Content uploaded by Alex Gunagwera
Author content
All content in this area was uploaded by Alex Gunagwera on May 14, 2022
Content may be subject to copyright.
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Abstract—The idea of autonomous vehicle platoons presents a
variety of social, economic, and safety benefits to the
transportation industry. However, implementing and deploying
autonomous vehicle platoons is still a challenge. In this paper, we
present a PID-based computationally cost-efficient controller to
aid in the longitudinal control of the inter-vehicle distance
between successive platoon members. The proposed approach is
facilitated by inter-vehicle communication. The algorithm was
implemented using the Robotics Operating System and Gazebo
simulation environment. In order to evaluate the performance
and applicability of the proposed approach, meticulous
simulations under numerous scenarios were performed using 3D
vehicle models so as to mimic the real-world. The algorithm
successfully maintains the longitudinal inter-vehicle distance
within the desired range, ensures that no collisions occur among
platoon members, and preserves the platoon formation.
Index Terms—Autonomous vehicles, autonomous vehicle
platoons, intelligent transport systems, longitudinal platoon
control, automated highway systems, PID.
I. INTRODUCTION
UTONOMOUS VEHICLE PLATOONS (AVPs) is a
trending multidisciplinary topic nowadays attracting
attention from researchers, practitioners, and governmental
bodies all over the world. AVPs are comprised of two or more
vehicles mechanically or electronically connected and
travelling closely together as a single unit with the same
lateral and, or longitudinal motion control. Autonomous
vehicle platooning promises a variety of social, economic and
safety benefits ranging from saving employee time [1],
optimal energy consumption [2], efficient road utilization [3]
ALEX GUNAGWERA, is with Department of Computer Engineering
Istanbul Sabahattin Zaim University, Istanbul, Turkey, (e-mail:
alex.gunagwera@izu.edu.tr).
https://orcid.org/0000-0002-0143-3743
AYDIN TARIK ZENGIN, is with Department of Computer Engineering
Istanbul Sabahattin Zaim University, Istanbul, Turkey, (e-mail:
tarik.zengin@izu.edu.tr).
https://orcid.org/0000-0002-0860-4509
Manuscript received Nov 1, 2021; accepted Jan 28, 2022.
DOI: 10.17694/bajece.1017623
to minimizing traffic accidents culminating from human error
[4].
Owing to the multidisciplinary nature of AVPs, a multitude
of work has been done pertaining to AVPs; their control,
analysis, deployment to mention but a few. Work from areas
such as control and analysis, the communication industry, and
energy department have contributed significant work to the
enhancement of AVP applications. For instance, studies on
Adaptive Cruise Control (ACC) [5], Cooperative Adaptive
Cruise Control (CACC) [6], [7], String Stability [8], [9]. The
cruise control model proposed by [5] was implemented in
MATLAB SIMULINK. They used the velocity and inter-
vehicle distance deviation as inputs to their controller.
Communication plays a vital role in the overall success of
autonomous vehicle platoons. In fact, AVPs incorporating
communication in their architecture register better results in
comparison to those that operate without communication
among platoon members [10]. The quality, ease of
communication and type of information shared significantly
affects the performance of an AVP. Furthermore, the
efficiency of the communication methodology employed, and
the amount of information communicated in the platoon also
have the ability to enhance the success of the entire platoon.
Steven E. Shladover et al. [6] provided the essential
definitions and distinctions among the different types of
CACC and the various communication types employed by
platoons. Robust Vehicle to Vehicle (V2V) communication,
such as VANET, DSRC [11], [12], Vehicle to Infrastructure
(V2I) or even both (V2X) communications facilitate the
functioning of AV platooning. Shen, Z. et al. [13] discussed
the effects of communication reliability and latency on the
performance of vehicle systems using 5G V2X hardware
prototypes and the 802.11 communication protocol. In our
study, we utilize the Robot Operating System (ROS) [14]
messages and topics for communication within the platoon.
Inter vehicle communication is wireless and is based on Wi-
Fi(IEEE 802.11) network.
String stability is another important feature of an
autonomous vehicle platoon. Cremer, D. [9] provided a string
stability criterion which only depends on the error in velocity
of each vehicle in comparison to the velocity of the Leading
Vehicle (LV). Cremer’s standards did not rely on the inter-
vehicle distance. Seiler et al. defined platoon stability as the
error between the desired and the actual inter-vehicle spacing
[15]. Swaroop et al. [16] presented string stability
requirements that depend on the inter vehicle distance in two
Longitudinal Inter-vehicle Distance Control of
Autonomous Vehicle Platoons Subjected to
Internal and External Disturbances
Alex Gunagwera and Aydin Tarik Zengin*
A
75
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
categories: the strong sense and the weak sense. In the strong
sense, the presented string stability conditions requires that the
maximum inter vehicle distance error of the ith vehicle should
either be equal or less to that of the i - 1th vehicle. String
stability in a weak sense has a requirement that just the
maximum inter-vehicle distance errors should be less than or
equal to those of the first follower (F) vehicle.
