B.M.S.

COLLEGE OF ENGINEERING,
Bengaluru-560019.
Autonomous Institute, affiliated to VTU

A Project Report on

AUTOMATED TOOL FOR REGRESSION TEST

Submitted in partial fulfilment of the requirements for the award of the degree of

MASTER OF TECHNOLOGY
in
DIGITAL COMMUNICATION

by
MAHALAKSHMI V (1BM20LDC09)

Under the guidance of

Dr. PRERANA GUPTA PODDAR

Associate Professor
Department of Electronics and Communication Engineering, BMSCE

2021 – 2022

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

B.M.S. COLLEGE OF ENGINEERING

Bengaluru-19
 BMS COLLEGE OF ENGINEERING
Autonomous Institute, affiliated to VTU
Bull Temple Road, Bengaluru – 560 019

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

CERTIFICATE

Certified that the project entitled AUTOMATED TOOL FOR REGRESSION

TEST is a bonafide work carried out by MAHALAKSHMI V (1BM20LDC09)
in partial fulfillment for the award of Master of Technology in Digital
Communication of the Visvesvaraya Technological University, Belagavi during the
academic year 2021 -22. It is certified that, all corrections/suggestions indicated for
internal assessment have been incorporated in the report deposited in the departmental
library. The project report has been approved as it satisfies the academic requirements
in respect of project work prescribed for the said degree.

Guide Head of Department

Dr. Prerana Gupta Poddar Dr. Siddappaji
Associate Professor Professor and Head
Department of ECE Department of ECE
BMS College of Engineering BMS College of Engineering

Principal
Dr. S Muralidhara
BMS College of Engineering

External Viva
Name of the Examiners Signature with Date
1.
2.
 DECLARATION

I, MAHALAKSHMI V (1BM20LDC09), hereby declare that the dissertation

work entitled AUTOMATED TOOL FOR REGRESSION TEST is a bonafide
work and has been independently carried out by me under the guidance of D r .
PRERANA GUPTA PODDAR, ASSOCIATE PROFESSOR, Department of
Electronics and Communication Engineering, BMS College of Engineering,
Bengaluru, in partial fulfillment of the requirements of the degree of Master of
Technology in Digital Communication of Visvesvaraya Technological
University, Belagavi.

I further declare that, to the best of my knowledge and belief, this project has not
been submitted either in part or in full to any other university for the award of any
degree.

Place: Bengaluru MAHALAKSHMI

V
Date: 1BM20LDC09
 ACKNOWLEDGEMENTS

This project thesis has been done to brief the work titled “AUTOMATED
TOOL FOR REGRESSION TEST” done as a part of M.Tech program.

I take this opportunity to express my profound gratitude and deep regards to

Dr. S Muralidhara, Principal, BMSCE and Dr. Siddappaji, H.O.D of Department of
ECE, BMSCE for their cordial support.

I heartfully thank my guide Dr. Prerana Gupta Poddar for her guidance and
constant encouragement throughout the course of this project without which this
project would not be successful.

I take this opportunity to thank Mr. Vikas Hegde, Associate Senior manager, Veoneer
India for his support throughout the work.

Last but not the least, I thank my family, friends and colleagues for their
encouragement and help in accomplishing the objective of the research work.
 ABSTRACT

Active safety is fundamentally concerned with the prevention of vehicle crashes or the
mitigation of their severity. Blind spot detection, lane change assistance, rear cross
traffic alert and safe door open are few of the active safety features to prevent
accidents by alerting the driver and indicating presence of an object near the car. False
positive alerts (alerts that are given at the point where no alerts are required) also
occur in some cases which must be debugged, corrected and tested. The authors
propose and test an automated tool in this work for the regression testing of data
collected for active safety features of a vehicle. It can monitor the old issues which are
fixed, with respect to new and updated code.
Table of Contents
Abstract
List of figures i
List of abbreviations ii

1. INTRODUCTION..................................................................................................................
2. LITERATURE REVIEW.......................................................................................................
3. PROBLEM STATEMENT....................................................................................................
4. RADAR IN AUTOMOTIVE
INDUSTRY……………………………………………….9.
5. METHODOLOGY...............................................................................................................
6. SIMULATION RESULTS...................................................................................................
7. CONCLUSION....................................................................................................................
8. FUTURE SCOPE.................................................................................................................
REFERENCES..............................................................................................................................
APPENDIX…………………………………………………………………………….iv
 List of figures

Figure Description of Figure Page

no. no.
1. Radar signal filtering and creation of detection list 11
2. Track list creation from the detection list 11
3. Scenario for Blind spot detection 15
4. BSD Execution Flow 17
5. Scenario for Lane change assist 18
6. LCA Execution Flow 19
7. Scenario for rare cross traffic alert 20
8. RCTA Execution Flow 21
9. Scenario for safe door open 22
10. SDO Execution Flow 23
11. Scenario for Free space detection 25
12. Proposed algorithm for the tool 27
13. Files in Automated tool for regression test folder 32
14. Path configuration to fetch .mat files and Issue list 32
15. List of .mat files in given path 33
16. Issue list with issue start and end cycle for BSD false positive 34
17. Issue list with issue start and end cycle for BSD false negative 34
18. Issue list with issue start and end cycle for RCTA false positive 34
19. Issue list with issue start and end cycle for LCA false positive 35
20. GUI window during execution. 36
21. Issue list with updated output after simulation for BSD false 36
positive.
22. Issue list with updated output after simulation for BSD false 37
negative.
23. Issue list with updated output after simulation for RCTA false 37
positive.
24. Issue list with updated output after simulation for LCA false 37
positive.
25. Time taken to complete analysis using “Recording analysis tool” 38
26. Time taken to complete the analysis using “Automated tool for 39
regression testing”
 List of abbreviations
Abbreviations Full Form
BSD Blind Spot Detection
LCA Lane Change Assistance
RCTA Rare Cross Traffic Alert
SDO Safe Door Open
CLS Classifier
MTI Moving Target Indicator
RSP Radar Signal Processing
ADAS Advanced Driver Assistance System
RADAR Radio Detection and Ranging
FP False Positive
FN False Negative
1. INTRODUCTION

In recent years, there are thousands of accidents happening on the road in one
single day. Accidents are occurring mainly while changing lanes, while backing out
from a parking space and cyclists or pedestrians are injured by car doors when
opening the door without realizing the presence of someone near the car. This
happens because of blind spot region, the region which can’t be seen through the rear-
view mirror, sudden change of lane from one to another and a person in the car opens
the door unknowingly can cause accidents. To avoid such type of accidents,
automotive industries are trying their best with different methods. One such method is
by using automotive radar.

1.1 Importance of Automotive- Radar

Engineers have been developing new uses for radio-waves based on their
detection and reaching capabilities ever since they were found and their capacity to
reflect off of objects was understood. The Radar- (Radio Detction and Ranging)
technique operates on the theory that a source transmits a radio wave, which is
reflected by a hitting an object and then is picked up by and handled by a reciever
system. Automotive-applications, more notably ADAS systems that provide improved
comfort and safety, are one such area where RADAR is experiencing an unheard-of
growth.

