Effects of Connected Highly Automated Vehicles on the Propagation of Congested Patterns on Freeways

Abstract

Highly automated vehicles (HAV) use sensing technologies to take over the task of driving, while connected vehicles obtain and share information that can allow the driver/vehicle to make better driving decisions. Connected HAVs promise to offer significant improvements in traffic performance, emergence and propagation of congestion due to their capabilities of vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication. However, in short-term these vehicles operate along the manually driven vehicles. We employed the microscopic traffic simulation tool Vissim to model HAVs, with communication capabilities, and manually driven vehicles by implementing different behavioral models for car-following and to analyze if and to what extent these vehicles can influence the propagation of congestions along the freeway by means of changing the propagation speed and the throughput of the network.

Full text

EFFECTS OF CONNECTED HIGHLY AUTOMATED VEHICLES ON THE
1
PROPAGATION OF CONGESTED PATTERNS ON FREEWAYS
2
3
Nassim Motamedidehkordi
4
Chair of Traffic Engineering and Control
5
Technische Universitaet Muenchen
6
Arcisstrasse 21, 80333 Munich, Germany
7
Phone: +49-89-28922665; Fax: +49-89-28922333; Email: nassim.motamedidehkordi@tum.de
8
9
Martin Margreiter
10
Chair of Traffic Engineering and Control
11
Technische Universitaet Muenchen
12
Arcisstrasse 21, 80333 Munich, Germany
13
Phone: +49-89-28928586 Fax: +49-89-28922333; Email: martin.margreiter@tum.de
14
15
Thomas Benz
16
PTV Group
17
Haid-und-Neu-Strasse 15, 76131 Karlsruhe, Germany
18
Phone: +49-721-19651336; Fax: +49-721-9651299; Email: thomas.benz@ptvgroup.com
19
20
21
Word count: 4,528 words text + 8 tables/figures x 250 words (each) = 6,528 words
22
23
24
Submission Date: July 31, 2015
25
Revision Date: November 11, 2015
26

Motamedidehkordi, Margreiter, Benz 2
1
ABSTRACT
2
Highly automated vehicles (HAV) use sensing technologies to take over the task of driving, while
3
connected vehicles obtain and share information that can allow the driver/vehicle to make better
4
driving decisions. Connected HAVs promise to offer significant improvements in traffic
5
performance, emergence and propagation of congestion due to their capabilities of vehicle-to-
6
vehicle (V2V) and vehicle-to-infrastructure (V2I) communication. However, in short-term these
7
vehicles operate along the manually driven vehicles. We employed the microscopic traffic
8
simulation tool Vissim to model HAVs, with communication capabilities, and manually driven
9
vehicles by implementing different behavioral models for car-following and to analyze if and to
10
what extent these vehicles can influence the propagation of congestions along the freeway by
11
means of changing the propagation speed and the throughput of the network.
12
13
Keywords: highly automated vehicles, V2V communication, congestion propagation, traffic jam
14
ahead warning, adaptive smoothing method
15

