Download as pdf or txt
Download as pdf or txt
You are on page 1of 67

Degree Project in Vehicle Engineering

Second cycle, 30 credits

Modelling of a high-performance
vehicle in MATLAB/Simulink and
Canopy Simulations

RICKARD HÖGLUND

Stockholm, Sverige 2022


Modelling of a high-performance vehicle in MATLAB/Simulink and
Canopy Simulations

Rickard Höglund

Master of Science in Engineering


Master programme in Vehicle Engineering
KTH Royal Institute of Technology

Supervisor at Koenigsegg Automotive AB: Dragos-Mihai Postariu


Supervisor at KTH: Mikael Nybacka and Lars Drugge
Examiner at KTH: Mikael Nybacka

Date of presentation: 06/10/2022

TRITA-SCI-GRU 2022:247

KTH Royal Institute of Technology


School of Engineering Sciences
KTH SCI SE-100 44 Stockholm, Sweden
URL: http://www.kth.se/sci
Abstract
The car industry is constantly evolving. High demands from customers and governing bodies
force the manufacturers to speed up development of new vehicle models. This process used
to be a very costly, time consuming and not very ecologically sustainable because new and
expensive parts had to be made that are difficult to manufacture. Today, there are many
different computer softwares that can help the development process by allowing the manu-
facturers to move this process into a digital environment which can save money, time and at
the same time help to become ecologically sustainable.
This Master’s thesis is a collaboration with Koenigsegg Automotive AB and aims to evaluate
computer softwares and their vehicle models. Three different softwares are tested by looking
at the correlation between them and measurement data from a real vehicle test. This to build
a better understanding of how the softwares work and how they can be used in the develop-
ment process of vehicle dynamics and control. The tested softwares are MATLAB/Simulink,
CarMaker and Canopy Simulations. A longitudinal model and a lateral model, also known
as the dual track model is created in MATLAB/Simulink and it is compared to the vehicle
created in the table based software CarMaker and the vehicle created in the optimization
oriented software Canopy Simulations.
To test the different models against each other some standard maneuvers are used, such
as slalom and constant radius cornering. The models are compared by measuring the lateral
acceleration and yawrate. Track driving is also tested since Canopy Simulations is specifically
developed for track driving.
The tests show a good correlation between the vehicle models, especially at lower velocities.
The developed dual track model during this Master’s thesis is a good vehicle model and
can predict the lateral acceleration and yawrate and shows most accurate results at lower
velocities and during non transient maneuvers where the load transfer is minimal. The model
is heavily dependent on the tire model which has to be taken into consideration. CarMaker
shows similar results to the dual track model but this software is more suited for testing
different driver aids. The way Canopy Simulations works makes it more complex to perform
standard maneuvers and test the correlation, but initial simulations show a good correlation
and its capabilities when it comes to vehicle development. It is a great tool for optimizing a
vehicle for performance on track.
Sammanfattning
Bilindustrin utvecklas ständigt. Höga krav från kunder och regeringar tvingar tillverkarna att
skynda på utvecklingen av nya fordonsmodeller. Denna process var tidigare mycket kostsamt,
tidskrävande och inte särskilt ekologiskt hållbart eftersom nya delar behövde tillverkas som är
svåra och kostsamma att tillverka. Idag finns det många olika datorprogram som kan hjälpa
utvecklingsprocessen genom att låta tillverkarna flytta denna process in i en digital miljö som
kan spara pengar, tid och samtidigt vara ekologiskt hållbart.
Detta examensarbete är ett samarbete med Koenigsegg Automotive AB och syftar till att
utvärdera dessa datorprogram och deras fordonsmodeller genom att testa tre olika program
och titta på korrelationen mellan dem och mätdata från ett riktigt fordonstest. Detta för
att skapa en bättre förståelse för hur datorprogrammen fungerar och hur de kan användas
i utvecklingsprocessen av fordonsdynamik och reglering. De testade programmen är MAT-
LAB/Simulink, CarMaker och Canopy Simulations. En longitudinell modell och en lateral
modell, även känd som dubbelspårsmodellen skapas i MATLAB/Simulink och den jämförs
med fordonet som skapats i det tabellbaserade programmet CarMaker och fordonet som
skapats i det optimeringsorienterade programmet Canopy Simulations.
För att testa de olika modellerna mot varandra används några standardmanövrar såsom
slalom och kurvtagning med konstant radie. Modellerna jämförs genom att mäta sidoacceler-
ationen och girhastigheten. Bankörning testas också eftersom Canopy Simulations är speciellt
utvecklad för bankörning.
Testerna visar en god korrelation mellan fordonsmodellerna, speciellt vid lägre hastigheter.
Den framtagna dubbelspårsmodellen under examensarbetet är en bra fordonsmodell och kan
förutsäga sidoaccelerationen och girhastigheten och visar bäst resultat vid lägre hastigheter
och under icke-transienta manövrar där lastöverföringen är minimal. Modellen är starkt
beroende av däckmodellen som man måste ta hänsyn till. CarMaker visar liknande resul-
tat som dubbelspårsmodellen men det här programmet är mer anpassat för att testa olika
förarhjälpmedel. På det sätt Canopy Simulations fungerar gör att det är mer komplicerat
att utföra standardmanövrar och testa korrelationen, men initiala simuleringar visar en bra
korrelation och dess förmåga när det gäller fordonsutveckling. Det är ett bra verktyg för att
optimera ett fordon för prestanda på bana.
Acknowledgements
This work could not have been done without the help from some people. The author would like
to thank Associate Professor Mikael Nybacka for doing a great job supervising the Master’s
thesis and always being ready to assist if help was needed. The author would also like to thank
Associate Professor Lars Drugge for helping with the modelling in MATLAB/Simulink and
Fritiof Hegardt for helping with Canopy Simulations. Finally, the author would like to thank
Dr. Dragos-Mihai Postariu for giving the opportunity to work together with Koenigsegg
Automotive AB and providing useful data from the company.
Contents

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Problem formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Research questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Simulation tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.1 MATLAB/Simulink . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.2 CarMaker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.4.3 Canopy Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Report outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2 MATLAB/Simulink 5
2.1 Coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Longitudinal model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1 Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2 Slip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.3 Transmission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.4 Clutch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.5 Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.6 Driver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Lateral model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.4 Magic formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Traction circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

3 CarMaker 18
3.1 Car . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 Road . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

4 Canopy Simulations 20
4.1 Car . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.1 Chassis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.2 Suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.3 Tyres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.4 Aero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.5 Brakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.6 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.1.7 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.8 Powertrain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Weather . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4.1 Straight Sim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4.2 Quasi-Static Lap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.4.3 Dynamic Lap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5 Standard maneuvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.6 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

5 Tests 27
5.1 Slalom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.2 Sine sweep . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.3 Step steer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.4 Steady-state circular driving . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.5 Acceleration and braking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.6 Track driving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.7 Test Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.7.1 Volvo S90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.7.2 Koenigsegg Jesko . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.7.3 Data logging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

6 Results 33
6.1 Lateral model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
6.2 Acceleration and braking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
6.3 Steady-state circular driving . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.4 Slalom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.4.1 Volvo S90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.4.2 Koenigsegg Jesko . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.5 Track driving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.5.1 Volvo S90 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.5.2 Koenigsegg Jesko . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

7 Discussion 47

8 Conclusion and Future work 50


8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
8.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

Appendices 54

2
A Volvo S90 data 55
List of figures

2.1 The coordinate system used when developing the vehicle model [14]. . . . . . 5
2.2 The layout of the longitudinal model. . . . . . . . . . . . . . . . . . . . . . . 6
2.3 All forces acting on the vehicle in the longitudinal vehicle model [14]. . . . . 6
2.4 An overview of the motion block in MATLAB/Simulink. . . . . . . . . . . . 8
2.5 The forces and torques acting on the wheel. . . . . . . . . . . . . . . . . . . 9
2.6 An overview of the slip block in MATLAB/Simulink. . . . . . . . . . . . . . 9
2.7 An overview of the transmission block in MATLAB/Simulink. . . . . . . . . 10
2.8 An overview of the driver block in MATLAB/Simulink. . . . . . . . . . . . . 12
2.9 An overview of the entire longitudinal model in MATLAB/Simulink. . . . . 12
2.10 All forces acting on the vehicle in the lateral vehicle model. . . . . . . . . . . 13
2.11 A visualization of the lateral slip on the front left wheel. . . . . . . . . . . . 14
2.12 Friction coefficient as a function of slip in dry tarmac conditions. . . . . . . . 16
2.13 A visual representation of the traction circle. . . . . . . . . . . . . . . . . . . 17

3.1 The computer software CarMaker. . . . . . . . . . . . . . . . . . . . . . . . . 18

4.1 The coordinate system used by Canopy Simulations [14]. . . . . . . . . . . . 21


4.2 The modelled suspension setup in Canopy Simulations. The figure shows the
inside of the right side of the vehicle. . . . . . . . . . . . . . . . . . . . . . . 22
4.3 An illustration of a double wishbone suspension with coilover. . . . . . . . . 23

5.1 The slalom track created in Canopy Simulations. . . . . . . . . . . . . . . . . 28


5.2 Visualization of the track created at Lunda Airfield. . . . . . . . . . . . . . . 30
5.3 Visualization of the track created at the Koenigsegg airfield. . . . . . . . . . 30
5.4 The Volvo S90 used for the tests in this Master’s thesis. . . . . . . . . . . . . 31
5.5 A render of the Koenigsegg Jesko [14]. . . . . . . . . . . . . . . . . . . . . . 31

6.1 The first slalom test at 30 km/h with the Volvo S90. . . . . . . . . . . . . . 34
6.2 The slalom test at 50 km/h with the Volvo S90. . . . . . . . . . . . . . . . . 34
6.3 The slalom test at 70 km/h with the Volvo S90. . . . . . . . . . . . . . . . . 35
6.4 The sine sweep test with the Volvo S90. . . . . . . . . . . . . . . . . . . . . 35
6.5 The steady-state circular driving test with a radius of 12 m. The test was done
with the Volvo S90. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

4
6.6 The steady-state circular driving test with a radius of 45.5 m. The test was
done with the Volvo S90. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.7 The step steer test with the Volvo S90. . . . . . . . . . . . . . . . . . . . . . 37
6.8 The acceleration test with the Volvo S90. . . . . . . . . . . . . . . . . . . . . 38
6.9 The braking test with the Volvo S90. . . . . . . . . . . . . . . . . . . . . . . 39
6.10 The steady-state circular driving test simulation with the Volvo S90 in Canopy
Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.11 The steady-state circular driving test simulation with the Volvo S90 in Canopy
Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.12 The calculated optimal racing line for the slalom test at 30 km/h. . . . . . . 41
6.13 The slalom driving test simulation at 30 km/h with the Volvo S90 in Canopy
Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.14 The slalom driving test simulation at 50 km/h with the Volvo S90 in Canopy
Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.15 The slalom driving test simulation at 70 km/h with the Volvo S90 in Canopy
Simulations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.16 The calculated optimal racing line for the slalom test at 70 km/h. . . . . . . 43
6.17 Comparison of the lateral acceleration and yawrate calculated by Canopy Sim-
ulations and the ones measured on the vehicle. . . . . . . . . . . . . . . . . . 43
6.18 The racing line used for the track driving simulation. . . . . . . . . . . . . . 44
6.19 Comparison of the lateral acceleration and yawrate calculated by Canopy Sim-
ulations and the measured by the VBOX. . . . . . . . . . . . . . . . . . . . . 45
6.20 Comparison of the velocity and steering wheel angle calculated by Canopy
Simulations and the lateral acceleration measured by the VBOX. . . . . . . . 45
6.21 The optimal racing line calculated by Canopy Simulations. . . . . . . . . . . 46
6.22 Comparison of the lateral acceleration and yawrate calculated by Canopy Sim-
ulations and the ones measured on the vehicle. . . . . . . . . . . . . . . . . . 46

