Decision Making for Advanced Driver Assistance

Systems for Public Transport

Yazan Wassouf Egor М. Korekov Vladimir V. Serebrenny
2023 5th International Youth Conference on Radio Electronics, Electrical and Power Engineering (REEPE) | 979-8-3503-9952-3/23/$31.00 ©2023 IEEE | DOI: 10.1109/REEPE57272.2023.10086753

Bauman Moscow State Technical Bauman Moscow State Technical Bauman Moscow State Technical
University University University
Moscow, Russian Federation Moscow, Russian Federation Moscow, Russian Federation
[email protected] KorekovEM @mail.ru [email protected]

Abstract —This paper discusses the decision-making process warning. These systems are using cameras to improve the
for Advanced Driver Assistance Systems (ADAS) on level 0 for safety on the road. But it does not discuss the sensitivity of
public transport in Russia. ADAS systems improve the safety on cameras sensors for weather changes and noise which
the road by warning the driver about any potential collision. The influence on the system performance in general.
architecture of the proposed system is discussed. Two new
systems are proposed, which are Turn Assist System (TAS) and Many researches also discussed the problem of passenger
Lateral Clearance Warning (LCW). Three other systems were stability for ADAS systems. In [7] the limits for accepting
developed according to the used vehicle (KAMAZ bus), Russian acceleration and deceleration are determined. The researches
road laws, and weather conditions. These systems are Forward in [8], [9], [10] discuss the handling-comfort contradiction of
Collision Warning (FCW), Pedestrian Collision Warning vehicles. The conducted results can be obviously used for
(PCW), and Blind Spot Monitoring (BSM). For each system, the braking control and improving ride comfort [11].
process of decision-making is discussed. The presented systems
were tested both in MATLAB/Simulink and on the test site Much other research discussed the general architecture of
(landfill). The conducted results correspond to system advanced driver assistance systems. Considering the pros and
requirements and help to improve the safety on the road. cons of sensors, and proposing control algorithms, discussing
the most popular research in this field [12], [13].
Keywords—Advanced Driver Assistance System (ADAS),
Public transport, RADAR systems, Decision making, road safety. In [14], a review was conducted for autonomous shuttles
considering fully automated systems and researches in the
I. INTRODUCTION world. But this work does not consider the semi-automated
systems which are used in the level 1 and 2 for ADAS systems.
Advanced Driver Assistance Systems (ADAS) can be
divided into six levels by the SAE (Society of Automation Public transport manufacturer and operators confirmed
Engineers) from the level 0 to level 5. On level 0 there is no that ADAS technology is an important factor for improving
control of the vehicle, just giving a warning of a potential public transports and road safety [15].
collision. Therefore, on level 1 and 2 there is a partial control
of the vehicle where automation includes controlling speed, However, it can be noticed from the literature review that
brakes, and steering. At level 5 it is expected from the vehicle safe and careful consideration of many factors in these
to be fully automated [1]. systems is required. So in this work, a decision-making system
for ADAS is discussed considering weather, sensors, road
Officially, level 3 of ADAS was reached by many architecture, laws, control, vehicle design, and speed limits.
companies, like Waymo, Honda, and Mercedes. And research
is still going to reach the required goals of ADAS systems. B. Contribution
And it is expected that by 2030, the application of ADAS This research can be considered one of the first researches
systems could prevent more than 25000 accidents [2]. about advanced driver assistance systems for public transports
in Russia. Proposing two new systems: Turn Assist System
This research is subjected to develop a driver assistance (TAS) and Lateral Clearance Warning (LCW). And
system for public transport according to the Russian roads [3]. developing three existed systems which are: Blind Spot
This system can be installed on buses and electric buses. Monitoring (BSM), Forward Collision Warning (FCW), and
Continuously analyzes the environment and informs the driver Pedestrian Collision Warning (PCW).
about the potential danger of a collision on Level 0, or take
control of the steering or braking on level 1 and level 2. The paper is organized as follows: system architecture is
Therefore, this system leads to improve the safety on the described in Section II, which gives an introduction for the
roads. decision making process on the ADAS 0, as presented in
Section III. Tests, experiments, and results can be found in
A. Related Works Section IV. Section V presents a conclusion for this research.
In [4], a collision warning system has been developed for
buses. II. SYSTEM ARCHITECTURE
In [5], a study at Transport for New South Wales tested A. System structure requirements
ADAS systems and its effect on driving behavior. The system structure should consist of a software and
In China, a study was conducted to determine the effect of hardware complex that includes technical vision equipment
driver assistance systems on braking behavioral and accident (radars and cameras) to provide information about the
rate [6]. surrounding road scene. Also a computer control unit for the
whole system is required.
Previous studies consist of many systems like blind zone
monitoring, forward collision warning, and lane departure

