Vehicle Dynamics

Vehicle Dynamics
Compendium
2017
Bengt Jacobson et al
Vehicle Dynamics Group, Division Vehicle and Autonomous Systems,
Department Mechanics and Maritime Sciences, Chalmers University of Technology, www.chalmers.se
Preface 2017
This edition has various smaller changes and additions. Thanks to Fredrik Bruzelius (VTI), Tobias
Brandin (VCC), Niklas Fröjd (Volvo GTT), Assar Jarlsson (Kinnarps), Pär Pettersson (Chalmers), among
others. Thanks also to many students that have found and reported errors in the previous edition.
Bengt Jacobson, Göteborg, October 2017
Preface 2016
This edition has various changes and additions. Some of these are: Chapter 1: Control engineering, Chapter 2: Tyre models,
Driver models, Chapter 3: Propulsion systems, Varying road pitch, Non-reactive truck suspensions, Chapter 4: Track-ability,
Articulated vehicles, and Cambering vehicles.
Thanks to Cornelia Lex (TU Graz), Fredrik Bruzelius (VTI), Niklas Fröjd, Anders Hedman, Kristoffer Tagesson, Peter Nilsson,
Sixten Berglund (Volvo GTT), Tobias Brandin, Edo Drenth, Mats Jonasson (VCC), Mathias Lidberg, Artem Kusachov, Anton
Albinsson, Manjurul Islam, Pär Pettersson, Ola Benderius (Chalmers), Mats Sabelström, and Roland Svensson among others.
Bengt Jacobson, Göteborg, October 2016
Preface 2015
This edition has various changes and additions. Some of these are: brush model with parabolic pressure distribution, typical
numerical data for heavy vehicle, added “2.2.3 Tyre ”, “4.5.3.2 Example of explicit form model”, more about tyre
relaxation, introduction of neutral steering point, introduction of steady state roll-over wheel lift diagram. Thanks to Anton
Albinsson, Edo Drenth (VCC), Gunnar Olsson (LeanNova), Manjurul Islam, Mathias Lidberg, Mats Jonasson (VCC), Niklas Fröjd
(Volvo GTT), Ola Benderius, Pär Pettersson, and Zuzana Nedelkova among other.
Bengt Jacobson, Göteborg, 2015
Preface 2014
This edition has various small changes and additions. The largest changes are: Function definitions added and major update
of section 1.3, 3.1.1, 4.1.1, 5.1.1.
Thanks to Lars Almefelt from Chalmers, Jan Andersson from VCC, Kristoffer Tagesson from Volvo GTT and Gunnar Olsson
from Leannova and Karthik Venkataraman.
Preface 2013
This edition has various small changes and additions. The largest additions were in: Functional architecture, Smaller vehicles,
Roll-over, Pendulum effect in lateral load transfer and Step steer.
Thanks to Gunnar Olsson from LeanNova, Mathias Lidberg, Marco Dozza, Andrew Dawkes from Chalmers, Erik Coelingh from
Volvo Cars, Fredrik Bruzelius from VTI, Edo Drenth from Modelon, Mats Sabelström, Martin Petersson and Leo Laine from
Volvo GTT.
Preface 2012
A major revision is done. The material is renamed from “Lecture notes” to “Compendium”. Among the changes it is worth
mentioning: 1) the chapters about longitudinal, lateral and vertical are more organised around design for vehicle functions, 2)
a common notation list is added, 3) brush tyre model added, 4) more organised and detailed about different load transfer
models, and 5) road spectral density roughness model is added.
Thanks to Adithya Arikere, John Aurell, Andrew Dawkes, Edo Drenth, Mathias Lidberg, Peter Nilsson, Gunnar Olsson, Mats
Sabelström, Ulrich Sander, Simone Sebben, Kristoffer Tagesson, Alexey Vdovin and Derong Yang for review reading.
3
Preface 2011
Material on heavy vehicles is added with help from John Aurell. Coordinate system is changed from SAE to ISO. Minor
additions and changes are also done.
Preface 2007
This document was developed as a result of the reorganization of the Automotive Engineering Master’s Programme at
Chalmers in 2007. The course content has been modified in response to the redistribution of vehicle dynamics and power
train education.
These lecture notes are based on the original documents developed by Dr Bengt Jacobson. The text and examples have been
reformatted and edited but the author is indebted to the contribution of Dr Jacobson.
Rob Thomson, Gothenburg, 2007
Cover: Left column, from top: Steady state cornering tests with 3-wheeler prototype Trike in cooperation with Cycleurope.
High speed lane change tests with long heavy combination vehicle, an A-double heavy vehicle, tractor+semitrailer+steered
dolly+semitrailer. 3D printed scale models from BSc thesis. Right column: Figures from written examinations 2017.
This compendium is also available as fulltext pdf file at http://publications.lib.chalmers.se/
Paper version printed at Repro-centralen, Chalmers, Göteborg, Sweden, 2017
© Copyright: Chalmers University of Technology, Bengt Jacobson
4
Contents
* marks section with (Complete Vehicle Motion) Function
Preface 2017 3
CONTENTS 5
1 INTRODUCTION 9
1.1 Definition of Vehicle Dynamics 9
1.2 About this compendium 9
1.3 Automotive engineering 10
1.4 Attributes and Functions 10
1.4.1 Attributes 11
1.4.2 Functions 12
1.4.3 Requirements 13
1.4.4 Models, methods and tools 14
1.5 Technical References 14
1.5.1 Engineering 14
1.5.2 Modelling, drawing and mathematics 15
1.5.3 Mechanical engineering 23
1.5.4 Control engineering 30
1.5.5 Verification methods with real vehicle 34
1.5.6 Verification methods with virtual vehicle 34
1.5.7 (Computer) Tools & Methods 35
1.5.8 Coordinate Systems 38
1.5.9 Terms with special meaning 40
1.5.10 Architectures 43
1.6 Heavy trucks 45
1.6.1 General differences 45
1.6.2 Vehicle dynamics differences 46
1.6.3 Definitions 46
1.7 Smaller vehicles 46
1.8 Typical numerical data 48
1.8.1 For passenger vehicle 48
1.8.2 For heavy vehicle 49
2 VEHICLE INTERACTIONS 51
2.1 Introduction 51
2.1.1 References for this chapter 51
2.2 Wheel & Tyre Terminology & Design 51
2.2.1 Wheel angles 52
2.2.2 Tyre Design 53
2.2.3 Tyre Model Architecture 54
2.3 Longitudinal Properties of Tyres 57
2.3.1 Tyre Rolling and Radii 57
2.3.2 Rolling Resistance of Tyres 59
2.3.3 Longitudinal force due to slip 64
5
2.4 Lateral force due to slip 78
2.4.1 Tyre brush model for lateral slip 79
2.4.2 Lateral tyre (slip) stiffness 83
2.4.3 Empirical tyre models 84
2.4.4 Influence of vertical load 84
2.4.5 Relaxation in contact patch 85
2.4.6 Other lateral effects 86
2.5 Combined Longitudinal and Lateral Slip 88
2.5.1 Isotropic combined slip, any 𝑭 = 𝒇𝒔; 90
2.5.2 Isotropic brush model for combined slip 90
2.5.3 Anisotropic brush model, combined slip 91
2.5.4 An approximate combined slip model 93
2.5.5 Transients and Relaxation 94
2.6 Vertical Properties of Tyres 95
2.7 Tyre wear 95
2.8 Vehicle Aerodynamics 96
2.8.1 Longitudinal wind velocity 96
2.8.2 Lateral wind velocity 97
2.9 Driving and Vehicle Operation 97
2.9.1 Mission, Road and Traffic 97
2.9.2 Driver 97
3 LONGITUDINAL DYNAMICS 103

3.1 Introduction 103
3.2 Steady State Functions 103
3.2.1 Propulsion System 103
3.2.2 Traction diagram 112
3.2.3 Driving Resistance 114
3.2.4 Functions after start 115
3.2.5 Starting with slipping clutch 116
3.2.6 Steady state vertical force distribution over axles 117
3.2.7 Friction limit 118
3.2.8 Start functions 119
3.3 Functions over (long) cycles 120
3.3.1 Description formats of Vehicle Operation 120
3.3.2 Rotating inertia effects 122
3.3.3 Four quadrant traction diagram 124
3.3.4 Functions over cycles 125
3.3.5 Load Transfer with rigid suspension 127
3.3.6 Acceleration 129
3.4 Functions in (short) events 131
3.4.1 Typical test manoeuvres 131
3.4.2 Deceleration performance 131
3.4.3 Brake System 132
3.4.4 Pedal functions 133
3.4.5 Brake proportioning 134
3.4.6 Steady state heave and pitch, trivial linkage 137
3.4.7 Load Transfer with suspension compliance and linkage 139
3.4.8 Pitch functions at quick force changes 143
3.4.9 Acceleration and Deceleration 144
3.5 Control functions 146
6
3.5.1 Longitudinal Control 146
3.5.2 Longitudinal Control Functions 147
3.5.3 Longitudinal Motion Functionality in a reference architecture 152
4 LATERAL DYNAMICS 153

4.2 Low speed manoeuvers 153
4.2.1 Path with orientation 154
4.2.2 Vehicle and wheel orientations 154
4.2.3 Steering System 155
4.2.4 One-track models 159
4.2.5 Ideally tracking wheels and axles 160
4.2.6 One-track model for low speeds, with Ackermann geometry 160
4.2.7 Low speed functions 163
4.2.8 One-track model for low speeds, with non-Ackerman geometry 166
4.2.9 One-track model for low speeds, with trailer 168
4.3 Steady state cornering at high speed 168
4.3.1 Steady state driving manoeuvres 168
4.3.2 Steady state one-track model 170
4.3.3 Under-, Neutral- and Over-steering * 175
4.3.4 Normalized required steering angle 178
4.3.5 How to design for understeer gradient 180
4.3.6 Steady state cornering gains * 184
4.3.7 Manoeuvrability and Stability 187
4.3.8 Handling diagram 188
4.3.9 Lateral Load Transfer at Steady state cornering at high speed 190
4.3.10 High speed steady state vehicle functions 199
4.3.11 Roll-over in steady state cornering 201
4.4 Stationary oscillating steering 206
4.4.1 Stationary oscillating steering tests 206
4.4.2 Transient one-track model 207
4.4.3 Steering frequency response gains * 214
4.5 Transient handling 220
4.5.1 Transient driving manoeuvres * 220
4.5.2 Transient one-track model, without lateral load transfer 222
4.5.3 Transient two-track model, with Lateral load transfer 230
4.5.4 Step steering response * 239
4.5.5 Phase portrait 241
4.5.6 Long heavy combination vehicles high speed functions 242
4.6 Lateral Control Functions 244
4.6.1 Lateral Control Design 244
4.6.2 Lateral Control Functions 244
5 VERTICAL DYNAMICS 249

5.2 Suspension System 250
5.3 Stationary oscillations theory 251
5.3.1 Time as independent variable 252
5.3.2 Space as independent variable 255
7
5.4 Road models 257
5.4.1 One frequency road model 257
5.4.2 Multiple frequency road models 257
5.5 One-dimensional vehicle models 260
5.5.1 One-dimensional model without dynamic degree of freedom 260
5.5.2 One dimensional model with one dynamic degree of freedom 262
5.5.3 One dimensional model with two degrees of freedom 267
5.6 Functions for stationary oscillations 272
5.6.1 Ride comfort * 272
5.6.2 Fatigue life * 276
5.6.3 Road grip * 278
5.6.4 Other functions 278
5.7 Variation of suspension design 279
5.7.1 Varying suspension stiffness 280
5.7.2 Varying suspension damping 280
5.7.3 Varying unsprung mass 280
5.7.4 Varying tyre stiffness 280
5.8 Two dimensional oscillations 282
5.8.1 Heave and pitch 282
5.8.2 Heave and Roll 285
5.8.3 Three dimensional oscillations 285
5.9 Transient vertical dynamics 286
BIBLIOGRAPHY 289
INDEX 292
6 APPENDIX: REFERENCES FOR LECTURES 294

6.1 Course MMF062 Vehicle Dynamics 294
8
INTRODUCTION
1 INTRODUCTION
1.1 Definition of Vehicle Dynamics
Vehicle Dynamics is the engineering subject about vehicle motion in relevant user operations. It is an
applied subject, applied on a certain group of products, namely vehicles. Vehicle Dynamics always uses
theories and methods from Mechanical engineering/Machine design, but often also from
Control/Signal engineering and Human behavioural science.
1.2 About this compendium

This compendium is intended for first-time-studies studies of vehicle dynamics. At Chalmers
University of Technology this is in the course “MMF062 Vehicle Dynamics”. The compendium covers
more than included in the course; both in terms of subsystem designs, which not necessarily has to be
fully understood to learn the vehicle level; and in terms of some teasers for more advanced studies of
vehicle dynamics. Therefore, the compendium can also be useful in general vehicle engineering
courses, e.g. in the Chalmers course “TME121 Engineering of Automotive Systems”; and as an
introduction to more advance courses, which at Chalmers is the course “TME102 Vehicle Dynamics
Advanced”. The parts covered in the first course is indicated in Table 6.1.
The overall objective of the compendium is to educate automotive engineers in designing for, and
verifying, requirements on complete vehicle motion functions. The compendium focuses on road
vehicles, primarily passenger cars and commercial vehicles. Smaller road vehicles, such as bicycles and
single-person cars, are only very briefly addressed. It should be mentioned that there exist a lot of
ground-vehicle types not covered at all, such as: off-road/construction vehicles, tracked vehicles, horse
wagons, hovercrafts or railway vehicles.
The vehicle can be seen as a component (or sub-system) in a superior transport system consisting of
other road users, roads, and weather. A vehicle is also, in itself, a system within which many
components (or sub-systems) interact. The subsystems are one way to structure the area.
Organisation after vehicle motion directions is another, which is used in the compendium:
• Chapter 3: Longitudinal dynamics (also introducing subsystems for propulsion and braking)
• Chapter 4: Lateral dynamics (also introducing subsystems for steering)
• Chapter 5: Vertical dynamics (also introducing subsystems for suspension)
Chapter 1 is introduction about automotive industry and required background knowledge. Chapter 2
is about what interacts with the vehicles from outside, like aerodynamics, driver and tyre. The sub-
system tyre is equally important for all directions.
Chapters 3-5 leads all the way to “functions”. Functions are needed for requirement setting, design and
verification. The overall order within the compendium is that models/methods/tools needed to
understand each function are placed before the functions.
It is important to qualitatively understand the characteristics of the vehicle’s sub-systems and, from
this, learn how to quantitatively predict and analyse the complete vehicle’s behaviour. These skills
support the overall goal of engineering design for desired attributes and functions, see Section 1.3. The
subsystem sections are:
• Wheel and tyre in sections 2.2-2.7
• Propulsion System in section 3.2.1
• Brake System in section 3.4.3
• Steering System in section 4.2.3
• Suspension System in sections 5.2 (and some in 3.4.7 and 4.3.9)
9
INTRODUCTION
The reader of this compendium is assumed to have knowledge of mathematics and mechanics, to the
level of a Bachelor of Engineering degree. Previous knowledge in dynamic systems, e.g. from Control
Engineering courses, is often useful. There are no prerequisites on previous knowledge in automotive
engineering, because basic vehicle components (or sub-systems) are presented to the necessary level
of detail. References and suggestions for further reading are provided and readers are directed to
these sources of information for a deeper understanding of a specific topic.
1.3 Automotive engineering