A.1 The engine map used in MATLAB/Simulink. . . . . . . . . . . . . . . . . . . 55


List of tables

2.1 PID controller coefficients. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11


2.2 Magic Formula constants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.1 Volvo S90 suspension data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

6.1 Scalar results of the simulation. . . . . . . . . . . . . . . . . . . . . . . . . . 44

A.1 Volvo S90 data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55


Introduction

1.1 Background
The car industry is constantly evolving. The vehicles are becoming more and more complex
and high performing because of the competition, but there are at the same time demands from
both customers and governing bodies for safer and more sustainable vehicles. Developing a
high performance vehicle that is safe and sustainable can be a quite time consuming process,
especially if the setup and other parameters need to be changed between each test run. A
good idea to speed up this process is to use computer softwares and create a mathematical
representation of the vehicle. Then the engineers can experiment and change parameters of
the vehicle and optimize it in the simulation before making changes on the real vehicle and
thereby save time. This will also have a great impact on the costs associated with vehicle
development. A substantial amount of money can be saved if new parts do not have to be
produced every time something on the vehicle has to be modified. There are several different
ways the development stage can be improved with the help of computer softwares. There are
many computer softwares on the market today that are especially developed to allow engineers
to develop and improve their vehicles in many different applications. The applications can be
lap time simulations, autonomous driving, ride comfort and many more. There are multiple
ways of simulating a vehicle. Simpler models can be programmed in MATLAB or in other
similar computer softwares. Mathematical equations can be used to simulate a vehicle. There
are also table based softwares, such as CarMaker, CarSim and VI-CarRealTime. They can do
more complex simulations. There are block based softwares, such as MATLAB/Simulink that
uses differential equations. A more complex type of software is multi-body dynamics based
softwares, such as ADAMS/Car that allows for simulations with an entire vehicle including
all small joint and bushing elements. The last type of simulation software is optimization
softwares, such as Canopy Simulations which can calculate the theoretical best lap times on
track by setting it up as an optimization problem. The software allows the user to test how
different parameters impact lap times by doing several simulations, with different input data,
simultaneously.
Many people have created different vehicle models in MATLAB/Simulink, all slightly different
from each other. Vadiraj Patil [1] has created two vehicle models, one simpler and one more
complex model. The models include both longitudinal and lateral dynamics. Kiran Kone [2]

1
has also created a vehicle model in MATLAB/Simulink using the dual track model approach.
Gaspar Gil Gómez [3] did an extensive experiment comparing subjective assessments and
objective measurements and suggested how they can be correlated in the best way. The
PhD thesis also includes descriptions about open loop standard maneuvers which are good
to use when testing the dynamic capabilities of a vehicle. P. Shakouri et al. [4] developed
a longitudinal model in MATLAB/Simulink using a motion resistance equation to calculate
the velocity of the vehicle. They also used a tire design model developed by Hans B. Pacejka
called the Magic Formula [10]. Their final vehicle model showed a good correlation between it
and other known longitudinal vehicle models. D. Miloradović et al. [5] developed a nonlinear
vehicle model and came to the conclusion that if the vehicle is to be doing non stationary
driving, a nonlinear vehicle model should be used if similar results between the real vehicle
and the model are wanted. A. Lönnergård et al. [6] have developed tests to be able to evaluate
vehicle handling in winter conditions. They used a bicycle model together with the brush tire
model and came to the conclusion that a bicycle model with a simpler tire model is not an
accurate enough model when the slip angles are large. Raffaele Di Martino [7] has created
two different vehicles in MATLAB/Simulink with front wheel drive and rear wheel drive.
The vehicles include both a detailed longitudinal model and a lateral model and different tire
models. The results show a good correlation between the real vehicle test and simulations
with the longitudinal dynamics.
In this Master’s thesis the focus is on developing a vehicle model in three different computer
softwares and these are MATLAB/Simulink, CarMaker and Canopy Simulations. Canopy
Simulations is a relatively new simulation software that has not been tested before by KTH
Royal Institute of Technology. There is not much public information about the software, hence
most of the time will be dedicated towards Canopy Simulations. This work will hopefully
give a good demonstration on how moving vehicle development to a digital environment can
help the society have a economically and ecologically sustainable development. A substantial
amount of money and time can be saved on development by using these softwares. They can
also have a positive impact on the environment because fewer vehicle components need to be
manufactured during the development process.

1.2 Problem formulation


This Master’s thesis consist of two parts. The first part involves MATLAB/Simulink and
CarMaker, and the other part involves Canopy Simulations. The goals are to test the vehicle
model in CarMaker and create a vehicle model in both MATLAB/Simulink and Canopy
Simulations. The MATLAB/Simulink model puts more focus into the longitudinal and lateral
model of the vehicle while Canopy Simulations puts more focus on lap time simulations and
optimization of vehicle parameters.

1.3 Research questions


The Master’s thesis will answer these following research questions:

2
• For what circumstances can a dual track vehicle model be good enough for vehicle
dynamics developments of a sports car?
– What are the strengths and limitations of this modelling approach?
• How can lap time simulation software be used in vehicle development of sports cars
and can it also be used for open-loop driving scenarios?
– How does a lap time simulation model compare to a dual track vehicle model and
other vehicle dynamics simulation software?

1.4 Simulation tools


1.4.1 MATLAB/Simulink
The first tool that is used in this Master’s thesis is Matlab and its added simulation tool
Simulink. MATLAB became commercially available in 1984 and the first version of Simulink
was released in 2002. MATLAB is a computer software developed by MathWorks that is
developed for numeric computing. MATLAB/Simulink is a graphical block based simula-
tion tool that allows the user to simulate dynamic systems, such as vehicles. The software
uses model-based design and models are built in a graphical block based environment. This
software will be further discussed in Chapter 2.

1.4.2 CarMaker
CarMaker is a table based computer software which means that data is entered in tables
and thereby separated from the code. The software is developed to simulate vehicles in
different driving scenarios and it can also help develop autonomous systems for the vehicles
by co-simulating with MATLAB. This means that both softwares are coupled and simulated
together. CarMaker will be further discussed in Chapter 3.

1.4.3 Canopy Simulations


Canopy Simulations is a relatively new computer software specifically designed to help cus-
tomers to create the fastest vehicle possible. Using different types of simulations the software
can show how changing some parameters of the vehicle can affect the performance of the
vehicle and the lap times. The software uses its own optimization algorithm to calculate the
best lap times and highest cornering velocities that are possible to achieve with the modeled
vehicle. This software will be further discussed in Chapter 4.

1.5 Report outline


The Master’s thesis start with a introduction including a background and a formulation of the
problem. It also includes some information about the tools used and state of the art. After the
introduction comes the MATLAB/Simulink chapter which contains a detailed description of

3
the theory used and the vehicle model that is constructed in the computer software. Following
the MATLAB/Simulink chapter comes the CarMaker chapter with information about the
software. The next chapter is about Canopy Simulations. It includes a comprehensive look at
the custom vehicle model that is created in Canopy Simulations. After that comes the tests
chapter which explains the tests that have been done to validate the MATLAB/Simulink,
CarMaker and Canopy Simulations vehicle models. Then the results are presented followed
by a discussion of the results and a conclusion that answers the research questions.

4
MATLAB/Simulink

The MATLAB/Simulink model consists of two parts, the longitudinal dynamics part and
the lateral dynamics part. The longitudinal dynamics part calculates the behavior of the
vehicle in the forward and rearward direction while the lateral dynamics part calculates the
behavior of the vehicle in the perpendicular direction of the longitudinal. These two parts in
MATLAB/Simulink are created and modified to resemble a Volvo S90 but the model can be
parameterized and modified to resemble any vehicle that has two axles and four wheels.
There are a few assumptions that has to be made when creating the vehicle model in MAT-
LAB/Simulink. It is assumed that the road does not have a tilt in the lateral direction to
the vehicle and that there is no wind causing an increase in drag. It is also assumed that the
road conditions are dry and are constant throughout all tests. This means that the friction
coefficient of the road is also constant.

2.1 Coordinate system


It is important to be consistent with the coordinate system throughout the development of
the vehicle model. It is decided to go with the international standard coordinate system of a
vehicle and it can be seen in Figure 2.1 [8]. The x axis is along the vehicle in the longitudinal
direction and the y axis is in the lateral direction. The z axis is perpendicular to both the
other axis with zero at the road level and positive upwards.

Figure 2.1: The coordinate system used when developing the vehicle model [14].

5
2.2 Longitudinal model
The longitudinal model includes everything from driver to the longitudinal velocity, vx . The
overall layout of the longitudinal model can be seen in Figure 2.2. The model has a reference
velocity as input and the longitudinal velocity as output.

Figure 2.2: The layout of the longitudinal model.

2.2.1 Motion
When the vehicle is moving forward there are several forces acting on the vehicle. An illus-
tration of the forces acting on the vehicle can be seen in Figure 2.3.

Figure 2.3: All forces acting on the vehicle in the longitudinal vehicle model [14].

The motion block contains the overall equation to calculate the longitudinal acceleration, ax .
The equation is also called the motion resistance equation. The total force,

max = Fx − Fa − Fs (2.1)

where the force, Fx is the traction force that propels the vehicle forward. It is the force
from the tire to the road and it depends on the slip and normal force. The traction force is
calculated in the tire force block and it is further explained in Section 2.4. The force, Fa is
the aerodynamic drag and is calculated the following way,

6
1
Fa = ρCD Av 2 (2.2)
2
where ρ is the density of the air, CD is the total drag coefficient of the entire vehicle and A is
the frontal area of the vehicle. It is assumed to be normal air conditions hence the air density
is set to 1.225 kg/m3 . The velocity is looped back in this block because the drag depends on
the velocity. The last force, Fs depends on the slope of the road. If the vehicle drives on a
slope the mass of the vehicle will work against the direction of the motion and the vehicle
needs a larger force to accelerate it. The force,

Fs = mgsin(αr ) (2.3)

where αr is the angle of the sloped road. The force is non zero if the vehicle is driving on
a sloped road and is positive if the road has a positive gradient. All the forces above are
also used to calculate the normal forces on the front and rear axle. They are calculated from
moment equilibrium equations [4]. Depending on the magnitude and position of the forces,
the relation between the normal forces on the axles may change when driving. The normal
force on the front axle,

−Fa haero − max h − mghsin(αr ) + mgrcos(αr )


Fz12 = (2.4)
L
and the normal force on the rear axle,

Fa haero + max h + mghsin(αr ) + mgf cos(αr )


Fz34 = . (2.5)
L
The constant, haero is the vertical distance from the ground to where the drag acts on the
vehicle and the constant, h is the vertical distance from the ground to the center of gravity.
An overview of the motion block can be seen in Figure 2.4.

