effective from 1,500 meters upstream
of the merging area. The length of the speed harmonization
zone is obtained through multiple trial runs, which shows the
1,500-meter zone gives the best performance. The speed
harmonization algorithm is implemented as automatic speed
control of CAVs, and therefore VADs are not affected by speed
harmonization directly, but they may be harmonized by nearby
CAVs. Also, note that speed harmonization is only activated
when the merge area average speed is below 80 km/h (50 mph);
otherwise, the mainline delay will drastically increase.
Similar to the results in TABLE II, the improvement of the
cooperative merge, when speed harmonization is implemented,
is significant when the MP is at 100%. The throughput
improvement is 9.54% (from 2,934 vphpl to 3,214 vphpl) and
the delay reduction is 79.59% (153.46 seconds to 31.32
seconds). Other trends are also similar to those in TABLE II.
In the extreme case of 0% penetration rate, the performance is
exactly the same as in TABLE II because speed harmonization
does not affect VAD vehicles. Similarly, the reduction in
capacity compared with the CACC pipeline capacity is also
most significant when the CAV penetration rate is 100%, but
the cooperative merge is much better than the non-cooperative
merge scenario, implying the importance of cooperative merge.
The effect of speed harmonization can be reflected in
multiple aspects. Compared with Analysis 2 in TABLE III, the
average delay further reduced significantly in most of the cases
because of the smoothing and breakdown prevention effects.
Note that the effect of speed harmonization more significant
when the cooperative merge is not implemented. This is
intuitive because there is already not much space for
improvement after CACC and cooperative merge is
implemented. The same argument can be made to explain why
there is no further improvement for the 100% MP scenario.
Additionally, we notice that the changes in throughput after
implementing speed harmonization, with and without
cooperative merge, is quite limited, close to zero. This is in line
with the simulation result obtained in [53], indicating greater
delay improvement and small throughput benefits, unless the
algorithm is specifically designed for throughput enhancement.
The last two columns of TABLE III show the comparison of
overall system performance. These interesting results further
confirm the importance of bundling application together to
realize the greatest potential of each component application.
V. CONCLUSION AND FUTURE RESEARCH
The emergence of CAV technologies offers extensive
opportunities to advance safety, mobility, and reliability on the
US roadways. The market penetration of these vehicles is,
however, expected to be low in the next decade, and as such,
their potential benefits may not be fully realized. The use of
managed lane facilities can support the realization of these
benefits at early deployment stages. This study focuses on
deployment stages of low market penetration and evaluates how
the proposed integrated application of speed harmonization,
cooperative adaptive cruise control (CACC), and cooperative
merging is operated to improve existing system performance.
This study proposes an integrated algorithm for the
integrated CAV application. Through microscopic simulation,
this study estimates capacity shifts under CAV market
penetrations, providing operational insights and guidance for
traffic management centers to control the volume on the
managed lanes to maintain the desired speed. Particularly, the
study examines the effectiveness of CACC, CACC plus
cooperative, and the addition of speed harmonization, under
different penetration rates. Simulation results show the
effectiveness of the integrated application to enhance system
throughputs and reduce delay, even with low CAV penetration
rates. The speed harmonization only shows significant
effectiveness with high CAV penetration, but the potential
safety benefits it brings to the system, though not evaluated in
the study, can be quite significant.
12
virtual platoons reach the merge area and human-driven C
(a) Market Penetration = 30 percent
(b) Market Penetration = 70 percent
(c) Market Penetration = 100 percent
Fig. 7. Sensitive Analysis
0
50
100
150
200
250
300
350
400
0
500
1,000
1,500
2,000
2,500
3,000
3,500
Platoon Release,
σ = 4 s, critical
gap = 1.6s
Platoon Release,
σ = 8 s, critical
gap = 1.6s
Platoon Release,
σ = 12 s, critical
gap = 1.6s
Platoon Release,
σ = 15 s, critical
gap = 1.6s
Single Release
Only, σ = ∞,
critical gap =
1.6s
Platoon Release
Only, σ = ∞,
critical gap =
1.6s
Platoon Release,
σ = 12 s, critical
gap = 0.9s
Platoon Release,
σ = 12 s, critical
gap = 1.2s
Platoon Release
+ SH, σ = 12 s,
critical gap =
1.6s
Average Delay (s/veh)
Throughput (veh/h)
Throughput and Average Delay ( MP = 30%)
Throughput Average Delay
0
50
100
150
200
250
300
350
400
450
0
500
1,000
1,500
2,000
2,500
3,000
3,500
Platoon Release,
σ = 4 s, critical
gap = 1.6s
Platoon Release,
σ = 8 s, critical
gap = 1.6s
Platoon Release,
σ = 12 s, critical
gap = 1.6s
Platoon Release,
σ = 15 s, critical
gap = 1.6s
Single Release
Only, σ = ∞,
critical gap = 1.6s
Platoon Release
Only, σ = ∞,
critical gap = 1.6s
Platoon Release,
σ = 12 s, critical
gap = 0.9s
Platoon Release,
σ = 12 s, critical
gap = 1.2s
Average Delay (s/veh)
Throughput (veh/h)
Throughput and Average Delay ( MP = 70%)
Throughput Average Delay
0
40
80
120
160
200
0
500
1,000
1,500
2,000
2,500
3,000
3,500
Platoon Release,
σ = 4 s, critical
gap = 1.6s
Platoon Release,
σ = 8 s, critical
gap = 1.6s
Platoon Release,
σ = 12 s, critical
gap = 1.6s
Platoon Release,
σ = 15 s, critical
gap = 1.6s
Single Release
Only, σ = ∞,
critical gap =
1.6s
Platoon Release
Only, σ = ∞,
critical gap =
1.6s
Platoon Release,
σ = 12 s, critical
gap = 0.9s
Platoon Release,
σ = 12 s, critical
gap = 1.2s
Platoon Release
+ SH, σ = 12 s,
critical gap =
1.6s
Average Delay (s/veh)
Throughput (veh/h)
Throughput and Average Delay ( MP = 100%)
Throughput Average Delay
13
Future research directions have been identified. First,
additional simulation analysis should be conducted to
understand other impacts on the system, such as safety benefits.
