Advanced Driver Assistance Systems

(ADAS)

Maria Merin Antony and Ruban Whenish

Introduction

A driver is one of the “best sensors in the vehicle” and is the main responsible for
avoiding crashes. But still, a large proportion of crashes are attributed to the driver
errors. A survey was conducted to identify the critical reason for each crash, and the
national sample of US crashes from 2005 to 2007 was examined [1]. It was noted
that the driver error was the critical reason contributing to 94 percentage of crashes
as shown in Fig. 1. These errors included recognition errors, decision errors, perfor-
mance errors, and nonperformance errors. Recognition errors are the result of inat-
tention and inadequate surveillance of the driver; decision errors arise due to the
misjudgments of the driver; performance errors arise due to overcompensation,
poor directional control, etc.; and nonperformance errors arise due to sleeping and
fatigue [2].
The development and deployment of the new in-vehicle technologies to counter-
act these driver errors and hence to support the driver to prevent crashes is ongoing.
Advanced driver assistance systems (ADAS) are a group of vehicle technologies
that warn the drivers timely regarding the risky or hazardous situations to avoid
crashes. Some ADAS technologies actively and automatically intervene to avoid
hazardous situations or when the system detects that a crash is imminent.
ADAS technologies are the precursor to autonomous vehicles and, depending on
the combination of ADAS equipment installed in a vehicle, can allow level 1 to level
2 autonomous driving at the present time as represented in Fig. 2 (SAE
International 2014).

M. M. Antony (*)
Centre for Optical and Laser Engineering, School of Mechanical and Aerospace Engineering,
Nanyang Technological University, Singapore, Singapore
R. Whenish
Sri Krishna College of Technology, Coimbatore, India

© Springer Nature Switzerland AG 2021 165

M. Kathiresh, R. Neelaveni (eds.), Automotive Embedded Systems,
EAI/Springer Innovations in Communication and Computing,
[Link]
166 M. M. Antony and R. Whenish

Estimation of critical reasons for pre-crash event

Driver Vehicles Environment Unknown Cri cal Reasons

Fig. 1 Estimation of the critical reasons for pre-crash event. Data source: NMVCCS 2005–2007

Fig. 2 Levels of autonomous driving as per the SAE classifications and ADAS role in the
automation

ADAS Technologies

Driver assistance systems (DAS) were developed aiming at the development of sys-
tems that are designed to assist and improve the comfort level of the driver while
driving. Initial systems developed as driver assistance systems include parking
assistance system (PAS) and cruise control (CC). With the advancements in DAS
technologies, the focus got shifted on to the increased level of safety of both the
driver and pedestrians. Such advanced driver assistance systems (ADAS) refer to
electronics, light-based, or sound-based systems that are integrated together to the
development of technologies that automate, facilitate, and improve systems in the
vehicles in order to assist drivers for better and safer driving. There are numerous
components now developed to be known as ADAS technologies. A complete over-
view of the features of ADAS technology is shown in Fig. 3.
Various ADAS technologies in a vehicle are realized with the help of many types
of environment sensors such as radars, lasers, ultrasonic sensors, and cameras.
These systems are equipped with control algorithms that signal the drivers or
directly control the brakes or throttles. The control systems of the vehicle with
ADAS will behave based on the signals from the environment and according to the
state of the vehicle such as steering angle, GPS data, etc. or state of the driver. The
ADAS technologies used in modern cars are described in short as below:
Advanced Driver Assistance Systems (ADAS) 167