7
Figure 2.4: An overview of the motion block in MATLAB/Simulink.

2.2.2 Slip
To calculate the slip the rotational velocity of the wheels, ωw has to be calculated first. The
rotational velocity is calculated by integrating the angular acceleration, ω̇w which is given by
the equation:

Jtot ω̇w = Tw − Tb − Fx rr − Fr rr . (2.6)

It is an equation that incorporates the rotational inertia of the drivetrain which has to be
taken into account in the calculations [7]. The rotational inertia determines the required
torque to reach a specific angular acceleration and the larger the inertia is the slower the
mass will accelerate. The total inertia,

Jtot = 4Jw + Jt (Uf )2 + Je (Ui Uf )2 (2.7)

where Je is the inertia of engine, Jt is the inertia of the transmission and Jw is the inertia of
one wheel. The term Jw is multiplied with four since the vehicle has four wheels. The torque,
Tw is the wheel torque that comes from the transmission. The longitudinal model has an
all-wheel drive (AWD) system, meaning both axles get torque from the transmission. The
torque is split 50/50 in this model which means 50% of the torque, Tw goes to each axle. The
model has one Equation (2.6) for each axle. The term, Tb is the brake torque caused by the
brake clamping on the brake disc. There is a brake on every wheel. The force, Fx is the force
from the road to the tire working against the rotation of the wheel. It is the same force as in
Equation (2.1) and it is multiplied with the term, rr which is the rolling radius of the wheel.
The force, Fr is the rolling resistance force and is calculated the following way,

8
Fr = mgfr (2.8)

where g is the gravitational constant and fr is the rolling resistance coefficient. The gravi-
tational constant is assumed to be 9.81 m/s2 . The value of the rolling resistance coefficient
differs depending on the road and tire but a normal vehicle tire on a dry road can have a
value of around 0.02. The forces and torques acting on the wheel can be seen in Figure 2.5.

Figure 2.5: The forces and torques acting on the wheel.

The rotational velocity of the wheel is then used to calculate the longitudinal slip, κ, which
is needed to calculate the traction force from the wheel. The traction force goes then to the
motion block but is also looped back to Equation (2.6). The slip is the relative speed of the
wheel compared the the velocity of the vehicle and the longitudinal slip,

ωw rr − vx
κ= . (2.9)
max(vx , ωw rr )
The slip can only be a value between -1 and 1. An overview of the slip block can be seen in
Figure 2.6.

Figure 2.6: An overview of the slip block in MATLAB/Simulink.

9
2.2.3 Transmission
The transmission is the part of the drivetrain that has the engine torque as input and has a
different torque as output depending on the selected gear and its gear ratio. For the engine
to work, it can only operate within a certain rotational velocity range. The engine therefore
needs to be coupled, via a clutch, to a transmission with several gear ratios that would allow
the vehicle to reach different velocities without the engine overrevving or stalling. The output
torque is the wheel torque and it goes from the transmission to the wheels. The output torque,

Tw = Te Ui Uf ηd (2.10)

where Ui is the gear ratio of gear i and Uf is the gear ratio of the final drive. The final drive
is located in the differential which connects the drive shaft to the axle. The coefficient ηd is
the drivetrain efficiency and is assumed to be 0.9. The torque, Te is the engine torque and
comes directly from the engine.
The rotational velocity of the engine is also calculated in the transmission block. The entire
drivetrain is directly connected hence the rotational velocity of the engine can be derived
from the rotational velocity of the wheels. The rotational velocity of the engine,

ωe = ωw Ui Uf (2.11)

and it is then an input to the engine block. An overview of the transmission in MAT-
LAB/Simulink can be seen in Figure 2.7.

Figure 2.7: An overview of the transmission block in MATLAB/Simulink.

The model selects gear depending on the velocity of the vehicle, hence the velocity of the
vehicle is an input to the transmission block. The model has nine gears and a certain velocity
span corresponds to a specific gear.

10
2.2.4 Clutch
The clutch is the part of the drivetrain that connects the engine with the transmission. It
consists of two or more discs that can be pressed together to transfer the torque from the
engine to the transmission. In this model the clutch can only be either completely engaged
or disengaged. The position of the clutch depends on the engine torque input and the brake
pedal position. The clutch is engaged if the engine torque input is non zero and the brake
pedal position is zero.

2.2.5 Engine
The second block in the longitudinal model is the engine block. The engine provides torque
to the drivetrain. The engine in the vehicle model consists of a 2-D lookup table that gives a
specific engine torque, Te as output which depends on the accelerator pedal position and the
rotational velocity of the engine. This is because the engine cannot give maximum torque if
the speed of the engine is too low or too high. The engine map can be seen in Figure A.1.

2.2.6 Driver
The first part of the longitudinal model is the driver. The driver is a PID controller which
is a common controller used in many different areas. It consists of three different parts, the
proportional part, the integrator and the derivative. The controller has a variable for each
part that can be manually changed to change the behavior of the controller. The driver
block takes the velocity of the vehicle and compare it to the reference velocity and then send
out a signal between -100 and +100 depending on if the vehicle has to decrease or increase
its velocity. A positive value corresponds to an input to the engine while a negative value
correspond to an input to the brake. The coefficient in the PID controller can be found in
Table 2.1. The driver block also includes a low pass filter and a anti wind up loop to avoid
any big overshoot and delays. An overview of the driver block can be seen in Figure 2.8 and
an overview of the entire longitudinal model can be seen in Figure 2.9. It shows how all blocks
are interlinked.

Table 2.1: PID controller coefficients.

Kp Ki Kd
200 20 10

11
Figure 2.8: An overview of the driver block in MATLAB/Simulink.

Figure 2.9: An overview of the entire longitudinal model in MATLAB/Simulink.

2.3 Lateral model


There are several types of vehicle models with different degrees of freedom (DoF). There are
models with two DoF and 14 DoF and even more. One representation of a vehicle with four
wheels is the dual track model. It only has two DoF but gives similar results as a model with
14 DoF [9]. In the dual track model it is assumed that the vehicle is a single point of mass
and the mass has four contact points to the ground. This model can be used to calculate
the lateral dynamics of the vehicle, such as lateral acceleration and yawrate. It has some
limitations [9]. It does not take load transfer into account which is something most vehicles
experience during spirited driving with hard accelerations, braking and steering inputs, but
it can still give a good indication about the forces involved. Especially for high performance
vehicle that have very stiff suspension and not much body roll and load transfer. Another
limitation is that the lateral model is only valid at close steady state longitudinal dynamics,
such as low longitudinal accelerations. The dual track model can be represented as a state
space model where the derivative,

ẋ = Ax + b. (2.12)

The variable,
 
y
 ẏ 
x=
ψ  .
 (2.13)
ψ̇

12
The state, y is the lateral position of the vehicle in the global coordinate system and ẏ is
the lateral velocity of the vehicle. The state, ψ is the yaw angle of the vehicle in the global
coordinate system and ψ̇ is the rotational velocity or yaw rate of the vehicle. To derive the
state space model, a free body diagram of the dual track has to be noted and it can be seen
in Figure 2.10.

Figure 2.10: All forces acting on the vehicle in the lateral vehicle model.

There are forces acting on the vehicle and Newton’s second law of motion gives these following
equations:

→: may = (F1 + F2 )cos(δ12 ) + (F3 + F4 )cos(δ34 ) (2.14)

tw12 tw34
Jz ψ̈ = (F1 + F2 )f cos(δ12 ) − (F3 + F4 )rcos(δ34 ) + (F1 − F2 ) sin(δ12 ) + (F3 − F4 ) sin(δ34 ).
2 2
(2.15)
The forces F1 , F2 , F3 and F4 are the lateral forces on each wheel and δ12 and δ34 are the
wheel angles on the front and rear axle. The term, Jz denotes the inertia of the whole vehicle
around the z axis and tw12 and tw34 are the track widths on the front and rear axle. It is the
distance between the left and right wheel. The lengths f and r are the distances between the
front axle and center of gravity and center of gravity and the rear axle. Hence the wheelbase
is written as,

L = f + r. (2.16)

13
The calculated time derivative of the lateral velocity, v̇y is not the same as the lateral accel-
eration, ay which is used in Equation (2.14). The acceleration, ay consists of change in lateral
velocity as well as the addition from the yawrate. The lateral acceleration is thus written as,

ay = v̇y + ψ̇vx . (2.17)

The lateral forces depends on the lateral slip, α. It determines how large the lateral force is
between the tire and the road. The relation between lateral slip and lateral force is something
that is not simple to model. When the slip is very small there is a linear relation between
the lateral slip and lateral force but when the slip angle is larger, which can happen in more
aggressive driving, the relation is highly nonlinear. When the slip is small the force can be
estimated with a constant similar to the following equation.

F = −Cα (2.18)

where C is a constant and is known as the cornering stiffness. In this Master’s thesis a
nonlinear tire model is used and it is presented in Section 2.4. The lateral slip for the front
left wheel is visualized in Figure 2.11. The lateral slip for the other three wheels can be
derived the same way but will have changed signs and other constants.

Figure 2.11: A visualization of the lateral slip on the front left wheel.

The lateral slip for each wheel is as follow:


!
vy + ψf ˙
α1 = tan−1 − δ12 (2.19)
vx − ψ̇ tw12
2

14
!
vy + ψf ˙
α2 = tan−1 − δ12 (2.20)
vx + ψ̇ tw12
2

!
vy − ψr ˙
α3 = tan−1 − δ34 (2.21)
vx − ψ̇ tw34
2

!
vy − ψr ˙
α4 = tan−1 − δ34 (2.22)
vx + ψ̇ tw34
2

The final state space model is as follow:

   
ẏ x2
1

 ÿ  
 = m
(F 1 + F 2 )cos(δ12 ) + (F 3 + F 4 )cos(δ 34 ) − mx4 v x

.
ψ̇   x4 
1 tw12 tw34

ψ̈ Jz
(F1 + F2 )f cos(δ12 ) − (F3 + F4 )rcos(δ34 ) + (F1 − F2 ) 2 sin(δ12 ) + (F3 − F4 ) 2 sin(δ34 )
(2.23)
This model allows the user to have a longitudinal velocity and steering wheel angle as input
and y position, lateral velocity, yaw angle and yawrate as output. One drawback is that the
longitudinal velocity has to be non zero and the longitudinal acceleration needs to be kept
low and close to steady state, for the state space model to work.

