Content uploaded by Jiaqi Ma
Author content
All content in this area was uploaded by Jiaqi Ma on Jun 01, 2020
Content may be subject to copyright.
1
Developing Highway Capacity Manual (HCM) Capacity Adjustment Factors (CAF) for
1
Connected and Automated Traffic on Freeway Segments
2
3
Adekunle Adebisi
4
University of Cincinnati, 722 Baldwin Hall, Cincinnati OH 45221
5
E-mail: adebisae@mail.uc.edu
6
7
Yan Liu
8
University of Cincinnati, 722 Baldwin Hall, Cincinnati OH 45221
9
E-mail: liu3y9@mail.uc.edu
10
11
Bastian Schroeder, PhD
12
Kittelson & Associates, Inc., 272 North Front Street, Suite 410, Wilmington, NC 28401
13
Email: bschroeder@kittelson.com
14
15
Jiaqi Ma, PhD*
16
University of Cincinnati, 765 Baldwin Hall, Cincinnati OH 45221
17
E-mail: jiaqi.ma@uc.edu
18
19
Burak Cesme, PhD
20
Kittelson & Associates, Inc., 300 M Street SE, Suite 810, Washington, DC 20003
21
Email: bcesme@kittelson.com
22
23
Anxi Jia, PhD
24
Kittelson & Associates, Inc., 300 M Street SE, Suite 810, Washington, DC 20003
25
Email: ajia@kittelson.com
26
27
Abby Morgan
28
Kittelson & Associates, Inc., 200 SW 1st Avenue, Suite 1070, Fort Lauderdale, FL 33301
29
Email: amorgan@kittelson.com
30
31
32
Corresponding Author: Jiaqi Ma
33
Word count
34
Body: 8198
35
Tables: 2*250 = 500
36
Total: 8698
37
(this is the accepted version on May 4th, 2020)
38
39
2
ABSTRACT
1
Connected and Automated Vehicles (CAVs) will undoubtedly transform many aspects of
2
transportation systems in the future. In the meantime, transportation agencies must make
3
investment and policy decisions to address the future needs of the transportation system. This
4
research provides much-needed guidance for agencies about planning-level capacities in a CAV
5
future and quantify Highway Capacity Manual (HCM) capacities as a function of CAV penetration
6
rates and vehicle behavior of car following, lane change, and merge. Due to numerous uncertainties
7
on CAV implementation policies, the study considers many scenarios that include variations in
8
parameters, including CAV gap/headway settings, roadway geometry, and traffic characteristics.
9
More specifically, this study considers basic freeway, freeway merge, and freeway weaving
10
segments in which various simulation scenarios are evaluated using two major CAV applications:
11
Cooperative Adaptive Cruise Control and Advanced Merging. Data from microscopic traffic
12
simulation are collected to develop capacity adjustment factors for CAVs. Results show that the
13
existence of CAVs in the traffic stream can significantly enhance the roadway capacity (as high as
14
35-40% under certain cases), not only on basic freeways but also on merge and weaving segments,
15
as the CAV market penetration rate increases. The human driver behavior of baseline traffic also
16
impacts the capacity benefits, particularly at lower CAV market penetration rates. Finally, tables
17
of capacity adjustment factors and corresponding regression models are developed for HCM
18
implementation of the results of this study.
19
Keywords: Connected and Automated Vehicles, Highway Capacity Manual, Capacity
20
Adjustment Factors, Microscopic Simulation, Freeway Segments
21
3
INTRODUCTION
1
Transportation in recent years has witnessed the development of advanced technologies in the form
2
of Connected and Automated Vehicle (CAV) systems, with safety and mobility applications being
3
integrated into vehicles and infrastructure to enhance the performance of transportation systems.
4
While the majority of the technical aspects are being handled by the private sectors, state and
5
federal institutions are steadfastly working towards the deployment of this suite of emerging
6
technologies as part of the intelligent transportation systems, particularly for their transportation
7
systems management and operations (TSMO).
8
CAVs are set to undoubtedly transform many aspects of the transportation systems.
9
However, there are still many uncertainties in future implementation. Meanwhile, transportation
10
agencies must make investment and policy decisions to address the future needs of the
11
transportation system. Long-range transportation plans, municipal transportation plans, regional
12
and system plans, corridor studies, and even traffic impact analyses all rely on multi-year forecasts
13
of travel demands and roadway capacities. Both are presently in flux, as ride-share services are
14
already being linked to an increase in vehicle miles traveled (VMT) while changing CAV
15
headways could cause either an increase or decrease in capacity. But how much will capacity
16
change? If there will be a capacity increase, will an increase in mainline capacity (shorter
17
headways) be offset by decreasing ramp and merge capacities (shorter and fewer gaps)? Will the
18
capacity change be proportional for both freeways and arterial streets?
19
The impacts of CAVs have been documented in previous studies. Ye and Yamamoto (1),
20
showed that the gradual penetration of CAVs changes the traffic flow dynamics and increases
21
capacity. Wang et al. (2) studied the effect of connected automated driving on traffic capacity
22
using a cellular automata model and reiterated similar results from past studies stating that; when
23
penetration of CAVs is low, improvement is not significant, but with increasing penetration rate,
24
capacity increases with accelerative rates.
25
There have been studies also pointing out the capability of CAVs to stabilize and smoothen
26
out traffic flow. This majorly pronounces the benefits of coordination between CAVs in congested
27
traffic. For instance, Delis et al. (3) proposed two macroscopic approaches to modeling Adaptive
28
Cruise Control (ACC) and Cooperative Adaptive Cruise Control (CACC) dynamics in traffic
29
flows. The results of both approaches showed that ACC and CACC-equipped vehicles are capable
30
of stabilizing flows with respect to on-ramp perturbations. String stability analysis conducted by
31
Talebpour and Mahmassani (4) also provided the same results, stating further that CAVs are more
32
effective in preventing shockwave formation and propagation. The benefits of CAVs have not only
33
been proved in terms of capacity or stability, but also fuel savings, efficiency, and safety. CAVs
34
can potentially reduce fuel consumption by 20% (5), thereby reducing emissions. Using a more
35
targeted approach, Guo et al. (6) established the possibility of realizing up to 32% fuel savings at
36
signalized intersections. By equipping CAVs with speed harmonization capability (7), up to 67%
37
safety risk reduction can be obtained.
38
Of more interest to this study, we consider two freeway CAV applications, which will have
39
the most significant capacity impact: CACC and Advanced Merging (A.M.). The CACC was an
40
improvement on ACC through V2V and V2I communications to form platoons. The benefits of
41
4
these communication systems include a more efficient intersection throughput, travel time
1
reduction, as well as reduction in fuel usage and emissions (8). Further, with the data collection
2
potential of CAVs about the driving environment, studies have shown that automated driving can
3
potentially decrease traffic congestion by reducing the time headways, thereby enhancing the
4
traffic capacity (9). CACC can significantly improve the safety and operations on roadways. At
5
full market penetration, it is possible to realize between 25-35% reductions in travel time (10) and
6
obtain a 64% improvement in average vehicle delay (11). With CACC, throughput enhancement
7
of over 100% can be realized (12), and up to 90% and 100% capacity increase can be obtained for
8
basic freeway segments (13), and freeway merge segments (14), respectively.