The Global Positioning System (GPS) [17], is one of the
most valuable sensors in look-ahead systems. For example,
[18] illustrated a high integrity navigation system’s
development and implementation for usage in autonomous
land vehicle applications. They mainly used GPS and the
Inertial Measurement Unit (IMU) in their work. GPS is the
backbone of the approach we present in this study as well.
Inter-vehicle distance (IVD) is an essential metric in
autonomous vehicle platoons. How well it is kept and
maintained ensures safe, comfortable, and more efficient road
usage, among others. It is with this background that we
propose this study.
In this paper, we propose a computationally cost efficient,
PID-based algorithm for controlling the inter vehicle distance
between successive members of an autonomous vehicle
platoon. Our approach differs from other numerous studies
mainly by using only the onboard GPS sensors of the vehicles,
gazebo robot simulator and the Robot operating system
(ROS). Furthermore, our PID controller requires just the
current longitudinal inter vehicle distance to the preceding
vehicle, unlike most PID approaches that require the velocity
and acceleration information as well. The algorithm observes
maintenance of platoon formation and makes sure no
collisions occur amongst platoon members. Obtained results
are presented and 3D simulations of the system further carried
out using ROS and Gazebo platforms to demonstrate the
performance of the proposed approach.
This paper is organized as follows. Section II presents the
problem statement and followed platoon model. Section III
explains the scenarios considered during simulations. Section
IV presents the environment and conditions under which the
simulations were performed. Section V illustrates the results
obtained from the simulations, whereas Section VI discusses
the obtained results, the limitations of the proposed approach,
overall practicality, and applicability of the proposed
algorithm. Furthermore, how the proposed algorithm differs
from the current related works. Finally, in Section VII, we
conclude this work and present directions for the future work.
II. PROBLEM STATEMENT AND PLATOON MODEL
The controlled platoon comprises of four vehicles in total. The
platoon Leader Vehicle (LV) and three Follower (F) vehicles.
We design a PID controller to aid the control of the distance
between vehicles. It takes as input the current inter vehicle
distance between vehicles and returns as output a velocity
reference for the corresponding Fi vehicle in
order to achieve the desired inter-vehicle distance, (D) to the
preceding vehicle. We thus state the problem as:
(1)
Ei = D - di
(2)
and |Ei| ≤ Ethresh.
Ultimately, the major purpose of our PID controller is to
reduce the error, Ei, and drive it as close to 0m as possible. So,
the best-case scenario at any point in the simulation is to have
Ei = 0m, especially during the steady state.
Fi, i
{1, 2, 3}
where di is the ith inter-vehicle distance, D is the desired inter-
vehicle distance, and Ei is the error between the ith inter-
vehicle distance and the desired distance, D. Ethresh is the
maximum and minimum threshold value beyond which the
error should not exceed in order to guarantee safety. This
constraint ensures that F vehicles are not allowed to fall more
than Ethresh behind the preceding vehicle, i.e., Ei ≤ Ethresh. It
also ensures that F vehicles do not get more than Ethresh closer
to the preceding vehicle, i.e., Ei ≥ -Ethresh. Every F vehicle runs
its own instance of the PID control algorithm. The platoon
model presented in this work is based on the Predecessor-
Follower communication model [19]. In the simulations, the
desired inter-vehicle distance, D, is set to 12m and Ethresh is
considered as 5m.
Fig.1. Controller Model
Fig.2. demonstrates the platoon setup. LV, is the Leading
vehicle, also referred to as the root node of the platoon and is
generally indexed as the first member of the platoon. The F
labels depict the follower vehicles. The distance from one
vehicle’s center of mass to the preceding vehicle’s center of
mass is referred to as the inter-vehicle gap/distance in this
study.
In this study, the control algorithm aims at ascertaining a
constant inter-vehicle gap with all vehicles’ velocity
sufficiently approximately equal to that of the LV in the
platoon using only the distance measure from the data
provided by the onboard GPS sensors of the vehicles. Errors in
the inter-vehicle gap should be bounded, and there should be
always no collisions among platoon members in the worst-
case scenario during the simulation. Following the definition
of the platoon stability provided by [15], we can formulate the
steady-state error transfer function as
.
(3)
Followingly, platoon stability is guaranteed, locally, if
||H(s)||∞ ≤ 1, and h(t) > 0 where h(t) gives the impulse
response corresponding to H(s) as per the ζ2 norm [10]. ζ∞
extends this notion throughout the whole platoon to ensure
that overshoots do not occur as the signals propagate up the
string, hence global stability.