This section is about a larger area than Vehicle Dynamics. It is about the context where Vehicle
Dynamics is mainly applied.
OEM means Original Equipment Manufacturer and is, within the automotive industry, used for a
vehicle manufacturer. OEM status is a legal identification in some countries.
In the automotive industry, the word Supplier means supplier to an OEM. There are Tier1 suppliers,
Tier2 suppliers, etc. A Tier1 supplies directly to an OEM. A Tier2 supplies to a Tier1 and so on.
Primarily, suppliers supply parts and systems to the OEMs, but suppliers can also supply competence,
i.e. consultant services.
From an engineering view, an OEM does Product Development and Manufacturing. But it is good to
remember that there is also Purchasing, Marketing & Sales, After Sales, etc. However, Product
Development is the main area where the vehicles are designed. It is typically divided into Powertrain,
Chassis, Body, Electrical and Vehicle Engineering. Vehicle Dynamics competence is mainly needed in
Chassis, Powertrain and Vehicle Engineering.
On supplier side, Vehicle Dynamics competence is mainly needed for system suppliers that supplies
propulsion, brake, steering and suspension systems.
Except for OEMs and suppliers, Vehicle Dynamics competence is also needed at some governmental
legislation and testing as well as research institutes.
There are engineering associations for automotive engineering. FISITA (Fédération Internationale des
Sociétés d'Ingénieurs des Techniques de l’ Automobile, www.fisita.com) is the umbrella organisation
for the national automotive societies around the world. Examples of national societies are IMechE
(United Kingdom), JSAE (Japan), SAE (USA), SAE-C (Kina), SATL (Finland), SIA (France), SVEA
(Sweden, www.sveafordon.com) and VDI FVT (Germany). There is a European level association also,
EAEC.
1.4 Attributes and Functions

Development of a vehicle is driven by requirements which come from:
• Use case based needs of the customer/users
• Legislation from the authorities and Rating from consumer organisation, and
• Engineering constraints from the manufacturer’s platform/architecture on which the vehicle
should be built.
One way of organising most of those, is to define Attributes and Functions. The terms are not strictly
defined and may vary between vehicle manufacturers and over time. With this said, it is assumed that
the reader understand that the following is an approximate and simplified description.
In this compendium, both attributes and functions implicitly concern the complete vehicle; not the
sub-systems within the vehicles and not the superior level of the transportation system with several
vehicles in a road infrastructure.
Attributes and Functions are used to establish processes and structures for requirement setting and
verification within a vehicle engineering organisation. Such processes and structures are important to
enable a good overall design of such a complex product as a vehicle intended for mass production at
10
INTRODUCTION
affordable cost for the customers/users. Figure 1-1 tries to give an overview, with reference to the
well-known V-process, how a vehicle program generally is organised. The figure is idealized in many
ways. First, one have more levels of functions and subsystems. Secondly, it is difficult to keep such a
clean hierarchical order between functions and subsystems. Thirdly and most conceptually difficult, is
the fact that each subsystem gets requirements from many functions as showed by the dashed lines;
this makes the complexity explode.
Traditional V-process An extended V-process for a vehicle programme
for product development Product Customer/User
planning perception
Describe Complete Validate
Vehicle Attributes Complete Vehicle
Design loops on
Design Vehicle program level
(pre-series, virtual and real)
Set Requirements on Verify Requirements
Complete Vehicle Functions. on Complete Vehicle
Decompose in subsystem. Design loops per

Verify Requirements
Break down to Requirements on Subsystem. Subsystem
on Subsystems
Drawn as a W-process to mark
that multiple subsystems
Figure 1-1: V-process applied to a vehicle program. Note that it becomes a W-process, since multiple
subsystems.
1.4.1 Attributes
An attribute is a high-level aspect of how the user of the vehicle perceives the vehicle. Attributes which
are especially relevant for Vehicle Dynamics are listed in Table 1.1. The table is much generalised and
the attributes in it would typically need to be decomposed into more attributes when used in the
engineering organisation of an OEM. Also, not mentioned in the table, are attributes which are less
specific for vehicle dynamics, such as Affordability (low cost for user), Quality (functions sustained
over vehicle lifetime), Styling (appearance, mainly visual), etc.
Attributes can be seen as one way to categorise or group functions, especially useful for an OEM
organisation and vehicle development programs. Functions can, in turn, be seen as one way to group
requirements. Legal requirements are often, but not always, possible to trace back to primarily one
specific attribute. Requirements arising from OEM-internal platform and architecture constraints
are often more difficult to trace in that way. Hence, legal and platform/architecture can be seen as two
“attribute-like requirement containers”, beside the other user-derived attributes.
Table 1.1: Attributes
Attribute Description
This attribute means to maximize output from and minimize costs for transportation. Transport
output can be measured in 𝑝 𝑟𝑠𝑜𝑛 𝑘 𝑡𝑜𝑛 𝑘 or 𝑘 . The costs are mainly energy costs and
Transport Efficiency
time, but also wear of vehicle parts influence. The attribute is most important for commercial
vehicles, but becomes increasingly important also for passenger vehicles. The attribute is mainly
addressing long-term vehicle usage pattern, typically 10 min to 10 hours. There are different ways to
define such usages, e.g. (Urban / Highway / Mixed) driving cycles. So far, the attribute is mainly
required and assessed by the vehicle customers/users.
The attribute can also be seen to include “Environmental Efficiency”, which means low usage of
natural resources (mainly energy) and low pollution, per performed transport task. This is to a large
extent required and assessed by society/legislation.
11
INTRODUCTION
Attribute Description
Minimizing risk of property damages, personal injuries and fatalities both in vehicle and outside,
Safety
while performing the transportation. This attribute is to a large extent required and assessed by
society/legislation. In some markets, mainly developed countries, it is also important for vehicle
customers/users.
How the occupants (often the driver) experiences the vehicle during transport; from relaxed
transport (comfort) to active driving (sensation). This attribute contains sub-attributes as:
User Experience (Driver Experience)
• Ride comfort. Ride comfort often refers to vibrations and harshness of the occupants’ motion,
primarily vertical but secondly longitudinal. So, V and H in NVH (=Noise, Vibration and
Harshness) is included. If expanding to “comfort” it would include also N (noise) and
compartment air conditioning, but these sub-attributes are less related to vehicle dynamics.
• Performance describes how the vehicle can perform at the limits of its capabilities;
acceleration, deceleration or cornering. Most often, it refers to longitudinal limitations due to
propulsion and brake systems limitations.
• Driveability, Handling and Road-holding describes how the vehicle responds to inputs from
driver and disturbances, and how driver gets feedback from vehicle motion e.g. through
steering feel. It is also the corresponding response aspects for a “virtual driver”, i.e. a control
algorithm for automated driving. Driveability often refers to longitudinal (acceleration,
braking gear shifting). Handling and Road-holding often refer to lateral manoeuvres.
• Trust in automated driving becomes increasingly important and it balanced; high trust but
not over-trust.
This attribute is to a large extent required and assessed by the vehicle customers/users, both through
own experience but to a large extent indirectly via assessments by experts, e.g. in motor journals.
1.4.2 Functions
A function is more specific than an attribute. Ideally, it defines a measure which is possible to set
(quantitative) requirements on, see Section 1.4.3. The (complete vehicle) function does not primarily
stipulate any specific subsystem. However, the realisation of a function in a particular vehicle
programme, normally only engages a limited subset of all subsystems. So, the function will there pose
requirements on that subsystem. It is easy to mix up whether a function origins from an attribute or a
subsystem. One way to categorise functions is to let each function belong to the subsystem which it
mainly implies requirements on rather than the source attribute. Categorizing functions by
subsystems tends to lead to “carry-over” function realisations from previous vehicle program, which
can be good enough in many cases. Categorizing functions by source attribute facilitates more novel
function realisations, which can be motivated in other cases.
The word “function” has appeared very frequently lately along with development of electrically
controlled systems. The function “Accelerator pedal driving” in Section 3.5.2.1 has always been there,
but when the design of it changed from mechanical cable and cam to electronic communication and
algorithms (during 1990’s) it became much more visible as a function, sometimes referred to as
“electronic throttle”. The point is that the main function was there all the time, but the design was
changed. The change of design enabled, or was motivated by, improvement of some sub-functions, e.g.
idle speed control which works better in a wider range of engine and ambient temperature.
At some places, the compendium emphasizes the functions by adding an asterisk “*” in section
heading and a “Function definition” in the following typographic form:
Function definition: {The Function} is the {Measure} … for {Fixed Conditions} and certain
{Parameterized Conditions}.
The {Measure} should be one unambiguously defined measure (such as time, velocity or force) of
something the vehicle does. The {Measure} is ideally a continuous, objective and scalar physical
quantity, subjected for setting a requirement on the vehicle. The {Fixed Conditions} should be
unambiguously defined and quantified conditions for the vehicle and its surroundings. The keyword
“certain” identifies the {Parameterized Conditions}, which need to be fixed to certain (particular)
numerical values before using the Function definition for requirement setting. A requirement can be:
“{Measure} shall be =, , > or < {value*unit}”, see more in Section 1.4.3.
12
INTRODUCTION
Since the term “Function” is defined very broadly in the compendium, these definitions become very
different. One type of Function definition can be seen in beginning of Section “3.2.4.1 Top speed”,
which includes a well-defined measure. Another type of Function definition is found in Section “3.5.2.3
Anti-lock Braking System, ABS” and “4.3.3 Under-, Neutral- and Over-steering”. Here, the definitions
are more on free-text format, and an exact measure is not so well defined.
The compendium defines functions on complete vehicle level, i.e. not cascaded to requirements on
specific sub-systems.
1.4.2.1 Organise functionality in functions

There is, of course, no physically correct or incorrect way to organise the total functionality in a
vehicle into a finite list or structure of functions. However, defining separate functions is very helpful
when organising a combined hardware and software development process within a vehicle
programme or when inheriting functionality between vehicle programmes. Some examples of
categorisation of functions are:
• Functions assigned to the (sub-)system which the function mainly belong (purchase wise or
technically). Such categories can be, e.g., Brake functions and Steering functions.
• Functions sorted in “Customer functions” and “Support functions”. Support functions are
typically used by many customer functions, see Section “1.5.10.2 Vehicle function architecture”.
• Functions sorted in whether they are realized solely mechanically or if realization involve
electronics.
• Functions sorted in Standard and Option in the product to end customer.
• Seeing vs Blind functions, differentiated by whether the function acts on observed
environment to the vehicle or only on subject vehicle own states.
• Dynamic vs Driver-informing functions, differentiated by whether the function actuates
vehicle motion or only inform/warn driver.
1.4.3 Requirements
A requirement shall be possible to verify as pass or fail. A requirement on the complete vehicle is
typically formulated as:
“The vehicle shall … do something or have measure … < 𝑜𝑟 > 𝑜𝑟 ≈ … number [unit] …
… under certain conditions.”
Examples:
• The vehicle shall accelerate from 50 to 100 km/h on level road in <5 s when acceleration pedal
is fully applied.
• The vehicle shall decelerate from 100 to 0 km/h on level road in <35 m when brake pedal is
fully applied, without locking any rear wheel.
• The vehicle shall turn with outermost wheel edge on a diameter <11m when turning with full-
lock steering at low speed.
• The vehicle shall have a characteristic speed of 70 km/h (10 km/h) on level ground and high-
friction road conditions and any recommended tyres.
• The vehicle shall give a weighted RMS-value of vertical seat accelerations < 1.5 ⁄𝑠 when
driving on road with class B according to ISO 8608 in 100 km/h.
• The vehicle shall keep its body above a 0.1 m high peaky two-sided bump when passing the
bump in 50 km/h.
In order to limit the amount of text and diagrams in the requirements it is useful to refer to ISO and
OEM specific standards. Also, it is often good to include the purpose in a use case, when documenting
the requirement.
The above listed requirements stipulate the function of the vehicle, which is the main approach in this
compendium. Alternatively, a requirement can stipulate the design of the vehicle, such as “The vehicle
shall weigh <1600 kg” or “The vehicle shall have a wheel base of 2.5 m“. The first type (above listed)
13
INTRODUCTION
can be called Performance based requirement. The latter type can be called Design based requirement or
Prescriptive requirement and such are rather “means” than “functions”, when seen in a function vs
mean hierarchy. It is typically desired that legal requirements are Performance based, else they would
limit the technology development in society.
1.4.4 Models, methods and tools

The attributes, functions and requirements are top level entities in vehicle development process. But
in order to define/formulate them, design for them and verify them, the engineers need knowledge in
terms of models, methods and tools.
As mentioned above, some sections in the compendium have an asterisk “*” in the section heading, to
mark that they explain a function, which can be subject for complete vehicle requirement setting. The
remaining section, without an asterisk “*”, are there to give the necessary knowledge to understand
the function. It is the intention that the necessary knowledge for a certain function appears before that
function section. One example is that the Sections “3.2.1 Propulsion System”, “3.2.2 Traction diagram”
and “3.2.3 Driving Resistance” are placed before Section “3.2.4.1 Top speed *”. The functions only
appear in Chapter 3, Chapter 4 and Chapter 5.
1.5 Technical References

This section introduces notations. The section also covers some basic disciplines, methods and tools.
Parts of this section probably repeat some of the reader’s previous education.
1.5.1 Engineering
Engineering Design or Engineering (in Swedish “Ingenjörsvetenskap”, in German
“Ingenieurwissenschaften”), is not the same as Analysis, Inverse analysis and (Nature) Science, see
Figure 1-2. In education, it is often easiest to do Analysis (or Inverse Analysis). However, in a product
based subject, such as Vehicle Dynamics, it is important to keep in mind that the ultimate use of the
knowledge is engineering, which is to propose and motivate the design and actual numerical values of
design parameters of a product. The distinction between Analysis and Inverse Analysis can only be
made if there is a natural direction (input or output) of interface variables.
Input Output
System
Laws of Nature
Given Find Process
Input, Laws of Nature, System Output Analysis
Output, Laws of Nature, System Input Inverse Analysis
Input, Output, System Laws of Nature (Nature) Science (induction)
Input, Output, Laws of Nature System Engineering Design (synthesis)
Dixon, J.R., (1966) Design Engineering, Inventiveness, analysis and Decision Making
Figure 1-2: Distinction between Engineering and related activities. Picture from Stefan Edlund, Volvo
Trucks.
14
INTRODUCTION
Engineering design (synthesis)

Requirements (Required output)
Requirements Test the design, with Analysis with a model:

Create or Design
(Required
improve parameters
output for Resulting
design certain Model
given input) (incl. Laws of achieved Good yes design
given
input Nature) output enough? of the
System
no
Re-design through experience/understanding
Design parameters varied to find best design Other parameters varied in order
(optimization) or good enough design (satisficing). to make the design robust.
Figure 1-3: How Analysis can be used as a sub-process in Engineering design.
1.5.1.1 Vehicle Dynamics role in industry

The activity type that sets the pace in automotive industry are vehicle programme or projects. It
defines the technology to be developed, the time and cost aspects. The work is organised around
such programmes, both at vehicle manufacturers and their sourced subsystem supplier. One way to
exemplify such is Figure 1-4. Engineers with vehicle dynamics profile are typically active at
departments called Chassis, Complete Powertrain, Electrical or Vehicle Engineering. They are
responsible for deliverables to vehicle programme in the form of:
• Hardware: Geometry, Strength and function per subsystem
• Software: ECU and SW releases per subsystem, Functions such as ABS, ESC, ACC
• Requirement setting and verification: Handling, Driveability, Brake performance, Ground
clearance, Ride comfort, Energy consumption. Verification in real and virtual tests.
The subsystems are typically those mentioned in Section 1.2.
year1 year2
Pre-Series1 Pre-Series2 Pre- Start
(mule=old (new chassis) Series3 Of
chassis) (final Production
chassis)
Research
projects
Geometry package, Changing geometry, Changing bushings,
Strength of parts, Function content, Tuning SW,
Signal interface, … …
… within subsystems within subsystems
between subsystems (hi-&lo mu tests) (hi&lo-mu tuning)
Figure 1-4: An example of vehicle programme.
1.5.2 Modelling, drawing and mathematics

1.5.2.1 Stages in Model Based Engineering
For the analysis in Figure 1-3, models are needed. A model is a representation of something from real-
world during a certain event or course of events, such as the model of the longitudinal motion of a car
during acceleration from stand-still to 100 km/h. Models are always based on assumptions,
approximations and simplifications. However, when using models as a tool for solving a particular
problem, the models at least have to be able to reproduce the problem one is trying to solve. Also,
models for engineering have to reflect design changes (e.g. changed parameters or input variable) in a
representative way, so that new designs can be proposed which solve/reduce the problem. Too
advanced models tend to be a disadvantage, since they require and produce a lot of data.
The models typically used in vehicle dynamics can be called lumped dynamic models (many alternative
names: functional models, one-dimensional models, system models, DAE-model, ODE-models, etc.).
15
INTRODUCTION
The models are typically multi-domain type, involving mechanics, hydraulics, pneumatics, electric,
control algorithms, driver’s actions, etc. The models typically lump (or discretises) the system.
Examples of modelling methods which are more seldom used in vehicle dynamics are: Finite Elements,
Computational Fluid Dynamics, CAD geometry models, etc.
One can identify modelling in 3 stages in the overall process of engineering, Sections 1.5.2.1.1 to
1.5.2.1.7 and Figure 1-5. In this process we identify 3 modelling stages: Sections 1.5.2.1.2 Physical
modelling 1.5.2.1.3 Mathematical modelling, and 1.5.2.1.4 Explicit form modelling. The compendium
spends most effort on the first 2 of those. (The modelling and computation stages in the process can,
conceptually, be replaced with building a real prototype and run real prototype tests, respectively.)
Computation
/Simulation
Explicit form modelling
Evaluate model validity
Mathematical modelling
Theoretical world
Physical modelling
Real world
Formulate engineering Evaluate requirement fulfilment,
task (problem) Redesign/Propose final Design
Figure 1-5: Modelling and Real versus Theoretical world.