9
Advanced merging also takes advantage of vehicle communication capability to control
10
the flow at regular congestion locations such as merge and lane drops. Using V2I and V2V
11
technologies, CAVs can signal other vehicles about their intention to merge into the mainline
12
traffic using a communication medium. With this, vehicles trying to merge can identify acceptable
13
gaps on the mainline and make coordinated lane changes. The coordination occurs between both
14
the mainline traffic and merging vehicles, thereby minimizing any merging disturbance. Recent
15
developments in A.M. capability have established the future benefits to derive from implementing
16
A.M. on merge areas. For example, by integrating both lane change and trajectory optimization, a
17
recent study (15) obtained a 93% reduction in average delay. Letter and Elefteriadou (16) realized
18
as much as a 62% improvement in speed and total travel time in a freeway merge scenario.
19
Pueboobpaphan et al. (17) obtained 60% and 75% average travel time improvement under low
20
flow and high flow conditions of merge segments, respectively.
21
22
RESEARCH OBJECTIVES
23
With numerous emerging technologies, transportation agencies must make investment and policy
24
decisions to address the future needs of the transportation system. Therefore, agencies need to
25
know the potential capacity effects of CAVs to aid their decision-making process with future
26
investments. Highway Capacity Manual (HCM) is a valuable tool used by practitioners for
27
planning-level assessments of various facilities and corridors and remains widely accepted
28
throughout the industry as a credible source and benchmark for capacity estimation and analysis
29
guidance. However, agencies are faced with shortfalls in HCM guidance pertaining to CAVs since
30
the HCM is silent on the effects of CAVs. This research provides much-needed guidance for
31
agencies about planning-level capacities in a CAV future and quantifies HCM capacities as a
32
function of CAV penetration rates and allowable vehicle “aggressiveness.” The research explores
33
and tests CAV and non-CAV interaction and provides guidance to state and local agencies about
34
the sensitivity of key parameters on the final capacities.
35
This study aims to develop capacity adjustment factors (CAF) for CAVs on various
36
freeway facilities at different levels of traffic demand and market penetration to adapt the use of
37
HCM in analyzing CAV applications. The goal of this research is to not only quantify the effect
38
of CAVs on the basic freeway, merge, and freeway weaving segments, but also develop tables of
39
CAFs and corresponding a statistical capacity prediction model that can be easily used to assess
40
5
the impact of different future CAV implementation policies. The statistical models developed here
1
are expected to be easily adaptable to changes in various parameters as CAVs continually penetrate
2
the transportation systems.
3
The remaining of this paper includes the methodology section, which provides detailed
4
information on the steps taken to achieve the study objective. Following the methodology is the
5
results section providing the outcome of each experimental setup. The section also gives extensive
6
insights into the impact of CAVs on traffic flow. Also, CAF tables and the method for developing
7
the empirical models for predicting the capacity impacts of CAVs are provided. Finally,
8
conclusions are given along with possible future research questions.
9
10
METHODOLOGY
11
Base Model Development
12
Three freeway segments were considered in this study, as shown in Figure 1; a basic
13
freeway segment (BFS), a freeway merge segment (FMS), and a freeway weaving segment (FWS).
14
A well-calibrated hypothetical network that reflects the capacity of freeway segments relative to
15
the free-flow speed (FFS) as specified in the HCM allows the flexibility of roadway and traffic
16
characteristics for sensitivity analysis. In this study, we used FFS of 75 mph for all scenarios. For
17
the BFS, two hypothetical 3- and 2-lane networks were modeled in VISSIM. The FMS was a 2-
18
lane mainline network with an on-ramp introduced one mile downstream the start of the network.
19
The acceleration length was designed as 500-ft. Finally, the FWS was a 4-lane weaving segment,
20
indicating a 3-lane mainline, a single acceleration/deceleration lane, an on-ramp, and off-ramp
21
1000-ft apart.
22
23
Figure 1. Freeway Segments Considered
24
6
To investigate the effects of CAVs under varying base capacities before the introduction of CAVs,
1
BFS was tested with a base (also referred to as a “starting”) capacity of 2,400pcphpl, 2,100pcphpl,
2
and 1,800pcphpl. This test was conducted because not all freeway segments have the same
3
capacities, even with similar geometric features, due to the external factors such as differences in
4
driver behavior or weather conditions. However, these base capacities were only applied to the
5
BFS. The analyses for the FMS and FWS used starting capacities of 2,400pcphpl for all scenarios.
6
This was to limit the scope of the study to more common roadway configurations and to be able
7
to analyze more facility types. Also, the base capacity of 2,400pcphpl is consistent with the HCM
8
recommended value, and therefore results from this study can directly be applied to existing HCM
9
procedures. It should be noted that although the driver behavior used for the BFS, FMS, and FWS
10
targeted 2,400pcphpl as the base capacity, the actual starting capacities for FMS and FWS were
11
lower due to ramp and weaving disturbances.
12
Obtaining the desired capacity for a fully human-driven vehicle (HDV) traffic stream in
13
the microsimulation model requires systematic calibration. The base model was calibrated using
14
the VISSIM in-built driver behavior model developed by Wiedemann (18). In this study, the
15
Wiedemann ’99 parameters, which are more suitable for freeway networks, were adjusted for
16
calibration (18). It provides ten (10) calibration parameters CC0 to CC9, which controls the
17
switching behavior of drivers. The major parameters used were CC0, CC1, and CC2, representing
18
the standstill distance, the headway time, and the following variation, respectively. The targeted
19
starting capacities, as a measure of driver variability, were achieved by adjusting CC2. By
20
adjusting this parameter, the response of drivers to the preceding vehicle is tuned, which in turn
21
affects the prevailing capacity of the simulation network. More detail on these parameters can be
22
found in the PTV VISSIM user’s manual (19). For all scenarios, CC0 = 1.5 m, CC1 = 1.05 s, and
23
CC2 = 5.0 m, 11-m, and 16-m for base capacities 2,400 pcphpl, 2,100 pcphpl, and 1,800 pcphpl,
24
respectively.
25
For each scenario, five (5) simulation replications with different random seeds were
26
performed. The result for each scenario was averaged over the replications. The input volume was
27
gradually increased and then decreased in such a way that the simulation experiments can capture
28
the free flow and the breakdown regimes over the simulation period. This allows the system
29
throughput to level off, thereby providing a means to estimate the resulting segment capacity. The
30
first 15-minutes were used as a warm-up period to ensure the traffic has stabilized before data
31
collection. The VISSIM default random vehicle arrivals were used in this study as well. All vehicle
32
types assume similar speed distribution on entering the roadway network, and therefore the desired
33
speeds of HDVs and CAVs in the network are not entirely the same by following a pre-specified
34
desired speed distribution. However, the speed of a CAV during the operations may also be
35
determined by other factors, such as the CACC protocol, if this CAV becomes a platoon follower.