Fig. 3 Overview of ADAS technologies used in a vehicle

Adaptive cruise control (ACC) also known as radar cruise control is a system
developed to control or adjust the speed of vehicle to maintain an optimum distance
between the vehicles with the help of a radar or a laser system. In ACC, three radar
sensors are needed, two short-range radars used to detect objects in the adjacent lane
and one with a long range used to detect objects in-path.
Forward collision control with autonomous emergency braking alerts the driver
when the speed of the vehicle can result in a collision with a vehicle or an object in
front. Autonomous braking (AEB) will prevent collision by applying brakes when
there is a high risk of collision and the response of the driver is inadequate to avoid
the collision.
Lane departure warning systems give warnings to the driver while there is a risk
or straying across the lane markings. The feature uses the long-range radar sensors
for the sensing purpose. A Lane estimation algorithm is used to gather the informa-
tion for lane departure warning (LDW) and lane keeping rely features for ADAS
technologies.
Side- and rear-view assistant systems monitor surrounding areas to warn the
driver for the blind zones with the help of cameras or radar sensors.
Traffic sign recognition displays are seen on the instrument panel reminding
drivers of the current speed limit. The traffic signals are recognized with the help of
the same camera which recognize the speed limits. Even the navigation system of
the vehicle can be supported by storing the information regarding the speed limits
in unsigned roads.
Blind spot detection (BSD) helps the driver to view the blind spots while overtak-
ing. A short-range radar is used for monitoring the road area behind and warning the
driver while overtaking the vehicles.
Rear cross traffic alert (RCTA) uses two radars that can monitor at 120 degrees
and help in detecting accidents while reversing out in the parking space.
Pedestrian detection technology monitors and detects the pedestrians in the road
and provides the driver the necessary safety guidelines with the help of normal and
IR cameras.
168 M. M. Antony and R. Whenish

Fig. 4 Steps taken in the

ADAS technology

Fig. 5 Classification of ADAS technologies

In each ADAS technology, there are three steps taken as shown in Fig. 4.
1. Recognition
2. Judgment
3. Operation
The first step in the ADAS technologies is recognition with the help of sensors
such as radar, LiDAR, cameras, and ultrasonic sensors. Based on the signals from
the sensors, the control action is decided with the help of advanced imaging pro-
cessing or machine learning algorithm.
Advanced Driver Assistance Systems (ADAS) 169

ADAS systems are basically classified into information-based assistance sys-

tems and manipulation-based assistance systems [3]. The detailed classification is
shown in Fig. 5.
Information-based assistance systems provide only necessary information and
warnings to the human driver. They don’t implement any driving decisions and
physically operate the vehicles by small amount.
Manipulation-based assistance systems only operate on the vehicle in case of
emergency or when asked by the driver.

Information-Based Assistance Systems

During driving, a lot of concerns are present in the mind of a driver regarding the
correct route to be taken, the level of traffic congestion, and time delay in reaching
the destination on the right time. The information-based assistance systems address
all these questions, and such a system do not control the vehicle by itself. The
information-­based assistance systems are further classified as Advanced Traveller
Information Systems (ATIS) as shown in Fig. 5.
The ATIS system performs main functions such as dynamic rerouting and antici-
pation of traffic congestion and provides information about the surroundings. These
information helps the driver to take key decisions regarding the route to be taken.
The data for dynamic and historic information about the traffic is made available
with the help of the cameras at different locations at various roads. Such informa-
tion is used to anticipate the congestion and reroute the traffic.
Drivers are distracted by many expected and unexpected reasons such as sleep,
tiredness, fatigue, etc. These reasons can be identified by various indicators such as
eye movements (blinking frequency of the eyes), head movements (yawning), bio-
logical and neurological signals (such as EEG and ECG), and facial expressions
(movement of the eyes, lips, etc.)
The indicators of inattention such as distraction are measurable. The effects of
distraction can be measured based on movements of the eye, head, and face and
signals obtained from the driver’s body such as electroencephalography (EEG) and
electrocardiogram (ECG). EEG signals carry information about the brain waves and
brain activities. In contrast, the ECG signals gives information on heart changes
which can be used to interpret and measure the change in the thought process of the
human driver.
The general procedure for detecting inattention uses the machine learning tech-
niques. A procedure which is typically followed for the pattern recognition system
is shown in Fig. 6.
As the initial step, the noise is removed from the data collected. Data is then
segmented, and the feature is extracted for training and to create the machine learn-
ing model. The real-world data is then segmented, and the feature is extracted and
classified based on the training data and model for the driver to take decision.
170 M. M. Antony and R. Whenish