It can be noted that the process assumes physical modelling. As opposite to this, one can think of
formulating a mathematical model without having a physical model in mind, e.g. assuming a certain
polynomial. Without a physical model, the parameters are generally not interpretable to real design
parameters, sometimes their units can even be unknown. For vehicle design the vehicle model should
be physical, so that its real-world design parameters can be identified, but a driver model can be useful
also without strict physical model, since the driver is not to be designed.
1.5.2.1.1 Formulate the Engineering design task
Based on a problem description, one formulates the engineering design task, which describes the
required design decisions (if possible, appoint design parameters) for the system (an existing product
or a concept/drawing). Also, requirements on how the output of the system should react in certain
vehicle operations (or manoeuvres or situations) have to be identified. Requirements can be either
constraints (something<number) and/or optimization (some scalar to be minimized).
The conceptual idea with requirements in vehicle industry is that they are set independently of the
design solution. However, the general requirements are seldom neither complete for nor directly
applicable on each engineering design task; one often have to question the requirements and invent
new. The operation parameters often have a range/spread/stochasticity to design for, see Figure 1-3.
Examples of such parameters describe traffic situation, driver, tyre/road characteristics, weather, see
Section 1.5.2.5.1. The range can be searched for in feedback or logged data from customers and
accident statistics. But it is also very important with experienced engineering judgment, to forecast
how the problem will appear in a future context, with a future fleet of vehicles and a future road
infrastructure. Design methodology can be used to reason about parameters and categorize
parameters (e.g. Taguchi’s: “signal, response, noise and control factors”).
1.5.2.1.2 Physical modelling
In this stage, one should generate a physical model, which means sketches and text. Free-body
diagrams and assumptions (see Section 1.5.3.1) on operating conditions (see Section 1.5.3.2) are
important ingredients in the physical model. The physical model shall clarify assumptions for different
parts of the system, e.g. rigid/elastic and inertial/massless. How to model here should be based on
what physical phenomena is needed to be captured and what physical variations (e.g. which design
parameters) are intended for simulations with the model. A list of parameters and variables are
suitably introduced and defined, for reuse in mathematical model.
1.5.2.1.3 Mathematical modelling
In this stage, one should generate a mathematical model, which means equations. Basically, it is
about finding exactly the right number of equations; equally many as unknown variables. Here, the
16
INTRODUCTION
assumptions are transformed into equations. For dynamic systems, the equations form a “DAE”
(Differential-Algebraic system of Equations). It is seldom necessary to introduce derivatives with
respect to other independent variables, such as positions, i.e. one does seldom need PDE (Partial
Differential Equations). The general form of a DAE:
• 𝒇𝑫𝑨𝑬 (𝒛̇ 𝒛 𝑡) = 𝟎;
The 𝒛 are the (dependent) variables and 𝑡 is time (independent variable). The mathematical model is
complete only if there are equally many (independent) equations in 𝒇𝑫𝑨𝑬 as there are variables in .
Note that the DAE is generally on implicit form, as opposed to explicit form below. Note also that it is
not only a question of finding suitable equations, but also to decide parameterisation, which is how
parameters relate to each other. The parameterisation should reflect a “fair” comparison between
different design parameters, which often requires a lot of experience of the product and the full set of
requirements on the vehicle. Also, selection of output variables is important so that output variables
are enough to evaluate the requirements on the system. Selecting more might drive unnecessary
complex models.
The Mathematical model is declarative, i.e. describes relations needed to find the output variables.
How to find the solution is not described until next stage, the Explicit form model, which is imperative,
i.e. describes how the output variables is obtained.
1.5.2.1.4 Explicit form modelling
In this stage, one should generate an explicit form model, which means equations rearranged to
assignment statements, i.e. to an explicit form (algorithm) which outputs the state derivatives. You
probably recognize this formulation as “ODE” (Ordinary Differential Equation) or “IVP” (Initial Value
Problem). The general form is:
• 𝒙̇ = 𝒇𝑶𝑫𝑬 (𝒙 𝒖(𝑡) 𝑡); 𝒚 = 𝒈(𝒙 𝒖 𝑡);
The 𝒙 is the state variables, 𝒖 is the input variables, and 𝒚 the output variables. In the mathematical
model, there was no distinction between different dependent variables in . However, in the explicit
model, each variable in z can be identified as belonging to either of 𝑥 𝑦 or 𝑢.
In linear cases, the State Space Form can be used:
• 𝒙̇ = 𝑨(𝑡) 𝒙 + 𝑩(𝑡) 𝒖(𝑡); 𝒚 = 𝑪(𝑡) 𝒙 + 𝑫(𝑡) 𝒖(𝑡); where 𝑨 𝑩 𝑪 𝑫 are matrices.
A dataflow diagram (see Figure 1-6 and Figure 1-14) is a graphical representation of the explicit form.
It is drawn using blocks with input and output ports and arrows representing signals between;
integration is represented by integration blocks with 𝒙̇ as input and 𝒙 as output.
1.5.2.1.5 Computation
Computation, is to find numerical solution. Several methods can be used, e.g.:
• Simulation (e.g. Initial value problem, IVP or End value problem, EVP). There are many
advanced pre-programmed integration methods and tools which one can rely on without
knowing the details. It is often enough to know that the concept is similar to the simplest of
them, “Euler forward with constant time step”, in which the state variables are updated in each
time step, Δ𝑡, as follows:
𝒙̇ 𝒏𝒐𝒘 = 𝒇𝑶𝑫𝑬 (𝒙(𝑡𝑛𝑜𝑤 ) 𝒖(𝑡𝑛𝑜𝑤 ) 𝑡𝑛𝑜𝑤 ); (Explicit form model)
𝒙(𝑡𝑛𝑜𝑤 + Δ𝑡) = 𝒙(𝑡𝑛𝑜𝑤 ) + Δ𝑡 ∙ 𝒙̇ 𝒏𝒐𝒘 ; (Derivative approximation)
• Linearization. The model can sometimes be linearized to 𝒙̇ = 𝑨(𝒕) ∙ 𝒙 + 𝑩(𝒕) ∙ 𝒖; and
𝒚 = 𝑪(𝒕) ∙ 𝒙 + 𝑫(𝒕) ∙ 𝒖;. Such numerically known matrices are very useful for eigen-modes,
eigen-frequencies, step response, stability, or use as model base in model based control design
methods.
• Optimization. Either optimizing a finite number of defined design parameters or time
trajectories, e.g. 𝑢(𝑡). There are many optimization methods, ranging from trial-and-error to
mathematically/numerically advanced gradient based or evolutionary inspired methods. If
optimizing time trajectories, one often discretize the model in time, which converts the
17
INTRODUCTION
differential equations into difference equations, using a certain derivative approximation, such
as 𝒙𝒌+𝟏 = 𝒙𝒌 + Δ𝑡 ∙ 𝒙̇ 𝒌 = 𝒙𝒌 + Δ𝑡 ∙ 𝒇(𝒙𝒌 𝒖(𝑡𝑘 ) 𝑡𝑘 ) = 𝒙𝒌 + Δ𝑡 ∙ 𝒇(𝒙𝒌 𝒖𝒌 𝑘 Δ𝑡);. Typically, very
simple derivative approximations, compared to today’s integration methods for simulation,
should be used.
1.5.2.1.6 Evaluation and Redesign
Both model validity and requirement fulfilment should be evaluated.
Often, there is an old design which was the cause for the whole engineering task and then there might
be test data to validate model with. Such validity check will be for one certain design. If directly
comparing one simulation with one test it is an absolute model validity. If comparing the change
between two tests where something has been changed, it is a relative model validity, e.g. telling if the
model well reflects design parameter changes. The relative model validity is often more important
than the absolute. Since one often don’t know exactly how the model is going to be used, it can be a
good habit to include a boolean auxiliary variable for validity. One clear reason to set the validity false
is when the physical model assumptions are not met, such as if a wheel lifts from ground during
simulation (assuming equations to handle wheel lift is not included). In other cases, it can be more of
an engineering judgement on what threshold to set the validity false. Then, if validity is false during
only a tiny fraction of the whole studied time, it is again a judgement whether to disqualify the
simulation result or not.
Evaluation of requirement fulfilment (can be called Analysis, as opposed to Synthesis) involves
understanding and interpretation of the computation results to real world. New requirements can
appear, typically new constraints that becomes active with a new design is tested.
Redesign (or Synthesis, as opposed to Analysis) is normally a creative part where experience and
intuition is important. In some rare cases, one can inverse calculate the new design or numerically
optimize to a new better design.
1.5.2.1.7 Propose final Design
Propose design, meaning a concrete proposal of design, such as a numerical design parameter value
or a drawing or control algorithm.
1.5.2.1.8 Closed form or analytic solutions
Sections 1.5.2.1.1 to 1.5.2.1.7 assumes numerical solution. If instead aiming at a Closed form solution or
Analytic solution, [𝒙 𝒚] = 𝒇𝒂𝒏𝒂𝒍𝒚𝒕𝒊𝒄 (𝑡);, the stage Explicit form modelling and Computation becomes
much different.
Starting from the mathematical model it is it is often motivated to reduce number of equations by
eliminating variables. This can help to understand the behaviour of the whole system, but it is often
less clear which physical phenomena has been modelled. The time differentiation order can increase
when eliminating variables, so that 𝒛̈ 𝒛
⃛ ⋯ can occur in equations. It is always easy to go from a few
higher order differential equations to more first order differential equations. The opposite direction is
not as easy. Examples where closed form is possible are linear models with inputs 𝒖(𝑡) expressed as a
simple time functions, such as step functions or harmonic functions.
If closed form solutions can be found, they are superior for understanding of how design parameters
influence the requirement fulfilment; there can even be possibilities to do analytic optimization, see
(Ross, 2015).
1.5.2.2 Discrete dynamics

The bullet list above assumes only continuous dynamics, as opposed to also including discrete
dynamics. Discrete dynamics modelling can be used for computational efficient models of, e.g., ideal
dry friction or ideal backlash. The way the states evolve over time differs between continuous and
discrete dynamics: Continuous states, 𝒙𝒄 , are updated in every time step using a derivative
approximation, e.g. 𝒙̇ 𝒄 = (𝒙𝒄 (𝑡 + Δ𝑡) 𝒙𝒄 )⁄Δ𝑡 ;. Discrete states, 𝒙𝒅 , are updated only at the time
instants (not time step!) when one of the transition conditions, 𝒉, becomes true. These time instants
18
INTRODUCTION
are called events. When a system includes both types (hybrid dynamics) the states evolve as follows:
𝒙𝒄 (𝑡 + Δ𝑡) = 𝒙𝒄 + Δ𝑡 ∙ 𝒙̇ 𝒄 = 𝒙𝒄 + Δ𝑡 ∙ 𝒇𝒄 (𝒙𝒄 𝒙𝒅 𝒖(𝑡) 𝑡); and
𝑤ℎ 𝑛 𝒉(𝒙𝒄 (𝑡) 𝒙𝒅 (𝑡) 𝒖(𝑡) 𝑡) > 𝟎 𝑡ℎ 𝑛 𝒙𝒅 (𝑡 + ) = 𝒇𝒅 (𝒙𝒄 (𝑡) 𝒙𝒅 (𝑡) 𝒖(𝑡) 𝑡);.
Some continuous state variables can also be reinitialized (𝒙𝒄 (𝑡 + ) = ⋯) in some events, e.g. motivated
by impact dynamics. There is a risk that the numerical solution gets stuck in undesired high frequency
switching of discrete states which is called chattering. Chattering can be caused by the computational
method, but is very often also due to inconsistent physical modelling. Discrete dynamics is not as well
established in most basic engineering education as continuous dynamics. A special type of discrete
dynamics is time events, where 𝒉 = 𝒉(𝑡);. Time events often appears when modelling the sampling in
digitally computed control algorithms.
1.5.2.3 Simultaneous equations or Algebraic loops

One problem that can appear for a mathematical model is Simultaneous equations, also called
Algebraic Loop. A simple example of this is when two equations need to be solved before each other,
see Figure 1-6. The model example is a vehicle decelerating with locked wheels; Vertical forces are
needed to calculate friction forces, which are needed to calculate acceleration, which is needed to
calculate vertical forces.
Loops appear often between bigger chunks of equations than in Figure 1-6, namely between modules
in the model. The decomposition in certain modules is often desired for model modularity. A typical
other example where a loop occurs is when doing quasi-steady state assumptions about roll moment
distribution between axles (4.3.9.3).
Loops can always be conceptually solved by setting up an iteration in each evaluation of time
derivatives. This can sometimes be accepted, but it gives poor computational speed. Other ways are to
handle the loops by sacrificing simplicity in model is to model an additional elasticity or inertia
between the modules, or simply a (non-physical) filter. “Memory blocks” can work but are not
recommended, since the simulations will be dependent on solver. Best is if the loop can be
symbolically solved, as in Figure 1-6.
Simultaneous equations (Algebraic loop)
If we guess we can
̇ calculate all unknowns,
= + ; including .
ℎ
̇+ + = ;
If the guess is adjusted
̇ ⁄ ⁄ = ; until it gives same
̇ ℎ+
calculated we have
= ̇ 𝑡; an iterative solution.
3 equations, 3 unknowns ( ): Solved by symbolic manipulation of equations:

= + ; ̇= ;
̇+ + = ; = 1⁄ ̇ ℎ ; The loop could, in this case, be
symbolically solved, so that we
̇ ℎ+ ⁄ ⁄ = ; = ; get this explicit solution.
(with as state, = ̇ 𝑡 ;) ( = ̇ 𝑡 ;)
Figure 1-6: An example of Simultaneous equations, also called Algebraic loop.
1.5.2.4 High index

One problem that can appear for a mathematical model is High index. A simple example of High index
system of equations is when two inertias are rigidly connected to each other, see Figure 1-7, where a
working causality cannot be found. Similarly to simultaneous equations, it can be solved with equation
manipulations, but it has to include also differentiation of some equations.
Similar cures as for simultaneous equations. It can be tempting to try a “differentiating” block ( ̇ =
𝑑
; instead of = ̇ 𝑡 ; in Figure 1-7) but that will not work. Other examples where high index
𝑑𝑡
occurs is when modelling purely rolling wheels (3.3.2).
19
INTRODUCTION
High index No working causality:

̇ 2 in-ports for
̇ = ; = ̇ 𝑡 ; and 2 out/ports
for .
̇ ̇ = ;
̇
̇ = ; = ̇ 𝑡;
3 equations, 3 unknowns ( ): Solved by symbolic manipulations:

̇ = ; 1. Differentiate ” = ;” to ” ̇ = ̇ ;”
̇ = ; 2. Algebra gives:
= ; ̇ = ⁄ + ;
(with as states, = ̇ 𝑡; 𝑜𝑟 = 1 ) = ̇ ;
= ̇ 𝑡;
= ; and ̇ = ̇ ;”
Figure 1-7: An example of a High index model; a rigidly connected car and trailer during acceleration.
1.5.2.5 Constants, Parameters, Variables and Signals

Variables and Signals vary with time during an experiment with the model. Constants and
parameters do not. In the following, it is understood that all have physical interpretation, e.g. a
corresponding (physical) quantity with a known unit. Units are central for engineering. A
{𝑞𝑢 𝑛𝑡 𝑡𝑦} = 𝑠𝑢𝑟 [𝑢𝑛 𝑡]; such as { 𝑛 𝑡ℎ} = [ ]= [ ];. If using a consistent set of units,
such as SI-units, each equation can be seen as either a relation between quantities or relation between
measures. The compendium assumes SI-units, where else is not stated. It is recommended to not use
numerical constants directly in the equations unless they are dimensionless=unitless, such as 𝜋 𝑜𝑟 .
1.5.2.5.1 Constants and Parameters
A parameter can be changed between subsequent experiments. A constant does not even change
between experiments. Typical constant is 𝜋 or gravity constant . Typical parameters are vehicle
wheelbase and road friction coefficient.
From vehicle engineering point of view there is a significant difference also between Design
parameters and Operation/Manoeuvre parameters. The first ones are intended to be varied to find a
good value. The latter ones can also vary, but this variation is rather a disturbance to designed for.
It is often natural to have different parameter sets to describe the design and operation/manoeuvre
and another for use in the equations. For example, the kerb weight (design) and load (operation) are
good for description while the total weight (sum of kerb and load) is the parameter that appears in the
equations. The parameters that appears in the equations cannot always be categorized as either design
or operation/manoeuvre parameters. It is easy to forget that the relations between parameters are
often an equally important part of the engineering problem, as the relation between variables (the
equations) is for the dynamic system.
1.5.2.5.2 Variables
Variables are those who vary during the studied time interval. Consequently, their time derivative is
not allowed to be zero, which is important when going from DAE to ODE formulation since some DAE
equations might need to be differentiated.
It can be noted that, during parameter optimization, some model parameters are varied and could
therefore potentially be called variables, causing a name confusion. To avoid confusion, they should be
called either model parameters or optimization variables, depending of context.
A very important type of variable is a state variable. The variables used as (continuous) state
variables are given initial values and then updated through integration along the time interval studied.
Which variables to use as state variables is not uniquely defined by the physical (or mathematical)
model.
20
INTRODUCTION
Chose of state variables

For mechanical systems, one often uses positions and velocities of (inertial) bodies as state variables,
but it is quite possible and sometimes preferable, to use forces in compliances (springs) and velocities
of bodies as state variables. With mass-spring system as example, the two alternatives become:
• [ ̇= ⁄ ; = ∙ 𝑥; 𝑥̇ = ;] (or [𝑥̈ = ⁄ ; = ∙ 𝑥;]) and
• [ ̇ = ⁄ ; ̇ = ∙ ; 𝑥 = ⁄ ;], respectively.
The latter alternative is especially relevant when position is not of interest, typically in propulsion
systems. Then then the last equation (𝑥 = ⁄ ) can be omitted. It is also often easier to express steady
state initial conditions for a pre-tensioned system, e.g. wheel suspension where spring forces can be
states. A drawback might be that, when overdetermined pre-tensioned systems (e.g. mass suspended
in two springs: [ ̇ = ( + )⁄ ; ̇ = ∙ ; ̇ = ∙ ; 𝑥 = ⁄ ; 𝑥 = ⁄ ;]) that more states
than necessary will be used, which means that positions (𝑥) can be calculated in two ways (𝑥 =
⁄ ; 𝑥 = ⁄ ;) which might drift apart from each other. The drift can be eliminated by completely
disregard the spring constitutions (𝑥 = ⁄ ; 𝑥 = ⁄ ;) and instead use position as a state (𝑥̇ =
;). The mathematical system is then effectively:
• [ ̇= ⁄ ; 𝑥̇ = ; ̇ = ∙ ;].
One can see the latter alternative as if the states are the variables which represents the two energy
forms kinetic and potential energy, respectively, see Lagrange mechanics in Section 1.5.3.3.5.
1.5.2.5.3 Signals
Since vehicle dynamics so often requires models of the control algorithms, one often use the words
variable and signal interchangeable. It is suggested to be aware of that there is a difference.
Similarities and differences which can be relevant:
• A signal can represent a variable but has a prescribed causality. So, a variable cannot be
represented by a signal before modelling stage “Explicit form model”.
• A variable 𝑥 represents both 𝑥 and higher time derivatives of 𝑥 (𝑥̇ 𝑥̈ ⋯), while one signal for
each is needed. So, from DAE point of view, [𝑥 𝑥̇ 𝑥̈ ⋯ ] counts as only one unknown. From ODE
point of view, they count as one unknown each, and “𝑥̇ = 𝑥 ⁄ 𝑡” is counted as one equation.
1.5.2.6 Causality
Systems can be modelled with Natural causality. For mechanical systems, this is when forces on the
masses (or motion of the compliance’s ends) are prescribed as functions of time. Then the velocities of
the masses (or forces of the compliances) become state variables and have to be solved through time
integration. The opposite is called Inverse dynamics and means that velocities of masses (or forces of
compliances) are prescribed. For instance, the velocity of a mass can be prescribed and then the
required forces on the mass can be calculated through time differentiation of the prescribed velocity.
Cf. Analysis and Inverse Analysis in Figure 1-2.
1.5.2.7 Drawing
Drawing is a very important tool for understanding and explaining. Very often, the drawing contains
free body diagrams, see Section 1.5.3.1, but also other diagrams are useful. Beyond normal drawing
rules for engineering drawings, it is also important to draw motion and forces. The notation for this is
proposed in Figure 1-8.
It is often necessary to include more than just speeds and forces in the drawings. In vehicle dynamics,
these could be: power flow and signal or data flows. These can preferably be drawn as arrows, but of
another kind than the motion and force arrows.
When connecting components with signal flow, the resulting diagram is a data flow diagram. Physical
components and physical connections can be included in such diagram, and if arrows between them it
would represent data flow or causality. Physical components can also be connected by “Physical
connections”, which does not have a direction, see Section 1.5.2.7.1. It should be noted that a
(computation) flow charts and (discrete) state diagram represent something quite different from data
21
INTRODUCTION
flow diagrams, even if they may look similar; in state diagrams, an arrow between two blocks
represents a transition from one (discrete) state or operation mode, to another.
Any vector or Translation (e.g. translational speed or force) Quantity flow (e.g. Energy
flow = Power):
Along axis in plane of paper:
Perpendicular to paper, pointing out of paper: Signal flow or Data flow
Perpendicular to paper, pointing in to paper: Physical Connection
(no prescribed causal direction):
Rotation (e.g. rotational speed or moment/torque)
Around axis in plane of paper:
Parametrization
Around axis perpendicular to paper: (sometimes relevant to show):
Figure 1-8: Arrow like notation. Left: For motion and forces in drawings. Right: Other
1.5.2.7.1 Physical connections
When connecting physical components in a data flow diagram with “Physical connections”, there can
be two main concepts:
• Connecting in “nodes”, see Figure 3-2. Here, each node has typically potential and flow
variables (velocities and forces in Mechanical systems). And there are only two components
connected to each node. The node variables appear in both components’ equations.
• Connecting with “connections between connectors”, see Figure 3-13. Here, the same variables
are defined in connectors at the components. A connection can be made with lines between the
connectors at the components, possibly connecting more than two components. The equations
in each component is formulated in the connector’s variables. The connection itself generates
these equations: “Potential variables in all components are equal” and “Sum of flow variables
to all components are zero”.
The first is often the easiest (least number of equations and variables) for small models with a fixed
topology. The latter is a more systematic which easier handles larger models and models with varying
topology. The latter is supported in the standard modelling language Modelica, see Section 1.5.7.5 and
the open book on internet in Reference (Tiller, 2016 (continuously updated)).
1.5.2.8 Notation conventions

Generally, a variable is denoted 𝒙 or 𝒙(𝑡), where 𝑡 is the independent variable time. In contexts where
one means that variable’s value at a certain time instant, 𝑡0 , it can be denoted 𝒙(𝑡0 ). In contexts where
one wants to mark that the variable’s time history over a time interval (infinitely many values over an
infinite or finite time interval) is referred to, it can be denoted 𝒙( ).
𝑑
Differentiation (of x) with respect to time (t): 𝑑𝑡 = 𝑥̇ = 𝑟(𝑥).
Matrices (two-dimensional arrays) are denoted with bold letters, often uppercase 𝑨. Column and row
vectors (one-dimensional arrays) are denoted with bold, often lowercase 𝒃. When elements in arrays
are written, brackets are used: [𝑦𝑥 𝑦].
Geometrical vector (or physical vector, or spatial vector), denoted by ⃗. In order to use geometrical
vectors in calculations, they are often expressed as one-dimensional arrays but then it has to be clear
also which coordinate system they are expressed in, e.g. by a subscript, e.g. 𝒗𝒙𝒚𝒛 .
Multiplication symbol (* or ∙ or × or •) shall be used to avoid ambiguity, e.g. ∙ 𝑏𝑥, not 𝑏𝑥
(assuming 𝑏𝑥 is one variable, not two). Stringent use of multiplication symbol enables use of variables
with more than one token, cf. programming. Stringent use of multiplication symbol reduces the risk of
ambiguity when using operators, e.g. it shows the difference between (𝑥) and (𝑥). Compendium
denotes multiplication between scalars and matrices equally, using “ ”. Multiplication between
geometric vectors are more special and varying: Cross multiplication, denoted ⃗×𝑏⃗⃗ and scalar
multiplication, denoted ⃗• 𝑏⃗⃗.
22
INTRODUCTION
Parentheses shall be used to avoid ambiguity, e.g. ( ⁄𝑏) ∙ or ⁄(𝑏 ∙ ), and not ⁄𝑏 ∙ .
An interval has a notation with double dots. Example: Interval between and 𝑏 is denoted . . 𝑏.
An explanation, between two consecutively following steps in a derivation of equations is written
within {} brackets. Example: 𝑥 + 𝑦 = { + 𝑏 = 𝑏 + } = 𝑦 + 𝑥 or 𝑥 + 𝑦 = 7 ⇒ { + 𝑏 = 𝑏 + } ⇒
𝑦 + 𝑥 = 7.
An inverse function is denoted with superscript 1, e.g. 𝑦 = (𝑥); ⇔ 𝑥 = − (𝑦). There can be an
ambiguity whether − means inverse function or inverted function value (1⁄ ). So, when name on
inverse function is available, it should be used, e.g. arctan(⋯ ), and not tan− (⋯ ).
Fourier and Laplace transforms of function (𝑡) are denoted ℱ( (𝑡)) and ℒ( (𝑡)), respectively:
+∞
−𝑗 𝜔 𝑡
ℱ( (𝑡)) = ∫ (𝑡) 𝑡 ; 𝑤ℎ 𝑟 𝜔 ∈ 𝑅
−∞
∞
−𝑠 𝑡
ℒ( (𝑡)) = ∫ (𝑡) 𝑡 ; 𝑤ℎ 𝑟 𝑠 = 𝜎 + 𝑗 𝜔 𝑤ℎ 𝑟 𝜎 𝑛 𝜔 ∈ 𝑅
0
There are many practical rules for manipulation transformed differential equations, such as
ℱ ( ̇ (𝑡)) = 𝑗 𝜔 ℱ( (𝑡));.
1.5.2.9 Notation list

Table 1.2 shows notation of parameters, variables and subscripts used in this compendium. The
intention of this compendium is to follow International Standards (ISO 8855), but deviations from this
is sometimes used, especially motivated by conflicts with areas not covered by (ISO 8855). Some
alternative notations are also shown, to prepare the reader for other frequently used notation in other
literature.
The list does not show the order of subscripting. For example, it does not show whether longitudinal
(𝑥) force ( ) on rear axle (𝑟) should be denoted or . The intention in this compendium is to
order subscript with the physical vehicle part (here rear axle) as first subscript and the specification
(longitudinal, x) as second, leading to . If there are further detailed specifications, such as
coordinate system, e.g. wheel coordinate (𝑤), it will be the third subscript, leading to 𝑤.
1.5.3 Mechanical engineering

Vehicle dynamics originates from Mechanical engineering (in Swedish “Maskinteknik”, in German
“Machinenbau”). Therefore, it is important to be familiar with the following basic relationships:
• Energy = time integral of Power
• [Torque or Moment] = Force ∙ Lever
• Power = Torque ∙ Rotational speed or Power = Force ∙ Translational speed
• (Torque) Ratio = Output torque / Input torque
• (Speed) Ratio = Input speed / Output speed
• Efficiency = Output power / Input power or, in a wider meaning, Efficiency = Useful / Used
1.5.3.1 Free-body diagrams

In the physical model, see 1.5.2.1.2, a free-body diagram is often central. See example in Figure 1-9 and
Figure 1-10. Also, division into sub-systems is often practical, e.g. to implement moment less
connection points or other models of the vehicle internal behaviour. The sub-system split typically
goes through:
• Connection point between towing unit and towed unit (typical interface quantities: 2 positions
(with their derivative, velocities) and 2 forces).
• Driveshaft close to each wheel (typical interface quantities: 1 shaft torque, 3 forces, and 2
angles (steering and shaft rotation), sometimes 1 wheel camber angle).
23
INTRODUCTION
• Surface between driver’s hand and steering wheel (typical interface quantities: 1 angle and 1
torque).
• Sensed signals and request signals, which cuts out control algorithms as sub-system.
The free-body diagram is important as part of the physical model and it helps setting up the
equilibrium equations in the mathematical model. Relevant notes for this are:
• A short-cut that can be taken when drawing the diagram is to assume forces are already in
equilibrium, which makes one equilibrium equation unnecessary. An example of this could be
to use instead of 0 in Figure 1-9, which would make the vertical force equilibrium for the
front axle ( = ) unnecessary.
In the free-body diagram, one draws arrows and names corresponding vector quantities, typically
forces and velocities. You can either use a standard sign rule and let mathematics decide the sign ( ‘+’
or ‘-‘ ) of the variable, or, try to “feel” which direction the force or speed will have and define your
arrows and names according to this, to get positive numerical values. It is a matter of taste which is
best, but it is recommended that care is taken – it is very easy to make sign mistakes.
24
INTRODUCTION
Table 1.2: Notation
Notation
Categorization Recommended Alternatives
Engineering Sub category Subject for notation in code by hand Unit Description / Note
1. General Basic (Shaft) Torque T T M Nm
1. General Basic Coefficient of friction mu m mue, muh 1=N/N
N/(m/s) or
1. General Basic Damping coefficient d d c,k,D
Nm/(rad/s)
1. General Basic Density roh r
Ratio between
1. General Basic Efficiency eta h 1=W/W useful/output power and
used/input power
1. General Basic Energy E E W Nm=J
1. General Basic Force F F f N
1. General Basic Gravity g g m/s 2 9.80665 average on Earth
1. General Basic Height h h H