Also, this study assumes a 100% communication success rate,
and it is interesting to study the potential consequences when
the communication performance is compromised. This may not
only lead to safety concerns but also cause frequent status
changes between platooning and human-driven modes,
therefore resulting in unstable platoons and oscillatory traffic.
Fourth, all merging vehicles on the left-side dedicated ramps
are assumed to be CAVs to ensure their cooperative behavior.
This assumption is made for our purpose of creating high-
performance traffic streams on managed lanes but can be
relaxed in future studies. It is interesting to investigate the
system performance when the merging traffic, from either left-
side or right-side ramps, is a mix of conventional vehicles,
VAD vehicles, and CAVs. Non-CAVs may cause traffic
disturbances in the merging area, and it is worth investigating
how the integrated freeway CAV application handles the
disturbances and smooths the traffic. Fifth, it is critical to
evaluate the integrated application in a well-calibrated real-
world network to understand the effectiveness of complex
traffic and geometric settings. The research team is currently
conducting a case study using an Interstate 66 network, a 21-
kilometer (13-mile) freeway segment outside the Interstate 495
Beltway near Washington, D.C.
ACKNOWLEDGMENT
The authors want to thank a few other project team members
for their contributions to the paper: Steven Shladover, Xiao-
Yun LU, Hao Liu, and Robert Ferlis. The work presented in this
paper remains the sole responsibility of the authors.
REFERENCES
[1] Bhargava, Abhishek, Euridice Oware, Samuel Labi, and Kumares C.
Sinha. "Ramp Metering and High occupancy vehicle facilities: A
synthesis study." Joint Transportation Research Program (2006): 281.
[2] Scarinci, Riccardo, and Benjamin Heydecker. "Control concepts for
facilitating motorway on-ramp merging using intelligent
vehicles." Transport reviews 34, no. 6 (2014): 775-797.
[3] Mahmassani, Hani S. "50th anniversary invited articleautonomous
vehicles and connected vehicle systems: flow and operations
considerations." Transportation Science 50, no. 4 (2016): 1140-1162.
[4] Hussain, Omar, Amir Ghiasi, and Xiaopeng Li. "Freeway lane
management approach in mixed traffic environment with connected
autonomous vehicles." arXiv preprint arXiv:1609.02946 (2016).
[5] Liu, Hao, Xingan Kan, Steven E. Shladover, Xiao-Yun Lu, and Robert E.
Ferlis. "Impact of cooperative adaptive cruise control on multilane
freeway merge capacity." Journal of Intelligent Transportation
Systems 22, no. 3 (2018): 263-275.
[6] Van Mierlo, Joeri, Gaston Maggetto, Erik Van de Burgwal, and Raymond
Gense. "Driving style and traffic measures-influence on vehicle emissions
and fuel consumption." Proceedings of the Institution of Mechanical
TABLE II
SELECTED RESULTS WITH COMBINED CACC AND COOPERATIVE MERGE
MP
ML
Volume
(vphpl)
Merge
Volume
(vphpl)
Throughput
(Coop / Non-
coop, vphpl)
Average
Delay (Coop
/ Non-coop,
seconds)
Pipeline
Capacity
(vphpl)
Improvement of
Throughput
(%)
Reduction of
Average
Delay
(%)
Reduction
from Pipeline
Capacity
(%)
100%
2800
436
3214
31.32
3288
12.77
80.88
-2.25
2850
163.82
-13.32
70%
2700
289
3012
36.74
3192
19.61
67.15
-5.64
2518
111.87
-21.11
30%
2050
610
2616
25.48
2412
10.66
72.42
+8.46
2364
92.40
-1.99
0%
1970
382
-
-
2316
N/A
N/A
-10.49
2073
132.23
MP = market penetration rate. ML = managed lane. TABLE III
SELECTED RESULTS WITH BUNDLED CACC, COOPERATIVE MERGE AND SPEED HARMONIZATION
MP
(%)
ML Volume
(vphpl)
Merge Volume
(vphpl)
Throughput (Coop /
Non-coop, vphpl)
Average Delay (Coop
/ Non-coop, seconds)
Pipeline Capacity
(vphpl)
100
2800
436
3214
31.32
3288
2934
153.46
70
2700
440
2942
32.38
3192
2484
65.10
30
2050
617
2647
23.75
2412
2244
75.85
0
1970
382
-
-
2316
2073
132.23
MP
(%)
Improvement of
Throughput
(%)
Reduction of Average
Delay (%)
Reduction from
Pipeline Capacity
(Coop / Non-coop, %)
Throughput
Compared with
Analysis 2 (Coop /
Non-coop, %)
Average Delay
Compared with
Analysis 2 (Coop /
Non-coop, %)
100
9.54
79.59
-2.25
0.00
0.00
-10.77
+2.95
-6.32
70
15.57
50.26
-7.83
-2.32
-11.87
-22.18
-1.35
-41.81
30
15.22
68.69
+9.74
+1.18
-6.79
-6.97
-5.08
-17.91
MP = market penetration rate. ML = managed lane.
