Content uploaded by Ahmed Hussein
Author content
All content in this area was uploaded by Ahmed Hussein on Apr 28, 2018
Content may be subject to copyright.
Autonomous Cooperative Driving Using V2X Communications
in Off-Road Environment
Ahmed Hussein, Pablo Mar´
ın-Plaza, Fernando Garc´
ıa and Jos´
e Mar´
ıa Armingol
Intelligent Systems Lab (LSI) Research Group
Universidad Carlos III de Madrid (UC3M), Legan´
es, Madrid, Spain
Email: {ahussein, pamarinp, fegarcia, armingol}@ing.uc3m.es
Abstract— During the last decade, a great deal of attention
is received for the idea of designing autonomous vehicles,
with intelligent, cooperative and coordinated capabilities to
reach an overall optimal performance and robustness. Almost
all intelligent vehicles are currently equipped with computing
modules, communication modules and multiple sensors for
environment perception. Accordingly, it is a matter of utiliz-
ing these sensors and modules to enhance and improve the
transportation safety and efficiency. In this paper, autonomous
cooperative driving using various communications schemes in
the off-road environment is presented. Moreover, it highlights
the importance of Vehicle-To-Vehicle (V2V) communication
and shows the advantages of Vehicle-To-Pedestrian (V2P) and
Pedestrian-To-Vehicle (P2V) communications when added to
the system environment perception. Several experiments have
been performed for real-life scenarios to verify the results,
and evaluate the performance of the proposed approach. The
experiments outcomes prove the high performance of the
proposed approach in real-world application and emphasize the
importance of V2X communications in autonomous vehicles.
I. INTRODUCTION
In December 2015, the U.S. Department of Transportation
(USDOT) announced the Smart City Challenge, the idea was
mainly focused on creating an integrated smart city [1]. One
of the main tasks in the challenge is the application of Au-
tomated and Connected Vehicles (ACV) [2]. Transportation
system is one of the key factors in smart cities research, as
it should be able to accommodate massive volume of road
users, decreases the number of accidents due to human error,
and support many number of services required by the city [3].
Researchers in the Intelligent Transportation Systems
(ITS) field are investing more time on the autonomous
cars research and development. For instance, Google Inc-
orporation launched the self-driving car project back in
2011 [4]. In addition to several car manufactures, such as
BMW and Audi [5], Mercedes-Benz [6], Volvo [7] are one
step closer to the mass production of autonomous cars. Last
but not least, autonomous public transportation systems were
proposed in [8], [9].
This research growth in the ITS field, lead many research
institutes to implement various approaches to optimize and
improve the performance of multiple autonomous vehi-
cles [10], [11]. Consequently this paper studies the effects
of various communication schemes over autonomous co-
operative driving. Where Vehicle-To-Vehicle (V2V) comm-
unication is used to coordinate between the multiple vehicles,
in addition to Vehicle-To-Pedestrian (V2P) and Pedestrian-
To-Vehicle (P2V) communications are used to enhance the
system environment perception. The main contribution of
this paper is to verify the importance of V2X communica-
tions in autonomous vehicles cooperative driving in off-road
environment during real-life scenarios. This is knowing that
off-road environments are unstructured compared to traffic
roads, which add more challenges to the autonomous such
as more unpredictability of surroundings, variety of obstacles
and more obstructions.
The remainder of the this paper consists of five sections.
Section II describes the driverless platforms used in the
experimental work and section III shows the available comm-
unication schemes in each platform. Followed by section IV,
which illustrates the proposed cooperative driving approach
for a multiples vehicles and how the different communication
schemes affect the cooperative performance. The selected
scenario, performed experiments and results discussion are
presented in section V. Finally, section VI summarizes the
work and highlights future work.
II. PLATF ORM DESCRIPTION
The autonomous platforms used in this work, as shown in
Figure 1, are the vehicles of the Intelligent Campus Auto-
mobile (iCab) project. They are electric golf carts that can
navigate in off-road environment and perform experiments
for various scenarios under different conditions.
Fig. 1. iCab 1 and iCab 2 autonomous platforms
Each vehicle is modified electronically and mechanically
for automated motion. Additionally it is equipped with
multiple sensors, such as GPS and compass modules, stereo-
camera, laser-rangefinder, 3D LiDAR and optical encoders.
All sensors are connected to two on-board embedded com-
puters, which are connected to a 4G router with a constant
internet connection [12].