76
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Fig.2. Illustration of inter-vehicle distance, di, LV, and F
At the start of the simulation, GPS data measurements are
retrieved asynchronously if available from every vehicle’s
onboard GPS sensor. From this data, the relative inter-vehicle
gap is calculated and forwarded to the PID algorithm which in
turn returns the reference velocity with which the
corresponding F vehicle’s velocity is updated to achieve the
desired inter-vehicle gap between F and the preceding vehicle.
The desired inter-vehicle distance, D is set as the PID
algorithm’s setpoint whereas the inter-vehicle gap estimates,
calculated from the data measurements provided by the GPS
sensor, is provided as the feedback for the PID control
algorithm. We repeat these steps throughout the simulation
lifetime.
Fig.1. shows the general algorithm flow. Each
Fi vehicle runs this algorithm with vf being its effective
reference velocity.
III. CONSIDERED SCENARIOS DURING SIMULATIONS
In reality, systems are exposed to various conditions and
sensors do not perform as smoothly as intended. Due to
natural and mechanical phenomena, sensor readings are
generally, corrupted and affected by external noise.
Furthermore, there is exhibited by the vehicle onboard sensors
(sensor lag), lag within the vehicle control system and the
delay in V2V communication. These are all inevitabilities that
should be put into consideration if we wish to make the
simulations as realistic, and accurate as possible. In case, the
LV is either being manned by a human or is following a
human-manned vehicle in traffic, the human-factor may, to
some degree, not be negligible, i.e., the driver may suddenly
brake or accelerate for any uncertain period of time. We shall,
hereafter, refer to this phenomenon as the human factor(HF) -
which may or may not be present in a particular scenario.
In this section we present the performance of the proposed
controller with the above-mentioned criteria under
consideration. The simulated scenarios are divided into four
categories: In the first scenario, the platoon is only affected by
sensor lag. HF and other forms of delays discussed are not
present. In the second scenario, the platoon is subjected to
both the sensor lag and random occurrences of the HF, only.
In the third scenario, the platoon is subjected to the sensor lag,
random V2V communication, and vehicle control delays with
no HF occurrences. In the final scenario, the platoon is
subjected to the random HF, sensor lag, and random V2V
communication, and control delays. For all simulation
scenarios presented in this section, the LV action at any given
moment can be one of acceleration, deceleration or moving at
constant speed. The action is randomly generated with the
following limitations and constraints:
1) Reverse vehicle motion is not permitted within the
platoon. In case the LV velocity, during the deceleration phase
period, were to drop below 0m/s, the LV is programmed to
stop, i.e., LV velocity is set to 0m/s.
2) No action; acceleration, deceleration or constant speed
should be executed more than once consecutively. This way,
we ensure that the platoon performance under various
uncertain scenarios is observed.
3) The duration of each phase/action is randomly generated
and may last between 20 - 50 seconds with the exception of
the final phase - the phase during which the LV decelerates to
rest.
The first and final phases are a bit different. Since the platoon
starts motion from rest, the first phase/action has to always be
acceleration; the LV accelerates for a random period of time.
The final phase is always the deceleration of the LV to rest,
therefore its duration is not determined randomly. The vehicle
simply uniformly decelerates to 0m/s.
Table 1 provides a categorical summary of the scenarios
considered in this simulation. Follower vehicle controllers are
data from sensors. The sensors are subjected to reductions in
update frequency which leads to delays in the processed data.
This helps to portray unfavorable real-world conditions [21].
V2V communication frequency was set in such a way that
vehicles publish their information at an average of 33Hz,
which is the proposed maximum frequency by [20]. We vary
this frequency with a minimum being 20Hz and the maximum
being 50Hz to account for variations in platoon member
mechanical differences as well as miscellaneous occurrences
such as natural phenomena such as variations in wind speeds.
The time between messages varies with an average of 0.03s
with a minimum of 0.02s and a maximum of 0.05s.
TABLE I
SUMMARY OF THE SCENARIOS OF THE SCENARIOS CONSIDERED
DURING SIMULATIONS
Scenarios 1 and 2
Scenarios 3 and 4
sensor lag w/o HF
sensor lag w/HF
sensor lag, V2V communication delays,
w/o HF
sensor lag, V2V communication delays,
w/HF
IV. SIMULATION ENVIRONMENT
3D vehicles were designed and modeled using Gazebo
robotics simulator version 9 integrated with ROS 1 (Melodic).
Visuals of the 3D vehicle objects were designed using the
gazebo platform, whereas vehicle motion was handled and
controlled via topics and messages by the nodes implemented
using the ROS framework [14]. The vehicle model is based on
the Hyundai Genesis-2014 [20] which reproduces the major
dynamic characteristics such as friction, acceleration,
deceleration/braking, wheel radius, and weight making it
possible to critically analyze the movement of vehicles in the
platoon.