14
Engineers, Part D: Journal of Automobile Engineering 218, no. 1 (2004):
43-50.
[7] Vahidi, Ardalan, and Antonio Sciarretta. "Energy saving potentials of
connected and automated vehicles." Transportation Research Part C:
Emerging Technologies(2018).
[8] Talebpour, Alireza, and Hani S. Mahmassani. "Influence of connected
and autonomous vehicles on traffic flow stability and
throughput." Transportation Research Part C: Emerging
Technologies 71 (2016): 143-163.
[9] Li, Tianxin, and Kara M. Kockelman. "Valuing the safety benefits of
connected and automated vehicle technologies." In Transportation
Research Board 95th Annual Meeting, no. 16-1468. 2016.
[10] Siegel, Joshua E., Dylan C. Erb, and Sanjay E. Sarma. "A survey of the
connected vehicle landscapeArchitectures, enabling technologies,
applications, and development areas." IEEE Transactions on Intelligent
Transportation Systems 19, no. 8 (2017): 2391-2406.
[11] Guanetti, Jacopo, Yeojun Kim, and Francesco Borrelli. "Control of
connected and automated vehicles: State of the art and future
challenges." Annual reviews in control 45 (2018): 18-40.
[12] Elliott, David, Walter Keen, and Lei Miao. "Recent advances in connected
and automated vehicles." Journal of Traffic and Transportation
Engineering (English Edition) (2019)
[13] ITSJPO. "Connected Vehicle Applications", accessed August 31, 2019,
https://www.its.dot.gov/pilots/cv_pilot_apps.htm
[14] S. Shladover, D. Su and X.Y. Lu, "Impacts of cooperative adaptive cruise
control on freeway traffic flow," Transportation Research Record: Journal
of the Transportation Research Board, vol. 2324, pp. 63-70, 2012.
[15] Naus, Gerrit JL, Rene PA Vugts, Jeroen Ploeg, Marinus JG van De
Molengraft, and Maarten Steinbuch. "String-stable CACC design and
experimental validation: A frequency-domain approach." IEEE
Transactions on vehicular technology59, no. 9 (2010): 4268-4279.
[16] Ntousakis, Ioannis A., Ioannis K. Nikolos, and Markos Papageorgiou.
"Optimal vehicle trajectory planning in the context of cooperative
merging on highways." Transportation research part C: emerging
technologies 71 (2016): 464-488.
[17] Letter, Clark, and Lily Elefteriadou. "Efficient control of fully automated
connected vehicles at freeway merge segments." Transportation
Research Part C: Emerging Technologies 80 (2017): 190-205.
[18] Malikopoulos, Andreas A., Seongah Hong, B. Brian Park, Joyoung Lee,
and Seunghan Ryu. "Optimal control for speed harmonization of
automated vehicles." IEEE Transactions on Intelligent Transportation
Systems (2018).
[19] Learn, Stacy, Jiaqi Ma, Kelli Raboy, Fang Zhou, and Yi Guo. "Freeway
speed harmonisation experiment using connected and automated
vehicles." IET Intelligent Transport Systems12, no. 5 (2017): 319-326.
[20] Hall, Randolph W., Ali Nowroozi, and Jacob Tsao. "Entrance capacity of
an automated highway system." Transportation Science 35, no. 1 (2001):
19-36.
[21] Fitzpatrick, K., C. Fuhs, M. A. Brewer, D. Ungemah, T. Lindheimer, S.
T. Chrysler, R. Avelar, N. Wood, and C. Swenson. "NCHRP Web-Only
Document 224: Research Supporting the Development of Guidelines for
Implementing Managed Lanes." Transportation Research Board,
Washington, DC.
[22] Ghiasi, Amir, Omar Hussain, Zhen Sean Qian, and Xiaopeng Li. "A
mixed traffic capacity analysis and lane management model for connected
automated vehicles: A Markov chain method." Transportation Research
Part B: Methodological 106 (2017): 266-292.