979-8-3503-9952-3/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: PES University Bengaluru. Downloaded on October 12,2023 at 11:19:03 UTC from IEEE Xplore. Restrictions apply.
 The system control unit should consist of the following 2) Computing unit
modules: computational and logical. A block with the following characteristics should be used
The computing module processes data from radars. as a computing block:
Conversely, the logic module receives data from the vehicle 1. multi-core architecture support;
bus, performs diagnostics, receives control signals, and sends
system operation results. 2. LPDDR4 type RAM with a size of at least 4 GB;

The results of the system operation are transmitted to the 3. GNU/Linux operating system support;
vehicle's visual and audio notification systems. Visual and 4. Supported interfaces: 2xCAN, RG MII, SPI, USB 2.0,
audio notification systems can be combined with the vehicle's UART;
virtual dashboard. The general structure can be cleared in
Fig. 1. 5. 2 GPU cores with LPDDR4 type RAM of at least 4 GB;
It should be mentioned that radars are used. The radar is 6. Memory capacity from 64 GB;
the first choice because of its robust against the weather 7. Support 24 V mains power.
conditions, which perform a real challenge for ADAS in
Russia regarding the weather. Also, a camera is used for The decision-making process is executed in the logical
pedestrian collision warning system for its good classification block which takes decision depending on data from computing
characteristics, which make it non-replaceable. unit and send command to make the right decision (warn or
control: speed, braking, steering).
III. DECISION MAKING ON ADAS 0
On the first level of ADAS there are five systems were
discussed which take a decision in the most critical situations,
There are many parameters which influence on decision-
making and the system working. Some of these parameters
can be mutual between all the systems, like transmission
(gearbox position) and speed of the vehicle. The speed of the
vehicle and gearbox determines the transitions between the
system states as cleared on the next figure:

Fig. 1. General architecture of the supposed system.

B. General purpose of the logic block

The general functionality of the logic block is conducting
preliminary diagnostics, obtaining data, and processing the
algorithms of the system taking into account the received data.
The logic block is designed for diagnostics of all ISPS
systems and communication of the system. While the general Fig. 2. The transitions between the system states.
CAN-network of the car responses to diagnostic requests via
the UDS protocol. Transitions can be described as follows:
The logic unit must receive processed radars and camera (1-2) - the engine starts working.
data from the computing unit, as well as data from the car's
(2-1) - the ignition is switched off or there is a system
CAN network.
malfunction.
After receiving the data, the logic block must calculate the
(3-1) - the ignition is switched off or the entire unit has a
required values. After processing the data, it is necessary to
malfunctioning.
send messages and display the necessary information on the
dashboard of the car. Also warn the driver if there is a potential (2-3) - the vehicle speed is above minimum speed limits
collision. and below maximum speed limits in km/h (for every system)
or the gearbox is not in the "Rear" position.
C. Technical characteristics of the logical and computional
blocks (3-2) - vehicle speed below minimum speed limits km/h or
1) Logic Block above maximum speed limits km/h, or gearbox in the "Rear"
As a logic block, it is necessary to use a real-time position.
microcontroller with at least six cores and a functional security Minimum speed limits for Blind Spot Monitoring (BSM)
level of at least ASIL-D. The frequency of the cores should be and Lateral Clearance Decreasing Warning (LCDW) are 10
from 200 MHz. The block must support interfaces: km/h. And limits for Forward Collision Warning (FCW) and
1. SPI. 2. RGMII. 3.CAN Pedestrian Collision Warning (PCW) are 30 km/h. And the