Automotive radar is growing rapidily as a crucial technology for enabling

smart and self- directed features in modern automobiles, such as releasing drivers
from boring tasks, lowering driver anxiety, and adding life-saving involuntary
intrusions. Automotive radar has recently been installed in several high - end vehicles
to enable important safety and ease features such as adaptive-cruise-control and
automatic emergency brake systems, which allow a vehicle to slow abruptly without
the intervention of the driver to avoid a potential collision. Because of the expensive
expense of sensor technology, these functions have only been available in high- end
automobiles so far. This scenario is fast changing, and the predicted widespread
worldwide market diffusion of automotive radar technology has necessitated
collaboration between internatonal and locall regulatory establishments and the
automobile industry to produce adequate and consistent values.
 There are many situations while driving cars on the road especially on curve
roads where the driver will be completely in a Blind Spot. Lane assist is a safety
feature that helps keep your car safely in its lane on the motorway. It's designed to
stop accidents caused when a car drifts out of lane – perhaps due toa lapse of
concentration or drowsiness –opening the car door – by warning imminent doom.

Acceptance of advanced driver assistance systems is increasing as

autonomous driving increases popularity and supervisory authorities throughout the
world mandate the incorporation of specific security features. Currently, a variety of
detection technologies, together with radar, LiDAR, camera, and ultra-sonic- based
are being deployed. Each of these methods has benefits and drawbacks, and the
majority of ADAS systems in use today rely on a combination of technologies to get
trustworthy products to market. It's claimed that radar is an all -weather solution.
Radar based devices function without being hampered by environmental factors like
temperature and humidity. Radar's ability to operate flawlessly in both day and
nighttime lighting settings is one of its main advantages.

One of the biggest drawbacks of nearness sensor-based systems is their

inability to distinguish between stationary and moving objects with ease. Radar can
do this with effortlessness. Using radio -waves, it is considerably simpler to precisiely
determine an object's position, velocity, and the distance. The actual identification and
warning sign to the driver take more time than they should, necessitating considerable
intervention to speed up processing time and enable prompt warning and other
preventative measures. The actual identification and warning sign to the driver take
more time than they should, necessitating considerable intervention to speed up
processing time and enable prompt warning and other preventive measures.

1.2. Overview on Radar signal processing

Signal from radar keeps transmitting from radar transmiter and once the
signal encounters an object the signal travels back and captured by receiver. To
produce a base-band-signal that contains the differnce between the transmitt and
receive waveforms at any specified moment, the mixer down-converts the received
micro-wave frequency signal while ramping up the send signal. As fixed frequencies
that depend on the target's range and Dopler shift return, the ramping is cancelled off.
At a maximum range for sensing horizon of 150 m, the mixer's output will be at
77GHz. As a result, anti-aliasing Low pass filter can be provided before the analog-
to-digital (ADC) converter using conventional passive components and operational
amplifiers. Prior to being supplied to the radar processor for target detection and
tracking, the receiving process samples and digitalizes the analog baseband signal. To
distinguish between the range frequencies and the Doppler frequencies, auto radars
normally employ their signal processor.

The typical method involves sending off a number of quick chirps, collecting
the echoes, and then organizing the data in a matrix that is frequently depicted as a
two - dimensional array with the detected-frequencies of each chirp given in a single
column. Fast time refers to the contents of the columns, whereas slow time refers to
the contents of the rows. The speed and range for one or more targets are determined
by performing an FFT on the fast time entries first, followed by an FFT on the slow
time data. Applications for the automotive industry must also resolve targets in terms
of their angulaar distance from the radar sensor. Radars typically use between four
and sixteen antennas to do this. Each antenna output undergoes a fast and slow time
FFT. The resulting data is frequently represented as a cube with a Z axis that
represents the data for each of the antennas and X and Y axiis that are made up of the
fast and slow time data. This cube effectively portrays a three - dimensional map with
axis for azimuth, distance, and speed.
2. LITERATURE REVIEW

Vipin Kumar Kukkal et.al,[1] Authors briefed, in today's vehicles, advanced-

driver-assistance-systems (ADASs) have become a critical section for safety. They're
also a critical component of developing self--driving vehicles. The majority of current
ADASs rely on vision, but light detection and range (lidar), radio detection and
ranging (radar), and other advanced- sensing technologies are gaining admiration.
Radar systems use the Dopplereffect to estimate the speed anddistance of an object by
measuring the change in frequency of the reflected wave. Micro-waves can travel
much further than optical-light and identify objects at a greater distance due to their
longerwavelength. Radar, unlike lidar, is unaffected by foggy or rainyweather and is
relatively inexpensive. Radarsystems are classed as short range-(1.8–30 m), medium
range-(30–80 m), or long range-(80–200 m) depending on their operational distance
range. Short-medium-range radars are used for a variety of purposes, including cross-
traffic alerts and blind-spot identification. These systems are frequently found in the
vehicle's corners. Adaptive-cruise-control is a long-range radar device that is installed
behind the front grill or under the bumper.

Chintaka Premachandra et.al,[2] used a 360-degree-view camera(3DVC) to

detect and track moving objects at traffic intersections for driver assistance and
automated driving. The ability of a 3DVC positioned at a junction corner to picture
the entire intersection area has been experimentally proven. Using the 3DVC images,
we also created algorithms for recognizing and tracking moving objects near the
intersection. Experiments under a variety of situations revealed that the created
algorithms were quite effective in detecting and tracking moving objects.

[3] The 22GHz radar system identifies and warns of objects in the vehicle's
blind zone, assists with lane changes, alerts of rear cross traffic and rear crashes, and
has a trailer tow assembly. Increased detection points are possible with some high-
resolution radars, resulting in a better understanding of free space, better object
separation in multiple target scenarios, and more precise detections in blind spot
warning and rear cross traffic alert. Oncoming vehicles are detected using 77GHz
radar systems, which automatically manage lane changes during highway driving.

Guiru Liu et.al, [4] proposed a system that the system used was a standard
radar detection system, with two radars fitted inside the car on the left and right sides,
as well as right and left rear view radars. A separate local CAN network connects the
Micro Controller Unit (MCU) and two radars. By using the CAN network, the radar
delivers the detection result to the MCU. The signal is captured by the MCU. The
signal from the MCU was used to control the warning LED and buzzer. The CAN
network is used by the MCU to capture information about the car.

Hongjun lee et.al., [5] offered a new data augmentation approach based on
GAN. Using this framework, the detection performance of an existing BSD system is
increased in a low- light situation. Using a conditional GAN, the proposed approach
creates synthetic nighttime visuals. The GAN was trained using a public front-view
image library; however, the synthetic images were not a complete replication of
reality because the photos were from diverse views. Nonetheless, they are sufficient
for vehicle detection data augmentation. The detection performance of the augmented
BSD system was substantially doubled when applied to actual nighttime photos,
compared to training with solely daylight images.

Jungme Park et.al, [7] Used two radars and one wide-angle camera, a robust
sensor fused RCT detection system. The planned RCT detection system's novelty is
that the RCT detection system's radar and camera sensors are integrated. The radar
points are transferred to camera images by translating the radar coordinate system into
the image coordinate system via sensor fusion. Furthermore, the information from the
two radar sensors on the overlay image is used to distinguish between static and
moving objects.

Shizhe Zang et.al, [8] discussed the impact of bad weather conditions on
state-of-the-art sensors such as lidar, GPS, camera, and radar. Characterization of the
impact of rain on vehicle radar, taking into account both attenuation and backscatter
effects are also considered. Under extreme rainfall conditions (150 mm/h), the
detection range of a mm-wave radar can be reduced by up to 45 percent, according to
simulated data. Furthermore, the backscatter effect plays a larger role for a close-
range target with a tiny RCS area, causing additional performance deterioration.