36
To estimate the resulting capacity, we used the prebreakdown flow rate defined by the HCM as
37
the 15-minute average flow rate immediately before the breakdown event. This was averaged over
38
all simulation replications.
39
40
CAV Modelling
41
7
The CACC car following models developed in this study was based on a well-accepted
1
study by Milanes and Shladover (20), which has been previously used (21, 22, 29). Interested
2
readers should refer to those studies for model details. We adapted their model in VISSIM
3
implementation to include testing various settings of intra-platoon gaps for sensitivity analysis.
4
We also developed additional CACC protocols in VISSIM API for operations of CACC vehicles
5
to form or leave platoons and perform lane following under various conditions, as shown in Figure
6
2. We assumed a maximum platoon length of 10 vehicles. This eliminates disturbances which
7
could hinder the performance of the algorithm at on- and off-ramps (if any) due to the necessary
8
lane changes. If the platoon length is too high, it would make merging difficult and causes
9
unreliable communication between the leader and vehicles toward the end of the platoon, and if it
10
is too low, it reduces the capability of the CACC implementation. In a previous study by the
11
California PATH’s program (link below), researchers tested the impact of various platoon lengths
12
of 5, 10, 15, 20, and 25 on the mainline throughput at freeway on-ramp and off-ramp area. It was
13
found that the best maximum platoon length should be between 10 and 15, although 10 is
14
recommended so as to lower the probability of traffic breakdown for challenging cases where the
15
on-ramp demand may be very high (23). A basic introduction of the logic is presented for the
16
completeness of the presentation and more detail can be found in Milanes and Shladover (20). All
17
model assumptions and parameters used here are within the ranges recommended in the same
18
study.
19
As shown in Figure 2, the CACC protocol consists of two modes (speed regulation and
20
gap regulation) in which the switching conditions within each mode is based on the regulation of
21
the speed and gap between consecutive CAVs. The purpose of the speed regulation mode is to
22
maintain the user-desired speed when the preceding vehicle is beyond a pre-set gap (i.e., a time
23
gap larger than 2 seconds from the preceding vehicle). In this case, the controller uses the vehicle
24
acceleration model defined in Equation 1 to control the speed. The control gain (assumed as
25
0.4 s-1) is the difference between the free flow speed and the subject vehicle’s current speed,
26
is the acceleration recommended by the controller to the subject vehicle (m/s2), is the current
27
speed of the subject vehicle (m/s), and is the free-flow speed (m/s).
28
29
If the preceding vehicle is an HDV, the subject CAV will switch to the ACC mode to
30
regulate the driving behavior. If the subject CAV is too close to the preceding vehicle (i.e., the
31
detected gap is smaller than a given minimum following threshold, i.e., a time gap smaller than
32
1.5 seconds from the preceding vehicle), it will switch to the ACC gap regulation mode to maintain
33
a safe following time gap , as shown in Equation (2). Otherwise, the CAV will repeatedly
34
implement previous control logic to ensure consistent driving behavior.
35
(2)
where and are control gains on following distance difference and
36
speed difference, respectively (Liu et al., 2018). The headway , preceding vehicle length , and
37
preceding vehicle speed are considered in Equation (2).
38
8
If the preceding vehicle is a CAV, the subject vehicle will switch to the CACC mode and
1
communicate with the preceding vehicle to exchange critical information (e.g., speed, location,
2
platoon size). Equations (3) and (4) are used for CACC following. If the length of the previous
3
CACC platoon is less than the maximum allowable platoon length, the subject CAV will catch up
4
with the preceding CACC platoon and become a platoon follower; therefore the intra-platoon gap
5
(i.e., different values used in this study for aggressive, normal, and conservative CACC
6
following capability) is applied to tightly follow the preceding CAV. Otherwise, the subject CAV
7
becomes a CACC platoon leader and applies the inter-platoon gap (1.5 seconds in this study) to
8
follow the preceding CAV. The specific regulation mode depends on the actual time gap between
9
the subject CAV and its preceding CAV. If the time gap is larger than a given threshold (2
10
seconds), the subject CAV will apply speed regulation mode, as shown in Equation (1). Otherwise,
11
it will apply the CACC gap regulation mode to keep a safe following distance with the determined
12
following gap (i.e., inter-platoon gap or intra-platoon gap) by implementing Equations 3 and 4,
13
where and (assumed as 0.45 s-1 and 0.0125 respectively) are the control gains for adjusting
14
the time gap between the subject vehicle and the preceding vehicle, is the time gap error defined
15
as , ,
16
and is the constant time gap between the subject vehicle and the last vehicle in the preceding
17
CACC platoon (assumed as 1.1 sec in this study).
18
19
20
Due to the linearity of the above models, the vehicles cannot handle emergency braking to
21
avoid collisions. The forward collision warning algorithm (Kiefer et al., 2003) developed by the
22
Collision Avoidance Metrics Partnership (CAMP) is included in the C/ACC car following modes
23
to determine whether the gap between the subject vehicle and the preceding vehicle is sufficient
24
for safe car following. If the crash warning is activated, it implies that a crash will happen if both
25
the subject vehicle and the preceding vehicle keep their current acceleration speeds for the next
26
few seconds. The algorithm will use a conventional car-following model (e.g., Wiedermann 99)
27
that guarantees collision-free to generate emergency deceleration commands until the crash
28
warning is deactivated.
29
The Advanced Merging (A.M.) algorithm used in this study is adopted from the VISSIM
30
11 advanced merge function and described next. The objective of the A.M. algorithm is to
31
coordinate the mainline and merging traffic using V2V and V2I technologies. When a merging
32
vehicle is detected (no matter it is an HDV or a CAV), a gap is created on the mainline that can
33
accommodate the merging vehicle. The system informs the mainline vehicles to cooperatively
34
change to another lane apart from the merging vehicle’s targeted lane or slow down slightly to
35
create the required gap. A.M. can be a standalone capability of CAVs in this study such that the
36
effects of A.M. only can be evaluated. The CAVs can also be equipped with CACC and A.M.
37
capabilities at the same time, referred to as “CACC + A.M.”. HDVs are only regular drivers, and
38
we do not assume that they have A.M. or other advanced driving capabilities.
39
9
1
Figure 2. CACC Protocol
2
Simulation Test Scenarios
3
As previously discussed, the simulation network was first calibrated to match the HCM
4
capacity values. Capacity, being the performance measure of interest in this study, was estimated
5
for different scenarios. For all scenarios tested, the market penetration rate (MPR) was varied from
6
0% to 100% at 20% increments.