Motamedidehkordi, Margreiter, Benz 3
INTRODUCTION
1
Nowadays people tend to travel more and more and traffic congestion has become a major
2
challenge for people across the world. In most countries, the construction of new transport
3
infrastructure is not an appropriate option any more, leading to the need of a more efficient way
4
of using the existing road capacities. A lot of research is conducted on how to reduce traffic
5
congestion and many studies propose the possibilities to reach this goal; however, many of them
6
only relocate the bottleneck from one location to another instead of targeting the cause of the
7
problem and optimizing the traffic flow.
8
Interest has increased in recent decades in the automation of the road transport in order to
9
overcome the enormous costs of congestion, pollution and traffic accidents of road transport (1),
10
(2). Automated vehicles could increase the segment capacity through lower time headways and
11
therefore reduce the temporal and spatial size of recurring congestions (3). Connected vehicles
12
could similarly reduce the duration and size of non-recurring congestions resulting from a
13
temporary bottleneck like an incident by reducing the frequency of disruptions (4). Connected
14
automated driving is expected to reduce the congestion because of the capability to predict the
15
near future downstream traffic state. The potential benefit of automation functions like adaptive
16
cruise control (ACC) and cooperative adaptive cruise control (CACC) has been shown in other
17
studies (5), (6), (7). Other than CACC, inter-vehicle communication of automated vehicles
18
provides the possibility of automated merging of vehicles, platooning of cars or trucks and traffic
19
jam ahead warning (TJAW) application.
20
TJAW application is assumed to avoid rear-end collisions and may reduce the traffic
21
congestion problem by taking advantage of communication capabilities of the vehicles as
22
wireless communication provides real-time information of proceeding vehicles in case of a traffic
23
breakdown. Thus, the vehicles can start reacting to the congestion faster and smoother than
24
conventional vehicles and beyond the line of the sight.
25
This paper focuses on simulating mixed traffic consisting of connected highly automated
26
and manually driven vehicles with the microscopic traffic simulation tool Vissim. This is done in
27
order to understand better their impact on freeways in case of traffic congestion because large-
28
scale field experiments are scarcely possible. In section 1, we examine existing literature on the
29
simulation of HAVs and the assessment of their impact. Thereafter, the simulation parameters
30
used in this research to simulate the connected HAVs are explained in detail. The results of the
31
simulation of different penetration rates and the method called “Adaptive Smoothing Method”
32
(ASM) which was used for the reconstruction of spatio-temporal data and generating speed
33
contour plots are explained in section 4. In section 5 we draw some preliminary conclusions and
34
present the potential further research topics.
35
36
LITERATURE REVIEW
37
One of the earliest research actions, which pioneered practical self-driving technology on roads,
38
was the PROMETHEUS-Project (1987-1995) in the EUREKA program (8). Another step in
39
automation, the platooning, was motivated by intelligent highway systems and road infrastructure
40
in 1997 within the PATH project in California (9). Some years later, the DARPA Grand
41
Challenge starting in 2004 and 2005 on a 350 km off-road track in the Mojave Desert encouraged
42
several producers of automated vehicles to test their developments in a real environment and in
43
competition to other teams. Two years later, in 2007, the DARPA Urban Challenge already
44
included an urban route and therefore containing interactions and cooperation with other
45
surrounding vehicles. This was followed up in 2010 by the automated vehicle ‘Leonie’ in
46
Braunschweig, Germany, being one of the first vehicles worldwide to fully autonomously drive in
47
a real urban traffic environment (10). Since then a vast number of also long-range test drives of
48