77
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Fig.3. Gazebo simulation environment
Table 2 presents some of the major vehicle model parameters
and their values. TABLE II
VEHICLE INFORMATION
Vehicle Attribute
Value
Vehicle Mass
Vehicle Length
Vehicle Width
Vehicle Height
Wheel Radius
Wheelbase
Wheel Width
Drag coefficient
1823.0kg
5m
1.89m
1.480m
0.34m
2.95m
0.225m
0.27 cd
Fig.3. shows the environment in which the simulations
were performed and monitored. The vehicles move forward
along the road during simulation to preserve the presented
platoon formation and the desired inter-vehicle distance. We
performed simulations under the assumptions and constraints
that:
1) The roads are straight and have got no slope so that
longitudinal control of the platoon remains the focus of the
platoon
2) Communication is wireless, that is, over a Wi-Fi (IEEE
802.11) and each vehicle is only allowed to communicate to
the preceding vehicle.
3) Overtake and reversing manoeuvres are not allowed in the
platoon.
V. RESULTS
In this section we present the results obtained from the
performed simulations. For each simulation scenario, two
study cases are presented. We manually tuned the PID
controller. The parameter gains of the controller that yielded
the presented results are listed in Table 3. We conclude from
the observed results that the proposed approach indeed does
guarantee the following effect while keeping the inter-vehicle
distance within the desired range. The algorithm also ensures
platoon formation preservation and no collisions amongst
platoon members since the error, Ei ≥ -Ethresh at all times
throughout the simulation. However, much as the vehicles in
the platoon obtain vehicle stability [23], overall platoon
stability is not guaranteed. The proposed controller also shows
best performance at steady state when the LV travels with
uniform velocity – as can be most explicitly observed in the
first scenario of the simulations. All error values presented
include the transient errors.
TABLE III
PID CONTROLLER PARAMETERS
Parameter
Value
P
I
D
0.05
0.00001
0.03
Long et al. [24] proposed a distributed model predictive
model for the longitudinal control of truck platoons. Their
model considers the state of the LV. In their study, platoon
members transition from cruise control (CC), adaptive cruise
control (ACC), and cooperative cruise control(CACC). ACC
and CACC constitute the most advanced platoon controllers
that exist today. They use MATLAB to evaluate the
performance of their model. We present a summarized
comparison of the performance of our approach and the ACC
phase of their study. Throughout the simulations, Long et al.
had the LV accelerate to ≈16m/s, then move with constant
velocity after that. They experienced a maximum transient
response error of ≈37m before the system finally settled and
had the range error converge to 0m. Our proposed controller
had a maximum transient error of 4.9m, after which it
converged to or close to 0m at steady state especially
whenever the LV was moving with constant speed.
A. Scenario 1
In this scenario, platoon members were only subjected to
sensor lag. This lag is assumed to occur in the GPS sensors in
this study. This scenario registered the best performance. In
this scenario, global platoon stability can seldomly be
observed at steady state especially when the LV travels with
constant speed.
In the first case study of scenario 1 (Fig.4.), there is the LV
velocity throughout the simulation illustrated in the first
subfigure. Platoon real-time IVD (di) is presented in the
second subfigure. The third subfigure presents the error (Ei) in
the IVD in comparison to the desired IVD (D). The fourth
subfigure presents platoon members’ real-time velocity
profiles during the simulation. In the first and second
subfigures, blue denotes the IVD (d1) and error (E1) in IVD
between the LV and F1, respectively. Similarly, red and green,
denote the IVD (d2, d3) and their corresponding errors (E2, E3),
respectively. In the fourth subfigure, v1, depicted by blue,
shows the LV velocity while v2(red), v3(green) and v4(cyan)
represent the velocities of the F vehicles (1, 2, and 3),
respectively.
During this case study, a minimum error of -2.8m and a
maximum error of 3.0m in IVD were obtained with maximum
standard deviation and variance of 1.94m and 3.7m2,
respectively. During this scenario, local platoon stability is not
guaranteed since, for instance, E3/E2 > 1.
Similarly, in the second case study of the first scenario (Fig.
5.), there is the LV velocity throughout the simulation. Platoon
real-time IVD (di) is presented in the second subfigure. The
78
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Fig.4. Scenario 1, first case study
third subfigure presents the error (Ei) in the IVD in
comparison to the desired IVD (D). The fourth subfigure
presents platoon members’ real-time velocity profiles during
the simulation. In the first and second subfigures, blue denotes
the IVD (d1) and error (E1) in IVD between the LV and F1,
respectively. Similarly, red and green, denote the IVD (d2, d3)
and their corresponding errors (E2, E3), respectively. In the
fourth subfigure, v1, depicted by blue, shows the LV velocity
while v2 (red), v3 (green) and v4 (cyan) represent the
velocities of the F vehicles (1-3), respectively. In this
particular case study, a minimum error of -3.5m and a
maximum error of 4.2m in IVD were obtained with maximum
standard deviation and variance of 1.9m and 4.9m2,
respectively. During this scenario, local platoon stability is not
guaranteed since, for instance, E2/E1 > 1.