The on-board embedded computers use Robot Operating
System (ROS) architecture to carry out numerous technolo-
gies. The ROS-based architecture is divided into three layers.
1. The reactive layer for the low-level commands, such as
communicating with sensors and actuators. 2. The sequencer
layer for the communication between the other two layers
and apply the logic of decomposing the complex task to
simple tasks. 3. The deliberative layer for the high level
commands, which is mainly responsible for environment
perception, obstacle detection and classification, intelligent
navigation, path planning and simultaneous localization and
mapping algorithms [12], [13].
III. COMMUNICATION SCHEMES
In vehicular communication systems; vehicles, pedestri-
ans, and infrastructures are the communicating nodes. They
provide each other with data, such as safety warnings, traffic
information and more [14]. The iCab platform architecture
allows various communication schemes, such as Vehicle-To-
Infrastructure (V2I), V2P, P2V and V2V communications.
Each of these schemes serve a role in enhancing the coop-
erative driving of multiple vehicles, the description of each
scheme can be summarized as follows:
•V2I: in this scheme, the infrastructure plays the role of
the observer, by gathering information from all vehicles
in the system for monitoring and inspection [15]. For
iCab project, the infrastructure node is a webserver with
a graphical user interface, which displays the system
information such as vehicle location, battery level, status
and tasks list, as shown in Figure 2.
Vehicle Icon Battery
Level Status Tasks
iCab 1 Busy
a2vulov0w1
zf0pfp8bpv
iCab 2 Busy
ovn2rujzh6
uiwdxqzx8i
l05c6mt5ls
ID:
l05c6mt5ls
From:
Sabatini North
To:
Library
Passengers:
2
Date & Time:
201701251315
Status:
Pending
Copyrights © Intelligent Systems Lab (LSI)
Fig. 2. Webserver graphical user interface for V2I communication
•V2P & P2V: both schemes involve sending and re-
ceiving messages between the Vulnerable Road Users
(VRU) and Intelligent Vehicles (IV). In the iCab project,
there is a proposed smartphone application that esti-
mates possible collision point between vehicles and
pedestrians, then send warning signals based on cal-
culated danger index. This increases the awareness of
the VRU and enhance the perception system of the IV.
The proposed approach is crucial to reduce number of
casualties due to distractions of the VRU or blind-spots
or malfunctions of the IV sensors [16].
•V2V: this scheme allows vehicles to talk to each other
and share pertinent information. In the literature, several
schemes are being investigated for inter-vehicle comm-
unication, based on the IEEE 802.11 standard, mainly
for short-range handshaking communication [17], [18].
However for a vast off-road environment, these ap-
proaches are not genuinely feasible. Accordingly, the
iCab project utilizes the concept of Internet of Things
(IoT), by creating a Virtual Private Network (VPN)
with unique certificates and keys for secure connections.
Each platform joins the VPN via 4G internet connection
with the proper authentication. By the means of the
designed ROS architecture, the platforms utilize ROS
multi-master system, proposed in [19], to send and
receive data directly to each other. Accordingly each
vehicle operates independently from other vehicles in
the system, meanwhile shares and collects data related
to other vehicles location, status, assigned tasks list,
environment perception and obstacles information.
Consequently, the iCab platforms utilize all above men-
tioned technologies and schemes to apply multiple au-
tonomous vehicles cooperation and coordination approaches.
These cooperation and coordination are modeled as Multi-
Robot Task Allocation (MRTA) problem, where numerous
optimization algorithms are proposed based on the work
in [20]. Thus achieving the overall objective of autonomous
cooperative driving.
IV. COO PER ATI ON APP ROACH
Cooperative driving aims at the coordination of multiple
vehicles, in order to increase the safety and efficiency of
road traffic [21]. One of the cooperative driving approaches
is flexible platooning of autonomous vehicles. Platooning is
a vehicle following technique, which have been widely used
in several IV applications, especially as a part of Advanced
Vehicle Control Systems (AVCS). There are other techniques
for vehicle following in the IV field; such as adaptive cruise
control systems and stop-and-go traffic systems [22]. In this
section, a model for the cooperative vehicles following is
described, and the algorithm with detailed steps that shows
how the vehicles benefits from V2X communications to
achieve an efficient cooperation between them.
A. Modeling
When people drive vehicles in a road, they take in con-
sideration the desired path to their destination point, the
surrounding vehicles position and speed, and other env-
ironment perception in taking decisions. Therefore for an
efficient cooperative driving among autonomous vehicles,
same concepts should be applied.