Authorized licensed use limited to: PES University Bengaluru. Downloaded on October 12,2023 at 11:19:03 UTC from IEEE Xplore. Restrictions apply.
maximum limit is 90 km/h. For Turn Assist System (TAS)
speed is between 0-30 km/h.
When the ignition is switched off, the system switches to
the «Off» state. In the event of a malfunction in operation, the
system switches from any state to the «Off» state, after
sending an error message.
In the "Waiting" state, the system monitors the vehicle's
speed and gearbox position. If the vehicle speed is within the
limits at which the system is capable of operating and the
gearbox is in the "Drive" position (all positions except
"Rear"), the system switches from the "Waiting" state to the
"Active" state.
The system enters the "Waiting" state from the «Off» state
if the ignition cycle is completed and the engine is working.
The system switches to the "Waiting" state from the
"Active" state if the conditions for being in the active state are
not met: if the vehicle speed is not within the limits at which Fig. 3. Dangerous situations in the lateral interval.
the system is capable of operating, or the "Rear" gearbox
position is selected. In general, a warning arises according to the next equation:
In the "Active" state, warnings are generated to the driver
 (( Speed _ Bus & Gear & X _ position &
as soon as the conditions for generating a warning occur. 1,
LCWwarning =  Y _ position) = True) (2)
Beside the speed of the vehicle and gearbox transmission, 
 0, else
there are many parameters which influence on the system
behavior. Every system, performance, and the decision
This warning happens when the next condition is satisfied:
making process can be illustrated as the next:
speed bus between 10-90 km/h, and gearbox is not in rear
A. Blind Spot Monitoring System position (value 1), and the lateral distance is less than the safe
The critical situations of this system are described in [16]. distance, and the vertical distance of the object is according to
For decision making related to this system in public transport the length of the vehicle.
it should consider the next parameters: C. Turn Assistance System
Turn Assistance System warns the driver if there is any
 (( Speed _ Bus & Gear & X _ position & obstacles or object around the bus while turning. The
1, (1)
BSM warning =  Y _ position & Obj _ velocity ) = True) considered critical situation is presented in Fig. 4. A warning
 0, else arises according to the next equation:


A warning arises if all conditions are satisfied as: speed  (( Speed _ Bus & Gear & X _ position &
bus between 10-90 km/h, and gear boxes is not in rear position 1,
TAS warning =  Y _ position & Theta _ turning ) = True) (3)
(value 1), and the object is located in the blind spot related to

rectangle zone on the left and the right side of the bus, and also  0, else
the object should not be static so its velocity is not 0.
A warning arises if all conditions are satisfied as: speed
B. Lateral Clearance Decreasing Warning System bus between 0-30 km/h, and gearbox is not in rear position
Lateral Clearance Decreasing Warning System monitors (value 1), and the object is in the warning region, according to
the lateral distance to the object and warn the driver if this longitudinal and lateral borders of this region, and the last
distance is less than a safe value. The dangerous situations can parameter is the turning angle which should be more than 30
be presented in the next figure: degrees (can be read from the CAN bus according to the
driving roll turning).

Authorized licensed use limited to: PES University Bengaluru. Downloaded on October 12,2023 at 11:19:03 UTC from IEEE Xplore. Restrictions apply.
 Explanations are presented in Table I:

TABLE I. SYMBOLS TO FIG. 5.