Motamedidehkordi, Margreiter, Benz 4
automated vehicles have been carried out, like the VisLab Intercontinental Autonomous
1
Challenge, which covers 13,000 km from Italy to China with no human intervention (11).
2
Additionally, lots of research focuses until now on simulating and determining the
3
impacts of such cooperative systems (supported by the IEEE 802.11p standard) and connected
4
automated vehicles in different penetration rates on the overall traffic efficiency. In the German
5
research project simTD a field operational test (FOT) as well as a traffic simulation on cooperative
6
vehicles with V2V and V2I capabilities were conducted. The project showed, that cooperative
7
assistance systems could improve the traffic efficiency and the traffic safety for example on
8
motorways (12), (13). Also M. Gouy et al. (14) shows in a simulator study how the use of V2V
9
communication increases the capacity of the existing road infrastructure. The effects on the road
10
capacity of an increasing penetration rate of CACC equipped vehicles were also discussed by
11
J. Van der Werf et al. (7). He concluded that high rates could potentially double the capacity of a
12
highway.
13
Not just cars, but also cooperative trucks are in the focus, like in the CHAUFFEUR 2
14
project (15), where a V2V truck platooning capability was developed and tested using the
15
microscopic traffic simulation models Vissim and FARSI. The result was a higher road capacity
16
and up to 20 % reduction in fuel consumptions due to more stability in the traffic flow of the
17
truck platoons. This effect especially showed up at night or on road sections with low traffic
18
volumes.
19
In contrast to most of the studies which focus on the impact of connected automated
20
vehicles on the throughput and the changes in the capacity of the network, W. Schakel et al. (16)
21
analyzes the effects on the traffic flow stability and the occurrence and characteristics of
22
shockwaves. It was found out, that at a 50 % penetration rate of vehicles equipped with CACC
23
the damping of shockwaves is more quickly and the shockwave speed decreases from -4.4 m/s in
24
the base scenario (using an adapted version of the Intelligent Driver Model) to about -18.5 m/s
25
with CAAC.
26
Also B. Van Arem et al. (6) concentrates on the effects on shockwaves using the traffic
27
flow simulation model MIXIC to simulate a highway merging scenario from four to three lanes.
28
The results show an improvement of the traffic flow stability and a slight increase in traffic flow
29
efficiency as well as a drastically decrease of the number of shockwaves with higher penetration
30
rates of CACC-equipped vehicles. In contradiction to W. Schakel et al. (16) no significant change
31
in the shockwave speeds could be observed. A decrease in the number of shockwaves and
32
therefore an improvement in the traffic stability could also be found in the CarTALK2000
33
project, also using the simulation model MIXIC for cooperative vehicles (17).
34
There are several steps towards automated driving focusing on automated highway
35
systems (AHS), platooning and cooperative adaptive cruise control (CAAC) which were and still
36
are in the focus of research in the past years. What most of the studies have in common is the
37
realized reduction in the vehicles time headway (THW) as one of the key parameters of
38
automated vehicles, which have an influence on the traffic flow. In literature, several values can
39
be found:
40
 THW of 0.5 s when following another CACC equipped vehicle and 1.4 s when following
41
a non-CACC equipped vehicle (6).
42
 Automated vehicle platoons with short following distance with THW of 0.3 s and large
43
following distance with THW of 1.4 s (14).
44
 THW of 0.5 s for 100 % penetration rate of CACC equipped vehicles (7).
45
 Truck platooning with THW ranging from 0.3 s to 0.6 s at a speed of 80 km/h (15).
46

Motamedidehkordi, Margreiter, Benz 5
 CACC with THW of 1.2 s including two different time headway distributions: 1.2±0.15 s
1
and 1.2±0.3 s (16).
2
3
SIMULATION FRAMEWORK
4
In order to develop the understanding of the effect of HAVs on freeways, we present the result of
5
a simulation study, which models a small part of the freeway network in Germany with traffic
6
consisting of HAVs and conventional vehicles. In this section, we detail on the simulation
7
software used and the behavior models and parameters used in the models.
8
The Vissim software uses two car-following models developed by Rainer Wiedemann. In
9
the Wiedemann 74 model, the minimum safety distance is calculated based on formula 1 (18).
10
11
   󰇛   󰇜 (1)
12
13
 Speed of the slower vehicle [m/s].
14
15
: Is a value in the range [0,1] which is normally distributed around 0.5 with a standard
16
deviation of 0.15.
17
18
 Average standstill distance, which defines the average desired distance between two
19
vehicles.
20
21
: Additive part of the following distance, which allows adjusting the time requirement
22
values.
23
24
 Multiplicative part of the following distance, which allows adjusting the time
25
requirement values.
26
27
28
29
FIGURE 1 Car-following model and driving states (19).
30
31

