DEGREE PROJECT IN VEHICLE ENGINEERING,

SECOND CYCLE, 30 CREDITS

STOCKHOLM, SWEDEN 2020

Connected Tyres
Real-time Tyre Monitoring System for Fleet and
Autonomous Vehicles with Tyre Wear Estimation
through Sensor Fusion

XINYU LIN

MINGZHE HE

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF ENGINEERING SCIENCES
Connected Tyres

Real-time Tyre Monitoring System for Fleet

& Autonomous Vehicles with Tyre Wear
Estimation through Sensor Fusion

XINYU LIN
MINGZHE HE

Master in Vehicle Engineering

Date: October 27, 2020
Supervisor: Lars Drugge
Examiner: Lars Drugge
School of Engineering Sciences
Host company: Robin Worth, Volvo Cars
Swedish title: Anslutna Däck
 i

Acknowledgement
We would like to thank those who have helped and supported us to com-
plete the master thesis. This has been an unforgettable experience for us.
Firstly, we give our sincere gratitude to Volvo Cars where we have been
offered the opportunity to carry out this Master’s Thesis. This was possible
thanks to Robin Worth, our Supervisor. It also includes the whole Vehicle
Motion and Climate department for providing what we needed. Moreover, we
specially thank Prakhar Tyagi for being our technical supporter who helped us
during the entire process of our thesis.
Secondly, we are grateful to KTH for an amazing international study period
during the last two years. We thank Lars Drugge, our academic supervisor and
examiner, for his valuable advice during the Master’s thesis. Furthermore, to
School of Engineering Science, we appreciate those professors and assistants
who have taught and helped us during our master’s program at KTH.
Lastly, we specially thank our family and friends in China, who have been
a constant support and for providing with everything.
Thank you all!
ii

Abstract
Tyres are one crucial part for vehicles, as they are the only contact point
between the vehicle and the road. Intelligent tyres are a trending new subject
in the tyre industry. They are designed to monitor various tyre states and send
this information to both drivers and remote servers. The master thesis focuses
on the proposal of a real-time tyre monitoring system for fleet and autonomous
vehicles. It includes developing a tyre wear model and analysis of the current
tyre pressure monitoring functionality by leveraging the connectivity of fleet
vehicles equipped with a Volvo web cloud service. The tyre wear model in-
directly monitors the tread depth of the vehicles all four tyres by identifying
characteristics between worn and fresh tyres. The two characteristics are iden-
tified by monitoring and analyzing vehicle speed and braking signals. The two
characteristics is input to a voting scheme which decides when a worn tyre is
detected. The test vehicle was a Volvo XC40 with three types of tyres: win-
ter tyres, summer tyres and worn summer tyres. The wear model gives 90 %
accuracy to 10 set of test data, randomly selected from all dataset at Hällered
Proving Ground (Sweden). The connectivity realizes the data transmission
from the raw data of onCAN and FlexRay signals stored in a Volvo web cloud
service to the tyre monitoring fleet system. The signals are filtered and re-
sampled, leaving the required signals of the tyre pressure monitor system and
the tyre wear model. Two signals, Calibration Status and iTPMS Status, are
used to perform a statistical analysis on tyre pressure by categorizing the cal-
ibration status and the tyre pressure conditions.The project outcome is an in-
terface built on MATLAB GUI for demonstration of vehicle identification and
tyre health conditions, with the embedded tyre wear model and connectivity.
 iii

Sammanfattning
Däck är en viktig del för fordon, eftersom de är den enda kontaktpunkten
mellan fordonet och vägen. Intelligenta däck är ett trendigt nytt ämne i däckin-
dustrin. De är utformade för att övervaka olika däcktillstånd och skicka denna
information till både förare och fjärrservrar. Examensarbetet är inriktat på ett
förslag till ett däckövervakningssystem i realtid för fordonsflottor och autono-
ma fordon och inkluderar en däckslitagesmodell och anslutning. Det inklude-
rar att utveckla en slitagemodell och analys av den aktuella däcktrycksövervak-
ningsfunktionen genom att studera Volvos fordonspark som är utrustade med
Volvos webbmolntjänst. Däckens slitagemodell övervakar indirekt slitbaned-
jupet på alla fyra däck genom att identifiera egenskaper mellan slitna och nya
däck. De två egenskaperna identifieras genom att övervaka och analysera for-
dons hastighet och bromssignaler. De två egenskaperna är inmatade i ett röst-
ningsschema som avgör när ett slitet däck upptäcks. Testfordonet var en Vol-
vo XC40 med tre typer av däck, vinterdäck samt nya och slitna sommardäck.
Modellen ger 90 % noggrannhet för 10 uppsättningar testdata, slumpmässigt
valda från alla dataset på Hällered provbana (Sverige). Anslutningen genom-
för dataöverföringen av rådata från onCAN och FlexRay-signaler lagrade i
en Volvos webbmolntjänst till däcksövervakningssystemet. Signalerna filtreras
och samplas på nytt för att skapa de nödvändiga signalerna till däcktrycksö-
vervakningssystemet och däckslitagemodellen. Två signaler, kalibreringssta-
tus och iTPMS-status, används för att utföra en statistisk analys av däcktrycket
genom att kategorisera kalibreringsstatus och däcktrycksförhållanden. Projek-
tets resultat är ett gränssnitt byggt på MATLAB GUI för demonstration av
fordonsidentifiering och däcktillstånd. med inbäddad däckslitagemodell och
anslutning.
Contents

1 Introduction 1
1.1 Problem statement . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Research significance and motivation . . . . . . . . . . . . . . 5
1.3 Research objectives . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Tyre Wear Model 8

2.1 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.1 Test site . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Test vehicle . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.3 Test tyres . . . . . . . . . . . . . . . . . . . . . . . . 10
2.1.4 Test plan . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Features of worn tyres . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Speed analysis . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Braking analysis . . . . . . . . . . . . . . . . . . . . 15
2.2.3 Frequency analysis . . . . . . . . . . . . . . . . . . . 20
2.3 Tyre tread depth indicator . . . . . . . . . . . . . . . . . . . . 20

3 Connectivity 22
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 WICE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3 Measured signals . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Data processing . . . . . . . . . . . . . . . . . . . . . . . . . 23

4 User interface 25
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 UI design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2.1 Main interface . . . . . . . . . . . . . . . . . . . . . 26
4.2.2 Interface of statistics on tyre pressure . . . . . . . . . 26

iv
 CONTENTS v

5 Results 28
5.1 Test on TWM . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 Trial run of RTMS . . . . . . . . . . . . . . . . . . . . . . . 29
5.2.1 RTMS . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2.2 Result of trial run . . . . . . . . . . . . . . . . . . . . 30

6 Discussion 33
6.1 Tyre Wear Model . . . . . . . . . . . . . . . . . . . . . . . . 33
6.2 Connectivity and UI . . . . . . . . . . . . . . . . . . . . . . . 34