Xiao Juan Yang el.al, [9] The issue with automated driving systems is that in
instances when unpredictable occurrences such as weather and false positive alert
may cause fatal crashes, sensors that monitor the vehicle's surroundings will be unable
to take corrective action. The issue of false positive failure detections emerges, in
which the monitor incorrectly classifies the commander's output as malfunctioning.
The failure mode system of automobile active safety components and electronic
control system, which provides early warning and remote diagnosis of automobile
active safety components and electronic control system failure, is formed by studying
the composition and working principle of automotive active safety components,
electronic control system and software used. Steering vibrator motor fault, forward
identification camera fault, steering vibration ECU fault, airbag spiral harness short
circuit fault, relevant circuit and wire connection fault are all common lane departure
alarm system defects which are the given typical examples to show the issues.

[10] The captured data from the sensor is given to a low pass filter to remove
noise and then passed to the analog to digital converter. The converted data along
with the captured video using camera are kept in a .cvw file. These files are analyzed
by using different tools, one of them is DANVIEW. This tool is used to monitor the
issue generated with the features while driving the car. Danview gives the animated
video for the captured data. Here, we can analyze at what cycle the error/issue
occurred and what cycle it ended. Hundreds of data will be captured from the sensor
and can be analyzed in it.

Muhammad Hassnain et.al, [11] discussed about the tester can concentrate on
testing the functionalities of a web service using an automated regression testing tool.
The regression testing and verification in uncommon or complex contexts is more
reliable with automated regression testing solutions. When data is readily available,
measurements and metrics can be collected and assessed. Despite the many benefits
of employing automated regression testing tools, several critical elements related to
the quality and breadth of automated tests are missing. Automated regression tests
that are statically utilized perform the same activities every time. This means that
when the test was run in the automated testing software, it followed the same path.
Using the tests repeatedly does not generate additional code coverage.

Inhwa Choi et.al, [12] said, since testing is such an important part of software
quality assurance, numerous studies are currently being conducted in areas including
test method design, test case generation, test case management, GUI test automation,
and integrated test automation to build more efficient and cost-effective tests.
Regression testing is used to determine whether a change or addition to a software
module has resulted in a defect or error in existing software. The major features of a
regression test automation system are the ability to dynamically generate test cases
during the regression test stage and execute the test scenarios generated automatically.

Shruti Malve et.al, [13] discussed that software testing is a broad field, which
are divided into two types: manual and automated testing. Manual testing involves
manually executing (running) test cases without the use of tools/scripts, whereas
automated testing involves using tools, scripts, and software to execute (run) test
cases. Software testing is costly, time- consuming, and time-consuming in the
software development life cycle. In software testing, there has always been a need to
reduce testing time. The method of testing (manual or automated) is determined by
several criteria, including the project's requirements, budget, timeline, expertise, and
suitability. The limitation of manual testing is that, for a tester, performing manual
testing is a tedious task. It must check each link on the site as well as all data about
the goods being tested. Testing necessitates a significant amount of time. We must
repeat the same operations over and over. Regression testing is used to see if defects
introduced by changing one part of the software have a detrimental impact on the
other part of the product. Manual testing takes time, is prone to errors, and
necessitates a significant amount of infrastructure and manpower.

3. PROBLEM STATEMENT
 To avoid the maximum time taken for comparison from new software to old
software - Sometimes the radar automated feature can give false positive alerts
which may cause customers to lose their interest in the feature and tend to avoid it
even if the real alert is given. So, to avoid it, analyzing and running through the
regression for all the sets of recorded data is necessary. After a regression test, to
check if the updated code resolves the issue and does not implement any other
new issue, we will have to cross check manually which may consume a lot of time
and human effort. This can be avoided by creating a tool to check if the issue is
solved and no errors are generated to all the data at once.

4. RADAR IN AUTOMOTIVE INDUSTRY

4.1. RSP - cluster track formation

Blind spot-monitoring, object-recognition, and colision warning and

mitigation are some of the major uses of radar in-ADAS. While colision mitigation
systems utilise the information from the radar-based system and take preventive
measures like counter steering and decelerating to dodge accidents, blind-spot-
monitoring, and object identification both rely on radar technology.

Radar that is fixed to the car keeps radiating radio signal/radar signal. Once
the signal hits any object, the signal is reflected and captured by the receiver in radar
and calculates the distance and relative speed of the approaching object and alerts the
driver to make a safe move based on given input data.

Target detection, which serves to distinguish an echoing target signal from

background signals, is the first step in the processing of a radar signal. This is
accomplished by setting a threshold and comparing the output power of the matching
filter with it. If a target is identified when the measured power amplitude exceeds the
threshold. The related signals are processed further to produce measurements of the
target's azimuth angle, range, and range rate. As opposed to that, the signals are
viewed as thermal noise or other forms of interference (such, if the power amplitude
is less than the threshold, ground and clutter) The field of signal processing for radar
systems is large and exciting, covering a wide range of techniques and application
areas. Since then, it has developed into a long-range sensor that is necessary in all
weather conditions. The main forces behind radar development have often been
security and military uses. Radar has, however, recently emerged as a crucial
technology for civil uses, such as air, sea, and ground traffic management, as well as
urban sensing and indoor monitoring. Radar influences not only the present but also
the future of humanity. The traditional radar mission is to detect and identify objects,
as indicated by its term radio detection and ranging.

The ability of a radar system to detect, track, and image objects or features of
interest in the environment depends on a number of factors, including how those
factors are reflected in the received signals that can be processed as well as the
characteristics of the objects, the environment, and the radar itself. The signal-to-
interference ratio and resolution are two of the most fundamental and significant
signal quality parameters. The majority of the fundamental radar signal processing
described focuses on enhancing SIR and resolution because of how important they
are.

Although many follow a wide range of signal processing methods, most of

them are based on a few fundamental concepts. These include target phase history
modelling, bandwidth expansion, coherent and noncoherent integration, and
maximum likelihood estimation.

Signal-to-Interference ratio and integration - The SIR for continuous time

signals is calculated as the difference between the power of the desired signal and the
interference. If s[n] is deterministic, the peak signal value is typically taken to be the
signal power, which may consequently happen at a particular time t. The average
signal power might be employed in some predictable situations instead. Consider a
discrete-time signal x[n] consisting, the sum of a desired signal s[n] and an interfering
signal w[n] is given in the equation 1:

x [ n ] =s [ n ] + w [ n ] . [1]

The interference is almost invariably modelled as a random process; hence

its power is the mean-square. If the interference is 0 mean, as is very often the case,
then the power also equals the variance of the interference, σw2. If the desired signal
is also modelled as a random process, then its power is also taken to be its mean-
square or variance. For amplitude A, the SIR of their sum x[n] is given in equation 2,

A2
x [n ]= . [2]
σ 2ω

Resolution and a resolution cell, two terms that are closely related, will come
up frequently. When two equal-strength scatterers split into two distinct signals at the
system output rather than merging to form a single undifferentiated output, they are
said to be resolved. In range, cross-range, Doppler shift or velocity, and angle of
incidence, the concept of resolution arrival.
 Figure 1. Radar signal filtering and creation of detection list

Figure 2. Track list creation from the detection list

Figure 1. and Figure 2. Are radar signal filtering and creation of detection list
and Track list creation from the detection list respectively. In radar signal filtering,
once the signal is received by radar receiver, the received is given as input to the
Matched filter. Matched filters are commonly used in radar, in which a known signal
is sent out, and the reflected signal is examined for common elements of the out-going
signal.