1. General Basic Imaginary unit j j i - j*j=-1
1. General Basic Mass m m M kg
2 Subscript can be given to

1. General Basic Mass moment of inertia J J I kg*m
denote around which axis
1. General Basic Moment (of forces) M M T Nm Not for shaft torque

1. General Basic Power P P W=Nm/s=J/s
Ratio between input and
u, Nm/(Nm) or output rotational speed,
1. General Basic Ratio r r
i (ISO8855) (rad/s)/(rad/s) or output and input
torque, in a transmission
1. General Basic Rotational speed w w w rad/s
1. General Basic Shear stress tau t N/m2
1. General Basic Stiffness coefficient c c C, k, K N/m or Nm/rad
1. General Basic Strain eps e m/m
1. General Basic Translational speed v v V m/s
1. General Basic Wave length lambda l m
Can be track width,
1. General Basic Width W W w m vehicle width, tyre width,
…
1. General Dynamics (Time) Frequency f f 1/s=periods/s
1. General Dynamics Angular (time) frequency w w w rad/s
1. General Dynamics Angular spatial frequency W W rad/m
Dependent variables in a dynamic Both state variables and
1. General Dynamics z z <various>
system output variables
1. General Dynamics Input variables in a dynamic system u u <various>
1. General Dynamics Mean Square value MS MS
1. General Dynamics Output variables in a dynamic system y y <various>
1. General Dynamics Power spectral density G F PSD

1. General Dynamics Root Mean Square value RMS RMS
Spacial frequency as radians per
1. General Dynamics W W rad/m
travelled distance
1. General Dynamics Spatial frequency fs fs 1/m=periods/m
1. General Dynamics States variables in a dynamic system x x <various>
The independent variable

1. General Dynamics Time t t t, T s
in a dynamic system
<not
1. General Operators Transfer function H H
applicable>
2. Vehicle 1. General Air resistance coefficient cd cd Cd 1
Cornering stiffness or lateral tyre dF y /d a = dF y /ds y at
2. Vehicle 1. General Cy Cy Ca N/rad or N/1
stiffness a =s y =0
2. Vehicle 1. General Longitudinal tyre stiffness Cx Cx N/1 df yw /ds x at s x =0
25
INTRODUCTION
Notation
2. Vehicle 1. General Mass m m M kg
2. Vehicle 1. General Rolling resistance coefficient f f 1=N/N
2. Vehicle 1. General Track width W W m
2. Vehicle 1. General Tyre stiffness C C c N/1 dF/ds at s=0
Kus, kus, N/(N/rad) =
2. Vehicle 1. General Understeer gradient Ku K u 2
U (ISO8855) rad/(m/s )
If no subscript, undefined
2. Vehicle 1. General Vehicle side slip angle b b b rad
or CoG
2. Vehicle 1. General Wheel base L L l, lf+lr, WB m
2. Vehicle 1. General Wheel radius R R r, Rw m

2. Vehicle 2. Road Curvature kappa k roh, r rad/m=1/m road or path
2. Vehicle 2. Road Curve radius R R r, roh, r m road or path
Positive when right side
2. Vehicle 2. Road Road bank angle pxr j xr of ground is lower than
left side ground
2. Vehicle 2. Road Road inclination angle pyr j yr rad Positive downhill
2. Vehicle 2. Road Vertical Position of road zr zr z, Z m
Vehicle acceleration, in interial ax, ay, decomposed in vehicle
2. Vehicle 3. Motion a x, a y, a z m/(s*s)
system az coordinates direction
dervx, der(vx),
Vehicle acceleration, in vehicle ̇ ̇ ̇ decomposed in vehicle
2. Vehicle 3. Motion dervy, der(vy), m/(s*s)
system coordinates direction
dervz der(vz)
r=[rx,ry,rz] Often refering to Center
2. Vehicle 3. Motion Vehicle position x,y,z x, y, z m
or [X,Y,Z] of Gravity, CoG
decomposed in vehicle
2. Vehicle 3. Motion Vehicle velocity, in interial system vx,vy,vz v x, v y, v z u,v,w m/s
coordinates direction
From ground, air, towed
Fx, Fy, F x , F y, F z, units, colliding objects,
[N, N, N, Nm,
2. Vehicle 4. Forces Forces and moments on vehicle Fz, Mx, M x, M y, etc. May appear also as
Nm, Nm]
My, Mz Mz reduced, e.g. for in road
plane: F=[Fx, Fy, Mz]
Subscript "i" is
placeholder for particular
Forces on one wheel, axle or side Fix,Fiy,F wheel, axle or side. May
2. Vehicle 4. Forces F ix , F iy , F iz lowercase f N
from ground iz appear also extended
with moments: [Fx, Fy, Fz,
Mx, My, Mz]
Forces on one wheel, axle or side Fixv, F ixv , F iyv , in vehicle coordinate
2. Vehicle 4. Forces lowercase f N
from ground Fiyv,Fizv F izv system
Forces on one wheel, axle or side Fixw,Fiyw F ixw , F iyw , in wheel coordinate
2. Vehicle 4. Forces lowercase f N
from ground ,Fizw F izw system
Sign convention might be

positive when either of
2. Vehicle 5. Angles Camber angle g g gamma, g rad "rotated as wheel roll
angle" or "leaning
outward vs vehicle body".
The angle required for a

Low speed or Ackermann speed
2. Vehicle 5. Angles dswA d swA rad certain vehicle path
steering wheel angle
curvature at low speeds
2. Vehicle 5. Angles Euler rotations y, q, j rad Order: yaw, pitch, roll
Angles, not rotations.
px, py,
2. Vehicle 5. Angles Roll, pitch, yaw angle j x , j y , j z j, q, y rad Roll and Pitch normally
pz
small
wx, wy,
2. Vehicle 5. Angles Roll, pitch, yaw angular velocities w x, w y, w z ̇ ̇ ̇ ̇ ̇ rad/s
wz
May refer to steering
2. Vehicle 5. Angles Steering angle d d delta rad
wheel, road wheel or axle
Normally an average of
2. Vehicle 5. Angles Steering angle of road wheels drw d rw RWA rad
angles on front axle
26
INTRODUCTION
Notation
2. Vehicle 5. Angles Steering angle of steering wheel dsw d sw SWA rad
2. Vehicle 6. Slip Tyre (lateral) slip angle a a alpha rad a=arctan(sy)
2. Vehicle 6. Slip Tyre lateral slip sy sy 1=(m/s)/(m/s) sy=tan(a)
2. Vehicle 6. Slip Tyre longitudinal slip sx sx k, -k 1=(m/s)/(m/s)
2. Vehicle 6. Slip Tyre slip s s 1=(m/s)/(m/s)
used e.g. as
<not
2. Vehicle 7. Subscript axle a a Ca=cornering stiffness
applicable>
for any axle
COG, cog, <not
2. Vehicle 7. Subscript centre of gravity CoG CoG
CG, cg applicable>
Often used as double
<not
2. Vehicle 7. Subscript front f f F, 1 subscript, e.g. fr=front
applicable>
right
<not Often means with
2. Vehicle 7. Subscript inner i i
applicable> respect to curve
<not
2. Vehicle 7. Subscript left l l L subscript, e.g. fl=front
applicable>
left
<not Often means with
2. Vehicle 7. Subscript outer o o
applicable> respect to curve
<not
2. Vehicle 7. Subscript rear r r R, 2 subscript, e.g. fr=rear
applicable>
right
<not
2. Vehicle 7. Subscript road r r
applicable>
<not E.g. drw=road wheel
2. Vehicle 7. Subscript road wheel rw rw RW
applicable> steering angle
<not
2. Vehicle 7. Subscript sprung mass s s
applicable>
SW, <not E.g. dsw=steering wheel
2. Vehicle 7. Subscript steering wheel sw sw
H (ISO8855) applicable> angle
<not
2. Vehicle 7. Subscript unsprung mass u u us
applicable>
used e.g. as F1v to
<not denote wheel force on
2. Vehicle 7. Subscript vehicle v v
applicable> wheel 1 in vehicle
coordinates
used e.g. as F1w to
<not denote wheel force on
2. Vehicle 7. Subscript wheel w w
applicable> wheel 1 in wheel
coordinates
3. Subsystem
1. Propulsion Engine torque Te Te Nm
in vehicles
3. Subsystem
2. Brake
in vehicles
3. Subsystem
3. Steering Steering wheel torque Tsw T sw SWT Nm
in vehicles
3. Subsystem
4. Suspension Caster angle CA rad
in vehicles
3. Subsystem
4. Suspension Caster offset c m
in vehicles
3. Subsystem
4. Suspension Kingpin inclination angle KPI rad
in vehicles
3. Subsystem
4. Suspension Pneumatic trail t m
in vehicles
1.5.3.2 Operating conditions

In physical and mathematical model one does assumptions that influences the operating conditions. In
the physical model, it is often a question about drawing or not drawing a mass or an elasticity. In the
mathematical model, it is often a question about taking in the corresponding terms, ̇ or ̇ ⁄ ,
respectively, into the equations. The operating conditions of a vehicle lend expressions from general
dynamics, as listed below.
• Static conditions, meaning that vehicle is standing still, are seldom relevant to study. Static
means that the all velocities are zero, i.e. that all positions are constant.
• Steady State operation means that time history is irrelevant for the quantities studied. Seen as
a manoeuvre over time, the studied quantities are constant. If refering to a certain
27
INTRODUCTION
mathematical model, it means that the derivatives of all the variables, that else could have
been state variables, are prescribed, to zero or as explicit functions of time.
• Transient (or Transient State, as opposed to Steady State) manoeuvres means that time
history is relevant; i.e. there are delays, represented by “state variables” when simulated.
• Stationary (oscillating) dynamics is a special case of transient, where cyclic variations
continue over long time with a repeated pattern. This pattern is often modelled as harmonic,
meaning that the variable varies as sinus and cosine with constant amplitudes and phases. An
example is sinusoidal steering with small enough steering amplitude.
• Quasi-steady state and Quasi-static are terms with more diffuse meaning. It has to refer to a
certain mathematical model and has no strict meaning for only a certain real vehicle
operation/manoeuvre. Refering to a certain mathematical model, it means that the derivatives
of some of the variables, that else could have been state variables, are prescribed, to zero or as
explicit functions of time.
1.5.3.3 Equation types

The step from physical model (see 1.5.2.1.2) to mathematical model (1.5.2.1.3) means basically to
identify variables and parameter and find the relationship between them, i.e. the equations. Models for
Vehicle Dynamics always includes mechanics, and for these parts we can identify 3 main types of
equations: Equilibrium, Compatibility and Constitution.
1.5.3.3.1 Equilibrium
Equilibrium gives relations between forces (including moments). For a static system, we can use the
(static) equilibria: Sum of forces in any direction is zero: ∑ 𝑭 = 𝟎; and Sum of moments around any
axis is zero: ∑ 𝑴 = 𝟎;
In a dynamic system, there is typically inertia effects. Inertia effects can be seen as “fictive forces” and
“fictive moments” (or “d’Alembert forces or moments”). Typically, the fictive forces are 𝑐𝑡 =
(where =translational acceleration) and the fictive moments are 𝑀 𝑐𝑡 = 𝐶𝑜𝐺 𝜔̇ . These are counter-
directed to and 𝜔̇ , respectively. We can find the equations either as “Dynamic equilibria”:
• Sum of forces in any direction is zero: ∑𝒊𝒏𝒄𝒍.𝒇𝒊𝒄𝒕 𝑭 = 𝟎;
• Sum of moments around any axis is zero: ∑𝒊𝒏𝒄𝒍.𝒇𝒊𝒄𝒕 𝑴 = 𝟎;
or as “Equations of motion” or “Newton’s 2nd law”:
• Sum without fictive in direction have to be equal to : ∑𝒆𝒙𝒄𝒍.𝒇𝒊𝒄𝒕 𝑭 = 𝒎 𝒂;
• Sum without fictive around axis through CoG in direction of 𝜔̇ have to be equal to 𝐶𝑜𝐺 𝜔̇ :
∑𝑪𝒐𝑮 𝒆𝒙𝒄𝒍.𝒇𝒊𝒄𝒕 𝑴 = 𝑱𝑪𝒐𝑮 𝝎̇ + 𝝎×(𝑱𝑪𝒐𝑮 𝝎);
The 𝐶𝑜𝐺 is the mass moment of inertia matrix in CoG along the same coordinate axes as 𝜔 is expressed
in. In this compendium, the alternative with Dynamic equilibria is mainly used. The fictive forces are
introduced in the free-body diagrams with dashed arrows, see Figure 1-9 and Figure 1-10. The general
form for the fictive forces are:
𝑟
⃗ 𝑐𝑡 = = ̇ ⏟ 𝜔× ⏟ 𝜔×(𝜔×𝑟) ⏟ 𝜔̇ ×𝑟 ;
𝑡 𝐶𝑜 𝑜 𝑠 𝑜 𝑐𝑒 𝐶𝑒𝑛𝑡 𝑢 𝑜 𝑐𝑒 𝐸𝑢 𝑒 𝑜 𝑐𝑒
Here, 𝑟 = 𝑝𝑜𝑠 𝑡 𝑜𝑛 𝑡𝑜𝑟. The Coriolis and Euler forces can often be assumed as zero in examples in
this compendium due to no motion within the vehicle and rigidity of rotating bodies. The general form
for the fictive moments are:
⃗⃗⃗
𝑀 𝑐𝑡 (𝑟𝑜𝑡 𝑡 𝑜𝑛 𝑙 𝑜 𝑛𝑡𝑢 ) = ( 𝐶𝑜𝐺 𝜔
= ⃗⃗⃗⃗⃗) = 𝐶𝑜𝐺 𝜔̇ + 𝜔×( 𝐶𝑜𝐺 𝜔);
𝑡 𝑡
The term 𝐶𝑜𝐺 ×( 𝐶𝑜𝐺 𝜔) can often be assumed as zero in examples in this compendium due to
symmetries and rotation in one plane at the time.
(A special case of equilibrium is a coordinate transformation between force quatities.)
28
INTRODUCTION
acceleration
m*acceleration air resistance