7 Conclusions and Future Work 36

7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

References 39

8 Appendix 42
8.1 Abbreviation list . . . . . . . . . . . . . . . . . . . . . . . . . 42
Chapter 1

Introduction

Tyres are essential for vehicles, as tyres are the only component that contacts
the roads. The longitudinal force generated by tyres affects the accelerating
and braking performance, while the lateral force, also called cornering force,
offers the capacity to resist sliding sideways when in a corner. Besides, tyres
support the load of the vehicle and absorb shocks when driving in uneven
roads, which improves the ride comfort. Therefore, the quality of tyres has a
crucial impact on the driving experience.
In fact, a lot of research have been conducted on modeling of tyres. As a
result, the linear model, brush model and magic formula have been proposed
for many years and they are very typical and useful tools to study the charac-
teristics of tyres. For example, the magic formula is a classic mathematical
model of tyres, which only uses one set of formulas to completely express the
mechanical characteristics of tyres under pure working conditions [1]. With
sensor fusion, the tyre-road contact forces are possible to be estimated in elec-
tric control units (ECUs) [2]. One of the most important systems, the tyre
pressure monitoring system (TPMS) is installed in the vehicle to monitor the
tyre pressure and temperature. TPMS is able to monitor the tyre condition
in real time through electronic sensors installed inside the tyre. An alert is
given to the driver when TPMS detects low pressure of any tyre. The National
Highway Traffic Safety Administration, U.S. Department of Transportation,
stipulates that there must be an alert from TPMS for the driver when the tyre
pressure is less than 75 % of the standard value [3]. The tyre manufacturer sug-
gests that when the tyre pressure drops by 30 % of the standard value, TPMS
needs to give an alarm to remind the driver.
TPMS has proved as a very effective method to reduce the tyre-related
accidents over the last decade. It is a standard configuration for many countries

1
2 CHAPTER 1. INTRODUCTION

and regions now. For instance, U.S. law requires that all passenger cars and
light trucks sold in the U.S. must be equipped with a TPMS since August 2007.
Shortly after, European and Chinese regulations have also been issued since
November 2014 and January 2020 respectively. Nowadays, almost every new
car is equipped with a TPMS.
Up till now, two kinds of TPMS are installed in the vehicles. They are
direct TPMS (dTPMS) and indirect TPMS (iTPMS) respectively.
The dTPMS uses a pressure sensor, usually located on the inside wall of the
valve, to measure the pressure of each tyre. Then the tyre pressure information
is transmitted to ECUs by a wireless transmitter. With temperature sensors,
the tyre temperature information can be obtained as well. Finally, pressure
and temperature of every tyre are displayed in the user interface of the vehi-
cle. When any tyre has the characteristic of excessively abnormal pressure or
temperature, there will be a warning signal in the car central control interface.
As the dTPMS uses extra hardwares to measure the tyre pressure, it in-
creases the vehicle production cost. Therefore, the vehicle companies prefer
iTPMS, which does not require extra sensors, but only some algorithms. There
have been a lot of research and study projects on iTPMS. The iTPMS relies on
two parts: vibration analysis and wheel radius analysis [4]. Vibration analysis
monitors the vibration frequency from the torsional spring-damper systems
and detects abnormal values of the frequency. Wheel radius analysis uses the
fact that the effective rolling radius of the wheel becomes smaller if the tyre
pressure is low. Applied with a voting scheme, two feature analyses have made
iTPMS reliable.
Despite the wide application of TPMS, among other safety centric features,
there are still 1.35 million death in traffic accidents every year. Between 20 and
50 million people are injured in different serious situations according to the
statistic on road traffic injuries of World Health Organization [5]. Road safety
is still a concern and needs to be improved. It is necessary and urgent to look
for more effective and creative methods to monitor more factors of tyres, such
as tyre wear conditions. Besides, the automotive industry is undergoing a rev-
olution. The most frequently-mentioned changes are: autonomous, connected,
electric and shared (ACES) [6]. In order to keep up with the industry trends
in this transformation, many car companies and even Internet companies have
already moved towards the four new directions.
During the Corporate Side Presentations at the 2019 Fleet Europe Sum-
mit, the experts working on connectivity and mobility from all over the world
presented their latest ideas and future visions [7]. Yaël Bennathan, the Head of
Arval Mobility Observatory, says that all new cars produced by 2025 will be
 CHAPTER 1. INTRODUCTION 3

connected [7]. As more and more cars are becoming connected, it promotes
the development of connectivity of tyres. In fact, many well-known tyre com-
panies have already conducted several studies and experiments on this topic
and some cases of connected tyres have already been put into practice.
The ContiConnect, shown in Figure 1.1, designed by Continental monitors
tyre information with built-in sensors. The data are transmitted to a web portal
through a yard reader station. The yard reader station is a communication
bridge between data from the tyre sensors and back-end. The data transmission
can only happen when there is a yard reader station. Usually the yard reader
stations are distributed in gas stations and some other public places where cars
often go. Otherwise, the driver can use the hand-held tool to import tyre data
to the web portal when no yard reader stations are nearby. The drivers can
enter the web portal to see the tyre conditions of their vehicles using their
smartphones and computers. The drivers will be informed by email and text
message with some recommended measures when there is great risk of tyre-
related breakdowns [8].

Tyre sensor Yard reader station

Hand-held tool
Web portal

Figure 1.1: ContiConnect components [8].

The intelligent tyre system developed by Hyundai Motor Co, Mando Corp
and Corechips uses the sensors mounted in the inner liner of tyres to measure
the tyre deformation, which is transmitted to the surface acoustic wave sensor
4 CHAPTER 1. INTRODUCTION

by passive radio communication. After that, the data are imported to the ECU
of the intelligent tyres and converted into tyre force information finally by a
force conversion algorithm [9].
Michelin developed Pilot Sport Cup 2 tyres with Track Connect technology
in 2018. The tyre pressure and temperature are measured by the sensors placed
in each tyre. The measured data are transferred to the vehicle by a radio signal
and then transmitted to a phone by Bluetooth [10]. The data are visualized in
the interface of the APP in the end so that the drivers can read the information
easily. This technology tells the driver the information about the tyre pressure
and temperature all the time during driving. Besides, it is also possible to tell
the driver’s driving behavior in the case of turning by analyzing the pressure
balance between front and rear tyres [10].
Similarly Pirelli once tried to introduce a mobile APP Connesso to display
the data gathered from senors. Now the company is working on Cyber Tyres
system, shown in Figure 1.2, combined with 5G technology as 5G improves
the data transmission speed greatly. The special sensor is mounted inside the
wheel and a radio transmitter inside is used to transmit the data to other elec-
tronic devices in the car [11]. With the Pirelli Cyber Tyre system, it is not only
possible to monitor the basic signals from tyres, but also has the potential to
identify the tyre grip, tread wear and the road information.

Figure 1.2: Pirelli’s Cyber Tyres describe the road surface [12].
 CHAPTER 1. INTRODUCTION 5

1.1 Problem statement

Nowadays, connected tyres are applied into practice on some vehicles, most
being buses and heavy duty trucks. One reason is that because losses and
injuries caused by accidents related to buses and trucks are more serious than
passenger cars. The other reason is the cost of connected tyres. As equipped
with data transmit devices and sensors in tyres, the cost of connected tyres is
much higher than normal tyres.
Meanwhile, the autonomous driving is one popular subject. In order to
guarantee safety, tyre health information of autonomous cars must be moni-
tored. Intelligent tyres on the market are one solution to establish connectivity
between cars and servers, but are often too expensive for car companies.