7
For CACC simulation, three levels of the intra-platoon gap were used in our study;
8
aggressive (0.6-sec), normal (following a distribution), and conservative (1.1-sec). All these gap
9
settings were applied in the BFS network evaluation. To control the total number of simulations
10
runs, only the “normal” gap setting was used for FMS and FWS because this is the most realistic
11
scenario. For the “normal” gap settings, we adopted intra-platoon gap distribution specified by
12
Nowakowski et al. (24), where the drivers in a survey test chose a time gap of 0.6 s for the 57% of
13
the time they were in car-following, 0.7 s for 24% of the time, 0.9 s for 7% of the time, and 1.1 s
14
for 12% of the time.
15
Specifically, for BFS, we tested two different lane configurations; 2- and 3-lane mainline
16
and starting capacities of 2,400pcphpl, 2,100pcphpl, and 1,800pcphpl. Additionally, we evaluated
17
the impact of ACC-equipped vehicles, i.e., isolated AVs that adopt commercial automated
18
10
following behavior, with empirical models also calibrated in Goñi-Ros et al. (11), which states the
1
commercial ACC behavior is conservative and causes string instability. The ACC-equipped
2
vehicles in this study also have the same car-following logic as the CACC already described.
3
However, they do not form platoons due to the absence of vehicle communication. In essence, they
4
are stand-alone vehicles. For the FMS, we tested 2-lane mainline with a single lane on-ramp and
5
volume-to-capacity (v/c) ratio of 0.8 and 1.0 for mainline traffic. The on-ramp volume was varied
6
from 300 vph at 200 vph increments until a stable capacity was reached for each scenario. In the
7
case of the FWS, we used the 3-lane weaving segment, as stated earlier, and performed tests using
8
volume ratios (V.R.) of 0.2, 0.3, and 0.4. The V.R. is the ratio of weaving traffic to non-weaving
9
traffic, given in the HCM as follows:
10
11
13
12
where the subscripts indicate the direction of flow, for example, from ramp to freeway, denoted as
14
RF or from the freeway to the ramp, denoted as FR. All the scenarios for FMS and FWS were
15
evaluated with and without the advanced merging algorithm.
16
17
Capacity Adjustment Factor (CAF) Estimation
18
The CAF is estimated as the ratio of the capacity of the evaluated scenario to that of the
19
base capacity. The HCM exhibit 12-6 (25) provides the relationship between the base segment
20
capacity and CAF in Equation 6 as
21
22
24
23
where = adjusted capacity (pc/hr/ln)
25
= base segment capacity (pc/hr/ln), and
26
= capacity adjustment factor (unitless)
27
Conventionally in HCM, capacity estimates resulting from the impacts of recurring or non-
28
recurring events are usually lower than the base capacity since base capacity reflects ideal
29
conditions (e.g., only passenger cars, clear day, level terrain, etc.). Therefore, the resulting CAF is
30
typically less than 1.0. However, in the case of CAVs, it is expected that the gradual penetration
31
will improve the traffic conditions rather than worsen it; therefore the expected CAF would be
32
greater than 1.0.
33
In the final part of this study, we developed a simple but efficient empirical model that
34
accepts certain inputs and predicts the capacity as a function of the inputs. The resulting value is
35
the same as the CAF, which can be used as a multiplier term similar to the one provided in Equation
36
2. The regression model can be easily integrated with any existing software for the HCM methods.
37
Both CAF tables and regression models enable policymakers to make a quick but reliable
38
estimation of the future capacity of the roadway segment based on selected factors.
39
11
RESULTS
1
Effects of CACC on Traffic flow
2
The fundamental diagram (FD) has been used to understand traffic flow for decades. It
3
spells out the basic principles behind the operations of freeway traffic and can also serve as a
4
means for capacity estimation. While past studies (4, 7) have studied the effect of CAVs on the
5
FD and established that there is the potential of removing the congestion region due to CAV
6
stability and coordination, we investigate the question from a different perspective. With different
7
roadway starting capacities, we hypothesize that even though CAVs can remove the congestion
8
region, the nature of human drivers in the traffic stream will have an effect on what CAV
9
penetration remove the congested regime. To test this, we conducted a simulation of the freeway
10
merge segment using a lower base capacity to compare the behavior of the FD between a high
11
starting capacity and a low starting capacity. The results for each starting capacity at each CAV
12
penetration is provided in Figure 3. It confirms earlier findings that CAVs can smoothen out
13
congestion. However, the results provide additional insights and indicate that the smoothing effect
14
of CAVs are only equal across all roadway scenarios when the market penetration is 80% due to
15
the dominant existence of CACC vehicles in the traffic stream.
16
17
18
12
1
Figure 3. Fundamental Diagrams for each CACC MPR with different mainline starting
2
capacities
3
Effects on Basic Freeway Segments
4
Figure 4 shows the capacity of BFS relative to the MPR of CACC-equipped vehicles. First,
5
the results show that for all the gap-setting used in this study (e.g., normal, conservative), the
6
capacity increases with respect to the MPR. More interestingly, they all follow a quadratic trend,
7
which indicates the capacity increase is faster as the MPR of CACC becomes larger. Similar
8
insights have been established in past studies (22) as well. However, the aggressive CACC intra-
9
platoon gap results in higher capacity impacts. This is logical because tighter headways between
10
vehicles directly impact capacity positively. The conservative scenario, on the other hand, has the
11
lowest impact. Comparing the capacity values for different gap settings for each MPR, at 20%
12
MPR, the CACC impacts are not far different from each other, with the aggressive scenario only
13
1.2% higher in capacity than the conservative scenario. As the MPR increases, we observe a
14
gradual increase in the margin between the two extreme CACC settings, and at 100% MPR, a 17%
15
margin is obtained. It should be noted that these results are for the scenarios with a base capacity
16
of 2,400pcphpl.
17
18
(a)
19
13
1
(b)
2
3
Figure 4. Results for Basic Freeway Segments; (a) Base Capacity = 2,400pcphpl,
4
(b) Base Capacity = 2,100pcphpl, and (c) Base Capacity = 1,800pcphpl
5
On further exploring the capacity impacts relative to the base capacities, we obtain even
6
more interesting insights. For example, the results from the 1,800pcphpl base capacity show that
7
the initially quadratic trend of capacity improvement smoothens out to become a much linear trend.
8
This means that due to lower base capacity, the effect of CACC is relatively constant as MPR
9
increases, and the capacity benefits are more pronounced even under low MPRs.
10
At the 1,800pcphpl base capacity scenario, we examine the difference between the capacity
11
impacts of different gap settings. It is interesting to find out that instead of the 1.2% difference
12
obtained from the 2,400pcphpl base capacity scenario for 20% MPR, there is a 7% difference in
13
the 1,800pcphpl scenario. Moreover, at 100% MPR, the capacity difference for the two extreme
14
gap settings for the two starting capacities are the same (17%). This further establishes the earlier
15
statements; CACC has higher benefits at lower MPR for roadways with lower capacities.