Matched filtering is the best type of filtering since it acquires the required signal while
suppressing others. Using chirp signals for radar detection, matched filtering can offer
greater range resolution. A method for finding a known signal or wavelet that is
covered by noise is matched filtering. The frequency response of the matched filter
will be proportional to the complex conjugate of the input signal’s spectrum which is
given in equation (1) [17]. For a fixed input signal to noise (energy) ratio, it
capitalizes on the output peak signal to mean noise (power) ratio i.e.,

¿ ( j 2 πf t 1)
H ( f ) =Ga S ( f ) e . [3]

Where, H(f) is the frequency response function.

S(f) is the voltage spectrum of input signal.

S¿ ( f ) is complex conjugate of S(f).

t 1 is the fixed value of time where the signal is at its maximum.

G a is maximum filter gain.

It is assumed that the noise that surrounds the signal is stationary and has a
uniform spectrum (white noise). It doesn't have to be gaussian. The North filter,
conjugate filter, or more often known as the matching filter is the filter whose
frequency-response function is described by the equation above. Another name for it
is the Fourier transform criteria. Contrast it with the impedance matching idea from
circuit theory, which maximizes power transmission rather than signal-to-noise ratio.

Target detection in radar system is a simple and crucial job where the received
signal is compared to a threshold. Therefore, setting a suitable threshold is the main
task involved in detection. Generally speaking, the threshold depends on both the
likelihood of detection and the likelihood of a false alarm. After a target is initially
detected, the radar must continue to detect the target and estimate target parameters
from radar observations like position, size, motion etc. Multiple observations are used
to develop a more accurate filtered estimate and get the detection list. The detection
list from figure 1 is given as input to the radar screening shown in figure 2. Screening
is used for detecting the position and the velocity of a distant object. It also indicates
the range, bearing or speed of a particular target or the status of a radar function. The
radar tracks the target by tracking one of the three parameters: Range, Angle and
Doppler Frequency shift. Consider that the tracks are created using the Doppler
Frequency shift. The Doppler shift describes the changes in the frequency of any
sound or light wave produced by a moving source with respect to an observer. The
tracker uses the doppler frequency shift technique and create a track from the source
vehicle and the target vehicle. After creating tracks, maintaining those tracks is the
process in which a decision is made about whether to end the life of a track or to
maintain it. Next step is target association where, the process of identifying the target
with multiple detections are combined and selected as one track.

If a detection was not associated with a plot in association phase, then there
is a chance that the target may no longer exist. Next step is predicting and updating
the movement of the target vehicle for which Kalman filter is used. The Kalman filter
is a discrete time linear dynamic system, and it has two phases - Predict and Update.
Predicting is where it Predicts the state estimate of the current time using the state
estimate of the previous time while Update phase combines the current time state
estimate with the current measured values and updates the current state estimate.
Kalman Filter is an optimal recursive data processing algorithm, where optimal
depends on the criteria chosen to evaluate the performance and recursive since it does
not require storage / reprocessing of all previous data [16]. This system depends on
the equations (4) and (5):

X k = A X k−1 . [4]

Z k =B X k . [5]

where, 𝑋𝑘 is the state of the system at time instant k. It is based on the state of the
system at time k-1. A is a matrix used to calculate the current state of the system from
the previous state with assumption of constant acceleration (the rate of change of
velocity of a body remains constant over time). 𝑍𝑘 is the measured value of the
system, and it relates to the calculated value 𝑋𝑘 and B is an identity matrix. From
predicting and updating the tracks using Kalman filter, track list will be provided. The
track list will further be combined with detection list and given to screening to
maintain the created tracks with new tracks.
4.2 Radar Features.

It is required that the test measuring system be totally independent from the
blind spot warning system under test: if the test target vehicle is to the rear of the
subject vehicle, it should be able to measure the longitudinal distance between the
trailing edge of the subject vehicle and the leading edge of the test target vehicle. If
the test target vehicle is in front of the subject vehicle, it should be able to measure the
longitudinal distance between the subject vehicle's leading edge and the test target
vehicle's trailing edge. If the test target vehicle is to the left of the subject vehicle, it
should be capable of measuring the lateral distance between the leftmost edge of the
subject vehicle and the rightmost edge of the test target vehicle. If the test target
vehicle is to the right of the subject vehicle, be capable of measuring the lateral
distance between the subject vehicle's rightmost edge and the test target vehicle's
leftmost edge. Be able to gauge how long it takes for a warning to be given when the
test target vehicle satisfies the necessary warning conditions and be able to calculate
the amount of time that passes between the point at which a warning is prohibited and
the point at which it is terminated.

According to International standard ISO 17387 [6], the lane-change warning-

function's warning-requirements are a mixture of the blind-spot warning- function's
warning requirements and the closing-vehicle warning-function's warning-
requirements. The blind spotwarning requirements for a given situation will result in
one of three outcomes: shall-give warning, might-give warning, or shall-not offer
warning. The closing vehicle warning- requirements will also produce one of the
following outcomes in the same situation: shall give warning, might give warning, or
shall not give warning. The system warning signal must be visual only in the warning
level 1 state. System warning indications at warning levels 2 and up may include
visible, audio, and/or haptic indicators. It is suggested that alerts specify the side on
which the target vehicle is present in some fashion. It is suggested that visual warning
indicators be put in areas that encourage the use of mirrors. Visual warnings should
stand out from the rest of the vehicle's visual information. If an auditory warning is
employed, it should be easily identifiable from the rest of the vehicle's aural signs. A
haptic warning should be easily distinguished from other haptic signals in the vehicle
if it is employed.

4.2.1. Blind spot detection - When the car is started after the technology is installed,
radar sensors on both sides of the rear radar and front radar emit a radar signal
backwards. The system will process the reflected signal when a vehicle in the blind
region is spotted approaching. When it detects an approaching object, it informs the
driver by emitting the appropriate sound or light signals. When the vehicle
approaches, the matching side of the LED lights will illuminate to indicate that the
vehicle's back is approaching. When the LED is lit, if the driver turns on the
corresponding turn signal, the LED will flash, and the buzzer will sound to indicate
that the turnaround is risky at this time. Figure 3 shows a Scenario for Blind spot
detection.

Figure 3. Scenario for Blind spot detection

BSD Sub-Layer: In the simplest sense, the BSD feature alerts when an object is
residing in the vehicle’s blind spot. BSD processing consists of executing the CLS
and the MTI algorithms, the results of which are made available to the BSD’s
“synthesis” algorithm.

1. CLS Manager AAU: The Classifier (CLS) Manager is responsible for overall
management of the CLS functions, including:

• CLS parameter initialization

• CLS Points initialization and handling

• CLS algorithm execution

The CLS Manager uses the Front Entry Detection AAU and the Stagnation Detection
AAU in a dual model approach to identify how front entry targets and stagnating
targets are moving with respect to the ground as well as the host vehicle. The resulting
data is passed both to the MTI AAU and to the BSD’s synthesis logic.

CLS Front Entry Detection AAU: The Front Entry Detection AAU uses Doppler
frequency peaks to classify front entry targets as vehicle or non-vehicle (e.g.,
guardrail).