1.2 Research significance and motivation

Tyres with good tread depth can reduce risks of aquaplaning and decrease the
braking distance when driving in wet road surface [13]. Intelligent tyres use
built-in sensors to calculate all the information about the tyres so as to decrease
the tyre-related driving risk, such as each tyre pressure, temperature, wear con-
dition, traction force, friction potential and slip condition [14]. Furthermore,
connectivity of intelligent tyres can be used to collect road information and
driving behavior. Through the fleet system, the user can monitor all tyre infor-
mation of various cars remotely. The data obtained from sensors can be used
for different organizations. For example, relevant government departments
can use the road information to monitor the urban road surface condition. Ex-
perts can optimize autonomous driving by analyzing the information related
to drivers’ driving behavior. These users do not need to go to testing grounds
and do tests by themselves; instead, they can sit in their offices or their labo-
ratory and extract the data they need from the fleet system. The data can be
from every vehicle with connected tyres.
To keep up with the development of autonomous driving and Internet of
Things (IoT), it is necessary to develop more advanced tyre monitoring sys-
tems.

1.3 Research objectives

This thesis aims to develop a real-time tyre monitoring fleet system. The infor-
mation is not limited to tyre pressure, but also including tyre wear condition.
6 CHAPTER 1. INTRODUCTION

Additionally, no extra sensors is needed for saving cost. In details, it consists

of three parts, the tyre wear model (TWM), the connectivity and the real-time
tyre monitoring system (RTMS).

• TWM: It indirectly estimates the tread depth of four tyres through iden-
tifying certain features of worn tyres. A warning indicator will be given
if worn tyres are detected. The two features are speed difference between
tyres and braking slip difference between worn and new tyres.

• Connectivity: Within Wireless Information Collection Environment (WICE),

data are collected and sent from vehicles to company portals through a
WICE Communication Unit (WCU). With data pre-processing, signals
are filtered and re-sampled for TWM and other applications.

• RTMS: The two parts, TWM and connectivity, are embedded in this
system. A user interface (UI) is designed by MATLAB GUI to display
tyre health information of vehicles on roads, including tyre pressure and
wear condition.

1.4 Thesis outline

The thesis is organised into the following chapters:

• Chapter 1 introduces the background and research significance of this

topic. Besides, it explains the framework of the report to give readers a
rough understanding of the whole report.

• Chapter 2 explores the possibility of indirectly detecting worn tyres

on a vehicle. It includes data analysing of two tests and how TWM is
constructed.

• Chapter 3 presents the data transmission of the on-board signals through

WCU.

• Chapter 4 demonstrates the UI design of RTMS as well as a data anal-

ysis interface of tyre pressure.

• Chapter 5 illustrates RTMS with embedded TWM and connectivity.

An example of RTMS is shown with the designed UI.

• Chapter 6 discusses the performance of RTMS appropriately. Further-

more, the results of the system verification test are discussed.
 CHAPTER 1. INTRODUCTION 7

• Chapter 7 summarises the background and results. This chapter also

highlights what can be done about RTMS in the future.
Chapter 2

Tyre Wear Model

In this chapter, the focus will be on the Tyre Wear Model. Beginning with the
tests that have been done, this chapter will show how each feature is identified
and calibrated.

2.1 Approach
According to the literature research in Chapter 1, tyre manufacturers are putting
lots of resources into intelligent tyres. However, the cost of those tyres are high
due to extra sensors and ECU installed within them. The possibility that the
tyre state can be indirectly estimated has not been fully explored. This section
will focus mostly on the wear degree of tyres. It only requires the sensors that
already exist in the vehicle. Tests have been done to identify such possibility.

2.1.1 Test site

There were two tests in total, one on February 26th and the other on May 19th
and 20th in 2020. The tests were done at Hällered Proving Ground. The first
test on February 26th was a sunny day during winter around 5 ◦ C and the tracks
were dry. The first day of the second test was sunny around 8◦ C but just after
raining. The tracks were wet during the morning and became dry in the after-
noon. The second day was sunny around 14◦ C and the tracks were dry. A top
view of Hällered Proving Ground is shown in Figure 2.1.

8
 CHAPTER 2. TYRE WEAR MODEL 9

Figure 2.1: Overview of Hällered Proving Ground[15].

In order to identify the feature of worn tyres, the following tracks were chosen:
• Track I: Country Road Track
This track contains a long straight road, where most of tests were done.
This represents one usual driving scenario, starting from standstill, driv-
ing straight with a constant speed and then braking until stop.

• Track II: High Speed Track

The track is an oval track with four lanes, where a car can run up to 250
km/h at the outside lane for licensed drivers. It imitates running on a
highway with high speed.

• Track III: Handling Track II

The handling track is a complicated track. It has uphills and downhills,
turning and uneven surfaces. This track shows the driving experience
within mountain areas, which is common in Sweden.

2.1.2 Test vehicle

A Volvo XC40 (Figure 2.2) was used as the test vehicle for both tests. The
test vehicle was an on-going Volvo project for testing. No modifications were
made to the test vehicle except the tyres.
10 CHAPTER 2. TYRE WEAR MODEL

Figure 2.2: Test Vehicle [16].

The data collection were done through CANoe and FlexRay. CANoe is a
comprehensive software tool for development, test and analysis of individual
ECUs and entire ECU networks. It supports network designers, development
and test engineers throughout the entire development process from planning to
system-level test [17]. FlexRay is a scalable, flexible high-speed communica-
tion system that meets the challenges of growing safety-relevant requirements
in the automobile [18].
For the test vehicle XC40, the CANoe module for transmitting bus signals
was positioned underneath the driver’s seat. After connecting the two cables
named as FlexRay from the module and Vector VN7640, the access to bus
signals was established on the laptop. In the interface of CANoe, the database
of FlexRay was imported and chosen to be recorded from the real-time signals.

2.1.3 Test tyres

The two worn tyres used for tests were Pirelli summer tyres, 245/40R21, with
the tread depth of 4.9 and 5.1 mm respectively. The threshold of worn tyres,
not worn-out, is hence set as 5 mm. According Swedish law, the legal re-
quirement for worn-out tyres is 3 mm for winter tyres and 1.6 mm for summer
ones[19]. During the first test, it was winter, so winter tyres with the same
radius and tread depth of 7.1 mm were mounted on the test vehicle for testing.
For the second test, four fresh tyres, which had the same tyre size and brand,
were ordered and tested with the worn tyres.
The cold tyre pressure for all tyres were controlled as 2.3 kPa, which was
 CHAPTER 2. TYRE WEAR MODEL 11

the recommended pressure for these tyres. Since tyre pressure can greatly
affect the performance of tyres, it is essential to control the tyre pressure as
the same for all test tyres. There is also no difference in tyre temperature
between the four tyres on the test vehicle.

2.1.4 Test plan

The test goal was to find the features of worn tyres compared to fresh tyres.
Hence, different tyre combinations were planned for testing. For the first test
on Feb 26th, the tests were carried out according to Table 2.1.

Table 2.1: Tyre group of the first test (0226).

Tyre Group Mounted Tyres

I (worn) 4.9 mm summer tyre on the front left with three 7.1 mm winter tyres
II (reference) four 7.1 mm winter tyres

For each tyre group, same tests were done based on

1. Straight Case Driving on Track I at 40, 60, 80 and 100 km/h respec-
tively with two trials.

2. Braking Case Hard braking until stop on Track I from 50 and 60 km/h
respectively with two trials.

3. Handling Case Driving on Track III at 40, 60, 80 km/h and free-run
respectively with two trials.

The second test happened on May 19th and 20th in order to verify what have
been found in the previous test. The procedure were similar as Table 2.2.