16
Furthermore, there are two BFS configurations analyzed in this study; 2- and 3-lane
17
segments. The capacity improvements are also examined for both configurations. In summary, for
18
all the scenarios, at all MPRs, regardless of the base capacity, the result for the 3-lane BFS is
19
slightly lower than the 2-lane segment. This directs attention to the effect of lane changing activity
20
on freeway capacity. The per-lane capacity of the freeway segment decreases as the number of
21
lane increases. However, the difference between the 2-lane and 3-lane scenarios are quite small,
22
14
partially because of the capability of CACC strings to absorb disturbances caused by lane changes.
1
Also, this result sheds light on future CACC operations that it is preferable to limit or discourage
2
lane changes to maintain stable traffic flow and high capacity.
3
While this study focuses on CAV impacts, we also conducted experiments on ACC, and
4
their impact on capacity is shown in Figure 4. The motivation to include ACC results is to show
5
the importance of connectivity in traffic capacity enhancement. First, it is observed that, for a base
6
capacity of 2,400pcphpl, there is a decrease in freeway capacity as the percent of ACC increases.
7
This is because the ACC performs worse than HDVs in the traffic. ACC systems are built for
8
comfort and safety is a higher priority for them, which generally results in a more conservative
9
behavior. In the 1,800pcphln scenario, however, we observe an increase in capacity as penetration
10
of ACC increases. This is because even though the ACC systems designed to be conservative, their
11
headways are still lower than the headways for HDVs, leading to capacity improvements. This
12
suggests that ACC systems of this nature can perform better than HDVs under non-ideal
13
conditions, likely due to the deterministic behavior of ACCs that stabilizes the traffic flow. The
14
result for ACC in the case of 2,100pcphpl base capacity is quite interesting. We obtain a decrease
15
in capacity for the first few penetration rates, then an increase thereafter. The initial drop in the
16
capacity is a result of the added variance of driver behavior due to the distinct behavior between
17
ACC vehicles and HDVs. The eventual increase in capacity then occurs because of the stability
18
effect when ACC vehicles constitute most of the system.
19
20
Effects on Merge Segments
21
The results of merge segments are provided below in terms of CAV MPR, CACC
22
capabilities, advanced merging (A.M.) capabilities, the mainline demand, and the on-ramp
23
demand. Recall that the segment configuration is a 2-lane mainline and a single lane on-ramp. The
24
acceleration lane is 500-ft, starting 1.5-miles downstream of the segment starting point. Table 1
25
provides the general capacity estimates based on different scenarios for the segment. The “No-
26
onramp” scenario is the capacity of the segment under the scenario with zero ramp demand. This
27
is the same as the results obtained in the 2-lane BFS scenario. However, every other scenario was
28
simulated by varying the onramp demand gradually from low (300pcphpl) to high volumes. The
29
“CACC” scenario indicates implementing only the CACC application for the CAVs in the traffic
30
stream. The “CACC + A.M.” scenarios involve equipping the CAVs with both CACC and A.M.
31
capabilities. The “A.M.” scenario indicates equipping CAVs with only A.M. capability. Each of
32
these other scenarios was compared to the “No-onramp” scenario, and the percent difference is
33
provided in the table as well.
34
As expected, merging traffic causes disturbances, which translates into lower capacities for
35
the fully HDV traffic scenarios (0% MPR). Even with increasing CAV penetration, the resulting
36
capacity still falls below the “No-onramp” scenario. However, on reaching 80% MPR of CAVs,
37
different results are obtained for scenarios with CACC technology. At 80% MPR, the effect of
38
improved vehicle capabilities allows much better coordination between mainline traffic and
39
merging traffic, thereby offsetting the reduction in capacity as a result of the initial merging
40
disturbance. This is because of the CACC following behavior (i.e., control algorithms) can make
41
vehicles react faster and stably to absorb disturbances from the downstream traffic. More
42
specifically, the 7% reduction in capacity from merging disturbance was removed, and even at
43
15
100% MPR, the mainline was able to accommodate about 2% more vehicles merging from the
1
ramp. In essence, as the MPR increases, the effect of merging disturbance reduces as a result of
2
CACC coordination.
3
The only exception to these results is the “A.M.” scenario, which involves CAVs with only
4
advanced merging capabilities (i.e., no CACC). Although there are improvements from increased
5
MPR of CAVs, the benefits are still low to offset the capacity reduction from the initial merging
6
disturbances. This further establishes the potential benefits that can be obtained from CACC
7
vehicle operations.
8
With the effect of CACC already established, it is important to also examine the impact of
9
an added advanced merging capability. This is done by comparing the “CACC” and “CACC +
10
A.M.” scenarios. The effect of A.M. is more pronounced at lower MPRs. This may be a result of
11
more HDVs in the traffic stream, which provides more gaps for merging purposes. At high MPR,
12
CACC-equipped vehicles are already traveling at smaller gaps with more coordination, thereby
13
leaving not much room for advanced merging possibilities. The greatest improvement from A.M.
14
capability relative to CACC is 5%, which occurs at 20% MPR. It is also noted from the comparison
15
that the main capacity benefits from CAVs, as expected, comes from the stable platoons with
16
shorter headways.
17
Table 1. Capacity Results for Freeway Merge
18
Market Penetration Rate (MPR)
Capacity (vphpl)
0%
20%
40%
60%
80%
100%
No on-ramp (BFS)
2,416
2,466
2,586
2,734
2,938
3,244
CACC
2,206
2,242
2,371
2,556
2,932
3,296
% difference
(-9%)
(-10%)
(-9%)
(-7%)
(0%)
(2%)
CACC + A.M.
2,206
2,353
2,439
2,662
2,976
3,306
% difference
-9%
-5%
-6%
-3%
1%
2%
A.M. capacity
2,206
2,231
2,280
2,330
2,346
2,353
% difference
(-9%)
(-11%)
(-13%)
(-17%)
(-25%)
(-38%)
19
To further explore the effect of onramp demand on segment capacity, we analyzed the
20
variation in the resulting capacity as a measure of the onramp demand and enhanced vehicle
21
capabilities. It is reasonable to assume that different onramp demand provides a different segment
22
capacity (e.g., the turbulence effects of an onramp with a demand of 100 vehicles will vary
23
drastically than an onramp with 500 vehicles). Besides, changes in mainline traffic demand also
24
provide interesting results. For instance, by having the mainline demand as 80% of the estimated
25
capacity for each MPR, the unused portion of the roadway should be able to accommodate more
26
merging vehicles. Figure 5 indicates the capacity trends as onramp demand increases for each
27
CAV MPR. Both scenarios were simulated with and without advanced merging capabilities of
28
CACC-equipped vehicles.