[23] Xiao, Lin, Meng Wang, Wouter Schakel, and Bart van Arem.
"Unravelling effects of cooperative adaptive cruise control deactivation
on traffic flow characteristics at merging bottlenecks." Transportation
research part C: emerging technologies 96 (2018): 380-397.
[24] Rakouth, Heri, Paul Alexander, Walter Kosiak, Masao Fukushima, Lali
Ghosh, Chris Hedges, Henry Kong, Sven Kopetzki, Ramesh Siripurapu,
and Junqiang Shen. "V2X communication technology: Field experience
and comparative analysis." In Proceedings of the FISITA 2012 World
Automotive Congress, pp. 113-129. Springer, Berlin, Heidelberg, 2013.
[25] J. Ma and E. Leslie, "Managed lanes for early deployment of connected
and automated vehicle applications: concept of operations," submitted to
98th Transportation Research Board Annual Meeting.
[26] C. Nowakowski, J. O'Connell, S. Shladover, and D. Cody, "Cooperative
adaptive cruise control: driver selection of car-following gap settings less
than one second," presented at 54th Proc. Hum. Factors Ergon. Soc.
Annu. Meet., San Francisco, CA, USA, Sep. 27-Oct. 1, 2010.
[27] C. Nowakowski, J. O'Connell, S. Shladover, and D. Cody, "Cooperative
adaptive cruise control: driver acceptance of following gap settings less
than one second," in 54th Proc. Hum. Factors Ergon. Soc. Annu. Meet.,
San Francisco, CA, USA, 2010, pp. 2033-2037.
[28] Shladover, Steven E., Xiao-Yun Lu, and Delphine Cody. "Development
and evaluation of selected mobility applications for VII: concept of
operations." University of California, Berkeley (2008).
[29] S. Shladover, C. Nowakowski, X. Lu, and R. Ferlis, "Cooperative
adaptive cruise control (CACC) definitions and operating concepts,"
Transportation Research Record: Journal of the Transportation Research
Board, vol. 2489, pp. 145-162, 2015.
[30] Raboy, Kelli, Jiaqi Ma, Edward Leslie, and Fang Zhou. "A proof-of-
concept field experiment on cooperative lane change maneuvers using a
prototype connected automated vehicle testing platform." Journal of
Intelligent Transportation Systems (2020): 1-16.
[31] F.C. Chou, S. Shladover, "Coordinated merge control based on V2V
communication," in 2016 IEEE Vehicular Networking Conference (VNC),
2016.
[32] C. Fuhs, P. Brinckerhoff. "Synthesis of active traffic management
experiences in Europe and the United States," FHWA, Washington, D.C.,
USA, Rep. FHWA-HOP-10-031, 2010.
[33] A. Talebpour, H.S Mahmassani, and S.H. Hamdar. "Speed
harmonization: evaluation of effectiveness under congested conditions,"
Transportation Research Record: Journal of the Transportation Research
Board, vol. 2391, pp. 69-79, 2013.
[34] A. Ghiasi., J. Ma, F. Zhou, and X. Li, "Speed harmonization algorithm
using connected autonomous vehicles," in 96th Transportation Research
Board Annual Meeting, 2017.
[35] J. Ma, Li. Xiao, S. Shaladover, H.A. Rakha, X.Y. Lu, R. Jagannathan, and
D.J. Dailey, "Freeway speed harmonization," IEEE Trans. Intelligent
Vehicles, vol.1, no. 1, pp. 78-89, 2016.
[36] A. Hafez, M. Rahman, and H. Zamzuri. "Effect of leader information
broadcasted throughout vehicle platoon in a constant spacing policy." In
2015 IEEE International Symposium on Robotics and Intelligent Sensors
(IRIS), pp. 132-137. IEEE, 2015.
[37] W. Levine, and M. Athans. "On the optimal error regulation of a string of
moving vehicles." IEEE Transactions on Automatic Control 11, no. 3
(1966): 355-361.
[38] L. Peppard, "String stability of relative-motion PID vehicle control
systems." IEEE Transactions on Automatic Control 19, no. 5 (1974): 579-
581.
[39] M. Bertozzi, and A. Broggi. "GOLD: A parallel real-time stereo vision
system for generic obstacle and lane detection." IEEE transactions on
image processing 7, no. 1 (1998): 62-81.
[40] S. Sheikholeslam, and C. Desoer. "Longitudinal control of a platoon of
vehicles with no communication of lead vehicle information: A system
level study." IEEE Transactions on vehicular technology 42, no. 4 (1993):
546-554.
[41] H. Liu, X. Kan, S. Shladover, and X. Lu, "Using cooperative adaptive
cruise control (CACC) to form high-performance vehicle streams: final
report," FHWA, Washington, D.C., USA, 2018.
[42] V. Milanés., S. Shladover, "Modeling cooperative and autonomous
adaptive cruise control dynamic responses using experimental data,"
Transportation Research Part C, vol. 48, pp. 285-300, 2014.