Table 2.2: Tyre group of the second test (0519 & 0520).

Tyre Group Mounted Tyres

I 5.1 mm tyre at front left with three fresh tyres
II 5.1 mm tyre at front left and 4.9 mm at front right with two fresh tyres
III 4.9 mm tyre at rear left with three fresh tyres
IV four fresh tyres

1. Straight Case Driving on Track I at 40, 60 and 80 km/h respectively

with two trials.
12 CHAPTER 2. TYRE WEAR MODEL

2. High-speed Case Driving on Track II at 100 km/h with two trials.

3. Braking Case Hard braking until stop on Track I from 40, 60 and 80
km/h respectively with two trials.

4. Handling Case Driving on Track III at 60, 80 km/h and free-run respec-
tively with two trials.

For simplicity, the tyre location is abbreviated as FL (front left tyre), FR (front
right tyre), RL (rear left tyre) and RR (rear right tyre).

2.2 Features of worn tyres

In this section, two features will be shown. The two features were first found
during the first test and then verified in the second one. In the first test, winter
and summer tyres were compared, but the difference between tyres were not
only tyre tread depth, but also tread patterns. The second test was a test with
control on variables. The difference is only the tyre tread depth. The findings
of both tests will be shown, but it should be noted that the first test comes with
a bias on tyres, as what was found in the first test was intentionally found.

2.2.1 Speed analysis

The first and most obvious difference between worn and fresh tyres is the speed
difference of four wheels during the first test. Figure 2.3 shows the speed sig-
nals of the vehicle, worn and fresh tyres in the straight driving cases.

40 km/h 60 km/h
 CHAPTER 2. TYRE WEAR MODEL 13

80 km/h 100 km/h

Figure 2.3: Speed signals during straight driving cases of first test (blue line
as 5.1 mm tyre, orange line as 7.1 mm tyre and yellow line as vehicle speed).

The blue line in Figure 2.3 is the speed signal of FL tyre, which is the worn
tyre. During the straight driving cases, the worn tyre has a much lower speed
than the other three tyres, as well as the vehicle speed. Since the worn tyre was
also a summer tyre while the other three winter tyres, it is unclear which factor
is more significant. During the second test, the same worn tyre is compared
with new tyres of the same model.
Figure 2.4 shows the speed signals of four wheels compared to the vehicle
speed.

Figure 2.4: Speed signals during straight case of second test.

However, a clear difference cannot be found in those cases, and thus the
14 CHAPTER 2. TYRE WEAR MODEL

slip of the four wheels are calculated as

rω − v
slip = ∗ 100%, (2.1)
rω
where r is the radius of tyres, ω is the wheel rotation speed and v is the vehicle
speed. Taking the median value of the slip, each test case has four values of
the median slip of the four wheels respectively. While driving at a constant
speed, the median slip of a wheel has a certain relationship with the vehicle
speed. Different speed requires different driving force on tyres, which results
in different slip value. Figure 2.5 shows the relationship between slip and
vehicle speed. The consistent difference between tyre groups is the slope of the
linear fit. For Group IV, which all four tyres are fresh, the slope is the minimal
as shown in Table 2.3. The slip of four tyres at one case are determined by the
median value of the data.

Front Left Front Right

Rear Left Rear Right

Figure 2.5: Median Slip vs vehicle speed during straight cases (dot as data
points and lines as linear fit).
 CHAPTER 2. TYRE WEAR MODEL 15

Table 2.3: Result of linear fit for straight cases.

Slope of linear fit

Group
FL FR RL RR
I 0.1592 0.1637 0.1604 0.1663
II 0.0135 0.0184 0.0182 0.0242
III 0.0169 0.0158 0.0153 0.0185
IV 0.0003 0.0052 0.0024 0.0080

The test vehicle of groups I, II and III are installed with worn tyres (5.1mm
or 4.9mm), demonstrating a higher slope compared to group IV, around 10
times larger. The maximal slope as group I implies that road condition would
affect the relationship between vehicle speed and slip, as tracks were wet just
after rain, which is mentioned in section 3.1.1. Overall, the straight cases of
the second test supports the speed feature found in the first test. The vehicle
with worn tyres tend to have a larger slip compared to that with fresh tyres.
Moreover, worn tyres, no matter the number and location of worn tyres, will
influence the slip of all four wheels. The mean value of slopes of four wheels
is thus used to determine whether a vehicle has worn tyres or not.

2.2.2 Braking analysis

So far, estimating tyre pressure has proven to be practical in vehicles, which
is called iTPMS. The tyre state however includes more than just tyre pressure.
According to a United States patent, tyre wear state can be estimated through
a similar method, wheel speed signal feature extraction [20]. In details, it uses
the existing speed signals of four wheels during braking and extracts a first
and second features from them, median slip ratio and median slip rate. Then,
a support vector data classification algorithm is applied to find the wear state
of tyres. Inspired by this patent tests are hence carried out in field.
v − rω
SlipRatio = ∗ 100% (2.2)
v
dSlipRatio SlipRatio(n) − SlipRatio(n − 1)
SlipRate = = (2.3)
dT ime time(n) − time(n − 1)

Using Eq 2.2 and 2.3, speed signals are converted into slip ratio and rate.
Next, the values of slip ratio and rate is plotted with the density function in
Figure 2.6. The density function is a function providing a relative likelihood
that the value of the random variable would equal that sample [21].
16 CHAPTER 2. TYRE WEAR MODEL

Slip ratio

Slip rate

Figure 2.6: Density function of slip ratio and rate for braking cases of first test.

The worn winter tyre in test 1 tends to have a peak shift to left and a higher
peak compared to the other tyres. In order to apply a classification algorithm,
the median value (location of peaks in Figure 2.6) of these two features are
extracted and plotted in Figure 2.7.
 CHAPTER 2. TYRE WEAR MODEL 17

Figure 2.7: System diagram showing features of tyres in test 1.

The data implies a linear discrimination between two different tyres. Com-
pared to fresh tyres, the worn tyre has a smaller median slip ratio and median
slip rate. However, due to the limited number of data points, the classification
algorithm cannot be determined. More tests are done in the second test, and
results are shown in Figure 2.8 with the same method.

Figure 2.8: System diagram for features of worn and fresh tyres in second test.

The data are categorised into worn and fresh tyres, corresponding to 4.9/5.1mm
and newly-ordered ones. The same pattern is found in the second test as the
18 CHAPTER 2. TYRE WEAR MODEL

first one. There exists a discrimination between worn and fresh tyres, and a
proper classification algorithm is then to be fixed.
Statistical learning, also known as machine learning, is a highly powerful
and common tool in modern world. For this braking feature classification,
supervised learning is a proper method. In other words, there is a set of vari-
ables that might be denoted as inputs, which are measured or preset. These
have some influence on one or more outputs. For each example the goal is to
use the inputs to predict the values of the outputs [22]. In this case, inputs are
the median slip ratio and median slip rate, while only one output is the wear
state of tyres.
The Classification Learner toolbox in MATLAB, as shown in Figure 2.9,
allows an easy way to perform supervised learning. With the input of features
extracted from braking cases, the toolbox gives results of different classifica-
tion algorithms.

Figure 2.9: Classification learner toolbox in MATLAB.