V2V via
VPN
LeaderFollower
(,,) &
Laser
rangefinder
Stereo-
camera
Fig. 3. Leader-Follower platooning modeling diagram
For instance, autonomous vehicle platooning is selected
to test the cooperation algorithm. A simple diagram for the
leader-follower platooning model is depicted in Figure 3.
Accordingly the kinematics modeling that govern the
motion of the autonomous vehicles can be outlined in such
a way. The desired distance ddr as a gap between the two
vehicles is evaluated as shown in Equation (1).
ddr =v2
l−v2
f
2amax
+vf.tbr (1)
where vland vfare the forward velocities of both the leader
and follower vehicles respectively, amax is the maximum
acceleration of the follower vehicle and tbr is the follower
vehicle braking time, which is estimated according to the
vehicle braking model.
The separating distance dsr is evaluated by the fol-
lower vehicle sensors. Both sensors, stereo-camera and laser
rangefinder, are used to detect and classify the ahead leader
vehicle [13]. Accordingly the spacing distance error esd is
evaluated as per Equation (2).
esd =dsr −ddr (2)
Along these lines, a classical proportional differential (PD)
controller is used to regulate the follower vehicle velocity,
as shown in Equation (3).
vf(t) = kp.esd(t) + kd.d
dt esd(t)(3)
On the other hand, the V2V communication ensures shar-
ing the vehicles pose, position (x, y) and orientation (θ), with
the each others. In addition to sharing the vehicles status, to
acknowledge which vehicle is elected to be the leader. Con-
sequently by means of Equation (4), the waypoint pose with
respect to the follower vehicle is estimated. Afterwards, the
vehicle applies path planning algorithm to get the necessary
navigation commands to reach that point.
xp
yp
θp
=
xl
yl
θl
−
sin(θ) 0 0
0cos(θ) 0
0 0 0
.ddr −
xf
yf
θf
(4)
Utilizing the vehicles kinematics model, on-board env-
ironment perception sensors and communication schemes,
the follower vehicle cooperates with the leader vehicle to
follow it based on the described algorithm in the next section.
B. Algorithm
The selection of platooning mode is carried out in the
reactive layer of the proposed ROS architecture. Accordingly
upon the system acquires a complex task, which requires
multiple vehicles to navigate the same path, it decomposes
the task into simpler tasks for the reactive layer.
The platooning algorithm is divided into several steps,
starting with the two vehicles navigate to the initial point of
the path from their current location. Upon arriving there, each
vehicle uses V2V communication to check the other vehicle
pose. Thus the leadership election is achieved, according to
which vehicle is ahead of the other in the direction of the
initial point. Afterwards both vehicles run Algorithm 1.
Algorithm 1: Cooperative platooning algorithm
Input:poseown,poseother ,statusplatoon,poseinital ,
posegoal,velocityown,velocityother
Output:posewaypoint ,pathleader,distanceseparation ,
distancedesired
1statusplatoon ←getAheadVehicle(poseow n,poseother)
2while poseown ! = posegoal do
3if statusplatoon == LEADER then
4pathleader ←pathPlanning(poseown,poseg oal)
5setNavigationCommands(pathleader)
6else
7distanceseparation ←getSeparationDistance()
8distancedesired ←getDesiredDistance(
velocityown ,velocityother )
9posewaypoint ←getFollowerWaypoint(
distanceseparation,distancedesir ed)
10 setNavigationCommands(posewaypoint)
11 end
12 end
This algorithm applies different functions in the system,
which are described below:
•getAheadV ehicle() function takes two poses as inputs
and returns which one is ahead of the other.
•pathP lanning() function utilizes the optimization ap-
proach in [12].
•getSeparationDistance() function reads the percep-
tion sensors in allocating the leader pose utilizing [13].
•getDesiredDistance() function takes vland vfas
inputs and returns the desired distance based on (1).
•getF ollow erW aypoint() function takes dsr and ddr as
inputs and returns the waypoint pose based on (4).
•setNavigationCommands() function takes a list of
waypoints and apply the navigation approach in [12].
During the vehicles navigation, the perception system is
active for obstacles detection, classification and avoidance.
Moreover, the leader vehicle has the V2P and P2V commu-
nications active, in case of VRU presence, thus it receives a
warning notification of possible collision point in advance.
Due to the cooperative driving architecture and V2V comm-
unication, the leader vehicle shares the information of the
presence of an obstacle ahead with the follower vehicle.