2.4 Magic formula


There are many tire models used today that are used to estimate the amount of force the
tire can transfer to the road. There are different nonlinear tire models, such as the brush tire
model and the Magic Formula. In this vehicle model the Magic Formula is used which is a
formula developed by Hans B. Pacejka [10]. The formula calculates the longitudinal force and
lateral force as a function of respective tire slip, α and κ, and the normal force, Fz . The tire
model may have uncertainties which may be a drawback [11], this depends on how good the
input data is and if one has access to tire test rig values. The longitudinal force is defined as,

Fx = f (κ, Fz ) = Fz µ = Fz Dsin(Ctan−1 (Bκ − E(Bκ − tan−1 (Bκ)))) (2.24)

and the lateral force is defined as,

Fy = f (α, Fz ) = Fz µ = Fz Dsin(Ctan−1 (Bα − E(Bα − tan−1 (Bα)))). (2.25)

15
The coefficient, µ is the friction coefficient and it depends on four constants which are different
depending on the type of road condition and tire. The constant, B is the the stiffness factor,
C is the shape factor, D is the peak factor and E is the curvature factor. The peak factor
determines the maximum amount of grip the tire can have and roughly the maximum lateral
acceleration. The exact values of the constants are difficult to find because they are often
protected intellectual property but some constants for different road conditions can be found
in Table 2.2 [12]. They are based on empirical tire data. In this model the conditions are
assumed to be similar to dry tarmac and the friction coefficient as a function of slip can be
seen in Figure 2.12.

Table 2.2: Magic Formula constants.

Conditions B C D E
Dry tarmac 10 1.9 1 0.97
Wet tarmac 12 2.3 0.82 1
Snow 5 2 0.3 1
Ice 4 2 0.1 1

Figure 2.12: Friction coefficient as a function of slip in dry tarmac conditions.

2.5 Traction circle


When a tire is in contact with the road there is limit to how big the contact force can be.
The maximum force is a function of the friction coefficient and the maximum total force,

F = µFz (2.26)

16
where µ is the friction coefficient of the road. The force includes both longitudinal and lateral
direction. This means that the tire can not have a maximum longitudinal and lateral force
at the same time [13]. This is the friction circle and a visual representation can be seen in
Figure 2.13.

Figure 2.13: A visual representation of the traction circle.

It is in reality a flat ellipse but can be simplified to a circle with constant radius. The radius
of the circle,
q
µFz = Fx2 + Fy2 . (2.27)

In the MATLAB/Simulink vehicle model, the longitudinal dynamics are calculated first and
depending on the magnitude of the longitudinal force, the lateral force is determined. The
lateral force,

q 2 
Fy = min µFz − Fx2 , Fy α, Fz (2.28)

to make sure that it is within the traction circle.

17
CarMaker

CarMaker is a computer software that is developed by IPG Automotive and is created to


test vehicle models. The software offers different vehicle models and possibilities to create
custom tracks where the vehicles can be tested. It is also possible to input a target velocity
and steering wheel angle which makes it possible to compare with data from a real vehicle.
To perform a simulation in CarMaker a car, a road and a maneuver have to be selected. An
overview over the computer software can be seen in Figure 3.1.
A big part of CarMaker is the ability to connect the computer software to MATLAB and
run co-simulations with it. This allows the user to test control functions that can be devel-
oped in MATLAB. Control systems, such as anti-lock braking systems (ABS) and traction
control systems (TCS) could be created using MATLAB/Simulink and tested together with
CarMaker which simulates the vehicle.

Figure 3.1: The computer software CarMaker.

3.1 Car
CarMaker has several different types of vehicles to choose from and it is also possible to modify
or create a custom vehicle. The user has the option to change many different parameters that
may have an effect on the behavior of the vehicle. In this Master’s thesis a vehicle is created
that resembles a Volvo S90 as good as possible. It is not possible to create an exact model
but the model is still very close to the real vehicle used in the tests with the correct power
output, mass, steering ratio, overall dimensions and spring stiffnesses.

18
3.2 Road
The user can create a custom road to test the vehicle on. The road can include multiple
lanes, junctions, road signs and traffic to test control systems. In this Master’s thesis a road
is created and it is a simple, long, wide, flat and straight road. It is big enough for the vehicle
to perform different standard maneuvers without running of the road during the tests.

3.3 Maneuver
CarMaker allows the user to create custom maneuvers and use telemetry as input. In this
Master’s thesis the maneuver is created using telemetry from the real tests. The telemetry
consist of a matrix with three columns which are time, longitudinal velocity and steering
wheel angle. The simulation will have these as inputs and if the velocity in the telemetry do
not start at zero the simulation will accelerate the vehicle to the starting velocity. The time
it takes to do the maneuver is the total time of the telemetry plus the initial acceleration
time. When the simulation is running, a 3D model of the vehicle is shown giving the user a
real-time visualization of the maneuver.

3.4 Post-processing
The computer software allows the user to choose what data that should be stored after each
test. When the maneuver is complete the computer software stores that data in a file that
can be downloaded and post-processed in MATLAB.

19
Canopy Simulations

Canopy Simulations is a web based simulation software especially designed for lap time
simulations and optimizations of vehicle parameters. The motorsport platform also offers
vehicle modelling and setup exploration. When performing a simulation the user has to
choose car, weather, track and type of simulation. A description about these can be found
below. The software uses cloud computing and a proprietary collocation solver to solve highly
nonlinear optimal control problems. A collocation solver uses a collocation method which is a
method to solve integral and differential equations numerically. The method involves choosing
a set of collocation points in the domain and the method will give an approximate solution
which satisfies the equation at those points.

4.1 Car
Two different vehicles have been created in Canopy Simulations for this Master’s thesis and
they are a Volvo S90 and a Koenigsegg Jesko. More information about the vehicles can be
found in Section 5.7. Canopy Simulations have a few preconfigured vehicles, most of them
similar to race cars. It is possible to create a custom car configuration, completely from the
start or based on the preconfigured vehicles, and that is done by manually changing the
values of some already set parameters. The parts on the vehicle that can be changed are:
• Chassis
• Suspension
• Tyres
• Aero
• Brakes
• Control
• Cooling
• Powertrain.

20
Every part has a list of several different parameters that can be manually changed. Most
parameters are already in the lists but the user has the option to add more parameters to
the vehicle if that is needed. The parameters describing the dimension of the vehicle and
the geometry of the suspension are based on the coordinate system seen in Figure 4.1. The
coordinates are rotated 180 degrees which means that the calculated lateral acceleration
and yawrate have to be multiplied with -1 to be comparable with the results from MAT-
LAB/Simulink.

Figure 4.1: The coordinate system used by Canopy Simulations [14].

4.1.1 Chassis
The chassis involves all general information about the overall vehicle that is needed. That
includes the mass, inertias, ride heights, weight distribution and steering.

4.1.2 Suspension
The suspension includes the suspension setup on the front and rear axle. It is possible to
change both the external and internal suspension. The external suspension is the pick up
points for the suspension and toe and camber settings. The toe angle is the angle the wheel
makes with the x axis of the vehicle. A positive toe, also known as toe in, means that the
front part of the wheel is closer to the center line. The camber angle is the angle the wheel
makes with the z axis of the vehicle. A negative camber is when the top of the wheel leans
inwards. The coordinates of the pick up points explain where the suspension is connected to
the vehicle body and wheels. The points describe the geometry of the suspension. The user
has the option to choose between several different types of suspensions, all with different
suspension parts and setups. Another important part that is included is the track-rod which
is part of the steering system. The internal part of the suspension includes the spring, damper
and anti roll bar. The spring is an important part of the suspension because it improves the
ride quality by allowing the wheel to move up and down with the road while the vehicle body
is still. The damper is also an important part since it dampen the movement of the vehicle
body. The anti roll bar is a cross bar that connect the left and right wheel and helps by
reducing the body roll while cornering.
The Volvo S90 has two different types of suspension setups. It has double wishbone suspension
with coilovers on the front axle and a multi link suspension on the rear axle. A double
wishbone suspension with coilovers consists of two A-arms in parallel connecting the wheel
to the chassis. An illustration of this can be seen in Figure 4.3. The coilover, which is a

21
combined spring and damper, is mounted on the lower A-arm. This suspension setup is
constructed in Canopy Simulations by specifying the pick-up points which determines the
exact geometry of the suspension. The modelled suspension setup can be seen in Figure 4.2.

Figure 4.2: The modelled suspension setup in Canopy Simulations. The figure shows the
inside of the right side of the vehicle.

The coilover is modified with a set spring coefficient. It is not possible to construct a multi
link suspension in Canopy Simulations, hence the suspension on the rear axle is replaced
with a double wishbone suspension with coilovers. The setup has some pick-up points that
are the same as the multi link suspension for the double wishbone suspension to have some
resemblance to the multi link suspension. Both axles have an anti roll bar each and they are
also constructed with pick-up points and are given a specific stiffness. It is a cross bar that
is located behind the axles. Coefficients of the suspension setup of the Volvo S90 can be seen
in Table 4.1.

Table 4.1: Volvo S90 suspension data.

Front spring stiffness 45 kN/m


Rear spring stiffness 80 kN/m
Front anti roll bar stiffness 126 kNm/rad
Rear anti roll bar stiffness 61 kNm/rad

The Koenigsegg Jesko has a double wishbone suspension setup on both the front and rear
axle. However, there are some differences on the setup compared to the Volvo S90. The
Koenigsegg Jesko has its coilovers attached to the the upper A-arm and the coilovers sit at a
larger angle pointing downwards and inwards to the chassis. Furthermore, the vehicle has a
coilover between the left and right wheel on both axles. That coilover helps with reducing the
rotation around the y-axis during heavy acceleration and braking. The anti roll bar on the
Koenigsegg Jesko is also different compared to the ones on the Volvo S90. The Koenigsegg
Jesko has a z-shaped anti roll bar which is located above the axle. When one wheel is pushed
upwards the z-shaped anti roll bar will cause the other wheel to also be pushed upwards.

22
Figure 4.3: An illustration of a double wishbone suspension with coilover.

4.1.3 Tyres
The tyres part is about the tire model. Canopy Simulations include a few different types
of tire models, including a complex version of the Magic Formula tire model explained in
Section 2.4. It is also possible to implement a thermal model of the tires. In this Master’s
thesis the standard Canopy Simulations tyre model is used because there is not enough data
to use the Magic Formula tire model in Canopy Simulations.

4.1.4 Aero
The aerodynamics may have a significant effect on the vehicle, especially on high performance
vehicles. The aero section involves the aerodynamics of the vehicle, both the vehicle body
and the tires. The aerodynamic model outputs an aerodynamic load in the front and rear
of the vehicle and lift and drag on the wheels using Equation (2.2). Canopy Simulations
uses polynomial aero maps to describe the aerodynamics of the vehicle. It shows the relation
between the geometrical and aerodynamic properties of the vehicle. Canopy Simulations can
also simulate a drag reduction system (DRS) which is a function some vehicles have that
allows them to temporarily remove some of the downforce and drag by, for example rotating
a wing. This may be useful when downforce is not that important, for instance on a straight
where top speed is crucial.