Distance Value
dmax Maximum detection distance
Minimum detection distance of a vehicle moving in
d2
front with a lateral displacement of less than 20%
d1 Minimum distance measured by the system
Minimum possible obstacle detection distance without
d0
distance measurement
WL Lane width

WV Width of the test vehicle

h1 Minimum detection height from the ground is 0.2 m

The maximum detection height from the ground surface
h
is 1.1 m
Fig. 4. Critical situation when the bus turns to the right or the left.
E. Pedestrian Collision Warning
D. Forward Collision Warning Next figure presents the process of pedestrian tracking
The purpose of the Forward Collision Warning (FCW) using a camera:
system is to generate warnings that will help drivers avoid or
reduce the severity of collisions with the vehicle moving
ahead. These warnings should be provided in time to help
drivers avoid the most common accidents by applying only
brakes. The time point should be chosen in such a way as to
provide early enough notification and help the driver avoid an
accident or reduce the severity of the consequences caused by
the accident.
Vehicles equipped with FCW must be equipped to
perform the following functions:
1) Determination of parameters regarding to the position Fig. 6. Pedestrian detection and tracking using a camera.
of the obstacles in front of the ego vehicle;
The centroid tracker combines tracker with the YOLO
2) Determination of the speed of a vehicle equipped with object detector using the OpenCV library [18]. The tracker
the proposed system; allows to track and distinguish pedestrians which every
3) Warning the driver, according to the required detection can be assigned to a unique ID. In each image, the
acceleration. A warning arises if the required acceleration is found pedestrians should be defined by a bounding box. And
more than 4.9 m/ . The required acceleration can be defined the position of the tracked objects can be predicted using
as the next equation [17]: Kalman filter.
Knowing the dimension of the bounding box and the
(Vr (t )) 2 (4) calibrated camera, and using Perspective n points (PnP)
Areq = ATV +
2( X c (t ) + X r (t )) algorithm, the distance to the object can be detected.
Using this distance and according to [19] a warning arises
where ATV is the acceleration of the bus, Vr (t ) is the as follows:
relative speed between the bus and the obstacle, X c (t ) is the
 (( In _ zone & required _ acceleration &
distance between the obstacle and the bus, and X r (t ) is the 1,
PCWwarning = &Vx ) = True) (5)
amount of reduction in the distance between vehicles during

the reaction time.  0, else
The obstacle can be a vehicle, motorcycle, or a static
where the object should be in the detection zone and the
object. This object should be located in the detection zone
required acceleration is more the 4.9 m/s2and the speed of the
which is presented in the next figure:
pedestrian is negative.
IV. TESTING AND EXPERIMENTS
ADAS systems for level 0 was tested both on
MATLAB/Simulink and on the test sites with real RADARS
and cameras. Modelling in Simulink was done according to
the next figure:

Fig. 5. Width and height of the detection zone.

Authorized licensed use limited to: PES University Bengaluru. Downloaded on October 12,2023 at 11:19:03 UTC from IEEE Xplore. Restrictions apply.
Fig. 7. The testing model in MATLAB/Simulink. C. TAS testing
On the test field, ARS 408 and SRR 208 radars were used. Turn assist system was tested in simulator because of the
ARS 408 [20] which is a long range radar used for FCW disability to estimate turn angel from CAN. In simulator a
system. SSR 208 [21] which is a short range radar used with decision about warning arises according the equation three
BSM, LCW, TAS. For PCW a camera from Milander with respond time no more than 1 second. The next figure
Company [22] also was used. clarifies the raised warning with the conditions:

A. BSM testing
System was tested by monitoring the object in the blind
spot and check if there is a warning happening or not. A
warning happens if the conditions satisfy to equation 1.
Simulation results can be presented in the next figure:

Fig. 11. Testing results for TAS in MATLAB/Simulink.