𝑭𝒇𝒙𝟎
Draw free body diagrams, 𝒇
acceleration selecting the free bodies
air resistance m*g 𝑭𝒇𝒛𝟎 𝑭𝒇𝒛𝟎

m*acceleration
𝑭 𝒛
𝑭𝒇𝒙𝟎
𝒇
Selecting: 𝑭𝒇𝒙
𝑭𝒇𝒙 Selecting: • vehicle body together with
𝑭 m*g 𝑭𝒇𝒛 • whole vehicle rear axle 𝑭𝒇𝒛
𝒛 air resistance
body with both • front axle alone 𝑭 𝒛
axles • remaining (ground & air) 𝑭𝒇𝒛
𝑭𝒇𝒙
Figure 1-9: Free body diagram. The dashed arrow is a “fictive force”. The star is a way to mark around which
point(s) moment equilibrium is taken in the later stage “Mathematical model”.
Free body diagram of Equilibrium for “sprung body”. Method with fictive
“sprung body”: forces, which allows any 3 equilibria in the plane. E.g.:
• Lateral: + ̇𝑠 = ;
Vehicle seen • Vertical: ̇ + = ;
𝜔̇ • Roll around any point, e.g. the star ( ):
from behind: ̇𝑠 +𝑀𝑠 𝜔̇ + Δℎ + ̇ 𝑠 Δℎ = ;
𝜔 ̇
Alternative equilibrium for “sprung body”. Method without fictive
forces, which only works with sum of forces in direction of 𝑠 and
𝑠 CoG
𝑀𝑠 and sum of moments around CoG:
• Lateral forces in 𝑠 direction: + = + ̇𝑠 ;
• Vertical forces in direction : + = ̇ ;
Δℎ • Roll moments around CoG:
𝑢 RC +𝑀𝑠 + Δℎ + Δℎ = 𝜔̇ ;
≈ Compatibility:
𝑀𝑠 • = ; and 𝜔 Δℎ = 𝑢 𝑠 ; and ̇ = 𝜔 ;
Constitution for suspension spring:
• 𝑀̇𝑠 = 𝜔 ;
Not used:
Actuation (assuming steering and good road grip):
Free body diagram 𝑀 • 𝑢 = 𝑢 𝑡 ; (=known function of time)
of (massless)
“unsprung axles”: 8 equations, 8 unknowns ( 𝑠 𝑢 𝜔 𝑀𝑠 ).
Equilibrium for “unsprung axles”:

Figure 1-10: Free body diagrams for combined translation and rotation. The figure also shows the two
different way of setting up equilibria, with and without fictive forces. Angle is assumed small.
1.5.3.3.2 Compatibility
Compatibility gives relations between motions (positions, velocities, accelerations, …). One example is
𝑆𝑝 = 𝑅 𝑢𝑠 𝑅𝑜𝑡 𝑡 𝑜𝑛 𝑙 𝑠𝑝 ; for a purely rolling wheel or 𝛽 ≈ ( + 𝑙 𝜔 )⁄ ; for body
side slip angle over front axle for a vehicle body moving in road plane.
(A special case of compatibility is a coordinate transformation between motion quatities.)
1.5.3.3.3 Constitution
Constitution are relations between forces (including moments) and motions, e.g.
𝑑
• For a (linear) spring: 𝑜𝑟 = 0 + 𝐷 𝑜𝑟 𝑡 𝑜𝑛; or 𝑑𝑡 𝑜𝑟 = 𝐷 𝑜𝑟 𝑡 𝑜𝑛𝑆𝑝 ;.
Metallic materials typically follow such linear behaviour, cf. Hooks law, if not deformed too
much or too fast.
• For a (linear) damper: 𝑜𝑟 = 𝐷 𝑜𝑟 𝑡 𝑜𝑛𝑆𝑝 ;. Shearing of thin liquid films typically
follows such linear behaviour, e.g. in lubricated bearing where shear force is proportional to
sliding speed.
• For a dry friction contact: 𝑜𝑟 = 𝐶𝑜𝑛𝑠𝑡 𝑛𝑡 𝑠 𝑛(𝑆𝑙 𝑛 𝑆𝑝 );. This is the most common
friction model in mechanical engineering, explained by adhesion between molecules and
29
INTRODUCTION
hysteresis when material is deforming over micro level asperities. So, the proportionality
constant depends on both cooperating bodies material and surface roughness. Note: When a
friction contact sticks, the equation switches to a compatibility equation (𝑆𝑙 𝑛 𝑆𝑝 = ;).
1.5.3.3.4 Controls
Via sensors and actuators, (control) algorithms can influence the mechanics. We can call this
mechatronics. Also, how the human driver controls the system can be modelled as algorithms which
connects the mechanical quantities. This category of equations becomes increasingly important for
vehicle dynamics but cannot be sorted under the traditional 1.5.3.3.1-1.5.3.3.3. Conceptually, any
quantities that can be sensed or actuated in, and outside of, the subject vehicle can occur in these
equations. (Finite) State machines are often useful when modelling algorithms and driver, see discrete
state machine in 1.5.2.7. A model of control algorithm can often be the same artefact as the design of it,
especially if using a modelling tool that allows automatic generation of real-time code, like Simulink.
1.5.3.3.5 Lagrange formalism
An alternative to Newton mechanics is Lagrange mechanics, which is sometimes an easier way to find
the same equations. Lagrange mechanics is basically to express kinetic and potential energy in
generalized coordinates. With Lagrange mechanics, partial differentiation of energy with respect to
these coordinates replaces the forces in the equilibria. The approach demolishes the categorization in
sections 1.5.3.3.1-1.5.3.3.3; the compatibility and constitution will be included from start. So, the
modularity corresponding to physical components in the system often gets lost. Also, energy
generating and dissipating components has to be treated with special care.
Lagrange mechanics formulates the equations in kinetic and potential energy in a way which has
similarities with using velocities of bodies and forces in compliances as state variables, as described in
Section1.5.2.5.2.1.
1.5.4 Control engineering

Vehicle dynamics is more and more influenced by electronics, where the algorithms are a main
artefact to engineer. For this reason, this section gives a brief introduction to some relevant theory and
methods and their connection to vehicle models as they are described in vehicle dynamics. From Figure
1-11, we realize that there are many other types of Vehicle Level Algorithms needed than just the
Vehicle Motion Controller. It should be underlined that the control structure in today’s vehicles are not
as clean and structured as in Figure 1-11, very much depending on that vehicle level control is
distributed in several ECUs, each belonging to its own subsystem.
Control algorithms can be designed without utilizing knowledge about the controlled system, such as
neural networks, tuned only on observations on how the system responds. However, in this section we
only consider Model based controllers. For those, the input and output signals, as well as parameters
inside, has a clear interpretation in vehicle motion quantities and units. For instance, the requests on
vehicle, 𝒚 𝒆𝒒, is typically forces on vehicle [ ; ; 𝑀 ] or accelerations of the vehicle [ ̇ ; ̇ ; 𝜔̇ ]. The
model base is helpful when tuning the controller parameters. Moreover, the model base helps to find a
consistency between derivatives order in the controller and the controlled system, such as if P, I or D
gains should be used in a PID-controller.
Four conceptual categorizations of vehicle control will be presented in Sections 1.5.4.1 to 1.5.4.4.
30
INTRODUCTION
Vehicle
Vehicle Level Algorithms
requests on requests on
Driver vehicle, 𝒚 𝒆𝒒 Vehicle actuators, 𝒖 𝒆𝒒 Environment
Driver Interp- Motion Actuated and
(road and
reter
Controller Sensing Vehicle
Switch traffic)
Virtual information,
𝒖𝒂𝒄𝒕𝒖𝒂𝒍 𝒖𝒄𝒂 𝒂𝒃𝒊𝒍𝒊𝒕𝒚
Driver
Vehicle and
Traffic
information
Estimator
Vehicle variables (positions, velocities, forces, etc.) and

parameters (mass, wheelbase, etc.)
(Open loop) Vehicle Motion (Closed loop) Vehicle Motion Controller (of gain
Controller type)
𝒚 𝒆𝒒 + request for 1 request,
Algorithm 𝒖
𝒚 𝒆𝒒 Algorithm 𝒖 𝒆𝒒 - increase, 𝒚 𝒆𝒒 Δ𝑡 𝒚̇ 𝒆𝒒 𝒆𝒒
information information
except 𝒚𝒂𝒄𝒕𝒖𝒂𝒍
𝒚𝒂𝒄𝒕𝒖𝒂𝒍 except 𝒚𝒂𝒄𝒕𝒖𝒂𝒍
Figure 1-11: Context for Vehicle Level Algorithms and Vehicle (Motion) Controller
1.5.4.1 Decision making or Execution

Nowadays, one can often here Decision-making algorithms as an expression. It means some kind of
rules for when to transit between discrete states, such as transit “from Follow lane to Change lane” in
automated driving or “switching from wheel torque to wheel speed control” for ABS/TC-like functions.
The complement to Decision making can be called Actuation execution and it would typically mean the
computation of continuous request signals, such as Steering Wheel Angle Request or Brake Pressure
Request, in the two above examples.
1.5.4.2 Momentaneous or Predictive control

If the controller considers the future (predicted) of its input 𝒚 𝒆𝒒 and/or output 𝒖 𝒆𝒒 , the controller
uses Predictive control. If the controller (or more correct, the controller designer) consider both its
input 𝒚 𝒆𝒒 and its output 𝒖 𝒆𝒒 to be at present time instant, the controller uses Momentaneous
control. This categorization is not strict, but there are cases where it makes sense. Since a controller is
normally implemented in a computer with sampled values, the future of the input cannot be known, so
future of inputs 𝒖 𝒆𝒒 should be seen as prediction e.g. via a time function or desired path ahead. The
future of the outputs is easier to understand, since a dynamic model can be used to tell how the system
evolves over future time.
A reason to do predictive control, as opposed to the simpler momentaneous control, is that the control
objective requires some strategic planning over future time. One example can be fuel consumption,
where the objective is an integral over future time, 𝑢 𝑙 = 𝑢 𝑙𝑅 𝑡 𝑡 ;, where a high 𝑢 𝑙𝑅 𝑡
can be good initially (momentaneously) to minimize the total integral. Another example is braking and
steering when entering a curve, where it is generally more strategically optimal to brake before
steering. Then, the objective is rather an integral 𝑜𝑠 𝑡 𝑜𝑛 = 𝑉 𝑙𝑜 𝑡 𝑠 𝑡 ;.
In Sections 1.5.4.3 and 1.5.4.4, momentaneous control will be assumed unless else is stated.
1.5.4.3 Open or closed loop control

As indicated already in Figure 1-11, a Closed loop control uses the actual value (𝒚𝒂𝒄𝒕𝒖𝒂𝒍 ) of the
requested quantity 𝒚 𝒆𝒒. Else it is an Open loop control.
31
INTRODUCTION
Since the driver is in the loop and can correct errors, Open-loop can often work well enough. Especially
if the driver is given good feedback, e.g. good steering feel. When automation of driving there is a
motivation to go towards more Closed-loop control. However, there can be architectural benefits to
close the loop in the Virtual driver instead of the Vehicle Controller, see Figure 1-11, so that same
algorithms can used in Vehicle Controller for manual and for automatic driving.
A more including definition of closed loop controller is that the controller uses any signal from the
controlled object, not exclusively whether it uses 𝑐𝑡𝑢 . With that definition, almost all controllers
end up as closed loop controllers. An example is an open-loop controller for a cruise control function:
𝑒𝑞 ← ̇ 𝑒𝑞 + 𝑒𝑠 ;. Here, the requested vehicle quantity is acceleration ( ̇ 𝑒𝑞 ) but the used
information about the vehicle is not ̇ 𝑐𝑡𝑢 but estimated mass ( ) and estimated driving resistance
force ( 𝑒𝑠 ). These estimated quantities are typically updated much slower than the request control
loop, but they are still used. With the initial definition, it is open loop controller, but with the more
including definition it would be a closed loop controller.
The cruise controller example above is an example of Inverse model controller, since it is an inverse
of the model: ̇ ← ( 𝑒𝑠 ) ;. The most common implementation of closed loop is when the
control error is calculated as an intermediate quantity. In the cruise control example, it would be
̇ 𝑒 𝑜 ← ̇ 𝑐𝑡𝑢 ̇ 𝑒𝑞 ; or with opposite sign it is rather a differential request Δ ̇ 𝑒𝑞 ← ̇ 𝑒𝑞
̇ 𝑐𝑡𝑢 (= ̇ 𝑒 𝑜 );. A Gain controller could then be: 𝑒𝑞 ← 𝐺 𝑛 Δ ̇ 𝑒𝑞 ;. This can be called a
P-control on acceleration or D-control on velocity. The model connection is not obvious since one
cannot directly see that 𝐺 𝑛 corresponds to mass and resistance force is not even present. A similar
gain controller could be formulated in another differentiation order, such as velocity instead of
acceleration: 𝑒𝑞 ← 𝐺 𝑛 Δ 𝑒𝑞 ;. This can be called P-control on velocity, I-control on acceleration
or D-control on position. The model ( = ̇ ;) combined with the controller ( 𝑒𝑞 ← 𝐺 𝑛
Δ 𝑒𝑞 ;) points out that 𝐺 𝑛 should be seen as 𝑠𝑠 𝑡 , where 𝑡 is the time to when 𝑒𝑞
should be reached. Interpretation of gains to physical quantities with units is encouraged, since it
facilitates tuning and reuse of the controller. Such interpretation requires a physical model.
It is very typical for automotive engineering, that algorithms are reused, and sometimes incrementally
developed, from one platform or vehicle programme to the next. Hence, given a certain controller, it is
very relevant to understand/derive its model base, for efficient parameter setting. However, it is
important to realize that it is not generally possible to derive the model base from a given controller.
So, the controller really has to be developed from a physical model. So, the documentation of which
physical model is used is very important output from the control design, beside the controller itself.
The documentation should include model equations and declarations of variables and parameters so
that next vehicle programme can inherit this as document.
1.5.4.4 Degree of over-actuation