Accordingly the necessary action of braking or maneuvering
is applied in both vehicles.
The cooperative driving architecture is implemented in
ROS. Therefore all shared topics and messages are inscribed
with time-stamp, which allows synchronization between the
vehicles systems, sensors and actuators.
V. EX PER IME NTAL WORK
To validate the efficiency and credibility of the proposed
work, a real-world scenario is selected to test the platooning
approach along with studying the effects of different comm-
unication schemes.
A. Scenario Description
The selected real-world scenario covers all testing pa-
rameters. The process is as follows, the system receives
transport request, from one point to another. This request
is for 4 passengers, since each vehicle has a maximum
capacity of 2 passengers only, the two vehicles have to carry
out this task. Accordingly, in order to achieve an optimal
cooperative driving, one vehicle should be responsible for the
main environment perception, path planning and autonomous
navigation from the starting point to the goal point, while
the other vehicle should communicate with the first vehicle
and follow it to the destination point. In other words, both
vehicles navigate to the starting point, then apply apply
platooning algorithm of a leader/follower mode till the
destination point. Figure 4 gives an overview example of
the selected scenario.
Meanwhile the vehicles are navigating to the destination
point, a distracted VRU is trying to cross their designated
path of the vehicles. This plot is directed to simulate a
real life scenario in a smart city, where the distracted user
is crossing the road without paying attention to the road,
and the autonomous vehicles are following each other to a
𝒅𝒔𝒓
VRU
Follower Leader
V2P
V2V
P2V
Fig. 4. Two autonomous vehicles platooning scenario with VRU crossing
their designated path
collision point with the user without any mean of detecting
the user with the on-board perception sensors.
The scenario studies the efficiency of V2V communication
and the platooning algorithm in cooperative autonomous
driving. Moreover, it reviews the advantages of V2P and
P2V communications in detecting the pedestrian in advance
and sending a warning message to the pedestrian and a
notification to the vehicle.
B. Results and Discussions
Real life experiments were performed inside the campus
vicinity as an off-road environment. Both iCab platforms
were used in these experiments as the autonomous vehicles to
test the cooperative driving approach. And several volunteers
participated in the experiments as the VRU. The scenario was
tested several times to obtain enough results for analysis. All
recorded data took place in the ROS architecture with time-
stamp.
-10 0 10 20 30 40 50
0
10
20
30
40
50
60
Global X-axis (meters)
Global Y-axis (meters)
iCab 1
iCab 2
VRU
t=35s
t=42s
t=42s
t=35s
t=0s
t=0s
Fig. 5. Trajectories of iCab 1 (Blue), iCab 2 (Red), and VRU (Green)
Figure 5 shows the navigated trajectories of the leader
vehicle (iCab 1) and follower vehicle (iCab 2) on XY plane,
along with the trajectory of the pedestrian (VRU). The
leader vehicle was set to drive at speed of 5 km/h and the
follower vehicle adjusted its speed and the spacing distance
accordingly. The figure also shows that the follower vehicle
is moving along the leader vehicle trail, while maintaining
the desired distance with minimal error.
Further details about the spacing distance error values
between the follower and the leader vehicles, in both X
and Y directions relative to the vehicle, are presented in
Figure 6 (a) and (b) respectively. It is seen from the figure
that the tracking errors in the lateral direction (vehicle
X-axis) osculates around the reference and settle to 0.65
meters. However, for the tracking errors in the displacement
longitudinal direction (vehicle Y-axis), it decays to almost
zero meters with a maximum peek of 1.3 meters error during
the turning part of the scenario.
0 20 40 60 8
0
-1.5
-1
-0.5
0
0.5
1
1.5
Vehicle X-axis Error (meters)
Reference
Error
0 20 60 8
0
-1
0
1
2
3
4
40
Vehicle Y-axis Error (meters)
Reference
Error
Time (seconds)
(a) Vehicle X-axis Error
Time (seconds)
(b) Vehicle Y-axis Error
Fig. 6. Spacing distance error between follower and leader vehicles
Figure 7 displays the results of the leader vehicle obstacle
detection algorithms against time, in addition to the moment
of which the warning notification message was sent through
the P2V communication. Since the pedestrian is crossing on
an orthogonal road, there is no possible way for the sensors
of detecting him. It is considered as a blind spot due to an
obstruction. Hence, the advantage of the developed system
is the anticipation of VRU crossing the vehicles path approx.