Motamedidehkordi, Margreiter, Benz 6
Some thresholds are calculated differently for the Wiedemann 99 car-following model than they
1
were for the Wiedemann 74 model. Nevertheless, the meaning of each threshold is the same. At a
2
certain speed, the minimum safety is calculated based on formula 2.
3
4
   (2)
5
6
 Defines the desired rear bumper to front bumper distance between stopped vehicles. This
7
parameter has no variation.
8
9
 Defines the time (in seconds) that the following driver wishes to maintain a certain speed.
10
The higher this value, the more conservative the driver drives.
11
12
A tool for the simulation of vehicle behavior and traffic properties, which would allow assessing
13
the impact and benefit of HAVs with different penetration rates in the vehicle composition, has
14
been implemented. For this purpose, the Vissim COM interface was used, which is versatile in
15
collecting vehicles information and modifying vehicles parameters during the simulation period.
16
Human drivers and connected HAVs are simulated on an 8.5 km stretch of the freeway A5 in
17
Germany. Travel demand, the number of vehicles entering the network, was derived directly from
18
the data recorded by detectors on the freeway A5. The simulation was used to generate one hour
19
of simulation data. The first 15 minutes of the simulation data, while the system initially
20
populates and the network is being filled, is not used. In order to trigger stop-and-go traffic that
21
ultimately leads to traffic congestion in the network, a temporary artificial bottleneck was created
22
in the network using a reduced speed area with very low speeds.
23
For manually driven vehicles, the Wiedemann 74 car-following model was assigned as
24
they have variations in their driving behavior. In order to calibrate the model, empirical detector
25
data and individual vehicle trajectories of the large-scale field operational trial at the German
26
research project simTD (20) were used (21). Vehicle fleet, desired speed, desired acceleration and
27
deceleration of the vehicles were derived directly from the empirical data. According to
28
calibration of the car-following model based on empirical data,  (standstill distance) was
29
determined with 1 m.  (additive part of safety distance) and  (multiplicative part of
30
safety distance) value were set to 3 and 5 respectively.
31
In contrast to the non-automated vehicles, HAVs have no variation in their behavior.
32
Therefore, the Wiedemann 99 model was assigned. Additional data about driving behavior are
33
needed to parameterize the model. The car-following behavior can be defined by setting
34
parameters of the Wiedemann model considering different stages. Since there is no data about the
35
behavior of HAVs in the simulation in hand, some assumptions about the parameters of these
36
vehicles have been made. The CC0 (standstill distance) was set to 1 meter. CC1 (desired time
37
between lead and following vehicle) was set to 0.5 seconds. CC2 (following variation oscillation)
38
was determined with the value of 4 meters. CC3 (time before a vehicle starts to decelerate to
39
reach safety distance) was set to -8 seconds. CC4 and CC5 values (following thresholds for
40
positive and negative speed differences during following state) were set to -0.1 and 0.1
41
respectively. CC6 (influence of distance on speed oscillation) was chosen to be 0. The values for
42
CC7 (acceleration during the oscillation process), CC8 (desired acceleration starting from
43
standstill) and CC9 (desired acceleration at 80 km/h) were set to 0.25, 3.5 and 1.5 m/s²
44
respectively. Besides, the safety reduction factor used for lane changing of HAVs was set to 0.75
45
and the maximum deceleration for cooperative braking was set to -3.5 m/s².
46
The resulting difference in the following behavior of the Wiedemann models for a HAV
47
and a conventional vehicle in exactly the same car-following situation can be seen in figure 2. A
48

Motamedidehkordi, Margreiter, Benz 7
conventional vehicle has a higher mean time headway while following its predecessor as well as
1
higher standard deviation of the THW.
2
3
4
5
FIGURE 2 Comparison of the time headway between a highly automated vehicle and a
6
conventional vehicle in the car-following mode.
7
8
The spatial propagation of information along the freeway for a distance of 1 km is simulated in
9
this research effort. In case the speed of a HAV dropped abruptly to a value below 40 km/h, the
10
TJAW transmission is activated. The communication between HAVs in this research is therefore
11
limited to transmitting and receiving the TJAW messages. The desired speed of the HAVs, which
12
received the warning, was set to 70 km/h. It has to be mentioned, that the desired speed of the
13
HAVs, in contrast to manually driven vehicles, does not have a distribution and has the value
14
equal to the speed limit.
15
The simulation is performed for mixed flow with 0 %, 5 %, 10 %, 20 % and 50 % of
16
HAVs randomly dispersed among manually driven vehicles. Three simulations with different
17
starting random seeds were run for each penetration rate and the results of the simulation runs
18
were averaged in order to overcome the stochastic nature of simulation.
19
20
SIMULATION RESULTS
21
With the help of the simulation tool, the data at the exact positions of the inductive loops on the
22
freeway were recorded with an aggregation interval of one minute. In reality, stationary detectors
23
provided minute-by-minute data of speed and flow. Usually, the traffic state cannot be directly
24
measured everywhere but needs to be estimated from incomplete, noisy and local traffic data.
25
Commonly, volumes or average vehicle speeds are measured at certain locations on the freeways,
26
for example by double-loop detectors. To estimate the overall traffic state from these point
27
measurements, an interpolation between the detectors is necessary. The speed contour plots for
28
different penetration rates of HAVs can be observed in the linearly interpolated graphs in figure 3.
29
For better illustration, the detector locations are marked with dashed lines. It can be observed that
30