After performing training to all algorithms, the linear support vector ma-
chine (SVM) yields the highest accuracy as 89.6%. SVM uses a technique
called the kernel trick to transform the data and then based on these transfor-
mations it finds an optimal boundary between the possible outputs [23]. In
this model, the linear SVM means that the boundary is a straight line that cuts
features into two groups, where the line is determined by SVM. The confusion
matrix of linear SVM is shown in Figure 2.10. Each class has 24 data. The
 CHAPTER 2. TYRE WEAR MODEL 19

linear SVM accurately predicts 20 for fresh tyres and 23 for worn ones, which
in total is 43 out of 48. The accuracy is then found as 89.6%.

Figure 2.10: Confusion matrix of linear SVM.

The linear SVM model is then exported and saved in a mat file. It can be
used for analysis of more test data. The linear SVM gives the classification
equation Eq 2.4, which is shown in Figure 2.11. The crosses are the mis-
classified points, while dots represent the correctly-classified ones.

y = −1.498 ∗ x + 53.99. (2.4)

Figure 2.11: Test points with linear SVM.

20 CHAPTER 2. TYRE WEAR MODEL

2.2.3 Frequency analysis

One method used in iTPMS is the vibration analysis, which considers the rub-
ber of tyres that behave like springs [4]. The tyre pressure affects the resonance
frequency of tyres significantly, and thus is used to predict tyre pressure. The
tread depth implies the thickness of rubber in tyres, which is an influencing
factor to the resonance frequency.
During the analysis of test data, the speed signals of four wheels were con-
verted into frequency domain through fast Fourier transform (FFT). The res-
onance frequency of tyres could be identified as the frequency where there
appears to be peaks. However, after two tests, this feature is found to be irrel-
evant to tyre tread depth, but to the tyre itself.

Figure 2.12: Plot of frequency response during the straight cases of test 1.

The blue line in Figure 2.12 represents the worn winter tyre during test 1.
Compared to the other three tyres, it has a lower peak frequency and ampli-
tude, which is also found in other cases in test 1. However, the worn tyres in
test 2 shows no such feature. Hence, it sums that the difference of frequency
peaks is only related to the tyre, not tyre tread depth.

2.3 Tyre tread depth indicator

In this section, the tyre tread depth indicator is constructed based on two fea-
tures in Section 2.2. Figure 2.13 shows the flow chart of the tyre tread depth
indicator. The function of the indicator is written in MATLAB, taking in data
 CHAPTER 2. TYRE WEAR MODEL 21

collected through CANoe. The output is an indicator showing whether any

tyre is worn or not.

Figure 2.13: Flow chart of tyre tread depth indicator.

For each feature analysis, the signals of four tyres are compared to the
reference data (tyre group IV) separately. The speed analysis compares the
slope of the linear fit between median slip and vehicle speed, while the braking
analysis uses the linear SVM for classification. The output of both features is
a flag ,ti (i = 1, 2), representing the result as

• ti =-1: Not enough data

• ti =0: No worn tyres detected

• ti =1: Worn tyre(s) detected

The voting scheme sums the two flags and the indicator shows different
colors for different situations. If either flag is negative (not enough data), the
indicator is gray, meaning the function unavailable. The sum of the two flags
then corresponds to green and yellow indicators based on

• Gray t1 < 0 or t2 < 0

• Green t1 + t2 = 0

• Yellow t1 + t2 > 0
Chapter 3

Connectivity

WICE collects vehicle signals through the dedicated component mounted on

company cars and saves the data in the cloud. In this chapter, the raw data
stored in WICE cloud are processed for further application in the UI. Connec-
tivity between raw data and the back-end is realised by python scripts.

3.1 Introduction
To collect vehicle signals and transmit the data to the fleet and other appli-
cations in real time, it is achieved by telematics technologies in most cases.
Nowadays this function is realized primarily by wireless communication, such
as trunked radio and cellular communication [24].
Tyre companies have started to use Radio Frequency Identification (RFID)
tyre tags to trace the tyre in its whole life. When a tyre with the RFID tag pass
through the reader areas, the tyre information and its data will be transmitted
to the fleet so that the operator can understand the tyre’s lifecycle better. Com-
pared with wireless local area network (WLAN), the RFID is more suitable for
large areas where the wireless network cannot completely cover. Therefore, it
is more widely applied [25]. Besides, bluetooth is also applied in data trans-
mission. Although it costs very little, it is limited by the transmission distance
to a high degree. So its application is not very wide.
Thanks to the rapid development of mobile communications, it provides
the opportunity to reach faster and more large-scale data transmission as long
as the place is covered by antennas coat cell towers [26]. Especially because
5G greatly increases the speed of data transmission, the application in aspect
of cellular communication is becoming more and more popular.

22
 CHAPTER 3. CONNECTIVITY 23

3.2 WICE
WICE is a system which collects vehicle signals in real time. The data of
measured signals are collected by the external Wireless Communication and
Data Acquisition Units installed in vehicles. The system supports the testing
and validation stages of automotive development by telematic technology, the
wireless local area network in the company in most cases. Each car is config-
ured with a Wireless Communication Unit (WCU), the special device to handle
the communication between the car and the portal. With the GSM/GPS/WiFi
antenna inside the car, the big data are stored in a network address.
To monitor one signal, the signal name should be added to the signal reader
through the WICE portal at first. The signal reader is a measurement module
that allows monitoring and logging of vehicle signals. Once the signals exist
in the signal reader, they will be recorded and saved automatically. The data
for every vehicle in every driving cycle will be downloaded and stored as a sy
data file in the network address. Every day a new folder which is named by
the date is created to store the data files.

3.3 Measured signals

In this thesis, all the test cars are equipped with iTPMS. The signals used to
indicate the status of tyre pressure and calibration are required to be measured.
The signals are shown in detail in Table 3.1. All the signals are measured from
the company cars mostly in Gothenburg area.

Table 3.1: Measured signals of iTPMS.

Signal names Description

CalSts Indicating calibration status
iTPMSSts Indicating tyre pressure warning or iTPMS malfunction

3.4 Data processing

The raw data consist of all signals in the signal reader. The data of redundant
signals should be removed at first. Another problem is that the sampling time
is not constant. Consequently, data processing is needed before utilising the
data. Then the processed data can be used for further application for different
24 CHAPTER 3. CONNECTIVITY

purposes. The data processing is done by python scripts. The data transmis-
sion process is shown in Figure 3.1.

Figure 3.1: Flow chart of data transmission.

The data processing is consisted of three parts, as listed:

• Signal filtering
The signals are filtered by specifying signal names so that only the data
of specific signals will be handled later.

• Time vector re-sampling

The sampling time is adjusted to be 1 second so that the data are recorded
at fixed time interval and the redundant data will be removed.

• Data extraction
Processed data are saved in an excel or a mat file in the local computer
for further application in MATLAB.

Through the three steps of data processing, only data of needed signals will
be saved in the new data file, which can be read by MATLAB directly. Ac-
cordingly, the connectivity between vehicle signals and back ends is achieved.
Chapter 4

User interface

The UI is a platform to demonstrate RTMS. It allows an easy interaction be-

tween users and RTMS, which is realised by categorising all date files by date
and visualising data of the chosen file. In addition, tyre pressure conditions
and calibration status are analysed in this chapter, such as the frequency statis-
tics of different scenarios in a month.