4.1.5 Brakes
The brakes includes the friction coefficient of the brake, the effective radius and area of the
disc brake. Canopy Simulations has also implemented a brake thermal model to calculate the
effect heat may have on braking performance.

4.1.6 Control
The control section allows the user to implement brake balance optimization which can be
useful when driving hard on track. There is a possibility to use a brake-by-wire system in
the simulations that can work together with an electric motor on the axle for regenerative
braking. This means that some of the energy that would have turned into heat during braking
can instead go back to the battery.

23
4.1.7 Cooling
When doing thermal simulations it can be good to test the cooling system in the vehicle.
Canopy Simulations enables the user to implement cooling circuits and radiator and simulate
them in a Dynamic Lap simulations, which is further explained in Section 4.4.3.

4.1.8 Powertrain
The engine section lets the user customize the engine, transmission and differential. Similar to
the engine configuration in Section 2.2.5, Canopy Simulations see the engine as a 2-D look up
table where a specific engine speed and accelerator pedal position results in a specific engine
torque. It is possible to add turbos to the engine and electric motors to the powertrain.

4.2 Weather
The weather is a parameter that has to be set before the simulation for it to work because
the weather may effect the aerodynamics of the vehicle and if the performance is heavily
reliant on its aerodynamics, the weather may have a compelling effect on the performance
of the vehicle. In this Master’s thesis the weather was always set to 25°C with normal air
pressure and low humidity.

4.3 Track
Canopy Simulations provide a wide selection of race tracks from around the world which
can be used in the simulations. It is also possible to create a completely custom race track
with speed limits which is useful for testing open-loop maneuvers. The speed limit is only
considered during one type of simulation which is further explained in Section 4.4.3.

4.4 Simulations
Canopy Simulations offer many different types of simulations to test the car in different sce-
narios. Below are the most useful simulations for this Master’s thesis listed. The simulations
show different driving times for the tests compared to the tests with the real vehicle because
the simulations will not always follow the speed limits that can be put along the driving
path.

4.4.1 Straight Sim


Straight Sim is a simulation that finds the static equilibrium state of the vehicle at different
velocities between 0 and 100 km/h. It is very useful if the user wants to know about ride
height, aerodynamics and mechanical balances at different velocities.

24
4.4.2 Quasi-Static Lap
The Quasi-static lap simulation finds the quasi-static equilibrium of the vehicle at every point
of the track. The simulation follows the racing line if it is predetermined but it is also possible
to let the simulation generate its own racing line. The simulation does not consider the speed
limits on the track. The results include telemetry of the car throughout the track.

4.4.3 Dynamic Lap


The Dynamic Lap simulation is a fully dynamic simulation and shows the maximum perfor-
mance of the vehicle including all of its dynamics. The simulation is an optimization problem
and the goal is to find the fastest lap time as possible. The simulation also generates an
optimal racing line and it is also possible to limit the velocity of the vehicle according to the
speed limits set on the track. If the track is short a special dynamic simulation type called
Dynamic Lap HD can be used. It will include more measurement points along the track which
results in much smoother data plots.

4.5 Standard maneuvers


Since Canopy Simulations is developed to simulate vehicles going round a race track, it is
more difficult to simulate standard maneuvers. It is not possible to set a reference velocity
and steering wheel angle which is done in the MATLAB/Simulink vehicle model, hence the
standard maneuvers have to be performed in other ways. One way to perform a standard
maneuver in Canopy Simulations is to create a custom track that has the same driving path
as the real vehicle had in the real test. By doing this the real vehicle has to first do the
standard maneuver while mapping its driving path. That driving path is then recreated as a
new track. One or several speed limits of minimum 10 m/s can then be set along the track to
prevent the vehicle going faster then what it did in the real test. The type of simulation has
to be a Dynamic Lap simulation because that is the only type of simulation that obeys the
speed limit. When the Dynamic Lap simulation is running it will automatically generate a
racing line along the custom track which the vehicle will follow. If the generated racing line
is not good enough, a custom racing line can be used instead. The custom racing line can
be used as track layout and racing line if it is retrieved from the Global Positioning System
(GPS) data. It is of great importance the GPS data is correctly scaled and is imported to
Canopy Simulations as x and y coordinates and in meters. Another possible way to perform
a standard maneuver in Canopy Simulations is by doing a Quasi-Static Lap simulations but
then there is a risk of the vehicle in the simulation going faster than the real vehicle. This
type of simulation requires a racing line which can be created or generated by the a racing
line generation simulation. The vehicle will follow that racing line during the Quasi-Static
Lap simulation.

25
4.6 Post-processing
When the simulations are completed the telemetry of the vehicle is shown. Canopy Simula-
tions also provides scalar results, such as lap time and the maximum velocity of the vehicle
on the track. It is possible to download the data for post-processing in MATLAB. The data
can be downloaded as a csv file which can be imported to MATLAB. The file can then be
converted to a matrix where the result is presented as a function of time or distance.

26
Tests

To make sure that the MATLAB/Simulink model and the Canopy Simulations model work
and give reasonable results the models have to be tested. There are several ways of testing
the models and one of them is the use of standard maneuvers. A standard maneuver has pre-
determined reference velocity and steering angle that should be followed when the tests are
performed. By letting a test driver perform these maneuvers in the real vehicle, the measured
velocity and steering wheel angle can be extracted and put into the model. Other measure-
ments can then be compared to the models and if they correspond to the measurements of
the real vehicle, the models are working. By having standards it is much easier to compare
vehicles. There are several different standard maneuvers and the ones used in this work is
listed below:
• Slalom
• Sine sweep
• Step steer
• Steady-state circular driving
• Acceleration and braking.
Most of the tests mentioned above are open-loop maneuvers. An open-loop maneuver does
not include a driver that can input an corrective action and thereby make the test repeatable
and independent of the driver [15]. The opposite of an open-loop maneuver is a closed-loop
maneuver and in those tests there need to be a driver involved to make sure the vehicle
follows the correct drive path.
When performing these maneuvers at least four variables have to be measured and they
are longitudinal velocity, lateral acceleration, yaw rate and steering wheel angle. The mea-
sured longitudinal velocity and steering wheel angle can then be used as input to the MAT-
LAB/Simulink model and CarMaker.
The test are performed in three different locations, Lunda Flygfält, Arlanda Test Track 1
and at the Koenigsegg airfield. Lunda Flygfält is an airfield located north east of Uppsala.
The runway is around 1.8 km long and 25 m wide. It also has a wider section that is used
for the steady-state circular driving. Arlanda Test Track 1 has a short track but also a large

27
circular area where the steady-state circular driving can be done. The radius of the circle is
45.5 m. Koenigsegg Automotive AB has access to an airfield that is 1.5 km long and has a
width of 40 m.

5.1 Slalom
The slalom test is a test where the driver is instructed to keep a constant velocity while
driving along a slalom shaped path created by evenly spaced cones. The cones are placed on
a straight line 20 m apart. The velocity should be at 100 km/h but test can also be done
at 20 km/h steps according to ISO 7401 [16]. This standard maneuver is developed to test
the vehicle in semi-static conditions. Depending on the velocity, the vehicle can be tested in
both linear regions and nonlinear region. The behavior of the vehicle will be more linear the
lower the velocity is. In this Master’s thesis the slalom test is done at three different velocity,
which are 30, 50 and 70 km/h. The slalom track created for the tests with the Volvo S90 in
Canopy Simulations can be seen in Figure 5.1. Every apex, which is the inner most point on
the curve, has an offset of 0.5 m from the center line to force the vehicle in the simulation to
follow the drive path of the real vehicle better.

Figure 5.1: The slalom track created in Canopy Simulations.

5.2 Sine sweep


The sine sweep test is similar to the slalom test but during this test the driver is instructed
to decrease the period after every left and right turn. This will result in a slalom shaped
driving path with an increasing frequency.

5.3 Step steer


The step steer test is a test that consist of driving straight ahead followed by a single hard
turn. When turning, the time between 10% and 90% of the steering input should not be
greater than 0.15 s according to ISO 7401. The test can be done at different velocities similar
to the slalom test seen in Section 5.1. This type of test is designed to see how the vehicle
behaves under transient maneuvers, hence the hard turn. The vehicle will have a nonlinear
behavior compared to when performing the slalom test.

28
5.4 Steady-state circular driving
The steady-state circular driving test, which is sometimes called constant radius cornering,
determines the steady-state circular driving behavior of a passenger vehicle according to ISO
4138 [17]. This tests involves three different methods. The first one is the constant radius
method, the second method is the constant steering wheel angle test method and the third
method is the constant velocity method. These are good tests to measure how the vehicle is
behaving in corners and to see if the vehicle is understeering, oversteering or has a neutral
steering. In other words, to see if the vehicle turns at the same rate as the wheels. In the
constant radius method the velocity is varied and the steering wheel angle is measured. In the
constant steering wheel angle method the velocity is also varied but the radius is measured.
In the constant velocity method the radius is varied while the radius is measured or vice
versa. In this Master’s thesis the test is done with a constant radius and a varying velocity.
The test is done with two different radii which are 12 m and 45.5 m.

5.5 Acceleration and braking


The acceleration and braking test consists of two parts, the first part is the acceleration of
the vehicle and the second part is the braking. It is a test to validate the longitudinal model
and compare the acceleration and braking time to the real vehicle. The vehicle will first
accelerate to 100 km/h and when the targeted velocity is reached the vehicle should stay at
that velocity for a few seconds. The vehicle should then brake as hard as possible until the
vehicle is at a stand still.

5.6 Track driving


The final test is track driving. The computer software Canopy Simulations is developed to
optimize lap times hence it is of great interest to test this by doing track driving in the
real world and in Canopy Simulations. The track driving is done, similar to the standard
maneuvers, at Lunda Flygfält and at the Koenigsegg airfield. A custom track is created by
placing cones along the track. The track at Lunda Flygfält has three corners, one small
bend, one small 180 degree turn and one larger 180 degree turn. It also has two straights
where the longer straight is 91.5 m long. The track can be seen in Figure 5.2. The track is
relatively simple compared to more famous race tracks but because it is simpler it is easier
to find the optimal racing line when driving. Then it is much easier to get a good comparison
between the real world test and the simulations in Canopy Simulations. The track is defined
in Canopy Simulations by points along the left and right side of the track. The coordinates
of these points are gathered from MATLAB. An over head picture of the track from Google
Maps is imported to MATLAB where the left and right side of the track is manually traced
and when the whole track is traced the coordinates are rescaled to a known length of the real
track, which is in this case the long straight. The airfield is almost completely flat hence the
road is assumed to have no gradient in MATLAB, CarMaker and Canopy Simulations.

29
Figure 5.2: Visualization of the track created at Lunda Airfield.

Some tests were also carried out at the Koenigsegg airfield. The custom track that was created
at the airfield can be seen in Figure 5.3. The track is designed to test the vehicle for straight
line speed and low and high speed cornering. The track has a width of 5 m and consists of a
400 m long straight followed by a 180 degree turn with a radius of 15 m. The other straight
has a 150 m long and 20 m deep fast corner and it is followed by another 180 degree turn
that leads back to the long straight. This track has also a relatively easily found optimal
racing line because of the simple layout.