Motamedidehkordi, Margreiter, Benz 8
the calculated speeds between the detectors are not realistic and the artifacts of the linear
1
interpolation can be obviously distinguished especially where the detectors are located far from
2
each other (more than 1 km).
3
In order to avoid the artifacts of this interpolation, which can be observed in the graphs,
4
the data was reconstructed using the adaptive smoothing method (ASM).
5
6
7
8
FIGURE 3 Speed contour plots for a) base scenario and b) 5 %, c) 10 %, d) 20 % and e)
9
50 % penetration rate of connected highly automated vehicles within the vehicle fleet.
10
11

Motamedidehkordi, Margreiter, Benz 9
ADAPTIVE SMOOTHING METHOD
1
The role of this two-dimensional spatio-temporal interpolation algorithm is to estimate at each
2
location  and time  the discrete speed  󰇛󰇜 as a result of the continuous function of
3
the local speed average 󰇛󰇜 ASM was developed by Treiber and Helbing (22) and yields to a
4
plausible reconstruction of the traffic state. This smoothing procedure is based on a two-
5
dimensional interpolation in space and time and takes into account the information about the two
6
typical velocities of underlying regimes in free flow and congested traffic (23):
7
1) In free flow traffic, perturbations move downstream (i.e., in the direction of traffic flow)
8
(24) with the propagation velocity , which is slightly lower than the local average
9
speed of the vehicles.
10
2) In congested traffic, however, perturbations travel upstream (against the movement of the
11
vehicles) due to the reaction of the drivers to their leading vehicle.
12
13
Based on these two traffic regimes two smoothed speed fields are considered.
14
15
󰇛󰇜 
󰇛󰇜
    
  (3)
16
17
󰇛󰇜 
󰇛󰇜
    
 (4)
18
19
20
󰇛󰇜: Smoothed speed in free flow traffic.
21
22
󰇛󰇜: Smoothed speed in congested traffic.
23
24
󰇛󰇜: Normalization of the weighting function in free flow traffic.
25
26
󰇛󰇜: Normalization of the weighting function in congested traffic.
27
28
: Kernel that includes all the data points .
29
30
: Perturbations propagation velocity in free flow traffic.
31
32
: Perturbations propagation velocity in congested situations.
33
34
The kernels 󰇛󰇜 and󰇛󰇜 of the linear homogeneous filters realize the required
35
smoothing and are particularly transmissible for perturbations propagating with the typical
36
velocities and  observed in congested and free flow traffic, respectively.
37
38
For the weighting kernel  the symmetric exponential function has been used in which and 
39
are the smoothing widths in the spatial and temporal coordinates. The exponential function
40
operates as a low pass filter.
41
42
󰇛    󰇜 󰇣󰇡