B. Scenario 2
In this scenario, the platoon members were subjected to sensor
lag and the LV was subjected to random HF.
In the first subfigure of the Scenario 2’s first case study(Fig.
6.), we have the LV velocity profile during the simulation.
Platoon real-time IVD(di) in the second subfigure followed by
the error(Ei) in the IVD in comparison to the desired IVD (D) ,
Fig.5. Scenario 1, second case study
in the third subfigure. Finally, the fourth subfigure shows the
entire platoon’s real-time velocity profiles during the
simulation. Just like in, Fig.4. and Fig.5., in the first and
second subfigures, blue denotes the IVD(d1) and error (E1) in
IVD between the LV and F1, respectively. Similarly, red, and
green, depict the IVD (d2, d3) and their corresponding errors
(E2, E3), respectively. In the fourth subfigure, v1, depicted by
blue, shows the LV velocity while v2(red), v3(green) and
v4(cyan) represent the velocities of the F vehicles (1-3),
respectively. In this case study, a minimum error of -3.8m and
a maximum error of 3.1m in IVD were obtained with
maximum standard deviation and variance of 1.9m and 3.7m2,
respectively. During this scenario, local platoon stability is not
guaranteed since, for instance, E3/E2 > 1.
In the first subfigure of Scenario 2’s second study(Fig.7.),
the LV velocity throughout the simulation is presented
followed by platoon real-time IVD(di) in the second subfigure.
The third subfigure illustrates the error(Ei) in the IVD in
comparison to the desired IVD (D). The fourth subfigure, on
the other hand, real-time velocity profiles of the platoon
during the simulation. In the first and second subfigures, blue
denotes the IVD(d1) and error (E1) in IVD between the LV and
F1, respectively. Similarly, red and green, denote the IVD (d2,
d3) and their corresponding errors (E2, E3), respectively. In the
79
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Fig.6. Scenario 2, first case study
fourth subfigure, v1, depicted by blue, shows the LV velocity
while v2(red), v3(green) and v4(cyan) represent the velocities
of the F vehicles (1-3), respectively. A minimum error of -
2.9m and a maximum error of 3.4m in IVD were obtained with
maximum standard deviation and variance of 1.4m and 2.0m2,
respectively in this study case. During this scenario, local
platoon stability is not guaranteed since, for instance, E3/E2 >
1.
C. Scenario 3
In this scenario, platoon members were subjected to random
V2V communication delays, control delays within the vehicle,
and sensor lags without the HF effect.
In the first case study of Scenario 3 (Fig.8.), there is the LV
velocity throughout the simulation. Platoon real-time IVD(di)
is presented in the second subfigure. The third subfigure
presents the error(Ei) in the IVD in comparison to the desired
IVD (D). The fourth subfigure presents platoon members’
real-time velocity profiles during the simulation. In the first
and second subfigures, blue denotes the IVD(d1) and error (E1)
in IVD between the LV and F1, respectively. Similarly, red and
green, denote the IVD (d2, d3) and their corresponding errors
(E2, E3), respectively. In the fourth subfigure, v1, depicted by
Fig.7. Scenario 2, second case study
blue, shows the LV velocity while v2(red), v3(green) and
v4(cyan) represent the velocities of the F vehicles (1-3),
respectively.
During this case study, a minimum error of -4.4m and a
maximum error of 4.9m in IVD were obtained with
maximum standard deviation and variance of 1.8m and 3.4m2,
respectively. During this scenario, local platoon stability is not
guaranteed since, for instance, E2/E1 > 1 almost everywhere.
In the second case study of this scenario (Fig.9.), we have the
LV velocity throughout the simulation. Platoon real-time
IVD(di) is presented in the second subfigure. The third
subfigure presents the error (Ei) in the IVD in comparison to
the desired IVD (D). The fourth subfigure presents platoon
members’ real-time velocity profiles during the simulation. In
the first and second subfigures, blue denotes the IVD(d1) and
error (E1) in IVD between the LV and F1, respectively.
Similarly, red and green, denote the IVD (d2, d3) and their
corresponding errors (E2, E3), respectively. In the fourth
subfigure, v1, depicted by blue, shows the LV velocity while
v2(red), v3(green) and v4(cyan) represent the velocities of the
F vehicles (1-3), respectively. During the second case study of
scenario 3, presented in Fig.9., a minimum error of -4.1m and
a maximum error of 4.2m in IVD were obtained with
maximum standard deviation and variance of 1.9m and 3.7m2,
80
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Fig.8. Scenario 3, first case study
respectively. During this scenario, local platoon stability is not
guaranteed since, for instance, E3/E2 > 1.