CLS Stagnation Detection AAU: The Stagnation Detection AAU identifies targets
that are moving relatively slowly with respect to the host vehicle.

2. MTI Manager AAU: The Moving Target Indicator (MTI) Manager is responsible
for overall management of the MTI functions, including:

• MTI Front AAU parameter initialization

• MTI Rear AAU parameter initialization

• MTI Points initialization and handling

• MTI algorithm execution

MTI Front AAU: The MTI Front AAU maintains probability scores for moving
versus stationary targets arising from front entry detections.

MTI Rear AAU: The MTI Rear AAU maintains probability scores for moving versus
stationary targets arising from rear entry detections.

3. BSD Logic Manager AAU: The BSD Logic Manager is responsible for overall
management of the BSD Feature, including:
• BSD Algorithm execution

• BSD Support functions (alert hysteresis, etc.)

• BSD Points initialization and handling

High-level view of the BSD flow of execution is shown in figure 4:

Figure 4. BSD Execution Flow

The algorithm consists of 4 major blocks: XY Filter, BSD Main, Decision Logic, and
Post BSD code. The XY Filter maintains filtered Cartesian target dynamics obtained
from radar measurements. These are used by the BSD Main code to manage each
target’s counters, namely the Rear Entry, Front Entry and Side Entry counters. As a
counter crosses different threshold, Decision Logic issues or clears BSD warnings for
the corresponding target. The Decision Logic also synthesizes the results with those
of MTI and CLS, for example, masking warnings coming from non-vehicle targets.
Finally, the Post BSD code exercises hysteresis logic on the alert as a function of
elapsed time and may also shut off the alert depending on the host vehicle’s speed.
4.2.2. Lane change assist - Changing lanes is always dangerous, especially when a
car in the adjacent lane is travelling faster. As we can see in figure 5, there's the issue
of the blind spot, which is a region behind the vehicle that isn't visible in the rear-view
mirrors and represented in red region. If there are vehicles in the area of vehicles
approaching quickly from behind, lane Change Assist technology informs the driver.
This information is incredibly useful, particularly if the vehicle abruptly changes
lanes. Vehicles travelling the wrong way or being overtaken are detected but not
signaled. The information is delivered to the driver via a signal smartly integrated into
the exterior rear-view mirror. The material is presented in a natural and intuitive
manner. Radar sensors are used in the lane change assistance system to scan the
surrounding region utilizing a variety of radar beams. The entire electronic radar has
no moving parts and can function in almost any weather condition.

Figure 5. Scenario for Lane change assist

LCA Sub-Layer: Lane Change Assist (LCA) is used to monitor the area behind the
host vehicle in adjacent lanes and issue an alert if a vehicle is moving towards the host
vehicle in such a way that a collision may occur due to a lane change maneuver. The
LCA Feature logic is predicated on the ability of the system to map tracked targets to
LCA zones and determine the threat imposed on the host vehicle. The LCA zones are
not static. Their placement and dimensions are adjusted as a function of the host
vehicle state (velocity, lane change status, previous path) and driving environment
(roadway radius of curvature, turn direction).

1. LCA Manager AAU

The LCA Manager is responsible for overall management of the LCA Feature,
including:

• LCA Algorithm execution

• Target Processing

• Alert Processing

• LCA Points initialization and handling.

High-level view of the LCA flow of execution is shown in figure 6:

Figure 6. LCA Execution Flow

2. LCA Capability Control AAU: For historical reasons, the lane tracking portion of
the LCA Feature is referred to as the Capability Control AAU. Using Vehicle
data, the Capability Control AAU estimates the Time to Collision and determines
adjacent lane boundaries.
3. LCA Target Process AAU: Target processing analyzes LCA targets and
computes an alert score for each track. It propagates the radar tracks to determine:

• Lateral Distance to the Host Vehicle

• Zone Classification of the Target
• Alert Score
The collection of alert scores for qualified tracks feeds into the LCA alert decision.
4. LCA Alert Process AAU: Alert processing uses the processed radar tracks from
Target processing to determine if LCA should alert. It coasts and maintains the
LCA Alert, using the current alert state, individual track alerting scores, vehicle
dynamic state, and hysteresis.

4.2.3. Rare cross traffic alert - A Rear Cross-Traffic Collision Warning is an

advanced driving assistance system that informs the driver if a vehicle is approaching
from the left or right when the driver’s vehicle is in reverse and is backing out of a
parking space. With this specific system, the alert icon will usually appear on the
appropriate side mirror with an audible warning and depending on the model, the alert
will also pop up on the dashboard. The vehicle includes two sensors on the left and
right sides of the back bumper that uses radar to determine if any vehicles are
approaching from either direction, similar to the Blind-Spot Collision Warning
driving assistance system.

RCTA Sub-Layer: Rear Cross Traffic Assist (RCTA) enhances a driver’s awareness
of rear traffic by monitoring objects located to the rear while the host vehicle is in
reverse gear as shown in figure 7. It alerts the driver if an approaching vehicle is
detected and within the warning limits as dictated by vehicle dynamics, parking angle
and time to collision.

Figure 7. Scenario for rare cross traffic alert

1. RCTA Manager AAU: The RCTA Manager controls the processing of the
RCTA feature. The RCTA Manager is responsible for overall management of the
RCTA Feature, including:

• Buffer processing

• Algorithm execution

• Alert Processing

• RCTA Points initialization and handling

2. RCTA Buffer Process AAU: The RCTA feature utilizes a FIFO memory
buffer in order to manage the tracked radar detection data for processing. The RCTA
Buffer Processing AAU manages the data that is collected for processing by the other
RCTA related AAUs.

High-level view of the RCTA flow of execution is shown in figure 8:

Figure 8. RCTA Execution Flow

3. RCTA Algo AAU: The RCTA Algorithm AAU is the main algorithm used in
RCTA. It implements the following functions:
• Dynamic Compensation
• Path Estimation
a. Continuing Track – Calculate host parking angle and target path distance.
b. Finished Track – Post process to update track information.
• Reference Track Updates
a. Continuing Track – Perform calculation for reference track method.
b. Finished Track – Update buffer for reference track.
Dynamic Compensation refers to the process of compensating for the path estimation
error when the host vehicle is moving, based on geometric relations relative to a
parking angle. It incorporates Host Vehicle Processing information which determines
the unbiased yaw rate and other vehicle dynamics parameters. When host vehicle
movement is detected, compensation for a previous part of the current track relative to
the parking angle is performed.

4. RCTA Alert Decision AAU: The RCTA Alert Decision AAU uses the previous
alert state, reference track alerting score, path estimate alerting score, and hysteresis
in determining the appropriate alert state to be reported.

4.2.4. Safe door open – When a passenger or driver of a car fails to notice moving
things in the rear as shown in figure 9, accidents occur frequently when they open the
vehicle's door. This feature is similar to blind spot detection. Once the object enters
the zone, people in the car get an alert indicating not to open the car door immediately
via light signal or acoustic signal.

Figure 9. Scenario for safe door open

SDO Sub-Layer: In the simplest sense, the SDO feature alerts when an object is
residing in the vehicle’s blind spot. SDO processing consists of executing the CLS
and the MTI algorithms, the results of which are made available to the SDO’s
“synthesis” algorithm which is very similar to blind spot detection (BSD).