7 seconds in advance. The leader vehicle then uses V2P
communication to warn the VRU, and V2V communication
to inform the follower vehicle of the presence of the VRU.
40 50 60 70 80
0
5
10
15
20
Time (seconds)
Distance to Pedestrian (meters)
Distance to Pedestrian
Detection
via P2V
Detection via
Vehicle Sensors
Fig. 7. Obstacle detection distance (Blue) and collision index (Red)
The qualitative results prove the viability of the proposed
approach for cooperative autonomous driving, in addition to
the advantages of the V2X communications in enhancing the
environment perception system.
Tables I and II present the quantitative analysis for the
overall system and compare the results with the use of V2X
communication and without them. The results are obtained
from performing the experiments on the same scenario and
under the same condition, once with the communication
schemes active and the other with the communication
schemes inactive, then calculate the average value of the error
over the whole path.
In Table I, VRU detection statistc are calculated based
on the Euclidean distance between the vehocle and the
pedestrian at the moment of detection. The leader vehicle
was able to detect the VRU in both cases, however with the
communication inactive, the detection was late which put the
decision of braking be more aggressive. On the other hand,
for the follower vehicle, it was not able to know that there
is a pedestrian in the route with the communication inactive.
TABLE I
VRU DETECTION STATISTICS [MET ERS ]
With
V2X
With
-out
Leader
Vehicle 16.13 8.79
Follower
Vehicle 21.32 N/A
In Table II, the leader vehicle errors are estimated based
on the deviation from the route obtained by the path planning
algorithm in the sequencer layer, therefore the values are the
same in both cases. However, for the follower vehicle errors,
the values are estimated based on the deviation from the trail
of path made by the leader vehicle. It is shown that the error
almost is doubled in the case of no communication.
TABLE II
TRAC KING E RROR S TATIST IC S [ME TE RS ]
With
V2X
With
-out
Leader
Vehicle 0.89 0.89
Follower
Vehicle 0.43 0.93
VI. CONCLUSION AND FUTURE WO RK
In this paper, an approach for autonomous cooperative
driving is proposed, which utilizes several V2X comm-
unication schemes to enhance the system performance and
environment perception. The main contribution of the work
is the proof of integrity for using various communication
techniques over autonomous vehicles in real-life scenarios.
Two autonomous vehicles were selected as the experi-
ments platforms. They are driverless golf-carts, from the
iCab project, which equipped with multiple sensors and
actuators that enable them execute autonomous driving and
full environment perception. Moreover the vehicles include
three different categories for communication. First category
is with infrastructure through V2I, to supervise the system
through a webserver. Second category is with pedestrians
through V2P and P2V, to communicate with VRU and
estimate possible collision point. Third category is with other
vehicles through V2V, to send and receive information for
the cooperation and coordination among the vehicles.
The proposed cooperation approach modeled the mul-
tiple vehicles in a platooning mode. The algorithm exe-
cutes the vehicle following task in the reactive layer of
the designed ROS-based architecture, after decomposing it
from a complex task of a transport request. Accordingly, in
order to validate the proposed work, a real-world scenario
was designed. Several experiments were performed with the
autonomous platforms and few distracted pedestrians, with
the communication schemes active and without them, for
results comparison.
These results indicate that the experiments with the comm-
unication schemes have a better overall system performance.
Moreover they prove the high performance of the proposed
cooperative driving approach in real-world application and
emphasize on the importance and the effects of V2X com-
munications in autonomous vehicles.
For future work, the model can be extended to include
more than two vehicles and study the vehicle following ap-
proach in a formation mode. Moreover, protocols for joining
or leaving the platoon should be studied. Finally, performing
several experiments on different real-world scenarios and
under different environment conditions.
ACKNOWLEDGMENT
This research is supported by Madrid Community project
SEGVAUTO-TRIES (S2013-MIT-2713) and by the Span-
ish Government CICYT projects (TRA2013-48314-C3-1-R,
TRA2015-63708-R and TRA2016-78886-C3-1-R).
REFERENCES
[1] USDOT, “Smart city challenge,” U.S. Department of Transportation,
2015.
[2] D. L. Bock, D. Kettles, and J. Harrison, “Automated, autonomous
and connected vehicle technology assessment,” Electric Vehicle Trans-
portation Center (EVTC), 2016.
[3] J. Jiong, J. Gubbi, S. Marusic, and M. Palaniswami, “An information
framework for creating a smart city through internet of things,” IEEE
Internet of Things Journal, vol. 1, no. 2, pp. 112–121, 2014.