D. Scenario 4
In this scenario, the platoon members were subjected to
random V2V communication and vehicle control delays,
sensor lag and random HF effect. In this scenario, platoon
stability is not guaranteed, as well, either globally or locally
since H(s) > 1 for all follower vehicles almost everywhere.
In the first subfigure of Fig.10. we present the LV velocity in
the simulation. Platoon real-time IVD(di) is presented in the
second subfigure. The third subfigure presents the error(Ei) in
the IVD in comparison to the desired IVD (D). The fourth
subfigure presents platoon members’ real-time velocity
profiles during the simulation. In the first and second
subfigures, blue denotes the IVD(d1) and error (E1) in IVD
between the LV and F1, respectively. Similarly, red, and green,
denote the IVD (d2, d3) and their corresponding errors (E2, E3),
respectively. In the fourth subfigure, v1, depicted by blue,
shows the LV velocity while v2(red), v3(green) and v4(cyan)
represent the velocities of the F vehicles (1-3), respectively.
During the case study presented in Fig.10. a minimum error
of -3.9m and a maximum error of 3.9m in IVD were obtained
Fig.9. Scenario 3, second case study
with maximum standard deviation and variance of 1.6m and
2.6m2, respectively. During this scenario, local platoon
stability is not guaranteed since, for instance, E3/E2 > 1.
Likewise, the first subfigure of Fig.11. demonstrates the LV
velocity throughout the simulation. Platoon real-time IVD(di)
is presented in the second subfigure. The third subfigure
presents the error(Ei) in the IVD in comparison to the desired
IVD (D). The fourth subfigure presents platoon members’
real-time velocity profiles during the simulation. In the first
and second subfigures, blue denotes the IVD(d1) and error (E1)
in IVD between the LV and F1, respectively. Similarly, red and
green, denote the IVD (d2, d3) and their corresponding errors
(E2, E3), respectively. In the fourth subfigure, v1, depicted by
blue, shows the LV velocity while v2(red), v3(green) and
v4(cyan) represent the velocities of the F vehicles (1-3),
respectively. In the second case study of the fourth scenario
(Fig.11.), a minimum error of -4.5m and a maximum error of
4.1m in IVD were obtained with maximum standard deviation
and variance of 2.0m and 3.9m2, respectively. During this
scenario, local platoon stability is not guaranteed since, for
instance, E3/E2 > 1.
81
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Fig.10. Scenario 4, first case study
VI. DISCUSSION
In this study, we provide a computationally less demanding
longitudinal inter-vehicle distance control algorithm for a
platoon of autonomous vehicles that only requires that
vehicles be equipped with GPS sensors and have a connection
to Wi-Fi. The algorithm computation cost is less than most
control algorithms described in the literature. For example,
physics-inspired control algorithms such as [22], most of the
other PID-based algorithms with different approaches like
taking as input to the PID the preceding vehicle’s velocity and
acceleration e.g., [5] and [23]. Furthermore, this study differs
from those that only perform numeric simulations such as [5]
by not only using generated GPS data, but also by applying
the algorithm to the 3D models of the platoon to mimic the
real world as closely as possible.
However, our approach does have limitations. The first
limitation of this algorithm stems from the fact that it is
mainly based on GPS sensors. GPS, in reality, is affected by
high-frequency faults culminating from multipath errors that
occur when signals bounce off surfaces before they can reach
the sensor receivers. The position fix, therefore, gets affected
as the signals are delayed. Another rarer cause of GPS faults
happens when one of the satellites used by the sensor receiver
Fig.11. Scenario 4, second case study
gets blocked and, as a result, has to be compensated by signals
received from a different satellite. The position fix estimated
by the GPS sensor is affected by the geometry of the satellites
from which the sensor gets signals. So, such changes in
configurations of the satellite observed by the sensor receiver
affect the position fix finally reported by the GPS receiver.
High-frequency faults and multipath make the accuracy of
GPS sensor, and ultimately, the efficiency of our algorithm
heavily environment-dependent, making it more accurate and
preferable in open space areas than in underground passages,
enclosed environments, or places with tall buildings such as
skyscrapers. The algorithm can be incorporated into indoor
environment or closed environments by replacing the GPS
technology with higher precision localization tools and, or
sensors such as beacon technology illustrated by [27] and [28].
Autonomous vehicle platooning applications in urban areas
involve high precision dependent maneuvers that require about
0.02m accuracy to guarantee safety, among other requirements
- such as lane-keeping/changing on busy streets, overtaking
operations, to mention but a few. In such applications, a 0.5m
error is pretty significant. Thus, expanding the applicability of
the proposed algorithm to all types of roads and environments
requires fusing data from other sensors such as LIDAR, the
Inertial Measurement Unit (IMU), and camera. Using sensor
fusion to enhance the applicability of the proposed algorithm
82
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
and make it more ideal for even more complex environments
is one of our future studies.