D. PCW testing
Fig. 8. Testing results for BSM. Pedestrian Collision Warning (PCW) was verified on
testing site. A warning a rises according to equation 5. It
B. LCW testing should be mentioned that pedestrian coordinates can be
In SIMULINK, a scenario was designed for a car moving estimated from camera data. The conducted results for
beside the bus. A warning arises according to equation 2. And warning satisfy system requirements of accuracy and response
as cleared on the next figure, warning arises according the time.
thresholding distance.

Fig. 9. Testing results for LCW in SIMULINK/MATLAB.

System was tested using SRR radar on the test field, bus Fig. 12. Testing results for PCW on the test site.
was moving toward and a warning arises if the lateral distance
to any object is less than 0.75 m. The next figure represents As mentioned above, the camera detects and recognize
the arising warning during tests when the lateral distance is objects and draw a bounding box and if a decision about
less the safe value (0.75 m). This value is determined by warning should happen a visual sign appears on the visual
considering Russian roads and by the results of tests. panel of the bus as cleared in Fig. 13.

Fig. 10. Testing Results for LCW estimated using SRR radar in the test field.
Fig. 13. The detection of the pedestrian using the camera and the proposed
warning on the visual panel.

Authorized licensed use limited to: PES University Bengaluru. Downloaded on October 12,2023 at 11:19:03 UTC from IEEE Xplore. Restrictions apply.
E. FCW testing [3] A. Merkuryeva, "Rules of the road of the Russian Federation,"
Moscow: Expo, 2017).
Forward Collision Warning system is based on long range [4] M. J., J. Spears, M. Lutin, Y. Wang, R. Ke, and S. M. Clancy, "Active
radar ARS, a decision about warning happens according to the Safety-Collision Warning Pilot in Washington State. Final Report for
required acceleration mentioned in equation 4, beside speed Transit IDEA Project 82," Transportation Research Board,
and gearbox condition and the object should be in the field of Washington, DC , 2017, p. 19.
vision for the system as was cleared on Fig. 5. The results of [5] James P. Thompson, Jamie RR Mackenzie, Jeffrey K. Dutschke,
the simulation show a good response as illustrated in the next Matthew RJ Baldock, Simon J. Raftery, and John Wall, "A trial of
retrofitted advisory collision avoidance technology in government fleet
figure: vehicles," Accident Analysis & Prevention, vol. 115, pp. 34-40, 2018.
[6] Lyu, Nengchao, Chao Deng, Lian Xie, Chaozhong Wu, and Zhihcheng
Duan, "A field operational test in China: Exploring the effect of an
advanced driver assistance system on driving performance and braking
behavior," Transportation Research Part F: Traffic Psychology and
Behaviour, vol. 65, pp. 730-747, 2019.
[7] Lawrence L. Hoberock, A survey of longitudinal acceleration comfort
studies in ground transportation vehicles. Council for Advanced
Transportation Studies, 1976.
[8] A. Muhammed, W. Yazan, and A. G. I, "Linear-control vs. ADRC for
automatic management of the handling-comfort contradiction of a
quarter-car system," International Journal of Heavy Vehicle Systems,
vol. 29, no. 2, p. 145-162, 2022.
[9] M. Alhelou, Y. Wassou, V. V. Serebrenny, A. I. Gavrilov, and E. S.
Lobusov, “Managing the handling-comfort trade-off of a quarter car
suspension system using active disturbance rejection control and
Vyshnegradsky equation,” Mekhatronika, Avtomatizatsiya,
Upravlenie, vol. 23, no. 7, pp. 367–375, 2022.
[10] M. Alhelou, Y. Wassouf, V. V. Serebrenny, A. I. Gavrilov, and
E. S. Lobusov, "The handling-comfort trade-off in a quarter-car
Fig. 14. Testing results for FCW in MATLAB/Simulink. system: automatic adaptive management via active disturbance
rejection control," Problemy Upravleniya, vol. 2, pp. 36-48, 2022.