[43] R.J. Kiefer, M.T. Cassar, C.A. Flannagan, D.J. LeBlanc, M.D. Palmer,
R.K. Deering, and M.A. Shulman, "Forward collision warning
requirements project: refining the CAMP crash alert timing approach by
examining "last-second" braking and lane change maneuvers under
various kinematic conditions," NHTSA, Washington, D.C., USA, Rep.
DOT HS 809 574, 2003.
[44] E. Miller-Hooks, C. Chou, L. Feng, and R. Faturechi. Concurrent flow
lanes-phase III. No. MD-11-SP009B4P. Maryland. State Highway
Administration. Office of Policy & Research, 2011.
[45] Kim, Seung-Jun, Wonho Kim, and Larry R. Rilett. "Calibration of
microsimulation models using nonparametric statistical techniques."
Transportation Research Record 1935, no. 1 (2005): 111-119.
[46] Park, Byungkyu, and Hongtu Qi. "Development and Evaluation of a
Procedure for the Calibration of Simulation Models." Transportation
Research Record 1934, no. 1 (2005): 208-217.
[47] Menneni, Sandeep, Carlos Sun, and Peter Vortisch. "Microsimulation
calibration using speed-flow relationships." Transportation Research
Record 2088, no. 1 (2008): 1-9.
[48] Liu, Pengfei, and Wei Fan. "Exploring the impact of connected and
autonomous vehicles on freeway capacity using a revised Intelligent
Driver Model." Transportation Planning and Technology 43, no. 3 (2020):
279-292.
15
[49] Miller-Hooks, E., M. TariVerdi, and X. Zhang. "Standardizing and
Simplifying Safety Service Patrol BenefitCost Ratio
Estimation." Report to the I-95 Corridor Coalition 14 (2012).
[50] Ma, Jiaqi, Edward Leslie, Amir Ghiasi, Zhitong Huang, and Yi Guo.
"Empirical analysis of a freeway bundled connected-and-automated
vehicle application using experimental data." Journal of Transportation
Engineering, Part A: Systems 146, no. 6 (2020): 04020034.
[51] Liu, Hao, Xingan David Kan, Dali Wei, Fang-Chieh Chou, Steven E.
Shladover, and Xiao-Yun Lu. "Using Cooperative Adaptive Cruise
Control (CACC) to Form High-Performance Vehicle Streams
Microscopic Traffic Modeling." (2018).
[52] PTV Group. "PTV Vissim 10.0 User Manual." PTV AG, Karlsruhe,
Germany (2018).
[53] D. Hale, T. Phillips, K. Raboy, J. Ma, P. Su, X. Lu, H. Rakha, and D.
Dailey. "Introduction of cooperative vehicle-to-infrastructure systems to
improve speed harmonization". No. FHWA-HRT-16-023. United States.
Federal Highway Administration. Office of Operations Research and
Development, 2016.
Yi Guo is a research assistant and Ph.D.
student at the University of Cincinnati. He
received his B.S. degree in information
management and information system at the
Beijing Jiaotong University in 2011 and
Ph.D. degree in management information
system at the Beijing Jiaotong University
in 2017. He focuses on intelligent
transportation systems, connected
automated vehicles, traffic modeling and simulation, and
machine learning. He is also interested in system design,
implementation, optimization, and application of Internet of
Things. From 2010 to 2016, he worked as a research and
development engineer and project manager in JPort team for
Guangzhou Port. Since 2011, he has been a research assistant
at the Beijing Jiaotong University until 2017.
Dr. Jiaqi Ma is the Director of the
Advanced Transportation Collaborative
at the University of Cincinnati. Prior to
that, he is a Project Manager and Research
Scientist at the Federal Highway
Administration Turner-Fairbank
Highway Research Center, and a
contractor researcher at the Virginia
Transportation Research Council of the
Virginia Department of Transportation
(DOT). He has led and managed many research projects funded
by U.S. DOT, National Science Foundation, and state DOTs,
covering wide range areas of smart transportation systems, such
as vehicle-highway automation, Intelligent Transportation
Systems (ITS), connected vehicles, shared mobility, and large-
scale smart system modeling and simulation, travel behavior
modeling and demand forecasting, and artificial intelligence
and advanced computing applications in transportation. He is a
Member of the Transportation Research Board Standing
Committee on Vehicle-Highway Automation, and Co-Chair of
IEEE ITS Society Technical Committee on Smart Mobility and
Transportation 5.0.
Edward Leslie of Leidos has a
background in research, design,
development, and testing of electrical and
computer-related equipment for scientific
use. He has led the design and test of
power supplies ranging in output from
100mW to 180kW for use in vehicle and
infrastructure power systems. Mr. Leslie
designs electronic circuits and
components for use in the field, including propulsion control
for ground vehicles. He has specified and installed
instrumentation for control and data acquisition on diesel-
electric hybrid gensets. He performs COTS hardware and
software integration at a higher-systems level, required to
control vehicle guidance systems, including control
engineering, optimization, testing and validation, and control
system design. He has supported numerous connected vehicle
applications, including Cooperative Adaptive Cruise Control,
Speed Harmonization, and Merge/Weave that take advantage
of V2V and V2I communication. Mr. Leslie has also tested and
validated the performance of Roadside Equipment from
multiple vendors against the USDOT RSU 4.0 Specification.