Fig. 6 Procedure used for pattern recognition problem

While detecting the images, the images are reduced in dimensionality using fea-
tures like scale invariant feature transform (SIFT) and speeded up robust transform
(SURF). Another possible approach is to use color filters for the detection of the
face. There are numerous techniques that are used for face detection and recogni-
tion. Modelling of blinking of the eyes, gaze, and head pose was recognized in order
to identify the inattention and fatigue in the drivers during driving.
The next task is to extract the features from the complete image, or signal is
broken down using dimensionality reduction techniques and then can be used for
classification. The techniques such as principal component analysis (PCA), inde-
pendent component analysis (ICA), and linear discriminant analysis (LDA) are used
for dimension reduction and extraction. Later, the classification is done with the
help of popular classifiers such as neural networks, support vector machines, fuzzy
inference systems, adaptive boosting algorithms, and decision trees.
Hybrid techniques are also used which include the combination of any of the
above discussed techniques (Figs. 7, 8, and 9). The commonly used multiple modal-
ities are discussed as follows:
(a) Data level fusion: Here the fusion is performed on the raw data of the different
modalities. The fused data is then passed through feature extraction and classi-
fication in order to help the driver make an appropriate decision.
(b) Feature level fusion: It is fused at the feature level, and then classification
is done.
Advanced Driver Assistance Systems (ADAS) 171

(c) Score and decision level fusion: A score function is generated after classifica-
tion from each data source by an integrator based on which decision is made.
Here, the generated score function is fused to form the final decision.
Measurement of the driver performance is important for transportation authority
to access the driver performance to extend or cancel license to drive. The systems
for measuring driver performance are built inside the vehicle to continuously rate
the driver’s performance. The best way to indicate the driver’s performance is by
assessing the distances of the vehicle in front, fluctuations in speed, and the ability
to make smooth lane changes and to maintain the lanes. Other parameters that can
be used to measure the condition of driver are the way of handling the steering
wheel, reversal speeds of steering, and pressure distribution on the seat of the driver.
All these parameters are helpful in detecting fatigue and drowsiness of the driver
which are continuously monitored by the ADAS technologies in the car.

Fig. 7 Data level fusion

Fig. 8 Feature level fusion

Fig. 9 Score and decision level fusion

172 M. M. Antony and R. Whenish

Manipulation-Based Assistance Systems

Manipulation-based assistance systems not only detect the state or risk involved in
driving but also manipulate the vehicle on behalf of the driver. These systems come
into action in three situations as follows:
• When emergency actions are needed and there is no time to warn the driver
• To perform certain regular and simple tasks which do not require human wisdom
• To perform certain precise and specialized tasks that are hard for humans to per-
form such as overtaking and parking
There are situations where the alerting or warning of the driver is ineffective due
to time constraints and there is a need of quick response in some situations to take
safe action. Such situation needs to be addressed by emergency braking systems. In
contrast, safety alert systems predict the risky situations and provide warnings to
take necessary action. The collision avoidance system is one of the common exam-
ples to explain these systems.
Another similar problem is the collision with pedestrians, especially the old-age
people and children who do not take much care while crossing the roads. There are
pedestrian detection systems which detect and alert the driver to avoid collision
and risks.
The adaptive cruise control (ACC) maintains the speed of the vehicle constantly
by the use of brake and throttle. The driver is relieved/freed from continuously mon-
itoring the speed by the use of ACC, and he just needs to control the steering. The
sensors such as radar, LiDAR, or ultrasonic sensors will continuously monitor the
distance from the vehicle and environment. While vehicle following is successful
with no risk and accidents, the ACC follows an elastic band. This band is always
maintained to be smooth and collision-free and tends to be the path for trajectory for
the vehicle to be followed as shown in Fig. 10.
Overtaking is desirable and helps the fast-moving traffic not to be penalized by
the slow-moving vehicles in the front. Overtaking assessment and assist helps in the
overtaking task. It has two important tasks as follows:

Fig. 10 Vehicle following

Advanced Driver Assistance Systems (ADAS) 173

• Deciding the feasibility of the overtake attempt: The feasibility of the overtaking
is based on the distance between the overtaking vehicles and the one being over-
taken and the space in the overtaking lane
• Actual overtaking: This task requires moving the vehicle overtaking from the
overtaking lane to the normal lane after overtaking as shown in Fig. 11.
One of the common problems faced by all drivers is to park the vehicle in an empty
slot correctly. This problem can be solved by an automated parking system. The prob-
lem of parking resembles close to that of a robot motion planning. Some number of
times, the vehicles goes back and forth before placing the vehicle perfectly in the
empty slot of the parking area as shown in Fig. 12. The commonly used motion algo-
rithms include rapidly exploring random trees, graph search, and many more [4].