Figure 5.3: Visualization of the track created at the Koenigsegg airfield.

5.7 Test Vehicles


To perform the tests two different vehicles were used, the Volvo S90 and the Koenigsegg
Jesko. The Volvo S90 is part of the KTH Royal Institute of Technology test fleet and the
Koenigsegg Jesko is a vehicle from Koenigsegg Automotive AB.

30
5.7.1 Volvo S90
The Volvo S90 was introduced in 2016 and is a sedan model with seating for four passengers.
The model is offered with several different engine configurations and the vehicle used in the
test has the D5 engine which is a four cylinder twin turbo diesel engine with a maximum
power of 173 kW at 4000 rpm, a maximum torque of 480 Nm between 1750-2500 rpm and
a maximum engine velocity of about 5000 rpm. There is no available complete engine map,
hence a custom engine map is made for the tests that is based on known engine data. The
vehicle has an AWD system, a weight of 1995 kg and has a wheelbase of 2.941 m. More data
of the Volvo S90 can be found in Table A.1 and a picture of the Volvo S90 can be seen in
Figure 5.4.

Figure 5.4: The Volvo S90 used for the tests in this Master’s thesis.

5.7.2 Koenigsegg Jesko


The Koenigsegg Jesko is a vehicle introduced by Koenigsegg Automotive AB in 2019. It is
a two seater high performance vehicle. A render of the Koenigsegg Jesko can be seen in
Figure 5.5.

Figure 5.5: A render of the Koenigsegg Jesko [14].

31
5.7.3 Data logging
To gather data from the tests an external data logger has to be used which is placed in the
vehicle. The data logger used in the Volvo S90 is a VBOX 3i dual antenna. It is a device that
can collect accelerations, velocities as well as CAN-bus data such as, steering wheel angle at
a rate of 100 samples per second. The VBOX is connected to several parts in the vehicle.
It is connected to an internal measurement unit (IMU) in the center of the vehicle and two
GPS antennas on the roof of the vehicle. Before the tests are performed the vehicle has to
drive around in a figure of eight for the antennas to calibrate and for the VBOX to know the
exact location of the vehicle. The VBOX is also connected to a computer in the vehicle that
shows the data. The data is gathered in a separate VBO file for each tests. That file can then
be imported to MATLAB and be filtered with a butterworth filter, to remove any spikes in
the data, if needed.

32
Results

In this chapter the results are presented. It includes results from MATLAB/Simulink, Car-
Maker and Canopy Simulations. The Volvo S90 from KTH Royal Institute of Technology was
used in most of the tests because of the existing knowledge of the vehicle and logged data
from previous tests. The Koenigsegg Jesko from Koenigsegg Automotive AB was included
in the later part of the Master’s thesis. The weather conditions were good when all the tests
were performed. The temperature was around 25°C.

6.1 Lateral model


During some of the test the longitudinal acceleration will be very small hence the longitudinal
slip and the longitudinal forces will be small. In this case the lateral forces can be maximum
according to the traction circle and this means that it is possible to isolate the lateral model
and only test this model. To test the lateral model some data from a test runs with the Volvo
S90 has been used. Four different tests has been done and they are slalom, constant-radius
cornering, sine sweep and step steer. The results can be seen in Figure 6.1-Figure 6.7. The
tests were carried out in MATLAB/Simulink and in CarMaker and then compared to the
data from the real vehicle test. The MATLAB code has steering wheel angle and longitudinal
velocity as input and calculates lateral acceleration and yawrate.

33
(a) Lateral acceleration (b) Yawrate

Figure 6.1: The first slalom test at 30 km/h with the Volvo S90.

The slalom test at 30 km/h were done twice. The high oscillations in the beginning of the tests
are because of the small values close to zero and the oscillations also depends on the solver in
MATLAB which in this case was the ordinary differential equation solver ODE45. Both slalom
tests at 30 km/h show that the the lateral model or dual track model can give reasonable
results for both the lateral acceleration and yawrate when the steering wheel angle is changes
gradually in both directions. Figure 6.1 also shows that CarMaker give reasonable values
when it has velocity and steering wheel angle as inputs. CarMaker gives some oscillations
in the results of the lateral acceleration but the values are still good compared to the data
from the VBOX. The oscillations may be due to the input data or the specific suspension
characteristics of the vehicle model.

(a) Lateral acceleration (b) Yawrate

Figure 6.2: The slalom test at 50 km/h with the Volvo S90.

Figure 6.2 shows that the lateral model is valid for a slalom test with a velocity of 50 km/h.
The largest differences between the data from the VBOX and the lateral model and CarMaker
are at the peaks and valleys where they give a slightly larger absolute value.

34
(a) Lateral acceleration (b) Yawrate

Figure 6.3: The slalom test at 70 km/h with the Volvo S90.

The biggest difference between the data from the VBOX and the lateral model and CarMaker
is in the slalom test at 70 km/h, which can be seen in Figure 6.3. The differences are quite large
at the peaks and valleys and the lateral model has a slight time delay. This test indicates that
the standard coefficients in the tire model are insufficient and result in small lateral forces.
The peak factor, D is increased from 1 to a more reasonable value of 1.2 to make the lateral
model be able to follow the drive path and give justifiable value for the lateral acceleration
and yawrate.

(a) Lateral acceleration (b) Yawrate

Figure 6.4: The sine sweep test with the Volvo S90.

The sine sweep test, presented in Figure 6.4, shows that the lateral model gives reasonable
results when the steering wheel angle is changed at a higher rate. This test was done at an
earlier stage when the Volvo S90 was equipped with another set of tires but the results show
that the older tires are quite similar to the new ones since the model gives similar results for
the lateral acceleration.

35
(a) Lateral acceleration (b) Yawrate

Figure 6.5: The steady-state circular driving test with a radius of 12 m. The test was done
with the Volvo S90.

The steady-state circular driving test can be seen in Figure 6.5. It is the test with the 12
m radius and the test shows an increase in lateral acceleration and yawrate during the test
because the longitudinal velocity is increased. It starts at 4.5 m/s and is gradually increased
to 11 m/s. The test has the same problem as the slalom test with high oscillations in the
beginning and in the end of the test. This is due to the near zero velocity. However, both the
lateral model and CarMaker gives similar values to the real vehicle at all velocities. There
are some differences for the calculated yawrate when the velocity is at its maximum.

(a) Lateral acceleration (b) Yawrate

Figure 6.6: The steady-state circular driving test with a radius of 45.5 m. The test was done
with the Volvo S90.

Figure 6.6 shows he results, from the constant-radius cornering test with the larger, 45.5
m radius. The velocity was increased from 5 m/s to about 22 m/s. The results show quite
large values for both the lateral acceleration and the yawrate. The peak lateral acceleration

36
is about -1.2 g which corresponds to the increased value of the peak factor, D which was set
to 1.2.

(a) Lateral acceleration (b) Yawrate

Figure 6.7: The step steer test with the Volvo S90.

In the step steer test, shown in Figure 6.7, the sharp turning was done in both directions
several times directly after one another. The lateral model and CarMaker have some issues
with the step steer test. Both softwares show too large values. The peak factor, D was also
increased in this test which may explain the large values, but lowering peak factor results in
the lateral model not being able to follow the drive path of the real vehicle.

6.2 Acceleration and braking


To get fair results of the acceleration and braking, the rolling resistant coefficient has to be
calculated first. This can be done using data from a coasting test. In this test the vehicle
is going at around 110 km/h, then the transmission is put into neutral which means that
the vehicle does not get any torque from the engine. The vehicle will then coast and slowly
decelerate because of the air drag and rolling resistance. The rolling resistance coefficient is
calculated by adding the equivalent mass to the Equation (2.1) and rearranging it, thus the
coefficient,

−(m + mj )ax − Fa − Fs
fr = . (6.1)
mg

The force Fx is in this case zero. Since the engine is disconnected from the drivetrain the
equivalent mass is different. The equivalent mass,

4Jw Jt (Uf )2
mj = 2 + . (6.2)
rr rr2

37
The longitudinal velocity and acceleration from the test is put in the equation and the rolling
resistance coefficient is calculated at every time step. The rolling resistance coefficient, used in
the acceleration and braking test, is the average throughout the coasting. The coasting test is
done twice, in opposite directions, to cancel out any possible effect from a road gradient. The
Volvo S90 has an average rolling resistance coefficient of 0.019 which is reasonable considering
the surface condition and tire model [18]. To test the acceleration of the Volvo S90, the vehicle
was accelerated from 0 to 100 km/h. The results from the test with the Volvo S90 can be
seen in Figure 6.8. The longitudinal model had to be slightly modified because the Volvo S90
has a torque limiter on the first gear. The torque is about 75% of the maximum torque in
first gear. The torque limiter is probably implemented in the vehicle because the engine has
quite a lot torque which can cause the wheels to lose the grip during hard accelerations. The
vehicle is presumably set up for smooth accelerations and not for fast acceleration times.

(a) Velocity (b) Acceleration

Figure 6.8: The acceleration test with the Volvo S90.

The result shows that it takes the real vehicle about 9.9 s to accelerate from 0 to 100 km/h
or 27.78 m/s while the MATLAB/Simulink model takes around 10.4 s to accelerate. The
MATLAB/Simulink model is slightly faster the first two seconds because the accelerator
pedal is immediately at 100% and the vehicle can start to accelerate right away. The real
vehicle has a slightly higher acceleration throughout the acceleration phase and thereby has
a moderately larger increase in velocity. The model has some oscillations in the beginning of
the acceleration but that is because of the solver in MATLAB/Simulink.
The braking test was also carried out with the Volvo S90. The vehicle was decelerated from
100 to 0 km/h as fast as possible and the results can be seen in Figure 6.9.

38
(a) Velocity (b) Acceleration

Figure 6.9: The braking test with the Volvo S90.

The model takes about 4.8 s to decelerate to 0 km/h while the real vehicle takes about 2.8
s. The real vehicle has also a higher deceleration compared to the model. This is expected
because the model does not have an ABS and thereby has to limit the maximum brake torque
to not lock the wheels too much.

6.3 Steady-state circular driving


This steady-state circular driving test was performed on the track with the constant radius of
45.5 m. This test was performed using a Dynamic Lap simulation since speed limits had to be
used. A custom circular track was created in Canopy Simulations with the same dimensions
as the circular track at Arlanda Test Track 1. The track was divided into equally sized
sections with a specific speed limit between 10 m/s and 22 m/s. To be able to increase the
velocity slowly, several consecutive laps had to be driven in the real test. It was not possible
to do a multi-lap simulation, hence the results from Canopy Simulations had to be rescaled
to compare it to the data from the VBOX. The velocity can be seen in Figure 6.10 and it
shows that the velocities are fairly similar. It also shows that the real vehicle has a slightly
higher top speed. The model has a top speed of 21 m/s while the real vehicle comes up to
22 m/s. The steering wheel angle can be seen in Figure 6.10. It displays that the steering
wheel angles of the real vehicle and the model are similar and the steering system is correctly
modeled. The lateral acceleration and yawrate of the Volvo S90 can be seen in Figure 6.11.