󰇢󰇤 (5)
43
44

Motamedidehkordi, Margreiter, Benz 10
1
2
FIGURE 4 Smoothing kernels for free flow and congested traffic. The slope of each kernel
3
represents different characteristic velocities cfree and ccong (23).
4
5
In the end, due to different smoothing in free flow and congested traffic, the average rate in the
6
formula below will be applied: The speed factor 󰇛󰇜 in the equation depends on the average
7
speeds  and  which takes    for congested traffic and   for free flow traffic.
8
The predictor 󰇛󰇜 leads to a better segregation of congested traffic from free flow traffic. The
9
parameter  is the transition between free flow and congested traffic.
10
11
󰇛󰇜 󰇛󰇜󰇛󰇜󰇟 󰇛󰇜󰇠 󰇛󰇜 (6)
12
13
󰇛󰇜
󰇟 󰇡
 󰇢󰇠 (7)
14
15
󰇛󰇜 󰇟󰇠 (8)
16
17
In table 1 the parameters used for the ASM in this study are illustrated:
18
19
TABLE 1 Parameters of the "Adaptive Smoothing Method" used for this study.
20
21
Parameter
Meaning
Value

Range of spatial smoothing in 
500 meters

Range of temporal smoothing in 
30 seconds

Propagation velocity of perturbations in free traffic
80 km/h

Propagation velocity for perturbations in congested traffic
-15 km/h

Crossover for free to congested traffic
60 km/h

Width of transition region
20 km/h
22

Motamedidehkordi, Margreiter, Benz 11
This smoothing method was applied on the simulation result of each scenario. The speed contour
1
plots for different penetration rates of connected HAVs after smoothing are illustrated in figure 5.
2
3
4
5
FIGURE 5 Smoothed speed contour plots for a) base scenario and b) 5 %, c) 10 %, d) 20 %
6
and e) 50 % penetration rate of connected HAVs within the vehicle fleet.
7
8
Although the reconstructed traffic state between the detectors looks more realistic than the non-
9
smoothed data, this low pass filter smoothed out perturbations observed in the graphs in figure 2.
10
Therefore, based on the application of the reconstructed data, the plausibility of using this filter
11
should be carefully considered.
12

Motamedidehkordi, Margreiter, Benz 12
From the graphs observation, it turns out that with increasing penetration rates of
1
connected HAVs, the congestion area becomes smaller and the queue length decreases. This is
2
primarily due to a reduction in time headways, as the vehicles are able to follow each other at
3
very short gaps. Moreover, stop-and-go instabilities caused by driver response lags and higher
4
acceleration rates of the vehicles are avoided. These parameters provide the possibility to pack
5
more vehicles within the existing infrastructure. On the other hand, as we have more HAVs in the
6
vehicle fleet, a more homogeneous traffic flow and fewer perturbations were observed. The
7
following parameters were carried out from the model to evaluate the network performance:
8
 Propagation speed of congestion.
9
 Network throughput.
10
Congested traffic is characterized by a high traffic density and a low average velocity (23). Since,
11
in contrast to the density, the velocity can be directly measured with stationary detectors, we
12
chose the velocity to determine the congestion. In order to calculate the propagation speed of the
13
congestion, an abrupt change of the speed recorded by detectors to values below 50 km/h (within
14
one minute) has been recognized and a linear regression analysis with fits the best line to these
15
points has been conducted. The slope of this line is considered the propagation speed of
16
congestion along the freeway. In the figure below the propagation speeds for different penetration
17
rates can be found. The negative values confirm that the congested pattern was propagating in the
18
opposite direction of traffic and a backward shockwave was forming.
19
20
21
22
FIGURE 6 Propagation speed of congestion for different penetration rates of HAVs.
23
24

Motamedidehkordi, Margreiter, Benz 13
The graph below shows the throughput (total outflow during one hour) of the network for the
1
different penetration rates of HAVs during the one-hour simulation run. The simulation results
2
showed that the throughput increase is non-linear and most of the change is observed before the
3
penetration rate of 20 %. 50 % of connected HAVs, which are randomly dispersed among the
4
conventional vehicles, will increase the throughput of the network by 5 % in comparison to the
5
base scenario.
6
7
8
9
FIGURE 7 Network throughput for different penetration rates of HAVs.
10