High-level view of the RCTA flow of execution is shown in figure 10:

Figure 10. SDO Execution Flow

The algorithm consists of 4 major blocks: XY Filter, SDO Main, Decision Logic, and
Post SDO code. The XY Filter maintains filtered Cartesian target dynamics obtained
from radar measurements. These are used by the SDO Main code to manage each
target’s counters, namely the Rear Entry, Front Entry and Side Entry counters. As a
counter crosses different threshold, Decision Logic issues or clears SDO warnings for
the corresponding target. The Decision Logic also synthesizes the results with those
of MTI and CLS, for example, masking warnings coming from non-vehicle targets.
Finally, the Post SDO code exercises hysteresis logic on the alert as a function of
elapsed time and may also shut off the alert depending on the host vehicle’s speed.
SDO Logic Manager AAU: The SDO Logic Manager is responsible for overall
management of the SDO Feature, including:
• SDO Algorithm execution
• SDO Support functions (alert hysteresis, etc.)
• SDO Points initialization and handling
4.3. Additional radar features

4.3.1. Autonomous Emergency Braking

If the vehicle is moving at a speed greater than 40 km/h and the automated emergency
braking senses that the distance to the vehicle in front of it or stationary is getting
critically close, it prepares the braking system for possible emergency braking. The
technology alerts the driver via an auditory and/or visual indication, followed by a
brief but visible pull of the brakes, if they do not respond to the hazardous situation.
The system then starts applying partial braking to slow down the vehicle and allow
the driver plenty of time to respond. The mechanism starts to support braking as soon
as the driver presses the brake pedal. To accomplish this, the system continuously
determines how much car braking is necessary to prevent the crash. If the system
determines that the driver has not applied enough brake pressure, it increases the
braking pressure to the necessary level so that the driver can try to stop the car before
an accident happens. Automatic emergency braking can automatically begin full
braking if the driver does not respond to the immediate risk of collision and the
system thinks that a rear-end accident is unavoidable. As a result, the vehicle is
moving at a substantially slower speed when it collides, lessening the impact on the
occupants of both cars.

4.3.2. Adaptive cruise control

Future generations of intelligent cars are expected to have significant amounts of

ACC technology. By keeping the best possible distance between vehicles and
minimizing driver mistake, the system improves passenger safety and convenience
while also expanding the capacity of the road. According to SAE International, cars
with autonomous cruise control are Level 1 autonomous vehicles. The car is regarded
as a Level 2 autonomous vehicle when paired with another driver assistance
technology like lane centering. Adaptive cruise control (ACC) systems can be
implemented as either Full Speed Range (FSRA) systems or Limited Speed Range
(LSRA) systems. LSRA systems are further divided into two categories based on
whether an automatic or manual clutch is required. The primary purpose of adaptive
cruise control is to offer longitudinal control for equipped vehicles while they are
travelling on highways (roads where pedestrians and non-motorized vehicles are
prohibited) in both free- flowing and, for FSRA-type systems, heavy traffic
conditions.

Other features, such as forward obstruction warning, can be added to ACC. For
FSRA-type systems, the system will make an effort to slow down behind an already
tracked vehicle within the limits of its deceleration capabilities and will be able to
restart the journey once the driver has given the system permission to do so. The
system is not required to respond to things that are stationary or moving slowly.

4.3.3. Free space detection

Its purpose is to report the amount, and quality of the free space on either side of the
car as shown in figure 11. This information can be used either internally by other
Features, and/or transmitted on the sensor external communication interface for use
by other ECUs in their decision making.

Figure 11. Scenario for Free space detection.

5. METHODOLOGY
All the features explained in the section 4.2 are deployed in the car and tested
to ensure the development and installation is done without any errors. These are
checked by giving a test drive with the feature installed. All the tested drive data is
captured and recorded in video are kept in .cvw file format for the future reference
and issue analysis if any issue/error (false positive or false negative alert) is found in
the data.
The captured data needs to be analyzed for recurring issues, then same should
be fixed and tested again. The new updated code should clear the captured issue and
does not introduce any other new issues. When the customer reports any issue, we
willhave to cross check what is the issue that exactly occurred and should be resolved.
For that, captured data along with recording is analyzed using a specialized tool
DANVIEW (This tool generates a simulated video that gives an idea of when the alert
started and when alert stopped).
Once we know the issue, the installed code for the feature must be debugged
for error code, which is one of the initial steps of verification. With the previous
method used, in initial step a before .mat file is generated from old version of
software, and which is given as one of the inputs to the automated tool. After
debugging and analyzing the error code, it mustbe fixed and run through the
regression to check if the issue is resolved or not. After a regression test of new
updated code one more .mat file is generated which can be named as after .mat file
and is given as one of the inputs to the automated tool.
These two .mat files are compared in an automated tool, and it generates a
report in whole for all the tested drive data files. In the latest software version if any
new issue is reported, then the issue is added to the reported issue list manually and
again the procedure is continued until the issues are completely resolved and no new
issues are introduced.
Here the .mat file contains array of captured data for each cycle (50ms),
function state and warn state that are explained in section 5.1. Comparing two mat
files for all the data array of each cycle inside it is a lot of time consuming. Thus,
ATRT Tool is developed to reduce the time taken to compare the drive data and test
results of old software to new software changes by providing a selected data array of
from specific time cycle where the issue has occurred, and it is explained in the
algorithm section 5.1.
5.1. ALGORITHM

Figure 12. Proposed algorithm for the tool

 The main advantage of using this method is to reduce the number of times
mat files generated before updating the software. In this particular procedure, only
false positive warning and false negative warning are considered to reduce the time
taken to check whether the updated software solved the issue or not. Referring to
figure 12, once the program starts, as an input, mat files and a manual input excel
sheet is given. Initially data from excel sheet is fetched and the data in it are read. The
manual input data in excel sheet includes list of features, base recording of test case,
software recorded version, corner that the sensor is placed, Issue start cycle and issue
end cycle. Each mentioned base recording in the excel sheet should be placed in the
folder of the same path with the same naming.
Next procedure is to classify multiple features. Each feature has different
variables to check for the warning such as, cycle array (Represents each cycle of the
recorded data), function state (Represents state of the sensor with inactive state=0,
active state=1 and fault state=2) and warning state (Represents at what cycle the
warning has been given. If no warning is present then the output is given as zero, if
warning is present then the output is given as 1). To be specific about the variables for
different feature each goes with the feature name prefix to variables such as,

 bsd_Cyclearr, lca_CycleNo, rctalog_Cyclearr, sdo_Cyclearr for Cycle array.