Sensors Used in ADAS Technology

Various sensors are used in implementing the ADAS technology. Some of the main
sensors are described in detail below:

Fig. 11 Overtaking

Fig. 12 Parking assistance with ADAS-based cars

174 M. M. Antony and R. Whenish

Camera vision: It acts as the eyes of the ADAS technology. This vision is affected
severely by weather conditions, low light, and glare. The image acquired is pre-­
processed using digital signal processors (DSPs). The camera sensors are typical
used for applications such as traffic sign recognition, parking assistance, and lane
departure warnings. Infrared cameras are used during nights which act as best night
vision cameras.
Radio detection and ranging (RADAR): Compared to cameras, radars are robust
in any weather and light conditions with good range. There are specific radars for
long-range and short-range applications.
Light detection and ranging (LiDAR): The LiDAR measures the distances based
on the light scattered w.r.t the laser irradiation that emits pulsed light. The main
advantage of the LiDAR technology is that the light scattered from the object gives
information distance as well as the nature of the object.
Ultrasound: These sensors work based on the principle of sound in determining
the distance of the nearby objects and to warn the driver regarding the obstacles.
Usually, a single car will be equipped with 12 ultrasonic sensors to completely
cover the surrounding environment. The range of ultrasonic sensors are very less
and is about 3 m.
The range and typical application of the sensors in the ADAS technology is tabu-
lated in the Table 1.

Reliability Modelling

A vehicle consists of active and passive safety systems and ADAS features. The
most important concept in ADAS is to maintain the safety by saving the lives. A
model is demonstrated here to represent various ways on how the safety of the

Table 1 Various sensors used with typical applications

Maximum range
Sensors (m) Typical applications
Short-range 20 Blind spot detection
radar
Long-range 150 ACC (adaptive cruise control)
radar
Camera vision 80 Pedestrian detection, sign recognition, parking assistance,
lane departure warning
Infrared 120 Night vision
Ultrasound 3 Park assist
LiDAR 120 Emergency braking, collision avoidance systems
(scanning)
Advanced Driver Assistance Systems (ADAS) 175

Fig. 13 RBD block diagram for a simple model

vehicle is maintained. The model represented is termed as reliability block diagram

(RBD) and is shown in Fig. 13. There are four different pathways to maintain the
reliability of the safety of the vehicle.
When the driver fails to operate the vehicle safely, the control systems alerts the
ADAS warning system to provide the necessary warning signals. When the failure
in ADAS warning systems result, it leads to the activation of ADAS crash avoidance
systems. Lastly, the failure of ADAS crash avoidance systems switches on the pas-
sive safety systems in the vehicles such as airbags to active state.

Vehicular Communication in ADAS

Vehicular communication system plays a vital role in ADAS by notifying useful

information through various channels like wireless communications and to imple-
ment the necessary actions especially regarding safe driving alerts. Safety is a key
factor of an automotive. Every year, the increase in number of road accidents can be
directly linked with increasing number of vehicles on the road. According to the
WHO, globally at least 1.2 million deaths are happening every year due to road
accidents. Traffic management is necessary to avoid possible accidents, to minimize
carbon emissions, to improve fuel efficiency, for navigating the fastest route, to
improve driver safety and comfort and to save resources. Several automobile manu-
facturers are doing research into the same to provide better future prospects for
automobile industry. Vehicles could communicate with vehicles and other elements
which substantially improvise the driving systems and safety prospects. Vehicle-2-­
vehicle (V2V), vehicle-2-infrastructure (V2I), vehicle-2-pedestrian (V2P),
vehicle-­2-cloud networks (V2N), and vehicle-2-everything (V2X) communication
systems allow the vehicle to communicate to its surroundings with short-range
wireless signals which is as depicted in Fig. 14.
Vehicular communication conveys to the driver about weather information, traf-
fic and signals, road conditions, accidents, and any harmful activities around
176 M. M. Antony and R. Whenish