4.1 Introduction
The UI helps the user to extract desired information and visualises the infor-
mation. The information is displayed in a certain pattern to allow people to
get the information which is relevant to their goals quickly [27]. While de-
signing a UI, there are many aspects that need to focus on, which are essential
to make the UI concise, such as the font size, the graphics and placement of
text and images, the colors and the page layout [28]. To create a good UI, its
function must match its design and it must have a clear interface and quickly
guide users.

4.2 UI design
In this thesis, as RTMS is based on MATLAB, the UI is also created by the
MATLAB GUI tool. The UI is designed to visualise the information including
tyre health conditions and vehicle identification, such as tyre pressure con-
ditions, calibration status, tyre wear estimation results, car alias and car ID.
Different colours are used to present different conditions, which is noted in
main interface. The UI has two interfaces: main interface and the interface of
statistics on tyre pressure. In the main interface, the operator can check any

25
26 CHAPTER 4. USER INTERFACE

document at any time to see its result. In the interface of statistics on tyre pres-
sure, tyre pressure conditions and calibration status of all cars in a month are
classified. The frequency of each scenarios is calculated and recorded. The
results are shown in line graphs, pie charts and tables in detail.

4.2.1 Main interface

Main interface is shown in Figure 4.1. There are mainly four colours which are
used to indicate different conditions for tyre pressure conditions, calibration
status and tyre wear estimation results. When the UI is opened, the operator
clicks the Log button and then choose the specific date. The filtered data files
with CalSts and iTPMSSts signals appear in the Datafiles list. When one data
file is clicked, the vehicle identification and tyre health conditions information
will be displayed automatically in the corresponding places. The switch of
data file under the same date can be achieved effortlessly by clicking another
data file. Similarly, it is straightforward to choose another date.

Figure 4.1: Main interface of UI for RTMS.

4.2.2 Interface of statistics on tyre pressure

Analogously in Figure 4.2, the list in the bottom left of the interface corre-
sponds with the month list in main interface. Once the Start button is clicked,
the month will be chosen accordingly. Furthermore, there are two line graphs
 CHAPTER 4. USER INTERFACE 27

and two pie charts. As for the line graphs, the top one is used to record the num-
ber of vehicles with iTPMSSts signals as well as the total number of vehicles.
The other one represents CalSts signals. Similarly, the pie charts demonstrate
the proportion of different signal values of the cars with those two signals in
the selected month individually. As some parts of pie charts are pretty small,
resulting that the values of the percentages coincide in the same place, all val-
ues are noted in the left of the pie charts so that users can see the values clearly.
If the user wants to see the results of another month, it can be done by changing
the month in main interface and then click Start button in the new data analysis
interface, while the previous interface will not be closed. By comparing the
results of two or more months, the user can obtain the total number of vehicles
with measured iTPMS signals and analyse the frequency of calibration.

Figure 4.2: The interface of statistics on tyre pressure.

Chapter 5

Results

Tests to validate the TWM and RTMS are conducted in this chapter.

5.1 Test on TWM

The accuracy of TWM is tested with the test data of the second test. In order
for two features to work at the same time, three test cases of the same trial
number are combined; in other words, there are totally eight groups of test
data (I.1, I.2, II.1, II.2, III.1, III.2, and IV.1, IV.2), as explained in Table 2.2.
The tyre wear model uses these groups to predict accurately results. Table 5.1
shows the results of TWM for data of second data.
Table 5.1: Results of TWM for data of second data.

Group Mounted Tyre Tyre Wear Indicator

I.1 FL worn Worn tyre detected
I.2 FL worn Worn tyre detected
II.1 FL & FR worn Worn tyre detected
II.2 FL & FR worn Worn tyre detected
III.1 RL worn Worn tyre detected
III.2 RL worn Worn tyre detected
IV.1 All fresh No worn tyre detected
IV.2 All fresh No worn tyre detected

In order to verify the model, a random sampling was made based on the
data of the second test to test the accuracy of TWM. The difference between
this method and the test in Table 5.1 is that a random sample can eliminate the
bias of sampling from the training data, as the test data is the training data.

28
 CHAPTER 5. RESULTS 29

Figure 5.1: GUI of test on TWM with random sampling.

In Figure 5.1 the number of sampling files is first chosen with the default
number being 10. After pressing the START button, random samples are cho-
sen and shown in the first column of the table. The second column demon-
strates the tyre conditions of each sample, while the third column shows the
prediction of TWM. The accuracy is then calculated and shown below the ta-
ble. One random validation is shown in Figure 5.2.

Figure 5.2: Result of random sampling to test TWM.

5.2 Trial run of RTMS

The TWM and UI are connected together to check the feasibility of RTMS.
The data of the car with all signals needed are analysed in the UI and the results
are displayed in the main interface.
30 CHAPTER 5. RESULTS

5.2.1 RTMS
The required signals for TWM are also added to the signal reader in WICE,
shown in Table 5.2. After the same data processing as iTPMS signals, the
required signals for TWM are saved in the same excel file and mat file. The
excel file is used to save the data of one specific data file while the mat file
is used to store the data in a month. In other words, the excel file is for the
main interface while the mat file is mainly for the interface of statistics on tyre
pressure. Through embedding TWM in the UI and calling the python scripts
by MATLAB, RTMS is realised.

Table 5.2: Input signals for TWM.

Signal names Description

FLWhlSpd Indicating the circumferential speed of front left wheel
FRWhlSpd Indicating the circumferential speed of front right wheel
RLWhlSpd Indicating the circumferential speed of rear left wheel
RRWhlSpd Indicating the circumferential speed of rear right wheel
BrkPedv Indicating the state of braking
VehSpdLgt Indicating vehicle longitudinal speed
VehSpdLat Indicating vehicle lateral speed

5.2.2 Result of trial run

When the UI is opened and the raw data file stored in WICE portal is selected,
the vehicle identification and tyre health information will be displayed in the
main interface after the data processing. According to the information dis-
played in the UI of RTMS which is shown in Figure 5.3, it can be seen that
for the selected car, all four tyres have good air pressure and wear conditions
and there is no calibration request. The interface of statistics on tyre pres-
sure shows the corresponding statistical results of the selected month in main
interface as shown in Figure 5.4.
From the interface of statistics on tyre pressure, it can be seen
that around one third vehicles equipped with WICE devices were logging
iTPMS signals. 95% of the cars had no warning of low tyre pressure. As for
the cars with low tyre pressure warning, 3 quarters of these cars experienced
common warning, which means more than one tyre had low air pressure. Only
1 car had the system failure. No cars had system unavailable and system un-
available SWDL. Regarding the calibration, 84.05% of the cars didn’t need
 CHAPTER 5. RESULTS 31

to do calibration and 10.58% of the cars successfully completed calibration.

5.37% of the cars were still experiencing calibration. No calibration failed in
May. Moreover, the number of cars with CalSts signal agreed with the num-
ber of cars with iTPMSSts signal according to the scatter plot shown in Figure
5.5. As the number of cars with calibration signal was increasing, the number
of cars with tyre pressure warning signal increased as well, which implies the
iTPMS inside company cars works well.

Figure 5.3: Result of RTMS.

32 CHAPTER 5. RESULTS

Figure 5.4: Result of statistics on tyre pressure.