VII. CONCLUSION
In this study, a computationally-cost efficient algorithm for the
control of the inter-vehicle distance of autonomous vehicle
platoons is presented. The proposed approach takes as input to
the PID controller, the updated inter-vehicle distance between
a follower vehicle and the preceding vehicle. This distance is
calculated from the data measured and provided by the
vehicles’ onboard GPS sensors. The controller returns the
reference velocity with which the follower vehicle should
move to achieve the desired inter-vehicle distance. 3D
simulations using gazebo and ROS are additionally used to
verify and monitor the performance of the system. The
proposed approach guarantees the following effect of the
platoon ensuring maintenance of platoon formation and no
collisions among platoon members. Furthermore, after the
transient response to the Leader vehicle’s acceleration, the
standard deviation of the inter-vehicle distance error was kept
under 14.7% of the desired inter vehicle distance throughout
the entire simulation period for all the scenarios. The system
seldomly achieved a 0m error at the steady state when the
leader moves with constant speed. However, the proposed
method is mainly suitable for open environments since GPS
accuracy is susceptible to High-Frequency errors resulting
from multipath and collision of GPS signals with surfaces
before they reach the receiver. Applicability of the approach
can be extended to closed and underground environments if
GPS is replaced with high precision localization equipment
such as position beacons installed in the target environments.
We are currently working on expanding the operability and
applicability of the proposed approach on more road types,
environments (urban, rural, to mention but a few).
Incorporating more sensors and sensor fusion techniques to
improve the accuracy of not only the inter-vehicle distance but
also the velocity of the platoon members is another direction
for our future work.
ACKNOWLEDGMENT
This work was supported by Istanbul Sabahattin Zaim
University’s Scientific Research Projects Programme (BAP) ”,
under Project No: BAP-1001-61.
REFERENCES
[1] R. Kimura, N. Matsunaga, H. Okajima, and G. Kotaki, “Design of
virtual platoon control system using augmented reality to assist welfare
vehicle users,” International Conference on Control, Automation and
Systems, vol. 2017-Octob, no. Iccas, pp. 330–335, 2017..
[2] F. Luo, J. Larson, and T. Munson, “Coordinated platooning with
multiple speeds,” Transportation Research Part C: Emerging
Technologies, vol. 90, pp. 213–225, 2018.
[3] R. Janssen, H. Zwijnenberg, I. Blankers, and J. de Kruijff, “Truck
platooning,” Driving The Future of Transportation, TNO, 2015.
[4] S. Belcher, E. Merlis, J. McNew, and M. Wright, “Roadmap To Vehicle
Connectivity,” funded by Crown Castle, Tech. Rep. September, 2018.
[5] V. V. Sivaji and M. Sailaja, “Adaptive cruise control systems for vehicle
modeling using stop and go manoeuvres,” International Journal of
Engineering Research and Applications, vol. 3, no. 4, pp. 2453–2456,
2013. [Online]. Available:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.433.9908&r=r
ep1&type=pdf
[6] S. E. Shladover, C. Nowakowski, X. Y. Lu, and R. Ferlis, “Cooperative
adaptive cruise control: Definitions and operating concepts,”
Transportation Research Record, vol. 2489, no. November 2014, pp.
145–152, 2015.
[7] G. Naus, R. Vugts, J. Ploeg, R. Van De Molengraft, and M. Steinbuch,
“Co-operative adaptive cruise control, design and experiments,”
Proceedings of the 2010 American Control Conference, ACC 2010, no.
1, pp. 6145–6150, 2010.
[8] D. Swaroop, “String Stability of Interconnected Systems – Automatic
Control, IEEE Transactions on,” IEEE Transactions on Automatic
Control, vol. 41, no. 3, pp. 349 – 357, 1996.
[9] M. Cremer, “On Convoy-Stable Control Laws for Automatically Driven
Vehicle Clusters,” 1992.
[10] S. Öncü, N. Van de Wouw, W. M. H. Heemels, and H. Nijmeijer,
“String stability of interconnected vehicles under communication
constraints,” in 2012 IEEE 51st IEEE conference on decision and
control (cdc). IEEE, 2012, pp. 2459–2464.
[11] T. L. Willke, P. Tientrakool, and N. F. Maxemchuk, “A survey of inter-
vehicle communication protocols and their applications,” IEEE
Communications Surveys Tutorials, vol. 11, no. 2, pp. 3–20, 2009.
[12] M. Jain and R. Saxena, “Overview of VANET: Requirements and its
routing protocols,” in 2017 International Conference on Communication
and Signal Processing (ICCSP). IEEE, 2017, pp. 1957–1961.