Dr. Zhitong Huang has over 9 years of
research experience and conducted over 15
research projects in the field of
transportation engineering. His main focus
is on transportation simulation and
modeling, connected and automated
vehicle (CAV) systems, and traffic
operation and management. Dr. Huang is
conducting several CAV research projects
at the Federal Highway Administration
Saxton Transportation Operations Laboratory, Tuner Fairbank
Highway Research Center. He is currently leading or
participating several CAV simulation and modeling projects.
He is also involved in testing Roadside Units (RSU) from
different vendors against the latest USDOT RSU Specification
(version 4.1). Dr. Huang is also working on another research
project, providing RSU engineering support and procurement
services for connected vehicle testbeds. He is familiar with
operations of, and standards for, DSRC Roadside Units and
Signal Phase and Timing (SPaT) and MAP DSRC Messages
and systems. In addition to DSRC standards, he has deep
understanding for ITS standards, such as NTCIP.
... Basically, platooning means cooperation among vehicles and it is therefore based on V2V communications. On a more ambitious scale, for instance if platooning is to be combined with other active traffic management (e.g., [20]) or logistics strategies (e.g., [21]), the involvement of the infrastructure and, thus, of vehicle-to-infrastructure (V2I) or vehicle-to-all (V2X) communications would be necessary. For an overview of these available communications and their desired requirements for platooning applications (e.g., latencies, safety protocols, etc.), readers can refer to [22], where the data processing phase is also explained. ...
... If multiple lanes are considered, some works establish a specific lane for platoons and, in some cases, this is defined as a dedicated lane, meaning that only vehicles within a platoon are accepted (e.g., [23]). Other researchers allow platoons to form in any lane and may even include more complex road stretches with on-or offramps (e.g., [20]). Finally, there are works focusing on the overall road network, especially those dealing with planning and routing (e.g., [21]). ...
... Other authors such as [20] and [23] worked on scenarios with dedicated facilities for CAVs, and combined platooning based on the predecessor-leader-following IFT with other active traffic management strategies. Particularly, [23] simulated an oval dedicated track of 3,965 m for CAVs, where no lane changes were allowed, with 10 points with on-and off-ramps. ...
Article
Full-text available
Platooning of connected automated vehicles (CAVs) on highways has attracted the interest of researchers, companies and administrations for years. This interest has yielded a vast literature on CAV platooning, which generally describes its benefits in terms of traffic efficiency, safety and environmental impacts, among others. However, the boundary conditions of these studies and the reported magnitude of these benefits vary greatly among the different contributions. In this context, this paper presents the result of a comprehensive literature review on CAV platooning on highways, organizing the existing research and establishing links between the different strategies proposed and their expected impacts. Starting points for future investigations are suggested, and the research gaps to be addressed are identified. The analysis performed shows that the current state-of-the-art has two major limitations. First, few CAV platooning strategies have been proven to be asymptotically stable, which should be a requirement for any feasible solution. Second, in most cases, the CAV platooning algorithms are simplistic adaptations from existing car-following models proposed for human driving that, therefore, unnecessarily transfer current traffic-related problems to future cooperative environments.
... Likewise, the algorithm proposed in refs. [91,92] integrated CACC, cooperative merge, and speed harmonization in platoons of equipped vehicles. Under different penetration rates, the effectiveness of the bundled application was examined, proving system throughput enhancement and reductions in delay even with low CAV penetration rates. ...
Article
Full-text available
Since Leonardo da Vinci’s creation of a self-propelled cart in the 1500s (Palmer. in Significant figures in world history p. 75--7, 2018), the evolution of Autonomous Vehicles (AVs) has aimed to revolutionize transportation. While AVs promise improved safety, traffic efficiency, and industrial optimization by reducing human intervention, ensuring their security remains paramount. This paper provides a thorough literature review spanning from historical milestones to contemporary advancements in AV technology. It delves into the significance of Vehicular Ad-hoc NETworks (VANETs) for safety applications and underscores the critical role of speed harmonization and string stability in safeguarding AV platoons. Furthermore, the paper addresses cybersecurity threats targeting platoon networks, advocating for research into encryption mechanisms, road-side units, control algorithms, hybrid communications, and on-board system security to bolster communication security within platoons. By advocating for a balance between AV technological advancements and robust security measures, this paper facilitates safe and reliable AV platooning operations.
... The latest advancements in V2X communication technology and Intelligent Transportation Systems enable intelligent agents to share information between each other [15][16][17][18][19]. Through multi-agent interactions, the limitations of individual agent perception, such as occlusion and long-range issues, can be alleviated, thereby enhancing the performance of camera-based 3D object detection. ...