Figure 5.5: Correlation coefficient of cars with CalSts and iTPMSSts signals.
Chapter 6

Discussion

6.1 Tyre Wear Model

Tyre tread depth is an important factor regarding safety of vehicles and cur-
rently is mostly measured with the help of physical tools. TWM is hence de-
veloped to identify whether any tyres of a vehicle is worn or not. TWM is
developed with its own feature analyses and a voting scheme, which is similar
to iTPMS in principle. Moreover, TWM uses the signals that already exist
in the bus signal, requiring no additional sensor. The one difference between
TWM and iTPMS is that TWM is written in MATLAB with input being the
converted data file through CANoe as written in Section 2.1.2. In other words,
TWM is not an exactly real-time application like iTPMS, which is installed on
vehicle ECUs, but can also be developed on cloud for processing and decision
making.
As for feature analyses, assumptions are made before building the model,
and each of them needs further testing for validation. Assumptions are listed
as

• Tyre pressure of four tyres on vehicles is in good condition, not an in-

fluence factor for tyre wear estimation.

• Different tyres and vehicles share the same features as the tested ones.

• Weather and road condition (temperature, road types, etc.) have little,
or no, influence on TWM.

• The threshold of being worn tyres is set as 5 mm, while the legal threshed
is 1.6 mm for summer tyres and 3 mm for winter tyres [19].

33
34 CHAPTER 6. DISCUSSION

The speed feature is inspired by the radius analysis of iTPMS [4], which
uses the difference between wheel speeds and vehicle speed to monitor the
effective rolling radius of tyres. A low tyre pressure leads to a smaller effec-
tive rolling radius. Tyre tread depth also contributes to the rolling radius, and
thus can be indirectly monitored through speed signals. However, it is unclear
whether TWM can tell a worn tyre from a tyre of low pressure.
The braking feature is found during the literature research. A worn tyre
behaves differently from a new tyre during braking [20]. The braking feature
is built on this. In this patent, it also mentions the braking force on a worn
tyre has a higher peak than that on a new tyre. The peak is caused by the anti-
locking braking system (ABS), as ABS will release the braking pad for a short
period, once the braking force is high enough to cause tyres to slip. The brak-
ing force can be measured through the pressure sensors of the braking system,
and can be added for determining worn tyres. However, the pressure sensors
in the test vehicle have a minimal sampling time of 1 s, while ABS responses
in milliseconds. The braking force thus are not included in the braking feature
analysis, but is worth looking into.
Overall, TWM shows a good accuracy for test data of two tests and is a
promising system. With further testing and development in the future, TWM
may become one useful function in vehicles like iTPMS, monitoring tyre tread
depth in real time.

6.2 Connectivity and UI

In this thesis, the connectivity between raw data in the cloud and the UI is
achieved by calling the python scripts in MATLAB. There are two kinds of
python scripts for the main interface and interface of statistics on tyre pressure
separately. One is used for handling the data of one specific data file while the
other is for processing monthly data and daily data.
Some abnormal problems occur when processing the latest data. In some
cases, the latest data file is available in the folder. However, it can not be
read. This kind of file will influence the performance of RTMS and leads to
error when filtering the signals and running the interface of statistics on tyre
pressure. Another problem is that the raw data may be unreasonable in some
data files. For example, the values should not be decimals but integer values.
This may be caused by too few samples.
In addition to the problems of raw data, there exist some methods to im-
prove the performance of RTMS in aspect of connectivity. For example, the
raw data can be saved in a database that MATLAB can be directly connected
 CHAPTER 6. DISCUSSION 35

to.
Another problem is that the UI can only display the result of one selected
data file. As there are some occasions that maybe the tyre health conditions
of more than one vehicle need to be shown in the UI, the system will be not
satisfying under that occasion.
In general, WICE collects the data of vehicle signals and saves the data in
the cloud, and then the connectivity realises the data transmission from WICE
cloud to the UI of RTMS. With embedded TWM in the UI, the vehicle identi-
fication and tyre health information can be visualised in the main interface of
the UI.
Chapter 7

Conclusions and Future Work

7.1 Conclusions
The background of this project is that many tyre manufacturers are spending
increasing resources into intelligent tyres. Compared to normal tyres, those
intelligent tyres are equipped with more sensors to monitor tyre condition,
such as tyre pressure, temperature and gripping force. In order to reduce the
cost, indirect monitoring of tyre conditions are studied. It means no addi-
tional sensors and uses the already-existed ones to predict tyre condition. The
previous studies on the indirect monitoring are mainly about iTPMS, which
has proven to be a robust system and replaced dTPMS. Another paper is US
patent 9610810, using feature extraction to estimate tyre state. Furthermore,
as the communication technology develops very fast in recent years, the perfor-
mance of telematic is significantly improved. One representative technology
is 5G which can greatly increase the speed and amount of data transmission.
Based on the telematics and new generation communication technology, the
potential to monitor big data of vehicle tyres in real-time is possible.
Based on the background information, two functions, TWM and connec-
tivity, have been developed in this project and combined together.
Firstly, the project has developed a Tyre Wear Model to detect worn tyres
on vehicles. It contains a speed analysis, braking analysis and a voting scheme.
The speed analysis extracts the median value of slip during driving and makes
a linear fit of median slip against vehicle speed. The slope of the linear fit is
compared to the reference data for classification. The braking analysis, based
on US patent 9610810, extracts the median value of slip ratio and slip rate
during braking, and applies linear SVM to classify the category of the data.
The voting scheme takes the idea of iTPMS and gives warning if either of the

36
 CHAPTER 7. CONCLUSIONS AND FUTURE WORK 37

analysis shows warning signals.

Secondly the connectivity is achieved by coordinating with the company’s
internal data extraction team to obtain data access permission. After testing the
connectivity of the system, valid data are successfully obtained since the mid
of May 2020. Three steps are processed for the raw data: data filter according
to signal names, data re-sample at fixed time intervals, data format conversion
and storage. While doing the data processing, the daily processing volume of
raw data is about 70GB. After that, MATLAB GUI is used to design the UI
where 6 key indicators are presented, including the information of car itself and
tyre health condition. The UI successfully demonstrates Cloud to MATLAB
application interaction.
Last, the two systems are combined together to be the RTMS. Through the
embedded TWM, tyre wear condition of cars is shown in the main interface of
the UI. Moreover, all driving cycle files in a month can be analysed and counted
to study the frequency of occurrence of different scenarios according to the
proportion of different signal values. This kind of information is presented by
pie charts and line graphs in the interface of statistics on tyre pressure.