Motamedidehkordi, Margreiter, Benz 14
1
CONCLUSIONS AND DISCUSSIONS
2
This study focuses on evaluating the potential effect of connected HAVs on the propagation of
3
the congestion caused by a bottleneck on a freeway by means of a microscopic simulation study.
4
HAVs were defined and implemented by selecting appropriate models and parameters describing
5
their driving behavior. For conventional vehicles, the Wiedemann 74 parameters were calibrated
6
based on data from local detectors. Wiedemann 99 was configured to fit the assumptions for the
7
car-following behavior of HAVs (e.g. headway distribution). Besides that, the driving behavior of
8
a HAV is changed in case it receives a TJAW from other HAVs which are experiencing a
9
downstream congestion. The impact of this setup has been evaluated by simulating 5 %, 10 %,
10
20 % and 50 % penetration rates of HAVs.
11
It has been shown that with the increase of the penetration rate of these vehicles within the
12
fleet the congestion length and the congestion area become smaller. The communication between
13
HAVs allows them to react earlier and safer on a congestion tail by receiving a warning whilst
14
approaching. On the other hand, keeping smaller time headways between these vehicles increases
15
the capacity of the existing road. As a result, having a higher penetration rate of HAVs
16
significantly reduces shockwave speeds from the value of 13.9 km/h for 0 % of HAVs down to
17
8.7 km/h with a 50 % penetration rate. Therefore, HAVs improve traffic flow and thus decrease
18
the traffic congestion whilst retaining driving comfort but their effect is highly dependent on the
19
penetration rate.
20
In this study, an assumption has been made about the driving behavior of HAV and of the drivers
21
of the surrounding manually driven vehicles. Extensive test drives with HAVs would be
22
preferable for calibrating their driving behavior and dynamics in the simulation models. The
23
identification and integration of the interaction and behavior of road users in the simulation of
24
automated vehicles is a future research need.
25
26

Motamedidehkordi, Margreiter, Benz 15
1
REFERENCES
2
3
1. Hoogendoorn, R., B. Van Arem, and S. Hoogerndoorn. "Automated Driving, Traffic Flow
4
Efficiency and Human Factors." Transportation Research Record: Journal of the
5
Transportation Research Board. 2014, pp. 113-120.
6
2. Malakorn, K. J., and B. Park. "Assessment of Mobility, Energy, and Environment Impacts
7
of IntelliDrive-based Cooperative Adaptive Cruise Control and Intelligent Traffic Signal
8
Control." IEEE International Symposium on Sustainable Systems and Technology.
9
Arlington, VA, May 17-19, 2010, pp. 1-6.
10
3. Motamedidehkordi, N., T. Benz, and M. Margreiter. “Shockwave Analysis on Motorways
11
and Possibility of Damping by Autonomous Vehicles”. Author: T. Schulze, B. Müller and
12
G. Meyer. Advanced Microsystems for Automotive Applications 2015: Smart Systems for
13
Green and Automated Driving. Switzerland: Springer International Publishing, 2015, pp.
14
37-52.
15
4. Bierstedt, J., A. Gooze, C. Gray, J. Peterman, L. Raykin, and J. Walters. Effect of Next-
16
generation Vehicles on Travel Demand and Highway Capacity. FT Think, 2014.
17
5. Schladover, S., D. Su, and X. Lu. "Impacts of Cooperative Adaptive Cruise Control on
18
Freeway Traffic Flow." Transportation Research Record: Journal of the Transportation
19
Research Board, Volume 2324. 2014, pp. 63-70.
20
6. Van Arem, B., C. J. G. Van Driel, and R. Visser. "The Impact of Co-operative Adaptive
21
Cruise Control on Traffic Flow Characteristics." IEEE Transactions on Intelligent
22
Transportation Systems, Volume 7, Issue 4. December 2006, pp. 429-436.
23
7. Van der Werf, J., S. E. Shladover, M. A. Miller, and N. Kourjanskaia. "Effects of
24
Adaptive Cruise Control Systems on Highway Traffic Flow Capacity." Transportation
25
Research Record: Journal of the Transportation Research Board, Volume 1800. 2002,
26
pp. 78-84.
27
8. Scholl, W. EUREKA - PROMETHEUS. http://www.eurekanetwork.org/project//id/45.
28
Accessed July 30, 2015.
29
9. Schladover, S. "The GM–PATH Platoon Scenario." Intellimotion: Research Updates in
30
Intelligent Transportation Systems, Volume 6, No. 3. 1997, p. 2.
31
10. Nothdurft, T., et al. "Stadtpilot: First Fully Autonomous Test Drives in Urban Traffic."
32
14th International IEEE Conference on Intelligent Transportation Systems. Washington,
33
D.C., USA, October 5-7, 2011.
34
11. Broggi, A., L. Bombini, S. Cattani, P. Cerri, and R. I. Fedriga. "Sensing Requirements for
35
a 13,000 km Intercontinental Autonomous Drive." IEEE Intelligent Vehicles Symposium.
36
University of California, San Diego, CA, USA, June 21-24, 2010.
37
12. Baur, M., F. Schimandl, S. Gabloner, M. Margreiter, and S. Hoffmann. "An Integrated
38
Approach for the Traffic-related Evaluation of Cooperative Systems in FOTs and Traffic
39
Simulations." 19th ITS World Congress. Vienna, Austria, October 22-26, 2012.
40
13. Schimandl, F., M. Baur, S. Hoffmann, M. Margreiter, and S. Gabloner. "Effects of
41
Cooperative Systems on Traffic Safety and Efficiency - Results of the German simTD
42
Project." mobil.TUM 2013 Conference - ITS for Connected Mobility. Munich, Germany,
43
June 18-19, 2013.
44
45