F. Discussions [11] J. Lee and S. Choi, “Braking control for improving Ride Comfort,”
All systems except pedestrian collision warning are based MATEC Web of Conferences, vol. 166, p. 02002, 2018.
on radar, which make them robust against weather condition [12] M. Ardelt, P. Waldmann, F. Homm, and N. Kaempchen, "Strategic
changes. According to tests in the test field, radar can detect decision-making process in advanced driver assistance systems," IFAC
Proceedings, vol. 43, no. 7, pp. 566-571, 2010.
objects with accuracy ± 10 cm. This leads to satisfying system
[13] Madhusmita Mishra and Abhishek Kumar, "ADAS technology: A
requirements and a warning arises in response of the presence review on challenges, legal risk mitigation and solutions," Autonomous
of a collision. Pedestrian collision warning is a camera-based Driving and Advanced Driver-Assistance Systems (ADAS), pp. 401-
system. Many detectors were used to detect and recognize 408, 2021.
pedestrians and the conducted results on the test field gave a [14] Calin Iclodean, Nicolae Cordos, and Bogdan Ovidiu Varga,
satisfying result with system requirement and response time. "Autonomous shuttle bus for public transportation: A review."
The main fault for camera system is the sensitivity against Energies 13, no. 11 (2020): 2917.
weather changes and for radar systems is the false alarms. [15] Cafiso, Salvatore, Alessandro Di Graziano, and Giuseppina
Pappalardo. "Road safety issues for bus transport management,"
These problems will be discussed in future works. Accident Analysis & Prevention, vol. 60, pp. 324-333, 2013.
V. CONCLUSION [16] GOST, R. "58808 2020 Motor vehicles." Blind spot monitoring
systems. General technical requirements and test method, 2020.
In this work, general structure of advanced driver [17] A. M. Ivanov, S. R. Kristalniy, N. V. Popov, M. A. Toporkov, and
assistance system is proposed. Two new systems are M. I. Isakova, "New testing methods of automatic emergency braking
proposed: Turn Assist System and Lateral Clearance Warning, systems and the experience of their application," in IOP Conference
and another three are developed. Decision making for the Series: Materials Science and Engineering, vol. 386, no. 1, p. 012019,
2018.
proposed system is discussed, highlighting the system
[18] A. Narayanan, R. Darshan Kumar, R. RoselinKiruba, and T. Sree
requirements and components. The proposed systems are Sharmila, "Study and analysis of pedestrian detection in thermal
modeled in Simulink environment, and different scenarios are images using YOLO and SVM," in 2021 Sixth International
developed to conduct verification using the scenario driving Conference on Wireless Communications, Signal Processing and
designer. Furthermore, real experiments on the landfill are Networking (WiSPNET), pp. 431-434, IEEE, 2021.
done, and the obtained results show better accuracy and [19] D. I. Mazin, "A system for warning road users about the presence of
response time satisfy system requirements. pedestrians at or near a pedestrian crossing," 2018.
[20] ARS 408. [Online]. Available: ARS-408-21_EN_HS-1.pdf (conti-
REFERNCES engineering.com) [Accessed: 22.06.2022].
[1] P. T. Blythe and A. M. Curtis, "Advanced driver assistance systems: [21] SRR 208. [Online]: Available:
Gimmick or reality, " 11th World Congress on ITS, Nagoya, 2004. 6144bc0e28060074ad9e1ef3219684ba.pdf (compotrade.ru).
[Accessed: 22.06.2022].
[2] Luke Blades, Roy Douglas, Juliana Early, Chun Yi Lo, and Robert
Best, "Advanced Driver-Assistance Systems for City Bus [22] Video camera «MIL-CAM». [Online]: Available:
Applications," SAE Technical Papers, 2020. https://device.milandr.ru/products/detektory-transporta-ustroystva-
asydd/videokamera-mil-cam/ . [Accessed: 14.11.2022].

Authorized licensed use limited to: PES University Bengaluru. Downloaded on October 12,2023 at 11:19:03 UTC from IEEE Xplore. Restrictions apply.