7.2 Future work

The purpose of RTMS proposed by this project is not only restricted to mon-
itoring tyres, but also having a potential of a new business model. For both
fleet and autonomous vehicles, companies can remotely monitor tyre health
information in real time. When tyres are in bad condition, such as completely
worn out, RTMS is alerted and sends warning to the corresponding company.
The company is then able to book a car service based on this information and
order fresh tyres ahead that the vehicle arrives. This reduces the risk of ac-
cidents due to bad tyre conditions, and more importantly saves money and
other resources for companies, such as cost of car inspection and time to wait
for fresh tyres being delivered. In order to make this reality, more studies are
needed and several parts are listed in this section.
Firstly, regarding the data processing, the raw data are suggested to be
stored in a database, which can be directly connected to MATLAB, such as
MySQL. Then the data processing can be done without conversing data to
excel files by Python. It can save the running time, reducing lagging time.
Secondly, further development of TWM should be investigated. It can be
done by validating the assumptions listed in section 6.1 through minimising
and controlling assumptions at first. The tests should use worn tyres with 3
mm tread depth, instead of 5 mm. A grinding machine is helpful for making
38 CHAPTER 7. CONCLUSIONS AND FUTURE WORK

tyres with desired tread depth. Other tyre parameters should be also tested
with TWM, especially tyre pressure. Moreover, it is possible to build a real-
time application of TWM in CANoe. This application can be used during the
test and save the time to convert data to MATLAB data files.
As mentioned in section 6.2, the main interface of UI should consider the
presentation of more than one vehicle, which is possible to be done by adding
more rows to display the information and counting the number of times of
file list clicks. Similarly with the interface of statistics on tyre pressure, an
analytical dashboard of tyre wear can be designed to provide information on
tyre orders for vehicle companies and the data can be sold to tyre suppliers.
Besides, the interface is made by MATLAB GUI in this project, but it can be
written as a web page, allowing to embed more functions.
As now the RTMS handles with static data (data files saved in a server), it is
only batch processing, which has a delay and is not a real-time data processing
system. Thus, future work needs to include how to deal with the streaming
data and collaborate with the data extract team in the company to get access
to streaming data.
References

[1] H. B. Pacejka and I. Besselink. Tire and Vehicle Dynamics. eng. Third
edition. GB: Butterworth Heinemann, 2012. isbn: 0080970168.
[2] K. Yi, K. Hedrick, and S-C Lee. “Estimation of tire-road friction using
observer based identifiers”. In: Veh. Syst. Dyn. 31 (1999), pp. 233–261.
[3] National Highway Traffic Safety Administration. “Federal Motor Vehi-
cle Safety Standards; Tire Pressure Monitoring Systems; Controls and
Displays”. In: Federal Register 67(108) (2002), pp. 1–2.
[4] N. Persson, F. Gustafsson, and M. Drevo. “Indirect Tire Pressure Moni-
toring Using Sensor Fusion”. In: SAE Transactions 17 (2002), pp. 1657–
1662.
[5] Road traffic injuries (2020, February 07). https : / / www . who .
int/news-room/fact-sheets/detail/road-traffic-
injuries. Last accessed June 29, 2020.
[6] M.W. Adler, S. Peer, and T. Sinozic. “Autonomous, connected, elec-
tric shared vehicles (ACES) and public finance: An explorative analy-
sis”. eng. In: Transportation Research Interdisciplinary Perspectives 2
(2019), p. 100038. issn: 2590-1982.
[7] J. Manning. Connected transformation: from tyres to telematics to car
sharing (2019, Nov 08). https://www.fleeteurope.com/en/
connected/europe/analysis/connected-transformation-
tyres - telematics - car - sharing. Last accessed June 17,
2020.
[8] ContiConnect. Digital Tire Monitoring. Anytime. Anywhere. https:
/ / www . continental - tires . com / transport / tire -
monitoring/conticonnect. Last accessed June 17, 2020.
[9] H. Young Jo et al. Development of Intelligent Tire System. eng. Tech.
rep. 2013. url: https://saemobilus.sae.org/content/
2013-01-0744.

39
40 REFERENCES

[10] MICHELIN Track Connect: connect your tires to improve your track
performance (2018, June 04). https://www.michelin.com/
en / news / michelin - track - connect - connect - your -
tires - to - improve - your - track - performance/. Last
accessed June 17, 2020.
[11] ADAS SYSTEMS: HI-TECH CARS AND CYBER TYRES (2016, March
03). https://www.pirelli.com/global/en- ww/road/
adas - system - hi - tech - cars - and - cyber - tyres. Last
accessed June 17, 2020.
[12] FROM THE TARMAC TO 5G: PIRELLI’S SMART TYRES DESCRIBE
THE ROAD SURFACE (2020, Jan 09). https://www.pirelli.
com / global / en - ww / road / from - the - tarmac - to -
5g - pirelli - s - smart - tyres - describe - the - road -
surface. Last accessed June 17, 2020.
[13] S. Jansen et al. Study on some safety-related aspects of tyre use: final re-
port. eng. Luxembourg: Publications Office, 2014. isbn: 9279431498.
[14] K. Singh. “Developing Enabling Technologies for Intelligent Tires”. In:
Jan. 2014.
[15] Volvo Cars. https://www.volvocars.com/se/om-volvo/
foretaget/karriar/kontor-och-fabriker/hallered.
Last accessed June 16, 2020.
[16] C. Davies. 2019 Volvo XC40 Inscription sees luxe crossover go sub-
tle (2018, Feb 19). https : / / www . slashgear . com / 2019 -
volvo-xc40-inscription-sees-luxe-crossover-go-
subtle-19520007/. Last accessed June 16, 2020.
[17] Vector Informatik. Testing ECUs and Networks with CANoe. https:
//www.vector.com/int/en/products/products- a-
z/software/canoe/. Last accessed August 20, 2020.
[18] Vector Informatik. Product Information CANoe.FlexRay. https://
assets.vector.com/cms/content/products/canoe/
_flexray / CANoe _ FlexRay _ ProductInformation _ EN .
pdf. Last accessed August 20, 2020.
[19] Däck. https://www.transportstyrelsen.se/sv/vagtrafik/
Fordon/Fordonsregler/dack/. Last accessed August 19, 2020.
 REFERENCES 41

[20] K. B. Singh. “Method of tire state estimation through wheel speed sig-
nal feature extraction”. US9610810B1. The Goodyear Tire & Rubber
Company. Apr. 4, 2017.
[21] C. M. Grinstead and J. L. Snell. Grinstead & Snell’s Introduction to
Probability. 2009. isbn: 161610046X.
[22] T. Hastie, R. Tibshirani, and J. Friedman. The Elements of Statistical
Learning. 2005. isbn: 9780387848587.
[23] J. Xu et al. “New incremental learning algorithm with support vector
machines”. In: IEEE Transactions on Systems, Man, and Cybernetics
49 (2018), pp. 2230–2241.
[24] A. Goel. Fleet Telematics Real-time management and planning of com-
mercial vehicle operations. eng. Operations Research/Computer Sci-
ence Interfaces Series, 40. 2008. isbn: 1-281-08768-8.
[25] J. T Yee and S-C Oh. “Case Study: Radio Frequency Identification”.
eng. In: 2013th ed. London: Springer London, 2013, pp. 237–268. isbn:
1447143892.
[26] Inter-Vehicle Communication at Intersections : An Evaluation of Ad-
Hoc and Cellular Communication. eng. KIT Scientific Publishing.
[27] J. Johnson. Designing with the mind in mind : simple guide to under-
standing user interface design guidelines. eng. 2nd ed.. 2014. isbn: 0-
12-411556-X.
[28] M. H. Miraz, P. S Excell, and M. Ali. “User interface (UI) design issues
for multilingual users: a case study”. eng. In: Universal Access in the
Information Society 15.3 (2016), pp. 431–444. issn: 1615-5289.
 Chapter 8

Appendix

8.1 Abbreviation list

ABS anti-locking braking system

dTPMS direct tyre pressure monitoring system

ECU electronic control unit

FFT fast Fourier transform

iTPMS indirect tyre pressure monitoring system

IoT Internet of things

RFID radio frequency identification

SVM support vector machine

RTMS real-time tyre monitoring system

TWM tyre wear model

UI user interface

WCU wireless communication unit

WICE wireless information collection environment

WLAN wireless local area network

42
TRITA -SCI-GRU 2020:343

www.kth.se