[4] S. L. Poczter and L. M. Jankovic, “The google car: driving toward a
better future,” Journal of Business Case Studies, vol. 10, no. 1, pp.
1–7, 2014.
[5] DailyNews, “Bmw, audi push self-driving cars closer to reality,” Daily
News - Autos, 2013.
[6] A. Bragman, “Mercedes-benz tech brings cars closer to self driving,”
USA Today, 2016.
[7] L. Laursen, “Volvo to test self-driving cars in traffic,” IEEE Spectrum
- Tech Talk, 2013.
[8] A. Y. S. Lam, Y.-W. Leung, and X. Chu, “Autonomous vehicle public
transportation system,” IEEE International Conference Connected
Vehicles and Expo (ICCVE), p. 571576, 2014.
[9] P. Khaligh and U. Weidmann, “A conceptual framework for the inter-
actions of autonomous public transport systems and urban planning
guideline,” Swiss Transport Research Conference, pp. 1–18, 2016.
[10] H. Rewald and O. Stursberg, “Cooperation of autonomous vehicles
using a hierarchy of auction-based and model-predictive control,”
IEEE Intelligent Vehicles Symposium (IV), pp. 1078–1084, 2016.
[11] E. Semsar-Kazerooni, J. Verhaegh, J. Ploeg, and M. Alirezaei, “Coop-
erative adaptive cruise control: An artificial potential field approach,”
IEEE Intelligent Vehicles Symposium (IV), pp. 361–367, 2016.
[12] A. Hussein, P. Marin-Plaza, D. Martin, A. de la Escalera, and J. M.
Armingol, “Autonomous off-road navigation using stereo-vision and
laser-rangefinder fusion for outdoor obstacles detection,” IEEE Intel-
ligent Vehicles Symposium (IV), pp. 104–109, 2016.
[13] P. Marin-Plaza, J. Beltran, A. Hussein, B. Musleh, D. Martin, A. de la
Escalera, and J. M. Armingol, “Stereo vision-based local occupancy
grid map for autonomous navigation in ros,” Joint Conference on
Computer Vision, Imaging and Computer Graphics Theory and Ap-
plications (VISIGRAPP), vol. 3, pp. 703–708, 2016.
[14] P. Papadimitratos, A. D. L. Fortelle, K. Evenssen, R. Brignolo, and
S. Cosenza, “Vehicular communication systems: Enabling technolo-
gies, applications, and future outlook on intelligent transportation,”
IEEE Communications Magazine, vol. 47, no. 11, pp. 84–95, 2009.
[15] E. Ndashimye, N. I. Sarkar, and S. K. Ray, “A novel network
selection mechanism for vehicle-to-infrastructure communication,”
IEEE International Conference on Dependable, Autonomic and Secure
Computing (DASC), pp. 483–488, 2016.
[16] A. Hussein, F. Garcia, J. M. Armingol, and C. Olaverri-Monreal,
“P2v and v2p communication for pedestrian warning on the basis of
autonomous vehicles,” IEEE International Conference on Intelligent
Transportation Systems (ITSC), pp. 2034–2039, 2016.
[17] S. Biswas, R. Tatchikou, and F. Dion, “Vehicle-to-vehicle wireless
communication protocols for enhancing highway traffic safety,” IEEE
Communications Magazine, vol. 44, no. 1, pp. 74–82, 2006.
[18] L. Tellis, F. Ahmed-Zaid, J. E. Stinnett, C. Nave, T. E. Pilutti, T. D.
Zwicky, J. A. Martell, and J. C. Ivan, “Vehicle-to-vehicle/vehicle-to-
infrastructure control,” IEEE The Impact of Control Technology, 2011.
[19] S. H. Juan and F. H. Cotarelo, “Multi-master ros systems,” Institut de
Robotics and Industrial Informatics, 2015.
[20] A. Khamis, A. Hussein, and A. Elmogy, Cooperative Robots and
Sensor Networks. Springer International Publishing, 2015, ch. Multi-
robot Task Allocation: A Review of the State-of-the-Art, pp. 31–51.
[21] J. Luo and J.-P. Hubaux, “A survey of inter-vehicle communication,”
Infoscience - LCA-REPORT-2004-013, pp. 1–12, 2004.
[22] S. K. Gehrig and F. J. Stein, “Collision avoidance for vehicle-following
systems,” EEE Transactions on Intelligent Transportation Systems,
vol. 8, no. 2, pp. 233–244, 2007.