Preprint
Multi-agent collaborative perception has emerged as a widely recognized technology in the field of autonomous driving in recent years. However, current collaborative perception predominantly relies on LiDAR point clouds, with significantly less attention given to methods using camera images. This severely impedes the development of budget-constrained collaborative systems and the exploitation of the advantages offered by the camera modality. This work proposes an instance-level fusion transformer for visual collaborative perception (IFTR), which enhances the detection performance of camera-only collaborative perception systems through the communication and sharing of visual features. To capture the visual information from multiple agents, we design an instance feature aggregation that interacts with the visual features of individual agents using predefined grid-shaped bird eye view (BEV) queries, generating more comprehensive and accurate BEV features. Additionally, we devise a cross-domain query adaptation as a heuristic to fuse 2D priors, implicitly encoding the candidate positions of targets. Furthermore, IFTR optimizes communication efficiency by sending instance-level features, achieving an optimal performance-bandwidth trade-off. We evaluate the proposed IFTR on a real dataset, DAIR-V2X, and two simulated datasets, OPV2V and V2XSet, achieving performance improvements of 57.96%, 9.23% and 12.99% in AP@70 metrics compared to the previous SOTAs, respectively. Extensive experiments demonstrate the superiority of IFTR and the effectiveness of its key components. The code is available at https://github.com/wangsh0111/IFTR.
... This era has witnessed the rapid development of intelligent transportation systems and autonomous driving and significant improvements in terms of driving safety, traffic efficiency, and energy consumption (Lin et al. [1], Kloukiniotis et al. [2], Elmorshedy et al. [3] and Meng et al. [4]). Among the autonomy techniques for transportation and mobility (Guo and Ma [5] and Guo et al. [6]), vehicle localization, one of the key modules of autonomous vehicles, has attracted substantial attention. Continuous and reliable localization is of critical importance for environmental perception, trajectory planning, decision-making, and motion control for autonomous driving or intelligent transportation systems (Raboy et al. [7], Ma et al. [8], Jeong et al. [9] and Valiente et al. [10]). ...
Article
Full-text available
The purpose of this study is to define the top-priority operational concept involving connected automated vehicle operation on managed lanes (CAV ML). The policies and scenarios associated with this concept should be chosen and designed to facilitate the rapid and successful deployment of a solution that has yet to be deployed anywhere. The proposed CAV operation involves platoons of equipped vehicles governed by an integrated set of cooperative adaptive cruise control (CACC), cooperative merging, and speed harmonization algorithms. The concept accounts for CACC vehicle behaviors in the vicinity of managed lane entry ramps. The concept further addresses platooning and cooperative merging on the managed lane and considers the managed lane parameters that affect CAV operation including design characteristics, physical characteristics, operational parameters, and technology requirements. The study also examines the organizational and institutional frameworks and stakeholder roles and engagement required to implement the proposed concept. The study provides a roadmap to leverage these common elements to support the CAV operational scenarios. Finally, the study identifies the goals and metrics of CAV operation that may align with and help further the goals of managed lane operations. It is considered that, for early CAV deployment, single-lane controlled access facilities, equipped with existing communication devices and roadside units, offer the opportunity to begin integrating CAVs into traffic. An example scenario is used to demonstrate a possibility of CAV ML implementation. The study provides agencies with a foundation for developing plans and system design for using ML for early CAV deployment and traffic management.
Article
Connected automation systems coordinate vehicle communications and automated vehicle control to improve transportation system mobility, safety, and sustainability. Real-world deployment of connected automated vehicle (CAV) applications and related traffic management technologies will require automation of all vehicle and traffic movements, including platooning, lane change, and merge maneuvers, to fully realize potential system benefits. It is also necessary to test all the essential CAV operations in the field with proof-of-concept experiments to understand and validate the performance of developed applications. This paper describes an effort in developing an innovative prototype vehicle testing platform to successfully conduct a proof-of-concept field experiment of a cooperative lane change maneuver driven by a simple algorithm, executed using automated longitudinal speed control, vehicle-to-vehicle (V2V) communications, and vehicle-based radar systems. Experimental results show the effectiveness of the proposed testing platform and the successful proof of the cooperative lane change operations. The effort also raises research questions to be addressed in the future. The work can benefit the literature by introducing a feasible vehicle control platform for other researchers’ use to test their research subcomponents of CAVs, such as various control algorithms.
Article
Connected-and-automated vehicles (CAV) hold the potential for substantial improvements to traffic safety, travel time reliability, driver comfort, roadway capacity, environmental impacts, and users’ overall travel experience. Numerous modeling and simulation studies have been conducted to evaluate these impacts. However, model accuracy and simulation assumptions limit the validity of evaluation results. These factors have resulted in the wide range of differences in effectiveness among studies examining the same CAV applications available in the literature. In this study, we propose a bundled CAV application that involves platoons of equipped vehicles governed by an integrated set of cooperative adaptive cruise control (CACC), cooperative merge, and speed harmonization applications. We implemented the bundled application in a fleet of five vehicles at the Saxton Transportation Operations Lab of the Federal Highway Administration. Experiments were conducted to collect and compare data on CAV and human-driven behavior. Based on the real experimental data, our results show that the performance of the CAV operations, including platooning and cooperative merging under varying Infrastructure-to-vehicle speed commands, demonstrate string stability. The results also present key behavioral parameters of the vehicles and strings. This will eventually help the research community, particularly the modelers, to come up with models with realistic performance to further understand the CAV impacts on traffic. The results can also serve as references for transportation agencies to make informed decisions on infrastructure and traffic management decisions.