 bsd_funcstate, lca_funcstate, sdo_functionstate for Function state.
 bsd_warnstate, lca_LCAWarning, rctalog_warnstate, sdo_warnstate for
warning state.
These are the output taken from mat files, variable name given should be same
as the name present in the fetched mat file.
Once the car with the deployed feature is on the road, all the radar sensors
keep transmitting radio signals, then function state of the radar is determined whether
it’s in active, inactive or fault state. For bsd_funcstate = 0, the radar is in inactive
state. For bsd_funcstate = 1, the radar is in active state. For bsd_funcstate = 2, the
radar is in fault state. When bsd_funcstate is, one then bsd_Cyclearr starts updating
until the bsd_funcstate becomes zero. Suppose any subject vehicle is near the target
vehicle, then the bsd_warnstate will become 1, which indicates some subject vehicle
is approaching it by giving some kind of alert such as LED blinking or acoustic alert.
And not to make any changes such as changing the lane.
 When subject vehicle moves away from the target vehicle, bsd_warnstate will
be zero and no warning is given. Similarly, all the other feature functions are done.
After feature classification, all the pre-recorded False positives are set to zero
and then the after_mat file is fetched from the added folder path which will be
checked one after the other. Initially, just for the confirmation, after_mat file is
compared with excel sheet to confirm whether the false positive issue is found or not,
by comparing the given name to it. If no false positive is detected next mat file is
checked. Confirming it continues with testing the start cycle of the issue found till the
end cycle which are mentioned in excel sheet. This particular task reduces time taken
to compare the cycle array.
Testing is followed by determining if the warning in the after_mat file
(warning_after) is greater than zero or equal to zero. For false positive issues, if the
warning_after is greater than zero, then the false positive has reoccurred hence
updating “Output = NOK” in the input excel sheet. If the warning_after is equal to
zero, then false positive has not reoccurred and hence updating “Output = OK” in the
same input excel sheet.
Similar concept goes with False negative, where warning in after_mat file
(warning_after) is greater than zero or equal to zero. If the warning_after is greater
than zero, then the false negative has not recurred and hence updating “Output = OK”
in the excel sheet. If the warning_after is equal to zero, then false negative has
recurred and updates “Output = NOK” in the same input excel sheet. After all the
after_mat files are compared then the program ends.

5.2 ADVANTAGES AND LIMITATIONS OF AUTOMATION

5.2.1 Advantages of Automation

 Decreases manual errors: Human error can happen during manual data
entering. Therefore, one of automation's primary advantages is that it may
perform manual data-entry chores on behalf of humans, which lowers the
possibility of human error. As a result, data-entry will always be of high
quality.
 Reduces workload for employees and increase productivity: Teams doesn’t
require to spend valuable time on tiresome administrative tasks by using
automation tools. In turn, this promotes productivity. Team members can
 produce more work if there are fewer processes and repetitive activities
holding them back.
 It promotes efficiency: One of the advantages of automation is unlocked only
by installing automation software. Here we get to analyze and improve the
processes since we set up the procedures and guidelines as the program to
follow. As a result, we can spot any inefficient or broken workflows. Also, we
can consequently experience increased operational effectiveness. Additionally,
we can increase the effectiveness of the crucial procedures.
 Quick responses: One more advantage of automation is that it enables quick
responses when they are most needed. Software for automating business
processes can parse incoming data from emails, web forms, social media,
system data, etc. After then, it can notify events like complaints or service
interruptions. Incoming message replies can also be handled via automation.
As a result, when something goes wrong, resolutions happen more quickly.
 Gaining of data-led vision: Today's businesses gather a lot of data. But using it
isn't always simple. Large chunks of it frequently go unused. Herein lies a
further advantage of automation. We can use automation tools to release the
insights that are locked up in the gathered data. We can also obtain insights
into more areas of organization with the use of tools like sentiment analysis
and the capacity to build intricate operational reports.
 Cut prices: Cost-cutting is another benefit of business process automation once
the automation tool is tested enough to make benefits. Software never gets old
or runs out. As a result, one can accomplish more with less resources. In other
words, getting more processes completed in less time for the cost of
automation software can be achieved.
 Clear liability: The enhanced transparency of followed procedures are one of
the advantages of automation. Company follows strict guidelines hence,
making sure that workflow adhere to legal requirements are simple.
Additionally, it allows for warnings, records, and audit trials. We can therefore
clearly see what, how, and when something goes wrong.

 Scalability: Software for automation scales with humans. Rules can be

changed, modified, and added to fit according to the needs. Therefore,
 regardless of where the company is or how much firm of the company
expands, company can make profit from automation.

5.2.2 Limitations of Automation

 Initial investment is high until the automation tool is tested enough to get the
benefits.
 Software malfunctions are one of the limitations where bugs and
vulnerabilities can affect the efficiency of the tool.
 Continuous maintenance of the code - Whenever a software is upgraded, it
requires a review of the automated codes.
6. SIMULATION RESULTS

Figure 13. Files in Automated tool for regression test folder

The above figure 13 shows the files used to develop the automated tool
which involves ATRT.m, IssueList.xlsx, and ATRT_GUI.m. ATRT.m will have
function for comparing the given cycle to check the issue. ATRT_GUI is developed to
configure path by browsing the address where after_mat file and input excel sheet is
stored. IssueList.xlsx will have list of issues reported along with issue start cycle and
end cycle.

ATRT_GUI.m file should be run to get the report analysis/Output. Once GUI
is started, window showed in Figure 14. Pops up which will have the browse option
for both excel sheet and mat files. List of mat files are shown in figure 15.
 Figure 14. Path configuration to fetch .mat files and Issue
list.
Figure 15. List of .mat files in given path
 Figure 16. Issue list with issue start and end cycle for
BSD false positive

Figure 17. Issue list with issue start and end cycle for
BSD false negative

Figure 18. Issue list with issue start and end cycle
for

RCTA false positive

 Figure 19. Issue list with issue start and end cycle for

LCA false positive

Figure 16, 17, 18, and 19. shows the excel sheet of issue list with features,
classification, reference sheet, base recording name (where mat files name should be
same as entered), sensor (which side the sensor is placed in the car i.e., rare right, rare
left, front right or front left), issue start cycle (The cycle where the false positive or
false negative alert has been started) and issue end cycle (The cycle where the false
positive or false negative alert has been stopped) of blind spot detection false positive,
blind spot detection false negative, Rare cross traffic alert false positive, and Lane
change assistance false positive respectively.
Figure 20. shows the window after browsing the addresses of both mat file
and excel sheet. Once start button is pressed entire execution starts and the status
changes into RUNNING in the place of START. After completion of execution the
status changes back to START.
 Figure 20. GUI window during execution.

Figure 21. Issue list with updated output after simulation for
BSD false positive.
 Figure 22. Issue list with updated output after simulation
for BSD false negative.

Figure 23. Issue list with updated output after simulation for
RCTA false positive.

Figure 24. Issue list with updated output after simulation for
LCA false positive.
Once the issue start cycle and end cycle is read from excel sheet and
checked with the base recording as explained in section 5.1. Output will be updated as
OK or NOK according to the changes occurred as shown in Figure 21, 22, 23, and 24.
For all the OK cases, comment will be updated as Solved with the prefix of that
feature to it. Similarly, for NOK cases, Reoccurred with the prefix of that feature. For
example, for the 40th test case in figure 23, the test status is OK hence the comment is
RCTAFP Solved. For the 41st test case in figure 23, test status is NOK and hence the
comment is RCTAFP Reoccurred.
 Figure 26 and figure 27 are given to compare the time taken to analyze and
generate the report.

Figure 25. Time taken to complete analysis

using “Recording analysis tool”

Figure 25. is the result generated after executing the recording analysis tool
that is discussed in chapter 5, which is the existing tool to compare between old
software and new software changes. In this tool, entire mat files are compared i.e., in
order to check the false positive and false negative issue or any unusual changes that
occurs after changing the software is done by comparing two entire mat files (for all
the cycle array) which are named as before mat file and after mat file. Before mat file
is the mat file generated using older software and after mat file is the mat file
generated using new software.
Figure 26. Time taken to complete the analysis using

“Automated tool for regression testing”

 As we can see in the figure 26. The time taken to complete the analysis using
“automated tool for regression testing” is 280.72 seconds (approximately 5 minutes)
whereas the time taken to complete the analysis using “recording analysis tool” is
727.79 seconds (approximately 12 minutes) as shown in figure 25. This shows that
the time taken to compare two versions of software using ATRT tool is reduced to
61.43% of the time compared to the existing tool.