Already before designing the controller, it is often clear what vehicle requests 𝒚 𝒆𝒒 and actuate-able
quantities 𝒖 𝒆𝒒 there are. The number (or dimension= ) of those enables one simple categorization:
• If (𝒖 𝒆𝒒 ) = (𝒚 𝒆𝒒 ), the vehicle is (formally) Neutral actuated.
• If (𝒖 𝒆𝒒 ) < (𝒚 𝒆𝒒 ), the vehicle is (formally) Under-actuated (𝒖 𝒆𝒒 is over-determined)
• If (𝒖 𝒆𝒒 ) > (𝒚 𝒆𝒒 ), the vehicle is (formally) Over-actuated (𝒖 𝒆𝒒 is under-determined)
In cases where actuators influence on vehicle motion is saturated in some way, or the vehicle motion is
constrained, the above definition is less useful. So, we have to include also a certain operating
condition for the vehicle, to be able to include such effects. Certain operating conditions means that we
think there is a certain model which describes the actuated vehicle motion, 𝒇𝑫𝑨𝑬 (𝒙 𝒖 𝒚 𝑡) = ;, or in
explicit form [𝒙̇ 𝒚] = [𝒇𝑶𝑫𝑬 (𝒙 𝒖 𝑡) 𝒈(𝒙 𝒖 𝑡)];. With such a model, we can include saturations and
constraints. Instead of dimension, we need to define the property “rank of the model” as the number of
dimensions that the vehicle motion can be influenced in by the available actuators. Using a
linearization of the model we can, find 𝑩 such that:
𝒚 ≈ 𝒈(𝒙𝟎 𝒖𝟎 𝑡0 ) + 𝑩 (𝒖 𝒖𝟎 ) = 𝒈𝟎 + 𝑩 (𝒖 𝒖𝟎 );
32
INTRODUCTION
𝜕
𝑤ℎ 𝑟 𝑩 ℎ 𝑠 dim(𝒚) 𝑟𝑜𝑤𝑠 𝑛 dim(𝒖) 𝑜𝑙𝑢 𝑛𝑠 𝑛 𝐵 𝑗 = ;
𝜕𝑢 𝑒𝑞𝑗
"𝑅 𝑛𝑘 𝑜 𝑜 𝑙" = rank(𝑩).
The controller should then aim at finding 𝒖 𝒆𝒒 so that 𝒚 𝒆𝒒 (𝒈𝟎 𝑩 𝒖𝟎 ) = 𝑩 𝒖 𝒆𝒒 ;. We identify
“virtual requests” 𝒗 𝒆𝒒 = 𝒚 𝒆𝒒 (𝒈𝟎 𝑩 𝒖𝟎 ); of same dimension as 𝒚 𝒆𝒒, so 𝒗 𝒆𝒒 = 𝑩 𝒖 𝒆𝒒 ;.
The rank of 𝑩 (or any matrix) is the number of non-zero diagonal elements in the middle matrix 𝑺 in
the singular value decomposition of 𝑩, 𝑠 (𝑩) = 𝑼 𝑺 𝑽𝑇 ;. Using the rank, we can define degree of
overactuation in a way which better helps us differ between different control design types, see also
Figure 1-12:
• If dim(𝒖 𝒆𝒒 ) = rank(𝑩), the vehicle is (effectively) Neutral actuated.
Conceptually, 𝒖 𝒆𝒒 could be calculated from 𝒗 𝒆𝒒, using the model, which would be an Inverse
model control design.
• If dim(𝒖 𝒆𝒒 ) < rank(𝑩), the vehicle is (effectively) Under-actuated (𝒖 𝒆𝒒 is over-determined).
Conceptually, there is no way to convert the problem to a Neutral actuated case (without
adding actuators or disregarding requests), but one can find and approximate solution through
optimization. The optima for an under-actuated vehicle is generally unique. The optimization
would be to minimize a cost function: 𝑜𝑠𝑡 = ‖𝒚 𝒆𝒒 𝒈(𝒙 𝒖 𝒆𝒒 𝑡)‖;. If we can use the linear
model, we can rewrite the 𝑜𝑠𝑡 = ‖𝒗 𝒆𝒒 𝑩 𝒖 𝒆𝒒 ‖ ;, which can be optimized. Optimizing in
Least Square sense gives 𝒖 𝒆𝒒 = (𝑩𝑇 𝑩)− 𝒗 𝒆𝒒 ;.
• If dim(𝒖 𝒆𝒒 ) > rank(𝑩), the vehicle is (effectively) Over-actuated (𝒖 𝒆𝒒 is under-determined).
Here, 𝑩𝑇 𝑩 will be singular, so there is no unique solution. A way to handle it would be to use
some stochastic way to find one 𝒖 𝒆𝒒 among all non-unique. This can lead to randomly varying
and jumping 𝒖 𝒆𝒒 in time, which is undesirable.
A second way is to convert the problem to a Neutral actuated case by simply prescribing
suitably many components in 𝒖 𝒆𝒒 : 𝑢 𝑒𝑞 = 𝑢𝑑𝑒𝑠 , where subscript 𝑠 means desired.
A third way is to to convert the problem into a Under-actuated case by adding is to add
equations for each and all components in 𝒖 𝒆𝒒 , typically: 𝒖 𝒆𝒒 = 𝟎. To use zero is difficult to
motivate directly from vehicle motion/actuation perspective but it can be thought of as a way
to guaranteeing unique mathematical solutions, which is good.
Combining second and third way gives an optimization problem with cost function as follows:
𝑜𝑠𝑡 = ‖𝑾𝒎𝒐𝒕𝒊𝒐𝒏 (𝒗 𝒆𝒒 𝑩 𝒖 𝒆𝒒 )‖ + ∑‖𝑊 𝑐𝑡 (𝑢 𝑒𝑞 𝑢𝑑𝑒𝑠 )‖ + ‖𝑾𝒔𝒎𝒐𝒐𝒕𝒉 𝒖 𝒆𝒒 ‖ ;.
The 𝑾 matrices are diagonal matrices with weight factors. For solving such mathematical
problems there are well established subroutines for Control allocation, CA, which can also
take constraints into account.
A visualisation of these cases and solution is given in Figure 1-12.
The above in the section 1.5.4.4 could be applied also for predictive control if we think of 𝒖 𝒆𝒒 and 𝒚 𝒆𝒒
as having 𝑁 times more elements, where 𝑁 is the number of time instants considered in the prediction.
The model can be applied 𝑁 times, and if it contains state variable, a time integration method is
needed to make the sequences of each variable consistent. If model is linear and optimization is to be
applied, there are cooking-book methods, such as Model Predictive Control (MPC), see e.g. Reference
(Ross, 2015) describes MPC. Simply adding more time instants adds equally many equations as
unknowns (𝒖 𝒆𝒒 ), so it does not change the degree of over-actuation. However, if the variation of
𝒖 𝒆𝒒 (𝑡) is parameterised (e.g. 𝒖 𝒆𝒒 (𝑡) = ; or 𝒖 𝒆𝒒 (𝑡) = 𝟏 + 𝟐 𝑡;), the problems becomes as under-
actuated, which is more well-conditioned and can be solved in a least square sense.
33
INTRODUCTION
Neutral actuated vehicle: 𝑏

𝑩=
dim=2 dim=2 dim=2
𝒗 𝒆𝒒 = 𝒚 𝒆𝒒 −
Actuated vehicle:
𝒚 𝒆𝒒 𝒗 𝒆𝒒 𝒖 𝒆𝒒 = 𝑩 𝒗 𝒆𝒒 ; 𝒖 𝒆𝒒
𝒈0 𝑩 𝒖 𝟎 ; 𝒇𝑫𝑨𝑬 𝒙 𝒖 𝒆𝒒 𝒚 𝑡 = 𝟎;
𝑏
𝑩=
Under-actuated vehicle:
dim=3 dim=3 𝒖 𝒆𝒒 = 𝑩"− " 𝒗
𝒆𝒒 ; or
dim=2
𝒗 𝒆𝒒 = 𝒚 𝒆𝒒 Actuated vehicle:
𝒚 𝒆𝒒 𝒗 𝒆𝒒 𝒖 𝒆𝒒
𝒈0 𝑩 𝒖 𝟎 ; 𝑢 𝒆𝒒 = ar min 𝑜𝑠𝑡 𝒖 𝒆𝒒 ; 𝒇𝑫𝑨𝑬 𝒙 𝒖 𝒆𝒒 𝒚 𝑡 = 𝟎;
𝒖 𝒆𝒒
𝑏
Over-actuated vehicle: 𝑩=
"− " 𝒗
𝑩 𝒆𝒒
dim=2 dim=2 dim=3
𝒗 𝒆𝒒 = 𝒚 𝒆𝒒 𝒖 𝒆𝒒 = 𝑢𝑑𝑒𝑠 ; or Actuated vehicle:
𝒚 𝒆𝒒 𝒗 𝒆𝒒 𝟎 𝒖 𝒆𝒒
𝒈0 𝑩 𝒖 𝟎 ; 𝒇𝑫𝑨𝑬 𝒙 𝒖 𝒆𝒒 𝒚 𝑡 = 𝟎;
𝒖 𝒆𝒒 = ar min 𝑜𝑠𝑡 𝒖 𝒆𝒒 ;
𝒖 𝒆𝒒
Figure 1-12: Visualisation of possible conceptual controllers for vehicles with different degree of over-
actuation. The actuated vehicles are represented by a very conceptual model. 𝑩"− " = (𝑩 𝑩)− 𝑩 .
Verification methods with real vehicle

1.5.5.1 In traffic
Driving on public roads in real traffic is the most realistic way to verify how a vehicle actually works. It
can be used for completely new vehicle models; or new systems, mounted on old models. The drivers
can be either ordinary drivers (FOT=Field Operational Test) or test drivers (expeditions). A general
existing vehicle population can also be studied by collecting data, e.g. as Accident Statistics Databases.
1.5.5.2 On test track

For vehicles and systems which are not yet allowed on public roads, or tests which are very severe or
need a high degree of repeatability, test are carried out at test tracks. There are specialized test tracks
for certain conditions, such as hot climate or slippery surfaces.
1.5.5.3 Augmented Reality, AR, manipulates environment

This is a new method. A typical example is: A real driver drives a real vehicle on a real road/test track.
Some additional (virtual/simulated) traffic objects are presented to driver, e.g. on a head-up display.
The same objects can be fed into the control functions, as if they were detected by the vehicles
camera/radar, which enables functions such as automatic braking to be triggered.
1.5.6 Verification methods with virtual vehicle

1.5.6.1 Testing with real driver
1.5.6.1.1 Simulator = Driver in the loop simulation
This is when a real human, not a driver model, uses real driver devices (pedals, steering wheel) to
influence a simulation model of the remaining system (vehicle and environment). The loop is closed by
giving the human feedback through display of what would be visible from driver seat, including views
outside windscreen. Feedback can be further improved by adding a motion platform to the driver’s
seat, sound, vibrations in seat, steering wheel torque, etc.
34
INTRODUCTION
The vehicle model run in Alternative solutions

the simulator can utilize Off-line Driving Real
HIL, SIL and MIL, from simulation simulator vehicle test
Section 1.5.6.1, for some Test new technology before it is built
+ - --
part of the vehicle model. (save time and money)
One can compare Design parameter
+! - --
sweep/optimization
simulator with other
Requirements
Repeatable testing + + -
ways to verify vehicles,
Safe testing + + -
see Figure 1-13.
Representative integration in vehicle - +! ++
w.r.t. to vehicle - - +!
Represen-
w.r.t. driver - +! +
tative
behaviour w.r.t. to surroundings - - (env. sens) - (env. sens) +!
(road & traffic)
Figure 1-13: Comparison of different tests to verify complete vehicle

functions.
1.5.6.2 Testing using driver model

Generally, this means that driver, vehicle and environment are modelled and simulated. To perform a
serious test there is also need to put effort on test scripts so that the model is run in the intended test
manoeuvre and result is pre-processed. When verifying functions partly built with algorithms, the HIL,
SIL and MIL below are different ways to represent the algorithm in the vehicle model.
1.5.6.2.1 HIL = Hardware in the loop simulation
The hardware is often one or several ECUs (Electronic Control Units). If several ECUs are tested, the
hardware can also contain the communication channel between them, e.g. a CAN bus. The hardware is
run with real-time I/O to simulation model of the remaining system (vehicle, driver and environment).
In some cases, there is also mechanical hardware involved, such as if the ECU is the brake system ECU,
the actual hydraulic part of the brake system can also be included in the HIL set-up, a so called “wet
brake ECU HIL”.
1.5.6.2.2 SIL = Software in the loop simulation
The software is often one or several computer programs (intended for download in electronic control
units). The software is run with synchronized time-discrete I/O to a simulation model of the remaining
system (vehicle, driver and environment).
The software is often used in compiled format (black box format) so that the supplier of the software
can retain his intellectual property.
1.5.6.2.3 MIL = Model in the loop simulation
The model, or more correctly, a control algorithm, is a conceptual form of the computer programs
(intended for download in electronic control units). The control algorithm is run with I/O to a
simulation model of the remaining system (vehicle, driver and environment).
The control algorithms can appear in compiled format so that the supplier of the control algorithms
can retain his intellectual property. Then it is hard to tell the difference between MIL and SIL.
1.5.7 (Computer) Tools & Methods

1.5.7.1 General mathematics tools
Examples of tool: Matlab, Matrixx, Python
We will take Matlab as example. Matlab is a commercial computer program for general mathematics. It
is developed by Mathworks Inc.
35
INTRODUCTION
Solve linear systems of equations, A*x=b

>> x=inv(A)*b;
Solve non-linear systems of equations, f(x)=0
>> x=fsolve('f',...);
Parameter optimization (may be used for trajectory optimization, if trajectory is parameterised)
>> (x,fval) = fmincon('f',x0, …)
Solve ODE (=”systems of Ordinary Differential Equations”) as initial value problems, dx/dt=f(t,x), given
x(0)
>> x=ode23('f',x0,...);
Find Eigen vectors (V) and Eigen values (D) to linear systems: D*V=A*V
>> [V,D]=eig(A);
Matlab is mainly numerical, but also has a symbolic toolbox:
>> syms x a; Eq='a/x+x=0'; solve(Eq,x)
ans = (-a)^(1/2)
-(-a)^(1/2)
>> diff('a/x+x',x)
ans = 1 - a/x^2
>> int('x^3+log(x)',x)
ans = (x*(4*log(x) + x^3 - 4))/4
Some simple Matlab code will be used to describe models in this compendium.
1.5.7.2 Block diagram based modelling and simulation tools

Examples of tools: Simulink, Systembuild
We will take Simulink as example. Graphical definition of non-linear systems of ODE, if known on
explicit form 𝒙̇ = 𝒇(𝑡 𝒙);. An example of how the corresponding graphical model may look is given in
Figure 1-14.
Simulink is designed for designing/modelling signal processing and control design. It can also be used
for modelling the physics of the controlled systems. There are no dedicated vehicle dynamics
tools/libraries from Mathworks (but there are in-house developed specific libraries in automotive
companies).
From this type of tools it is often possible to automatically generate real time code, which is more and
more used instead of typing algorithms. It can be used for rapid prototyping of control functions, or
even for generation of executable code for production ECUs.
1.5.7.3 Vehicle dynamics specialized simulation tools

Examples of tools: CarMaker, veDYNA, CarSim.
These tools are specialized for vehicle dynamics. They contain purpose-built and relatively advanced
models of vehicles, drivers and scripts for test manoeuvres. They are well prepared for parameter
changes, but less prepared for modelling conceptually new designs. However, they often have an
interface to Simulink, so that the user can add in their own models.
1.5.7.4 MBS tools

Examples of tools: Adams, Simpack, LMS Virtual Lab
These are general 3D mechanics modelling and simulation tools, so called MBS (Multi-Body
Simulation) tools. As one example, Adams contains libraries of general bodies, joints and force
elements. But there are toolboxes in Adams for vehicle dynamics, where template models and special
36
INTRODUCTION
components (such as tyre models and driver models) are available for vehicles dynamics. The models
are very advanced and accurate for 3D mechanics, and there are import/export interfaces to Simulink.
Figure 1-14: Graphical model of dynamic system 𝑥̇ = (𝑡 𝑥); using Simulink.
1.5.7.5 Modelica based modelling tools

Examples of tools: Dymola, Maplesim, System-Modeler, AMESim, Optimica Studio, Jmodelica,
OpenModelica
Modelica is not a tool but a globally standardized format for lumped dynamic models on DAE form (or
Mathematical form, see 1.5.2.1.3). There are several tools which supports the format. Specification of
Modelica is found at https://www.modelica.org/. When learning Modelica, http://www.xogeny.com/
is helpful. The model format is non-causal and symbolic. There is one declaration part and one
equation part in a model. An example of model is given in Figure 1-15.
Physical model:
Simulation in Dymola:
𝑥 (Mathematical) Model
in Modelica:
The tool manipulates the 3 equations to this

F + m*der(v) = 0; explicit form:
𝑟( )
𝑆𝑡 𝑡 𝐷 𝑟 𝑡 𝑠 = = ;
𝑟(𝑥)
𝑢𝑡𝑝𝑢𝑡𝑠 = = ;
Figure 1-15: Example of model in Modelica format (using the tool Dymola).
The model format is also object oriented, which means that libraries of model components are
facilitated. These are often handled with graphical representation, on top of the model code. There are
some open-source libraries for various physical domains, such as hydraulic, mechanics,
thermodynamics and control. There are also commercial libraries, where we find vehicle dynamics
relevant components: Vehicle Dynamics Library and Powertrain Library. Some simple Modelica code
will be used to describe models in this compendium.
37
INTRODUCTION
In one way, Mathematical modelling (Declarative modelling) is more efficient than Explicit form
modelling (Imperative modelling), since the engineer does not need to spend time on
symbolic/algebraic manipulation of the equations. This is especially true when the natural way of
physical modelling generates so called “higher index problems”, see Figure 1-16. In this compendium,
many models are only driven to Mathematical model, since it is enough if assuming there are modern
tools as Modelica tools available. One the other hand, an explicit form model has the value of capturing
the causality, i.e. the cause-to-effect chain. The causality can sometimes facilitate the understanding
and in that way help the engineer, which is why at least one and rather complete model is shown as
explicit form model, see Section 4.5.3.2.
1.5.7.6 FMI supporting simulation tools