Motamedidehkordi, Margreiter, Benz 16
14. Gouy, M., K. Wiedemann, A. Stevens, G. Brunett, and N. Reed. "Driving Next to
1
Automated Vehicle Platoons: How Do Short Time Headways Influence Non-platoon
2
Drivers’ Longitudinal Control?" Transportation Research Part F 27. 2014.
3
15. Benz, T., et al. "CHAUFFEUR 2: Final Report, Deliverable D24." DaimlerChrysler,
4
Stuttgart, Germany, 2003.
5
16. Schakel, W. J., B. Van Arem, and B. D. Netten. "Effects of Cooperative Adaptive Cruise
6
Control on Traffic Flow Stability." 13th International IEEE Conference on Intelligent
7
Transportation Systems. 2010, pp. 759-764.
8
17. Malone, K. M., and B. Van Arem. "Traffic Effects of Inter-vehicle Communication
9
Applications in CarTALK2000." 11th World Congress on Intelligent Transportation
10
Systems. Nagoya, Japan, 2004.
11
18. Wiedemann, R. Simulation des Strassenverkehrsflusses, Band 8. Karlsruhe, Germany,
12
1974.
13
19. PTV AG. Vissim 8 Manual. Karlsruhe, Germany, 2015.
14
20. Stuebing, H., et al. "simTD: A Car-to-X System Architecture for Field Operational Tests."
15
IEEE Communications Magazine 06/2010. 2010.
16
21. Motamedidehkordi, N., M. Margreiter, and T. Benz. "Shockwave Analysis on Motorways
17
and Possibility of Damping by C2X Applications." Mobil.TUM 2015 - Technologies,
18
Solutions and Perspectives for Intelligent Transport Systems. Munich, Germany, June 30
19
and July, 2015.
20
22. Treiber, M., and D. Helbing. Reconstructing the Spatio-Temporal Traffic Dynamics from
21
Stationary Detector Data. 2002, pp. 3.1-3.24.
22
23. Treiber, M., and A. Kesting. Traffic Flow Dynamics. Springer, Berlin, 2013.
23
24. Kerner, B., and H. Rehhborn. "Experimental Features and Characteristics of Traffic
24
Jams." Physical Review E, Vol. 53-1297. February 1, 1996.
25