7. CONCLUSION
 An automated tool to compare the test results of two versions of software is
developed and successfully evaluated with actual data. This ATRT tool increases the
efficiency by reducing the time taken to evaluate the issues by 61.43% as compared to
previous tool. This tool also maintains the list of previously captured issues, so
verifying this database with new software release will ensure the old reported issues
are not recurring with new update.

8. FUTURE SCOPE
 Utility of this tool can be enhanced in the future by analyzing other issues
such as ghost object detection, late warning and track drop, apart from false positive
and false negative issues.

REFERENCES
1. Vipin Kumar Kukkala, Jordan Tunnell, Sudeep Pasricha, and Thomas
Bradley, “Advanced Driver-Assistance Systems A path toward autonomous
vehicles.,” IEEE Consum. Electron.Mag., vol. 7, no. 5, pp. 18–25, 2018.
2. Chintaka Premachandra, Shohei ueda, and Yuya Suzuki, ‘‘Detection and
Tracking of Moving Objects at Road Intersections Using a 360-Degree
Camera for Driver Assistance and Automated Driving,’’ IEEE Access, vol. 8,
pp. 135652 - 135660, 2020.
3. “Radar Technology” – [online] Available: http://veoneer.prod-mid-
euw3.investis.com/en/radar
4. G. Liu, L. Wang and S. Zou, “A radar-based blind spot detection and warning
system for driver assistance,”2017 IEEE 2nd Advanced Information
Technology, Electronic and Automation Control Conference (IAEAC), 2017,
pp. 2204-2208.
5. Hongjun Lee, Moonsoo Ra and Whoi-Yul Kim, ‘‘Nighttime Data
Augmentation Using GAN for Improving Blind-Spot Detection,’’ IEEE
Access, vol. 8, pp. 48049 - 48059, 2020.
6. ISO. (1 May 2008). “Intelligent transport systems — Lane change decision aid
systems (LCDAS) — Performance requirements and test procedures”.
[Online] Available: ISO 17387 - Intelligent transport systems — Lane change
decision aid systems (LCDAS)Performance requirements and test procedures |
Engineering360 (globalspec.com).
7. MDIP – SENSORS (2021). “A Sensor Fused Rear Cross Traffic Detection
System Using Transfer Learning”. [online] Available: Sensors | Free Full-Text
| A Sensor Fused Rear Cross Traffic Detection SystemUsing Transfer
Learning (mdpi.com).
8. Shizhe Zang, Ming Ding, David Smith, Paul Tyler, Thierry Rakotoarivelo, and
Mohamed Ali Kaafar, “The Impact of adverse weather conditions on
autonomous vehicles.,” IEEE Vehicular tech. Mag., vol. 14, no. 2, pp. 103 -
111, 2019.
9. X. J. Yang, S. Quan Xv and F. J. Liu, "Research on Typical Faults of Active
Safety Components of Passenger Cars," 2020 4th Annual International
Conference on Data Science and Business Analytics (ICDSBA), pp. 166-168,
2020.
 10. Veoneer. (3 February 2020) – Danview data and CVW file format.
[Confidential].
11. Muhammad Hassnain and Anjum Abbas, “A Literature Survey on Evaluation
and Comparison of Regression Testing Techniques for Web Applications.,”
ITEE Journal., vol. 3, no. 3, 2014.
12. Inhwa Choi and Wonshik Na (2018), “Construction of a Regression Test
Automation System.,” [online] Available: (PDF) Construction of a Regression
Test Automation System (researchgate.net).
13. Shruti Malve and Pradeep Sharma, “Investigation of Manual and Automation
Testing using Assorted Approaches.,” IJSRCSE Research paper., vol. 5, no. 2,
pp. 81 – 87, 2017.
14. MATLAB Documentation - MathWorks India
15. Gregory L. Charvat (2014), “Small and Short-Range Radar Systems” [online]
Available: Small and Short-Range Radar Systems | Gregory L. Charvat |
Taylor & Fr(taylorfrancis.com)
16. Veoneer (22 August 2019) - Algorithm for Features [Confidential]
17. Radar Systems - Matched Filter Receiver – [Online] Available: Radar Systems
- Matched Filter Receiver (tutorialspoint.com)

APPENDIX A:

Plagiarism Report
APPENDIX B:

Program Outcomes (Pos) of the M. Tech Program:

1. Scholarship of Knowledge: Acquire in-depth knowledge of specific discipline or
professional area, including wider and global perspective, with an ability to
discriminate, evaluate, analyse and synthesise existing and new knowledge, and
integration of the same for enhancement of knowledge.

2. Critical Thinking: Analyse complex engineering problems critically, apply

independent judgement for synthesising information to make intellectual and/or
creative advances for conducting research in a wider theoretical, practical and policy
context.

3. Problem Solving: Think laterally and originally, conceptualise and solve

engineering problems, evaluate a wide range of potential solutions for those problems
and arrive at feasible, optimal solutions after considering public health and safety,
cultural, societal and environmental factors in the core areas of expertise.

4. Research Skill: Extract information pertinent to unfamiliar problems through

literature survey and experiments, apply appropriate research methodologies,
techniques and tools, design, conduct experiments, analyse and interpret data,
demonstrate higher order skill and view things in a broader perspective, contribute
individually/in group(s) to the development of scientific/technological knowledge in
one or more domains of engineering.

5. Usage of modern tools: Create, select, learn and apply appropriate techniques,
resources, and modern engineering and IT tools, including prediction and modelling,
to complex engineering activities with an understanding of the limitations.

6. Collaborative and Multidisciplinary work: Possess knowledge and

understanding of group dynamics, recognize opportunities and contribute positively to
collaborative-multidisciplinary scientific research, demonstrate a capacity for self-
management and teamwork, decision-making based on open-mindedness, objectivity
and rational analysis in order to achieve common goals and further the learning of
themselves as well as others.

7. Project Management and Finance: Demonstrate knowledge and understanding of

engineering and management principles and apply the same to one’s own work, as a
member and leader in a team, manage projects efficiently in respective disciplines and
multidisciplinary environments after consideration of economical and financial
factors.

8. Communication: Communicate with the engineering community, and with society

at large, regarding complex engineering activities confidently and effectively, such as,
being able to comprehend and write effective reports and design documentation by
adhering to appropriate standards, make effective presentations, and give and receive
clear instructions.

9. Life-long Learning: Recognize the need for, and have the preparation and ability
to engage in life-long learning independently, with a high level of enthusiasm and
commitment to improve knowledge and competence continuously.

10. Ethical Practices and Social Responsibility: Acquire professional and

intellectual integrity, professional code of conduct, ethics of research and scholarship,
consideration of the impact of research outcomes on professional practices and an
understanding of responsibility to contribute to the community for sustainable
development of society.

11. Independent and Reflective Learning: Observe and examine critically the
outcomes of one’s actions and make corrective measures subsequently, and learn from
mistakes without depending on external feedback.

APPENDIX C:
Research Publications