29
16
1
(a)
2
3
(b)
4
Figure 5. Capacity trend with increasing on-ramp demand for each CACC market
5
penetration; (a) 100% mainline demand, and (b) 80% mainline demand
6
From the 100% mainline demand scenario, the first interesting observation is that at low
7
CAV MPR, the segment is unable to maintain its capacity, but instead reduces under increasing
8
on-ramp demand until it reaches a stable value. However, at high CAV MPR, the segment can
9
maintain the same capacity longer under increasing onramp demand before reaching stable
10
capacity conditions. These results indicate that careful consideration should be put in place because
11
there are different segment capacities for different onramp demand, even in cases with fully HDV
12
traffic. The segment capacity is not constant across all onramp demand volumes. The capacity of
13
the merge segment can decrease when the demand from the on-ramp is very high. This is similar
14
to findings from past related studies (26, 27).
15
On the other hand, the 80% mainline demand confirms the initial expectation that the
16
unused portion of the mainline can accommodate more merging vehicles. As a matter of fact, the
17
20% volume that was removed from the mainline traffic was recovered from the merging traffic.
18
Even more importantly, if the 20% mainline demand was removed at 0% CACC MPR, the segment
19
can only accommodate about 300 vehicles, but if it was removed at 100% CACC MPR, the
20
segment can accommodate about 1,000 more vehicles due to CACC operations. The trend obtained
21
at low CAV MPR for 80% of mainline demand indicates that more vehicles are entering the
22
mainline under this condition. However, it should also be noted that similar to the 100% mainline
23
17
demand, the capacity benefits eventually fall and then reaches a stable value, also reiterating the
1
findings that different onramp demand can result in different segment capacities.
2
We established that at 80% of mainline traffic, the merge segment can accommodate more
3
vehicles from the ramp. To ensure the practicality of the result, a congestion analysis was
4
conducted on the mainline and on-ramp traffic. The results indicated that the on-ramp was not
5
suffering from any level of congestion as a result of oversaturation. The mainline traffic was only
6
having some level of delay, which is as expected due to merging disturbances.
7
Effects on Weaving Segments
8
Weaving segment results are provided in Figure 6. As stated earlier, three volume ratio
9
(V.R.) levels are tested in this study. Using each V.R., we tested the effect of MPRs of CACC-
10
equipped vehicles with and without advanced merging capabilities. This was simulated using the
11
“normal” gap settings for CACC-equipped vehicles.
12
13
(a)
14
15
(b)
16
Figure 6. Freeway Weaving Segment Results with varying V.R. (a) V.R = 0.2, 0.3, and 0.4,
17
(b) All V.R. combined (Without advanced merging)
18
The capacity increase obtained from FWS also follows a quadratic trend, which is similar
19
to all the other scenarios with similar simulation setup. However, the gradient of the curve is
20
steeper due to the lower base capacity of the weaving segment. It should be noted that, without
21
any weaving volume, the capacity of the segment is 2,400pcphpl, and with the weaving demand,
22
the capacity drops, even at 0% CAV penetration, due to the friction impacts. Therefore, the base
23
capacity reduces from the original 2,400pcphpl to 2,256pcphpl. As previously mentioned, the
24
simulation model was calibrated, such that the segment capacity with 0% MPR is similar to the
25
HCM capacity given for the weaving segments.
26
Furthermore, examining the impact of V.R. on capacity (as shown in Figure 6b), it is
27
observed that, similar to past studies (28), as the V.R. increases, the resulting capacity decreases.
28
18
A higher V.R. indicates a higher volume of vehicles trying to make lane changes from the freeway
1
to the ramp and vice versa. Lane changes directly impact capacity, and this is also established from
2
the BFS analysis. At 100% CACC, the reduction in capacity is as high as 8% as a result of
3
increasing V.R. from 0.3 to 0.4.
4
Advanced merging capabilities have generally been observed to increase capacity in this
5
study. The difference between the experiment here and the one conducted for FMS is that there is
6
more lane changing activities as a result of the weaving. Results obtained here indicate that the
7
A.M. capability is only effective between 20 to 80% CAV MPR, with the biggest benefits at 40%
8
and 60%. The A.M. requires qualified gaps to function properly. Most of the large gaps are only
9
available when HDVs are in the traffic stream. The tight gaps from CAVs from platooning actually
10
limit the advanced merging capabilities.
11
Capacity Adjustment Factor Results
12
The penultimate objective of this study is to obtain capacity adjustment factors (CAFs) for
13
different roadway configurations under varying CAV conditions. Table 2 shows the result of the
14
CAF for all the configurations tested in the simulation. Note that due to similar results of 2-lane
15
and 3-lane results, we combine them and take the average to represent the BFS capacity. The
16
“A.M.” indicates scenarios with only CAVs with A.M. capability in the network, and the MPR
17
represents the portion of CAVs with the A.M. capability. In the “CACC” scenario, none of the
18
vehicles have A.M., and the MPR indicates the portion of CAVs with the CACC capability. The
19
“CACC+A.M.” scenario is the same as the CACC scenario except that the CAVs in the network
20
now have A.M. capability. The “CACC+A.M.” MPR represents the percentage of vehicles with
21
both capabilities.
22
Finally, to derive empirical models that quantitatively establish the relationship between
23
the capacity impact of CAVs and different freeway configurations, we conduct a regression
24
analysis of the obtained results. Three different empirical relationships are provided for each
25
freeway segment under consideration. The provided variables are significant at a 95% confidence
26
interval on the CAF. The regression result further shows that even though our analysis shows slight
27
capacity decreases for 3-lane compared to 2-lane BFS, the number of lanes is still not a significant
28
predictor (p-value = 0.16) of the resulting capacity. Advanced merging is significant for both FMS
29
and FWS.
30
Table 2. CAF Results for all Freeway Configurations Tested
31
Basic Freeway Segment
MPR/S.C.
2400
2100
1800
0
1.00
1.00
1.00
20
1.02
1.02
1.15
40
1.07
1.10
1.27
60
1.13
1.25
1.40
80
1.22
1.37
1.60
100
1.35
1.53
1.82
Freeway Weaving Segment (without A.M.)
MPR/V.R.
0.2
0.3
0.4
0
1.00
1.00
1.00
19
20
1.03
1.04
1.05
40
1.08
1.08
1.09
60
1.15
1.15
1.13
80
1.23
1.22
1.20
100
1.37
1.37
1.34
Freeway Weaving Segment (with A.M.)
MPR/V.R.
0.2
0.3
0.4
0
1.00
1.00
1.00
20
1.05
1.05
1.08
40
1.11
1.13
1.14
60
1.17
1.20
1.18
80
1.25
1.26
1.24
100
1.37
1.38
1.35
Freeway Merge Segment
MPR
CACC
CACC+A.M.
A.M.