[13] Z. Shen, X. Zhang, D. Yang, I. I. S. Ensor, and S. H. N. Etwork,
“Performance Analysis of Extended Sensor Sharing in Vehicular Ad
Hoc Networks,” 2018 International Symposium on Antennas and
Propagation (ISAP), pp. 1–2, 2018.
[14] M. Quigley, “ROS: an open-source Robot Operating System,” in ICRA
2009, 2009.
[15] P. Seiler, A. Pant, and K. Hedrick, “Disturbance propagation in vehicle
strings,” IEEE Transactions on automatic control, vol. 49, no. 10, pp.
1835–1842, 2004.
[16] D. Swaroop, J. K. Hedrick, C. Chien, and P. Ioannou, “A comparision of
spacing and headway control laws for automatically controlled
vehicles,” Vehicle system dynamics, vol. 23, no. 1, pp. 597–625, 1994.
[17] S. Edelkamp, S. Jabbar, and T. Willhalm, “Geometric travel planning,”
IEEE Conference on Intelligent Transportation Systems, Proceedings,
ITSC, vol. 2, no. 1, pp. 964–969, 2003.
[18] S. Sukkarieh, E. M. Nebot, and H. F. Durrant-Whyte, “A high integrity
IMU/GPS navigation loop for autonomous land vehicle applications,”
IEEE Transactions on Robotics and Automation, vol. 15, no. 3, pp. 572–
578, 1999.
[19] O. Karoui, M. Khalgui, A. Koubâa, E. Guerfala, Z. Li, and E. Tovar,
“Dual mode for vehicular platoon safety: Simulation and formal
verification,” Information Sciences, vol. 402, pp. 216–232, 2017.
[20] E. T. S. Institute, “ETSI TR 103 299 V2.1.1 Intelligent Transport
Systems (ITS); Cooperative Adaptive Cruise Control (CACC);,” Tech.
Rep., 2019. [Online]. Available:
https://cdn.standards.iteh.ai/samples/45854/7b3e6d0d3c014cb995408e4a
8a0a87d5/ETSI- TR-103-299-V2-1-1-2019-06-.pdf
[21] R. Rajamani and S. E. Shladover, “An experimental comparative study
of autonomous and co-operative vehicle-follower control systems,”
Transportation Research Part C: Emerging Technologies, vol. 9, no. 1,
pp.15–31, 2001.
[22] I. Edmunds. (2014) Used 2014 hyundai genesis specs & features. Used
2014 Hyundai Genesis Specs & Features. [Online]. Available:
https://www.edmunds.com/hyundai/genesis/2014/features-specs/
[23] R. Rajamani, Vehicle dynamics and control. Springer Science &
Business Media, 2011
[24] M. Long, G. Tian, and H. Cheng, “Longitudinal control for truck
platooning,” in 2020 4th CAA International Conference on Vehicular
Control and Intelligence (CVCI), 2020, pp. 418–423.
[25] S.-Y. Yi and K.-T. Chong, “Impedance control for a vehicle platoon
system,” Mechatronics, vol. 15, no. 5, pp. 627–638, 2005. [Online].
Available:https://www.sciencedirect.com/science/article/pii/S095741580
5000279
[26] Z. Ali Memon, S. Jumani, and J. Larik, “Longitudinal Control of a
Platoon of Road Vehicles Equipped with Adaptive Cruise Control
System,” vol. 31, no. 3, pp. 475–494, 2012. [Online]. Available:
http://oaji.net/articles/2016/2712-1454749706.pdf
[27] D. Surian, V. Kim, R. Menon, A. G. Dunn, V. Sintchenko, and E.
Coiera, “Tracking a moving user in indoor environments using
83
BALKAN JOURNAL OF ELECTRICAL & COMPUTER ENGINEERING, Vol. 10, No. 1, January 2022
Copyright © BAJECE ISSN: 2147-284X http://dergipark.gov.tr/bajece
Bluetooth low energy beacons,” Journal of Biomedical Informatics, vol.
98, p. 103288, 2019. [Online]. Available:
https://www.sciencedirect.com/science/article/pii/S1532046419302072
[28] R. Siegwart and I. R. Nourbakhsh, Introduction to Autonomous Mobile
Robots. USA: Bradford Company, 2004.
BIOGRAPHIES
GUNAGWERA ALEX received his M.S.
degree in computer science from Istanbul
S. Zaim University, Istanbul, Turkey in
2017 where he is currently pursuing the
Ph.D. degree in computer engineering.
ZENGIN A. TARIK received his electrical
and electronics engineering B.S. degree
from Ege University, Turkey, in 2007.
Later on, received M.E. and Ph.D.
degrees from Department of computer
science and electrical engineering,
Kumamoto University, Japan, in 2010
and 2013, respectively. Currently, he is an
Asst. Prof. at Istanbul S. Zaim University. His research
interests include Autonomous systems and control theory.
84