39
(a) Velocity (b) Steering wheel angle

Figure 6.10: The steady-state circular driving test simulation with the Volvo S90 in Canopy
Simulations.

(a) Lateral acceleration (b) Yawrate

Figure 6.11: The steady-state circular driving test simulation with the Volvo S90 in Canopy
Simulations.

The results show that Canopy Simulations gives reasonable results for the lateral acceleration
and yawrate and that the Canopy Simulations tyre model is a good representation of the tires
on the vehicle.

6.4 Slalom
6.4.1 Volvo S90
The slalom test was performed using the Dynamic Lap HD simulation and the calculated
optimal racing line for the test at 30 km/h can be seen in Figure 6.12. The red arrow indicates

40
the positive y-direction in the track coordinate system, and the blue ball indicates the position
of the vehicle on the track. The racing line is the narrow blue line and the center of gravity
of the vehicle follows it. Canopy Simulations does not allow the front wheels on the vehicle
to go outside the track. It looks as if the wheel leave the track in the figure below but that
is because the visualized track is narrower than the one used in the calculation. The results
of the Volvo S90 can be seen in Figure 6.13 - Figure 6.15. They were all simulated on the
same track and the only thing that was changed between the tests was the speed limit on
the track. The track had to be a loop with the start and finish in the same position for the
simulation to work. That meant that the plots had to be manually aligned to be compared.
The data from the VBOX is shifted in the time domain to align the first turn in the slalom
course. The data from the rest of the loop in Canopy Simulations is cut out since it is not
part of the test.

Figure 6.12: The calculated optimal racing line for the slalom test at 30 km/h.

(a) Lateral acceleration (b) Yawrate

Figure 6.13: The slalom driving test simulation at 30 km/h with the Volvo S90 in Canopy
Simulations.

As previously mentioned, the minimum speed limit that can be set in Canopy Simulations
is 10 m/s or 36 km/h. It is not a good velocity to test Canopy Simulations but during this
test the real vehicle did go a bit faster than 30 km/h and because of that the figures show
similar magnitude on both the lateral acceleration and yawrate.

41
(a) Lateral acceleration (b) Yawrate

Figure 6.14: The slalom driving test simulation at 50 km/h with the Volvo S90 in Canopy
Simulations.

The slalom test at 50 km/h does also give similar results for the lateral acceleration and
yawrate but the data from the VBOX is slightly shifted to the right after the first turn and
it is because the velocity of real vehicle was a bit lower than 50 km/h.

(a) Lateral acceleration (b) Yawrate

Figure 6.15: The slalom driving test simulation at 70 km/h with the Volvo S90 in Canopy
Simulations.

The slalom test at 70 km/h is difficult to compare. Neither the real vehicle nor the simulations
had a velocity of 70 km/h and they had never the same velocity at a given timestamp. The
simulation went down to around 50 km/h. The high entry velocity resulted in another racing
line that was not a slalom shaped path hence the different number of peaks in results from
Canopy Simulations.

42
6.4.2 Koenigsegg Jesko
The Koenigsegg Jesko performed one slalom test at 70 km/h. The dimension of the slalom
track was the same as the one for the Volvo S90 but two turns were added and the cones
where placed in such a way that the apex did not need an offset in Canopy Simulations.
The vehicle used in the test was running on petrol which meant that the power output was
lower. The slalom test was done using the Dynamic Lap simulation and a speed limit of 70
km/h. The calculated racing line can be seen in Figure 6.16 and the results can be seen in
Figure 6.17. The values on the axis have been removed because it is information that can not
be shared.

Figure 6.16: The calculated optimal racing line for the slalom test at 70 km/h.

(a) Lateral acceleration (b) Yawrate

Figure 6.17: Comparison of the lateral acceleration and yawrate calculated by Canopy
Simulations and the ones measured on the vehicle.

The data from the real test is slightly shifted to the left because the real vehicle went a bit
faster than 70 km/h. The difference in velocity may also explain why the measured lateral
acceleration and yawrate from the real test are significantly larger compared to the results
from Canopy Simulations. Furthermore, the results can also indicate that the tyre model used
for the vehicle in Canopy Simulations is not close to the real vehicle tires in those scenarios.

43
6.5 Track driving
6.5.1 Volvo S90
The track driving was simulated using the Dynamic Lap HD simulation but with a predefined
racing line which can be seen in Figure 6.18. The grey arrow head in the figure shows the
direction the vehicle is moving and the start and finish line. Normally, Canopy Simulations
will calculate the optimal racing line when performing a Dynamic Lap simulation but it was
not possible to do this for this simulation. This will be further discussed in Chapter 7. A
predefined racing line had to be used which resembles the racing line taken by the real vehicle.
The racing line was retrieved from the GPS data from the VBOX.

Figure 6.18: The racing line used for the track driving simulation.

The most interesting scalar results of the simulation are the maximum velocity and the lap
time. They can be seen in Table 6.1. The results are compared to the real test. The track
driving with the real vehicle was done by an experienced driver and the Volvo S90 was driven
on its limits. Canopy Simulations simulates flying laps hence the measurements where taken
over several consecutive laps. The lap time of the real test is an average of several laps.

Table 6.1: Scalar results of the simulation.

Canopy Simulations VBOX


Maximum velocity 23.53 m/s 22.85 m/s
Lap time 17.02 s 18.08 s

The lateral acceleration, yawrate, velocity and steering wheel angle are compared and they
can be seen in Figure 6.19 - Figure 6.20. Since Canopy Simulations only simulates one lap
and the real test consisted of several laps, the simulation is only compared to one of the real
laps. The figures show that the measurements made by the VBOX are fairly similar to the
calculations made by Canopy Simulations. The maximum values of both the lateral accel-
eration and yawrate are similar. However, Canopy Simulations calculates a smaller lateral
acceleration and yawrate at the end of the big turn. The velocity profile from Canopy Simu-
lations shows that the powertrain and braking are correctly modeled in the software because
both the longitudinal acceleration and the braking are similar to the data from the VBOX.

44
It also shows that there is a possibility to go slightly faster on the longer straight and a bit
faster in the first corner.

(a) Lateral acceleration (b) Yawrate

Figure 6.19: Comparison of the lateral acceleration and yawrate calculated by Canopy
Simulations and the measured by the VBOX.

(a) Velocity (b) Steering wheel angle

Figure 6.20: Comparison of the velocity and steering wheel angle calculated by Canopy
Simulations and the lateral acceleration measured by the VBOX.

6.5.2 Koenigsegg Jesko


The custom track driven by the Koenigsegg Jesko was big enough for Canopy Simulations
to perform a standard Dynamic Lap simulation. The calculated optimal racing line can be
seen in Figure 6.21 and the lateral acceleration and yawrate can be seen in Figure 6.22. The
values on the axis have been removed because it is information that can not be shared. This
test was performed by an experienced driver from the company and the goal was to drive
around the track as fast as possible.

45
Figure 6.21: The optimal racing line calculated by Canopy Simulations.

The custom track in Canopy Simulations had to have smooth corners for the optimization
algorithm to find the optimal racing line hence the round corners in the fast corner. The real
vehicle had the same racing line as Canopy Simulations apart from the last corner where
Canopy Simulations suggests a wider line.

(a) Lateral acceleration (b) Yawrate

Figure 6.22: Comparison of the lateral acceleration and yawrate calculated by Canopy
Simulations and the ones measured on the vehicle.

The lateral acceleration and the yawrate show a good correlation between the measurements
and the results from Canopy Simulations. The measured lateral acceleration has some oscil-
lations but the absolute peak values of the acceleration are similar to Canopy Simulations
which shows that there is a good correlation in both low and high speed corners.

46
Discussion

The dual track model used in this Master’s thesis has some weaknesses when it comes to
vehicle dynamics development. The model has a limited number of parameters that can be
changed because it has only 2 DoF and it is not possible to use the model to see how a
different suspension setup may affect the handling. Moreover, the dual track model does
not include load transfer which may explain the differences in the results from the high
velocity slalom test. However, the results from the test of model show that even though the
model is limited it can give reasonable results compared to measurements from the real test
vehicle. The coefficients in the Magic Formula have a great impact on the results and they
are sufficient in the lateral model but there are some differences. The inconsistency can be
seen in Figure 6.11 when the lateral is at its maximum. In that region the lateral force on
the tire is no longer a linear function of the lateral slip and the magnitude of the force is
dependent on the nonlinear tire model.
The longitudinal model gives quite good results even though many assumptions had to be
made about the Volvo S90. The engine map, the exact torque split and the exact torque
limit were unknown. The acceleration time was within half a second to the real vehicle and
this shows that the model can be good for simulating the acceleration phase of a vehicle.
However, the model was not good at simulating a vehicle braking hard. An ABS has to be
implemented in the model for it to give better results since the Volvo S90 has such a system.
The coefficients in the Magic Formula used in this Master’s thesis is not recommended when
it comes to modelling of a high performance vehicle since the maximum friction coefficient
given by the formula is about one. A high performance vehicle usually have sticky tires
that has better grip and therefore a friction coefficient of more than one. A larger friction
coefficient results in a larger traction force and faster acceleration of the vehicle.
Canopy Simulations allows the user to simulate standard maneuvers in the software by allow-
ing the user to create custom driving scenarios and custom tracks with speed limits. The way
Canopy Simulations works makes it difficult to generate results that can be directly compared
to MATLAB/Simulink and CarMaker. Some measurement points give similar results which
can verify the vehicle model in Canopy Simulations but it is otherwise difficult to compare
Canopy Simulations with the results from the real tests. However, Canopy Simulations is
good at doing what the developers say they are good at, which is lap time simulations. The