0
1.00
1.00
1.0
20
1.02
1.07
1.01
40
1.07
1.11
1.03
60
1.16
1.21
1.06
80
1.33
1.35
1.06
100
1.49
1.50
1.07
The R-square values for the three regression models are obtained as 0.89, 0.86, and 0.97
1
for BFS, FMS, and FWS, respectively, indicating excellent fits. The relationship between the CAF
2
and the independent variables can, therefore, be expressed as
3
(7)
4
5
(8)
6
7
(9)
8
where is the percent of platoon/CACC-equipped vehicles in the traffic stream, is
9
the intra-platoon gap, is the starting capacity, is the ramp demand, is the volume
10
ratio, and is the percentage of vehicles with advanced merging capability. For the case where
11
follows a distribution, we recommend using the expected value of 0.71, which represents the
12
average of the distribution, also used in Liu et al. (21).
,
and
are the
13
capacity adjustment factors (CAFs) for basic freeway, freeway merge, and freeway weaving
14
segments respectively. Various variable interactions were tested in developing the empirical
15
model, and they do not improve the model performance.
16
17
20
CONCLUSIONS AND FUTURE RESEARCH
1
CAV technologies are set to revolutionize the Nation’s transportation systems in terms of
2
safety and operational features. Numerous research and testing are ongoing to specifically equip
3
vehicles and roadways with advanced technologies. CAVs will be able to communicate with each
4
other, thereby offering improvements in roadway performance in the near future. However, it is
5
essential to make wise decisions on future implementation strategies. The majority of CAV studies
6
are still based on simulation because of the cost of conducting a naturalistic study of such
7
magnitude. This usually requires a large amount of simulation that may be time-consuming. To
8
address this, the objective of this study is to evaluate different possible implementation scenarios
9
and provide decision-makers with a quick evaluation method for assessing the effect of different
10
implementation possibilities. In this study, we considered some of the most common geometric
11
and traffic characteristics of roadways, evaluate the future impacts, and then provide an empirical
12
model that can be used to assess the benefits of different future CAV implementation.
13
For the basic freeway segments, we analyzed the impact of CACC-equipped vehicles using
14
different starting capacities. The results confirm the findings of similar past studies and also
15
provided some new findings. The capacity impact on CACC follows a quadratic trend. However,
16
in cases of lower starting capacities, the trend begins to change to a linear trend. This infers that
17
the capacity impacts are not the same across all jurisdictions. Therefore, depending on the type of
18
existing roadway users, the capacity increase can sometimes follow a linear trend. We also
19
analyzed the effect of on-ramp demand on merge segment capacity. Results indicated that different
20
roadway capacities are achieved at different ramp demand levels. More interestingly, CACC
21
coordination can potentially reduce the effect of merging disturbance at on-ramps when the market
22
penetrations high enough. On weaving segments, results showed the capacity impacts of CACC
23
decreases with an increase in volume ratio. The weaving disturbances drastically reduces the
24
effects of CACC coordination. Even with an advanced merging capability, the effects of weaving
25
intensity were still pronounced.
26
Future work is required in this study. More complex freeway scenarios such as managed
27
lanes, higher weaving ratio, and two-lane on-ramps may be incorporated in future studies. Other
28
roadway segments, such as urban streets and arterials, can be investigated as well. Additionally,
29
the combined effect of other CAV applications that may potentially be implemented in the nearest
30
future may be considered in later studies.
31
ACKNOWLEDGMENTS
32
This study is supported in part by the Highway Capacity Manual Pooled Fund Study, led
33
by the Oregon Department of Transportation. The authors want to thank all the technical panel
34
members and other teammates (Anxi Jia, Paul Ryus, Yi Guo) for their insights throughout the
35
process of this work. The work presented in this paper remains the sole responsibility of the
36
authors.
37
AUTHOR CONTRIBUTIONS
38
The authors confirm contribution to the paper as follows: study conception and design: A.
39
Adebisi, J. Ma, Y. Liu, B. Schroeder, A. Morgan, and B. Cesme. data collection: A. Adebisi, J.
40
Ma, and Y. Liu; analysis and interpretation of results: A. Adebisi, J. Ma, Y. Liu, B. Schroeder, A.
41
21
Morgan, and B. Cesme; draft and manuscript preparation: A. Adebisi, J. Ma, Y. Liu, and B. Cesme.
1
All authors reviewed the results and approved the final version of the manuscript.
2
3
REFERENCES
4
1. Ye, L., and T. Yamamoto. Modeling Connected and Autonomous Vehicles in Heterogeneous
5
Traffic Flow. Physica A: Statistical Mechanics and its Applications, Vol. 490, 2018, pp. 269–
6
277. https://doi.org/10.1016/j.physa.2017.08.015.
7
2. Wang, Q., B. Li, Z. Li, and L. Li. Effect of Connected Automated Driving on Traffic
8
Capacity. Presented at the 2017 Chinese Automation Congress (CAC), Jinan, 2017.
9
3. Delis, A. I., I. K. Nikolos, and M. Papageorgiou. Macroscopic Traffic Flow Modeling with
10
Adaptive Cruise Control: Development and Numerical Solution. Computers & Mathematics
11
with Applications, Vol. 70, No. 8, 2015, pp. 1921–1947.
12
https://doi.org/10.1016/j.camwa.2015.08.002.
13
4. Talebpour, A., and H. S. Mahmassani. Influence of Connected and Autonomous Vehicles on
14
Traffic Flow Stability and Throughput. Transportation Research Part C: Emerging
15
Technologies, Vol. 71, 2016, pp. 143–163. https://doi.org/10.1016/j.trc.2016.07.007.
16
5. Ma, J., J. Hu, E. Leslie, F. Zhou, P. Huang, and J. Bared. An Eco-Drive Experiment on
17
Rolling Terrains for Fuel Consumption Optimization with Connected Automated Vehicles.
18
Transportation Research Part C: Emerging Technologies, Vol. 100, 2019, pp. 125–141.
19
https://doi.org/10.1016/j.trc.2019.01.010.
20
6. Guo, Y., J. Ma, C. Xiong, X. Li, F. Zhou, and W. Hao. Joint Optimization of Vehicle
21
Trajectories and Intersection Controllers with Connected Automated Vehicles: Combined
22
Dynamic Programming and Shooting Heuristic Approach. Transportation Research Part C:
23
Emerging Technologies, Vol. 98, 2019, pp. 54–72. https://doi.org/10.1016/j.trc.2018.11.010.
24
7. Ghiasi, A., X. Li, and J. Ma. A Mixed Traffic Speed Harmonization Model with Connected
25
Autonomous Vehicles. Transportation Research Part C: Emerging Technologies, Vol. 104,
26
2019, pp. 210–233. https://doi.org/10.1016/j.trc.2019.05.005.
27
8. Lazar, C., A. Tiganasu, and C. F. Caruntu. Arterial Intersection Improvement by Using
28
Vehicle Platooning and Coordinated Start. IFAC-PapersOnLine, Vol. 51, No. 9, 2018, pp.