Fig. 14 Vehicular communication system in ADAS

vehicle. In autonomous vehicles, having the navigation system adds extra features
when compared with traditional vehicles. One of the vehicular communication sys-
tems, V2X, has a smart working system which reports automatic payment for toll,
parking slot identification, weather forecast, shortening the distance towards the
destination, fuel station/restaurant notifications, etc. V2X communication has sev-
eral components which integrate information that connect vehicles to everything
(pedestrians, vehicles, roads, and cloud environments) aiming to reduce the number
of road accidents and building an intelligent transport system (ITS) [5].
Certain features are added with ADAS by considering safety concerns like
Emergency Electronic Brake Lights, Forward Collision Warning, Intersection
Movement Assist, Left/ Right Turn Assist, Blind Spot/Lane Change Warning, Curve
Speed Warning; by considering environment concerns like vehicle classification,
traffic management, intersection analysis, destination studies, emission control,
smart parking and by considering roadside mobility concerns like emergency com-
munications and evacuation, cooperative adaptive cruise control, wireless inspec-
tion, connection protection, etc. Adding all these features with ADAS, vehicular
communication will be a major advancement in the field of automotive by adopting
technologies like Internet of Things (IoT), information and communication tech-
nologies (ICT), and artificial intelligence (AI), leading to safe driving environ-
ment [6].
Advanced Driver Assistance Systems (ADAS) 177

Vehicular Communication Technologies Used in ADAS

Cellular systems, Bluetooth, microwave communication, millimeter wave commu-

nication, DSRC (dedicated short-range communications), ZigBee, WiMAX, and
long-term evolution (LTE) are key wireless communication technologies used in
vehicular ad hoc networks (VANETs) for effective vehicular communication.
Various vehicular communication technologies used in ADAS are listed in Table 2,
which could be used as wireless communication channels to connect the vehicle
with everything around it depending on the applications.

Pathway Towards Nextgen: ADAS in Global Automotive OEMs

The advancements in terms of safe driving and automation carried in the form of
ADAS had already arrived in OEMs globally. A leading Japanese car manufacturer
introduced driving safety support system (DSSS) to its products Prius, Crown
Majesta, Crown Royal, and Crown Athlete. DSSS have sensors and send messages
to the driver about detected pedestrians and vehicles that were not visible to the
driver’s sight and alert the driver. The main benefits of DSSS are reducing the traffic
accidents at intersections, increasing awareness, and reducing the driver’s burden.
DSSS features include pre-collision warning and braking, enhancing the road safety
during night drives. DSSS uses V2V, V2I, and V2X communication systems. DSSS
comes under the awareness driving phase which is the present trend in automotive
sectors. Another car manufacturer hailing from Japan, Honda, thrives to demon-
strate advanced V2P and V2V through DSRC communication technology. The ulti-
mate aim of this R&D work ensures the reduction in collisions between vehicles
and pedestrians. Next phase sensing and driving also make impact in the automotive
sector about hazardous things on the road. In 2017 Daimler introduced autonomous
driving with Mercedes E-class car using V2X communication. The sensing and
driving principles had been applied to identify weather conditions and road acci-
dents, intended to increase safety and reduce the stresses associated with driving.
Few other automotive OEMs are involved in R&D of self driving cars. The third
phase of advancement in ADAS comprises cooperative driving which enables V2V
communications effectively. This phase operates multiple driver functions simulta-
neously. The German-based automotive, Volkswagen, adopted V2V communica-
tions for its vehicles from 2019, using public wireless LAN (pWLAN) with IEEE
802.11p standards. This communication enables to connect with vehicles and infra-
structure which added some features like lane merger assistance, platooning, pedes-
trian safety, and parking reservation.
Further into the advancements in ADAS for next generation, the synchronized
cooperative driving provides the features like overtaking assistance and intersection
assistance. Most of the decisions would be made by the vehicle, but it requires the
driver to supervise the actions. The automation is limited in this phase in terms of