47
software is user-friendly and it is easy to set up track driving simulations and the results
from the simulations are easy to interpret. It should be mentioned that the simulation will
sometimes fail because there is something wrong with the configuration of the car or track
and since the calculations are not done on the user’s computer it can be quite difficult to
exactly know what causes the simulation to sometimes fail.
For an unknown reason it was not possible to do a normal Dynamic Lap simulation to simulate
the track driving. The simulation could not find a solution to the optimization problem. This
is not good because neither the optimal racing line nor the best lap time could be found for
this specific track and vehicle. It is possible that the simulation was on the limit of what
was possible to do with Canopy Simulations. The vehicle is quite big compared to the track
and the track had relatively tight corners. It was possible to simulate the vehicle on a bigger
track with bigger corners which indicates that a bigger track should be used for future track
tests.
It was possible to simulate the track driving when a predefined racing line was used and the
results from Canopy Simulations were comparable to the data from the VBOX. However,
there were some differences. There may be some reasons to why the results are different.
Firstly, the car created in Canopy Simulations is not an exact copy of the real vehicle. As
previously mentioned, the Volvo S90 has a multi link suspension on the rear axle which can
not be created in Canopy Simulations. This means that the rear suspension in the simulation
probably behaves and moves differently when the car is driving even though the overall
dimensions are the same. The vehicle has also an AWD system which means that that all
four wheels on the vehicle get torque from the engine. It was not possible to select AWD
for the vehicle in Canopy Simulations, hence the Volvo S90 was chosen to have rear-wheel
drive instead. The type of drive train will have an effect on the behavior of the vehicle in
corners. Secondly, the conditions of the surface of the track may have been different and not
as dry as assumed in the simulation in Canopy Simulations. A different surface condition can
have resulted in less grip for the real vehicle and thereby slower acceleration and cornering
velocities. Thirdly, the racing line used in Canopy Simulations may not have been an exact
copy of the racing line that the real vehicle used. The end of the racing line also had to be
slightly modified because the start and finish line had not the same coordinates. The last
position had to be moved for both lines to have the same position. This may explain why
there is a dip in lateral acceleration and yawrate at the end of the big turn.
It was decided to only do the constant radius cornering test, slalom test and track driving test
in Canopy Simulations because they were the only tests that could be properly recreated in
the software. The tests have well defined drive paths. Both the constant radius cornering test
and the track driving tests with the Volvo S90 had good correlation between the data from
the VBOX and the results from Canopy Simulations. The slalom test had good correlation
at the lower velocities but not at 70 km/h. There can have been some mistakes made when
the track was manually measured out with the measurement wheel. This may have resulted
in a different track in Canopy Simulations that was to difficult to follow at a high velocity
that is on the limit of what the vehicle can travel at.
The results from the track driving with the Koenigsegg Jesko also show a good correlation

48
between the real test and the simulations even though the simulation used its own generated
racing line. This means that the driver did a great job driving the vehicle on its limits and
the vehicle was correctly modelled. This also shows that the default tire model is good for
the vehicle. The results from the slalom test are unexpected. As previously mentioned, the
large values from the real test were probably caused by the high velocity and a tire model
that should have been tuned more.

49
Conclusion and Future work

In this chapter the conclusions and suggested future work are presented.

8.1 Conclusion
The goal of this Master’s thesis was to answer these following research questions:
• For what circumstances can a dual track vehicle model be good enough for vehicle
dynamics developments of a sports car?
– What are the strengths and limitations of this modelling approach?
• How can lap time simulation software be used in vehicle development of sports cars
and can it also be used for open-loop driving scenarios?
– How does a lap time simulation model compare to a dual track vehicle model and
other vehicle dynamics simulation software?
The dual track vehicle model can be a good enough model for sports car developments when
the vehicle will not be driven on its limits. The model is better at predicting behavior at
lower velocities and during non transient maneuvers. When the vehicle is driven on its limits
the other parameters, such as load transfer needs to be taken into account and the simple
dual track vehicle model does not include that. The tire model is a very important part that
also needs to be taken into consideration.
Lap time simulation software, such as Canopy Simulations can be used in sports car develop-
ments because the software allows the user to explore how different parameters of the vehicle
affect important aspects of track driving. The user can change the powertrain, suspension
setup, aerodynamic properties, chassis properties, grip and more and simulate the vehicle on
famous existing tracks or custom tracks and see how the lap time, top speed and cornering
speeds are affected by the changes. Since Canopy Simulations allows the user to create cus-
tom tracks, it is possible to test some open-loop driving scenarios. However, it can only be
done after a real world test. It requires a predetermined track layout or racing line which
can be acquired from GPS data. If the tests have to be done in the software first, the test

50
will not be open-loop because the collocation solver uses the track layout to determine the
steering wheel angle input similar to a closed-loop scenario.
A Canopy Simulations vehicle model has a good correlation to a dual track vehicle model
and a CarMaker vehicle model, especially at lower velocities. They all show similar values
for the lateral acceleration and yawrate when compared to data from real tests.

8.2 Future work


Future work include developing a more accurate clutch model in the longitudinal model to
better simulate the rotational velocity of the engine and transmission. The MATLAB/Simulink
model could also be improved by adding more DoF and a suspension model. A better suited
tire model is also recommended to be implemented. The coefficients in the Magic Formula
used in this Master’s thesis is good at predicting tire forces at lower velocities but they are
not suited for higher velocities and for high performance vehicles that heavily rely on the
tires.

51
Bibliography

[1] V. Patil, ”Generic and complete vehicle dynamic models for open-source platforms,”
M.S. thesis, Department of Applied Mechanics, Charlmers Univiversity of Technology.,
Gothenburg, Sweden, 2017. Accessed: Sept. 23, 2022. [Online]. Available:
https://publications.lib.chalmers.se/records/fulltext/251558/251558.pdf
[2] K. Kone, ”Lateral and longitudinal control of an autonomous racing vehicle,” M.S.
thesis, Department of Mechanical and Aerospace Engineering, Polytechnic University of
Turin, Turin, Italy, 2019. Accessed: Sept. 23, 2022. [Online]. Available:
https://webthesis.biblio.polito.it/11982/1/tesi.pdf
[3] G. G. Gómez, ”Towards efficient vehicle dynamics development: From subjective
assessments to objective metrics, from physical to virtual testing,” Ph.D., Department
of Aeronautical and Vehicle Engineering, KTH Royal Institute of Technology,
Stockholm, Sweden, 2017. Accessed: Sept. 23, 2022. [Online]. Available:
https://www.diva-portal.org/smash/get/diva2:1075945/FULLTEXT01.pdf
[4] P. Shakouri, A. Ordys, M. Askari, and D. S. Laila, ”Longitudinal vehicle dynamics using
Simulink/Matlab,” Faculty of Engineering, Kingston University, London, United
Kingdom, Jan., 2010. Accessed: Sept. 23, 2022. [Online]. Available:
https://www.researchgate.net/publication/261390259_Longitudinal_vehicle_
dynamics_using_SimulinkMatlab, DOI: 10.1049/ic.2010.0410.
[5] D. Miloradović, J. Glišović, N. Stojanović, and I. Grujić ”Simulation of vehicle’s lateral
dynamics using nonlinear model with real inputs,”, IOP Conf. Ser.: Mater. Sci. Eng. 659
012060, Faculty of Engineering, University of Kragujevac, Kragujevac, Serbia, 2019.
Accessed: Sept. 23, 2022. [Online]. Available:
https://iopscience.iop.org/article/10.1088/1757-899X/659/1/012060/pdf,
DOI: 10.1088/1757-899X/659/1/012060.
[6] A. Lönnergård and M. H. Asher ”Objective evaluation of vehicle handling during winter
conditions,” M.S. thesis, Department of Aeronautical and Vehicle Engineering, KTH
Royal Institute of Technology, Stockholm, Sweden, 2015. Accessed: Sept. 23, 2022.
[Online]. Available:
https://www.diva-portal.org/smash/get/diva2:892115/FULLTEXT01.pdf

52
[7] R. Di Martino, ”Modelling and simulation of the dynamic behaviour of the automobile,”
M.S. thesis, Faculty of Engineering, University of Salerno, Fisciano and Baronissi, Italy,
2012. Accessed: Sept. 23, 2022. [Online]. Available: https://tel.archives-ouvertes.
fr/file/index/docid/736040/filename/DIMARTINO_THESE_2005.pdf
[8] www.sis.se. https://www.sis.se/api/document/preview/914200/ (accessed Sept. 23,
2022)
[9] S. Na, J. Jang, K. Kim, and W. Yoo, ”Dynamic vehicle model for handling performance
using experimental data,” Department of Mechanical Engineering, Pusan National
University, Busan, Korea, 2015. Accessed: Sept. 23, 2022. [Online]. Available:
https://journals.sagepub.com/doi/pdf/10.1177/1687814015618126, DOI:
10.1177/1687814015618126.
[10] H. B. Pacejka and I. Besselink, ”Semi-Empirical Tire Models,” in Tire and Vehicle
Dynamics, 3rd ed. Oxford, United Kingdom: Elsevier Ltd, 2012, pp. 165.
[11] B. Xie, Y. Zhao, Y. Tian, and G. Nie, ”Lateral Dynamic Modeling and Verification of
Vehicles with Steering Characteristics,” College of Communication Engineering, Jilin
University, Changchun, China, 2019. Accessed: Sept. 23, 2022. [Online]. Available:
https://ieeexplore.ieee.org/abstract/document/8832806
[12] https://se.mathworks.com/help/sdl/ref/tireroadinteractionmagicformula.html
(accessed Sept. 22, 2022)
[13] O. Danielsson, I. Jonsson, E. Hansson, F. Mannerhagen, P. Molander, N. Olofsson, J.
Pettersson, A. Pluto, and G. Sahlin, ”Design of electrical powertrain for Chalmers
Formula Student with focus on 4WD versus RWD and regenerative braking,” B.S.
thesis, Division of Vehicle Engineering and Autonomous Systems, Chalmers University
of Technology, Gothenburg, Sweden, 2013. Accessed: Sept. 23, 2022. [Online]. Available:
https://publications.lib.chalmers.se/records/fulltext/191837/191837.pdf
[14] www.news.cision.com. https://news.cision.com/koenigsegg/r/
all-new-koenigsegg-jesko-announced-at-geneva-2019,c2881595 (accessed Sept.
23, 2022)
[15] R. Powell, ”RELATIONSHIPS BETWEEN LANE CHANGE PERFORMANCE AND
OPEN- LOOP HANDLING METRICS,” M.S. thesis, Clemson University, Clemson,
United States of America, 2009.
[16] Swedish Standards Institute, ”Road vehicles - Lateral transient response test methods
- Open-loop test methods,” 2nd ed. Stockholm, Sweden: SIS Förlag AB, 2003.
[17] Swedish Standards Institute, ”Passenger cars - Steady-state circular driving behaviour
- Open-loop test methods (ISO 4138:2004, IDT),” 3rd ed. Stockholm, Sweden: SIS
Förlag AB, 2007.
[18] J. Páscoa, F. Charrua Santos, F. Brojo, P. M. O. Fael, ”An innovative experimental
on-road testing method and its demonstration on a prototype vehicle,”

53
Electromechanical Engineering Department, Faculty of Engineering, University of Beira
Interior, Covilhã, Portugal, 2012. Accessed: Sept. 23, 2022. [Online]. Available: https:
//www.researchgate.net/publication/257774870_An_innovative_experimental_
on-road_testing_method_and_its_demonstration_on_a_prototype_vehicle, DOI
10.1007/s12206-012-0413-8.

54
Volvo S90 data

Table A.1: Volvo S90 data

Inertia of vehicle 3089 kgm2


Inertia of engine 0.4 kgm2
Inertia of transmission 0.3 kgm2
Inertia of wheel 1.8 kgm2
Weight 1995 kg
Wheelbase 2.941 m
Track width 1.634 m
Center of gravity height 0.575 m
Frontal area 2.3 m2
Wheel radius 0.3433 m

Figure A.1: The engine map used in MATLAB/Simulink.

55
TRITA – SCI-GRU 2022:247

www.kth.se

You might also like