29
136–141. https://doi.org/10.1016/j.ifacol.2018.07.023.
30
9. Yang, D., S. Zheng, C. Wen, P. J. Jin, and B. Ran. A Dynamic Lane-Changing Trajectory
31
Planning Model for Automated Vehicles. Transportation Research Part C: Emerging
32
Technologies, Vol. 95, 2018, pp. 228–247. https://doi.org/10.1016/j.trc.2018.06.007.
33
10. Melson, C. L., M. W. Levin, B. E. Hammit, and S. D. Boyles. Dynamic Traffic Assignment
34
of Cooperative Adaptive Cruise Control. Transportation Research Part C: Emerging
35
Technologies, Vol. 90, 2018, pp. 114–133. https://doi.org/10.1016/j.trc.2018.03.002.
36
11. Goñi-Ros, B., W. J. Schakel, A. E. Papacharalampous, M. Wang, V. L. Knoop, I. Sakata, B.
37
van Arem, and S. P. Hoogendoorn. Using Advanced Adaptive Cruise Control Systems to
38
Reduce Congestion at Sags: An Evaluation Based on Microscopic Traffic Simulation.
39
Transportation Research Part C: Emerging Technologies, Vol. 102, 2019, pp. 411–426.
40
https://doi.org/10.1016/j.trc.2019.02.021.
41
12. Liu, B., Q. Sun, and A. El Kamel. Improving the Intersection’s Throughput Using V2X
42
Communication and Cooperative Adaptive Cruise Control. IFAC-PapersOnLine, Vol. 49,
43
No. 5, 2016, pp. 359–364. https://doi.org/10.1016/j.ifacol.2016.07.139.
44
22
13. Shladover, S. E., D. Su, and X.-Y. Lu. Impacts of Cooperative Adaptive Cruise Control on
1
Freeway Traffic Flow. Transportation Research Record: Journal of the Transportation
2
Research Board, Vol. 2324, No. 1, 2012, pp. 63–70. https://doi.org/10.3141/2324-08.
3
14. Xiao, L., M. Wang, W. Schakel, and B. van Arem. Unravelling Effects of Cooperative
4
Adaptive Cruise Control Deactivation on Traffic Flow Characteristics at Merging
5
Bottlenecks. Transportation Research Part C: Emerging Technologies, Vol. 96, 2018, pp.
6
380–397. https://doi.org/10.1016/j.trc.2018.10.008.
7
15. Hu, X., and J. Sun. Trajectory Optimization of Connected and Autonomous Vehicles at a
8
Multilane Freeway Merging Area. Transportation Research Part C: Emerging Technologies,
9
Vol. 101, 2019, pp. 111–125. https://doi.org/10.1016/j.trc.2019.02.016.
10
16. Letter, C., and L. Elefteriadou. Efficient Control of Fully Automated Connected Vehicles at
11
Freeway Merge Segments. Transportation Research Part C: Emerging Technologies, Vol. 80,
12
2017, pp. 190–205. https://doi.org/10.1016/j.trc.2017.04.015.
13
17. Pueboobpaphan, R., F. Liu, and B. van Arem. The Impacts of a Communication Based
14
Merging Assistant on Traffic Flows of Manual and Equipped Vehicles at an On-Ramp Using
15
Traffic Flow Simulation. Presented at the 2010 13th International IEEE Conference on
16
Intelligent Transportation Systems - (ITSC 2010), Funchal, Madeira Island, Portugal, 2010.
17
18. Laufer, J., and M. Australia. Freeway Capacity, Saturation Flow and the Car Following
18
Behavioural Algorithm of the VISSIM Microsimulation Software. p. 15.
19
19. PTV Vissim User Manual. p. 1219.
20
20. Milanés, V., and S. E. Shladover. Modeling Cooperative and Autonomous Adaptive Cruise
21
Control Dynamic Responses Using Experimental Data. Transportation Research Part C:
22
Emerging Technologies, Vol. 48, 2014, pp. 285–300.
23
https://doi.org/10.1016/j.trc.2014.09.001.
24
21. Liu, H., X. (David) Kan, S. E. Shladover, X.-Y. Lu, and R. E. Ferlis. Impact of Cooperative
25
Adaptive Cruise Control on Multilane Freeway Merge Capacity. Journal of Intelligent
26
Transportation Systems, Vol. 22, No. 3, 2018, pp. 263–275.
27
https://doi.org/10.1080/15472450.2018.1438275.
28
22. Liu, H., X. (David) Kan, S. E. Shladover, X.-Y. Lu, and R. E. Ferlis. Modeling Impacts of
29
Cooperative Adaptive Cruise Control on Mixed Traffic Flow in Multi-Lane Freeway
30
Facilities. Transportation Research Part C: Emerging Technologies, Vol. 95, 2018, pp. 261–
31
279. https://doi.org/10.1016/j.trc.2018.07.027.
32
23. Liu, H. Liu, H., X. (David) Kan, S. E. Shladover, X.-Y. Lu. Using Cooperative Adaptive
33
Cruise Control (CACC) to Form High- Performance Vehicle Streams: Simulation Results
34
Analysis. California PATH (2018). p. 47.
35
24. Nowakowski, C. Cooperative Adaptive Cruise Control: Driver Acceptance of Following Gap
36
Settings Less than One Second. th ANNUAL MEETING, 2010, p. 5.
37
25. Highway Capacity Manual 2010. Transportation Research Board of the National Academies,
38
Washington, D.C., 2016.Pdf.
39
26. Kondyli, A., P. Gubbala, and L. Elefteriadou. The Contribution of Ramp Demand in the
40
Capacity of Merge Bottleneck Locations. Transportation Research Procedia, Vol. 15, 2016,
41
pp. 346–355. https://doi.org/10.1016/j.trpro.2016.06.029.
42
27. Shen, J., W. Li, F. Qiu, and S. Zheng. Capacity of Freeway Merge Areas with Different On-
43
Ramp Traffic Flow. PROMET - Traffic&Transportation, Vol. 27, No. 3, 2015, pp. 227–235.
44
https://doi.org/10.7307/ptt.v27i3.1566.
45
23
28. Chen, X., L. Yu, X. Jia, and H. Gong. Capacity Modeling for Weaving, Merge, and Diverge
1
Sections with Median Exclusive Bus Lanes on an Urban Expressway: Microsimulation
2
Approach. Transportation Research Record: Journal of the Transportation Research Board,
3
Vol. 2553, No. 1, 2016, pp. 99–107. https://doi.org/10.3141/2553-11.
4
29. Guo, Y., and J. Ma. "Leveraging existing high-occupancy vehicle lanes for mixed-autonomy
5
traffic management with emerging connected automated vehicle applications."
6
Transportmetrica A: Transport Science 16, no. 3 (2020): 1375-1399.
7
8