Article
Connected and autonomous vehicle (CAV) technologies are expected to change driving/vehicle behavior on freeways. This study investigates the impact of CAVs on freeway capacity using a microsimulation tool. A four-lane basic freeway segment is selected as the case study through the Caltrans Performance Measurement System (PeMS). To obtain valid results, various driving behavior parameters are calibrated to the real traffic conditions for human-driven vehicles. In particular, the calibration is conducted using genetic algorithm. A revised Intelligent Driver Model (IDM) is developed and used as the car-following model for CAVs. The simulation is conducted on the basic freeway segment under different penetration rates of CAVs and different freeway speed limits. The results show that with an increase in the market penetration rate, freeway capacity increases, and will increase significantly as the speed limit increases.
Article
Connected and automated vehicle (CAV) is a transformative technology that has great potential to change our daily life. Therefore, CAV related research has been advanced significantly in recent years. This paper does a comprehensive review on five selected subjects that lie in the heart of CAV research: (i) inter-CAV communications; (ii) security of CAVs; (iii) intersection control for CAVs; (iv) collision-free navigation of CAVs; and (v) pedestrian detection and protection. It is believed that these topics are essential to ensure the success of CAVs and need to be better understood. For inter-CAV communications, this paper focuses on both Dedicated Short Range Communications (DSRC) and the future 5G cellular technologies; for security of CAVs, this paper discusses both passive and active attacks and the existing solutions; for intersection control, this paper summarizes the pros and cons of both centralized and decentralized approaches; for collision avoidance, this paper concentrates on four subareas: maneuverability, vehicle networking, control confliction, and motorcycles; for pedestrian detection, this paper covers sensor, radar, and computer vision based approaches. Under each topic, this paper not only shows the state-of-the-art, but also unveils potential future research directions. By establishing connections among these subjects, this paper shows how they interact with each other and how they can be integrated into a seamless user experience. It is believed that the literature covered and conclusions drawn in this paper are very helpful to CAV researchers, application engineers, and policy makers.
Article
This paper addresses the problem of controlling the speed of a number of automated vehicles before they enter a speed reduction zone on a freeway. We formulate the control problem and provide an analytical, closed-form solution that can be implemented in real time. The solution yields the optimal acceleration/deceleration of each vehicle under the hard safety constraint of rear-end collision avoidance. The effectiveness of the solution is evaluated through a microscopic simulation testbed and it is shown that the proposed approach significantly reduces both fuel consumption and travel time. In particular, for three different traffic volume levels, fuel consumption for each vehicle is reduced by 19-22% compared to the baseline scenario, in which human-driven vehicles are considered, by 12-17% compared to the variable speed limit algorithm, and by 18-34% compared to the vehicular-based speed harmonization (SPD-HARM) algorithm. Similarly, travel time is improved by 26-30% compared to the baseline scenario, by 3-19% compared to the VSL algorithm, and by 31-39% compared to the vehicular-based SPD-HARM algorithm.
Article
Connected and automated vehicles (CAV) are marketed for their increased safety, driving comfort, and time saving potential. With much easier access to information, increased processing power, and precision control, they also offer unprecedented opportunities for energy efficient driving. This paper is an attempt to highlight the energy saving potential of connected and automated vehicles based on first principles of motion, optimal control theory, and a review of the vast but scattered eco-driving literature. We explain that connectivity to other vehicles and infrastructure allows better anticipation of upcoming events, such as hills, curves, slow traffic, state of traffic signals, and movement of neighboring vehicles. Automation allows vehicles to adjust their motion more precisely in anticipation of upcoming events, and save energy. Opportunities for cooperative driving could further increase energy efficiency of a group of vehicles by allowing them to move in a coordinated manner. Energy efficient motion of connected and automated vehicles could have a harmonizing effect on mixed traffic, leading to additional energy savings for neighboring vehicles.
Article
The calibration of traffic microsimulation models has received widespread attention in transportation modeling. A recent concern is whether these models can simulate traffic conditions realistically. The recent widespread deployment of intelligent transportation systems in North America has provided an opportunity to obtain traffic-related data. In some cases the distribution of the traffic data rather than simple measures of central tendency such as the mean, is available. This paper examines a method for calibrating traffic microsimulation models so that simulation results, such as travel time, represent observed distributions obtained from the field. The approach is based on developing a statistically based objective function for use in an automated calibration procedure. The Wilcoxon rank–sum test, the Moses test and the Kolmogorov–Smirnov test are used to test the hypothesis that the travel time distribution of the simulated and the observed travel times are statistically identical. The approach is tested on a signalized arterial roadway in Houston, Texas. It is shown that potentially many different parameter sets result in statistically valid simulation results. More important, it is shown that using simple metrics, such as the mean absolute error, may lead to erroneous calibration results.