Examples of tools: Dymola, Simulink, CarMaker, OpenModelica
FMI (Functional Mock-up Interface) is not a tool but a globally standardized format for dynamic
models on explicit form (see Section1.5.2.1). There are a number of tools which supports this format.
FMI enables model export/import between tools. It also allows to hide Intellectual Property (IP) by
using “black-box format”, i.e. models compiled for certain processors, which is important in relation
between OEMs and suppliers.
Physical model: Simulation in Dymola:
̇ ̇
(Mathematical) Model
in Modelica:
model ExampleModel_HigherIndex
parameter Real m1=0.5;
parameter Real m2=1.5;
parameter Real c=3;
Real v1;
Real v2;
The tool first realizes need for differentiating, to get another
Real F1;
Real F2(start=3); algebraic equation:
equation 1= ⇒ 𝑟( 1) = 𝑟( );
m1*der(v1)
F1 = -F1;
+ m1*der(v1) = 0; The tool then selects state variables and manipulates the
m2*der(v2)
F2-F1 = F1-F2;= 0;
+ m2*der(v2) equations to this explicit form:
v1=v2; 𝑟( 1) ( 1+ )
der(F2) = c*v2; 𝑆𝑡 𝑡 𝐷 𝑟 𝑡 𝑠 = = ;
𝑟( ) 1
end ExampleModel_HigherIndex;
1
𝑢𝑡𝑝𝑢𝑡𝑠 = = ;
1 1 ( 1+ )
Figure 1-16: Example of “higher index problem” model in Modelica format.
1.5.8 Coordinate Systems

A vehicle’s (motion) degrees of freedom are named as in marine and aerospace engineering, such as
heave, roll, pitch and yaw, see Figure 1-17. Figure 1-17 also defines the 3 main geometrical planes,
such as transversal plane and symmetry plane. For ground vehicles, the motion in ground plane is
often treated as the primary motion, which is why longitudinal, lateral and yaw are called in-ground-
plane degrees of freedom. The remaining degrees of freedom are referred to as out-of-ground-plane.
The consistent use of parameters that describe the relevant positions, velocities, accelerations, forces,
and moments (torques) for the vehicle are critical. Unfortunately, there are sometimes disparities
between the nomenclature used in different text books, scientific articles, and technical reports. It is
important to apply ISO coordinate system, Reference (ISO 8855). It is the predominant coordinate
system used nowadays. Historically, a coordinate system with other positive directions, Reference
(SAEJ670), has been applied.
The distinction of vehicle fixed and inertial (= earth fixed = world fixed) coordinate systems is
important. Figure 1-18 depicts the four most relevant reference frames in vehicle dynamics: the
38
INTRODUCTION
inertial, vehicle, wheel corner and wheel reference frames. All these different coordinate systems
allow for the development of equations of motion in a convenient manner.
or heave
vertical
ground, road
or horizontal ISO 8855
plane z
yaw y
transversal
plane
x
symmetry
plane
Figure 1-17: Left: Vehicle (motion) degrees of freedom and important planes. Right: ISO coordinate system
The orientation of the axes of an inertial coordinate system is typically either along the vehicle
direction at the beginning of a manoeuvre or directed along the road or lane. Road or lane can also be
curved, which calls for curved longitudinal coordinate.
Origin for a vehicle fixed coordinate system is often centre of gravity of the vehicle, but other points
can be used, such as mid front axle (ground contact or wheel centre height), mid of front bumper,
outer edge of body with respect to certain obstacle, etc. Positions often need to be expressed for centre
of lane, road edge, other moving vehicle, etc.
In ISO and Figure 1-18, tyre side-slip is defined so that it is positive for positive lateral speed. This
means that lateral forces on the wheel will be negative for positive side-slip angles. Some would rather
want to have positive force for positive angle. Therefore, one can sometimes see the opposite
definition of tyre side-slip angles, as e.g. in (Pacejka, 2005). It is called the “modified ISO” sign
convention. This compendium does not use the modified sign convention in equations, but some
diagrams are drawn with force-slip-curve in first quadrant. Which is preferable is simply a matter of
taste.
Often there is a need to number each unit/axle/wheel. The numbering in Figure 1-19 is proposed. It
should be noted that non-numeric notations are sometimes used, especially for two axle vehicles
without secondary units. Then front=f, rear=r. Also, to differentiate between sides, l=left and r=right.
Speed of
v wheel hub
lateral=y
yaw rate=wz=wz x
vy vx
y
n =course angle x y d=d=Road wheel)
CoG Steering angle
y
z
Inertial coordinate system: =pz=
=heading angle
NOTE: The ISO8855 tyre lateral side slip is positive for tyre sliding with
z x positive lateral speed, giving negative tyre lateral force. (Opposite tyre
lateral slip definition is sometimes used, but not recommended.)
Figure 1-18: Coordinate systems and motion quantities in ground plane
39
INTRODUCTION
𝑜𝑟 𝑜𝑟
Δ 𝑜𝑟
5 or 1,3,1 3 or 1,2,1 1 or 1,1,1
Δ
6 or 1,3,2 4 or 1,2,2 2 or 1,1,2
𝑜𝑟
Δ =Δ = = Dp1z=
𝑜𝑟
(rear) articulation angle of unit 1
axle 3 axle 2 axle 1
or 1,3 or 1,2 or 1,1
unit 2 unit 1
Figure 1-19: Proposed numbering of units, axles, wheels and articulation angle. Example shows a rigid truck
with trailer. If multiple units: 𝛥 𝑛 = 𝑛− 𝑛 ;
1.5.9 Terms with special meaning

1.5.9.1 Load levels
The weight of the vehicle varies through usage. For many vehicle dynamic functions it is important to
specify the level, which is these definitions are important to know and use.
Kerb weight is the total weight of a vehicle with standard equipment, all necessary operating
consumables (e.g., motor oil and coolant), a full tank of fuel, while not loaded with either passengers or
cargo. Kerb weight definition differs between different governmental regulatory agencies and similar
organizations. For example, many European Union manufacturers include a 75 kg driver to follow
European Directive 95/48/EC.
Payload is the weight of carrying capacity of vehicle. Depending on the nature of the mission, the
payload of a vehicle may include cargo, passengers or other equipment. In a commercial context,
payload may refer only to revenue-generating cargo or paying passengers.
Gross Vehicle Weight/Mass (GVW/GVM) is the maximum operating weight/mass of a vehicle as
specified by the manufacturer including the vehicle's chassis, body, engine, engine fluids, fuel,
accessories, driver, passengers and cargo but excluding that of any trailers.
Other load definitions exist, such as:
• “Design Weight” (for passenger vehicles, this is typically Kerb weight plus 1 driver and 1
passenger, 75 kg each, in front seats)
• “Instrumented Vehicle Weight” (includes equipment for testing, e.g. out-riggers)
• “Road-allowed GVW/GVM” is the maximum GVW/GVM for a certain road, limited by the
strength of the road and bridges. It is applicable for heavy trucks.
For vehicle dynamics it is often also relevant to specify where in the vehicle the load is placed because
it influences how vertical forces under the wheels/axles distribute as well as moments of inertia.
1.5.9.2 Open-loop and Closed-loop Test Manoeuvres

Two expressions used in vehicle dynamics are "Open-loop" and "Closed-loop" test manoeuvres.
An open-loop manoeuvre refers to the case where the driver controls (steering wheel, brake pedal
and accelerator pedal) are operated in a specific sequence, i.e. as functions of time. A typical case is a
sine wave excitation of the steering wheel. The time history of the steering wheel angle is defined as a
function that is independent of the road environment or driver input. This type of manoeuvre is a
practical design exercise to study the steering response of the system, but has little relation to a real-
40
INTRODUCTION
world driving case. Theoretical simulation and real testing with a steering robot are examples of how
such studies can be made.
A closed-loop manoeuvre refers to the case when (human) driver feedback via driver controls is
included. This represents real-world driving better. In real vehicle or driving simulator testing, a real
driver is used. This enables collection of the drivers’ subjective experience. In cases of theoretical
simulation, a "driver-model" is needed. A driver-model can have varying levels of complexity but in all
cases simulates the response of a human driver to different effects, such as lateral acceleration,
steering wheel torque, various objects appearing outside the vehicle, etc.
A test with real vehicle, carried out with a steering-robot (and/or pedal robot) can also be called
closed-loop if the robot is controlled with a control algorithm which acts differently depending on the
vehicle states, i.e. if the algorithm is a driver model.
The concepts of open- and closed-loop control are demonstrated in Figure 1-20. The important
difference to note is that in open-loop control there is no feedback to the driver. The figure also shows
a “Virtual driver” which represents today’s automatic functions based on environment sensing, such as
adaptive cruise control (section 3.5.2.2) and lane keeping aid (section 4.6.2.1). Closed loop control
incorporates the different types of feedback to the driver. Drivers automatically adapt to the different
feedback. Understanding driver response to different feedback, and expressing it in mathematical
descriptions, is an active area of research, particularly for driver support functions.
Position of road edges, lanes, traffic

Visual
signs, other road users, …
Environment
sensing & “Virtual Road irregularities, Road friction
driver” system (radar,
camera, GPS, WiFi, …)
Road banking
and inclination,
Requests (signals)
Collision forces
Steering
Suspension,
system, Surroundings /
Propulsion Vehicle Environment
Driver Linkages,
Steering system, Road wheel Forces, body Position, (air, road, other
wheel, Brake angles, Wheel Moments Speed road users, …)
Wheels
Pedals, … system torques, …
Warnings, Info
HMI (lamps,
sounds, …)
Sensed quantities
Noise, vibrations
Steering wheel torque
Inertial forces
Open-loop system:
Closed loop system: Whole diagram
Figure 1-20: Open and closed loop representations of vehicle dynamic systems
1.5.9.3 Objective and subjective measures

Two main categories of finding measures are:
• An Objective measure is a physical measure calculated in a defined and unique way from data
which can be logged in a simulation or from sensors in a real test.
• A Subjective assessment is a rating measure set by a test driver (e.g. on a scale 1-10) in a real-
vehicle test or driving simulator test.
41
INTRODUCTION
One generally strives for objective measures. However, many relevant functions are so difficult to
capture objectively, such as Steering feel and Comfort in transient jerks, so subjective assessments are
needed and important.
1.5.9.4 Path, Path with orientation and Trajectory

A path can be 𝑥(𝑦) or 𝑦(𝑥) for centre of gravity where 𝑥 and 𝑦 are coordinates in the road plane. To
cope with all paths, it is often necessary to use a curved path coordinate instead, 𝑠, i.e. 𝑥(𝑠) and 𝑦(𝑠).
The vehicle also has a varying orientation, (𝑥) or (𝑠), which often is often relevant, but the term
“path” does necessarily include this. In those cases, it might be good to use an expression “path with
orientation” instead.
A trajectory is a more general term than a path and it brings in the dependence of time, 𝑡. One typical
understanding is that trajectories can be [𝑥(𝑡); 𝑦(𝑡); (𝑡)]. But also, other quantities, such as steering
angle or vehicle propulsion force can be called trajectory: (𝑡) and (𝑡), respectively. The word
“trace” is sometimes used interchangeably with trajectory.
1.5.9.5 Stable and Unstable

Often, in the automotive industry and vehicle dynamics, the words “stable” and “unstable” have a
broad meaning, describing whether high lateral slip on any axle is present or not. In more strict
physics/mathematical nomenclature, “unstable” would be narrower, meaning only exponentially
increasing solution, which one generally finds only at high rear axle side-slip.
It is useful to know about this confusion. An alternative expression for the wider meaning is “loss of
control” or “loss of tracking” or “directional unstable”, which can include that vehicle goes straight
ahead, but road bends.
1.5.9.6 Subject and object vehicle

The subject vehicle is the vehicle that is studied. Often this is a relevant to have a name for it, since one
often studies one specific vehicle, but it may interact with other in a traffic situation. Alternative names
are host vehicle, ego vehicle or simply studied vehicle.
If one particular other vehicle is studied, it can be called object vehicle or opponent vehicle. A special
case of object vehicle is lead vehicle which is ahead of, and travels in same direction as, subject
vehicle. Another special case is on-coming vehicle which is ahead of, and travels in opposite direction
as, subject vehicle.
1.5.9.7 Active Safety, ADAS and AD

The expression Active Safety is used a lot in Automotive Engineering. There are at least two different
usages:
• Active Safety can refer to the vehicle’s ability to avoid accidents, including both functions
where the driver is in control (such as ABS and ESC, but also steering response) and functions
with automatic interventions based on sensing of the vehicle surroundings (such as AEB and
LKA). See http://en.wikipedia.org/wiki/Active_safety. Active Safety can even include static
design aspects, such as designing the wind shield and head light for good vision/visibility.
• Alternatively, Active Safety can refer to only the functions with automatic interventions based
on sensing of the vehicle surroundings. In those cases it is probably more specific to use
Advanced Driver Assistance Systems (ADAS) instead, see
http://en.wikipedia.org/wiki/Advanced_Driver_Assistance_Systems. ADAS does not only
contain safety functions, but also comfort functions like CC and ACC.
Functions that off-load the driver the direct tasks during driving can be sorted under Automated
Driving, AD. Fully automated driving, e.g. transportation without human driver on-board, is probably a
far future vision. On the other hand, it is already a reality that some portion of the driving tasks are
automated in the latest vehicle on the market, such as Adaptive cruise control, see Section 3.5.2.2, and
Lane Keeping Aid, see Section 4.6.2.1. If both those are active at the same time, we have already
42
INTRODUCTION
automated driving. Definitions of automation levels 0-5 is found in Reference (SAE_J3016, 2016).
Today’s version of these systems normally has a way to hand-back responsibility to driver rather
immediately in hazardous situations. Future automated driving functions will need to always have a
safe-stop function. The way to compete between vehicle manufacturers will probably be to avoid
hand-backs (maximize up-time) and to allow as long “hand-back times” as possible. So, vehicle
dynamics will be important in the development, especially for safety reasons for automated driving in
higher speeds; hazardous situations where human driver selects to take back the driving.
1.5.10 Architectures
Vehicles are often designed in platforms, i.e. parts of the design solutions are reused in several
variants. Typical variants may be different model years or different propulsion system. To be able to
reuse solutions, the vehicles have to be built using the same architecture.
A mechanical architecture may include design decisions about certain type of wheel suspension on
front and rear axle. An electrical and electronic architecture may include design decisions about
electric energy supply system (battery voltage etc.) and electronic hardware for computers (Electronic
Control Units, ECUs) and how they are connected in networks, such as Controller Area Network, CAN.
That the mechanical architecture influences vehicle dynamics functions is rather obvious. However, it
is noteworthy that also the electronic architecture also is very important for the vehicle dynamics,
through all electronic sensors, actuators and control algorithms. One example of this importance is the
ABS control of the friction brake actuators. Architectures for functions are therefore motivated, see
Section 1.5.10.2.
1.5.10.1 Subsystems
The architectures are dependent on the business model for how to purchase and integrate subsystems
to a vehicle. Hence, it is relevant to define the subsystems. For vehicle motion functionality, the
relevant subsystems (or “motion support devices”) are typically:
• Propulsion system
• Brake system
• Wheel suspension
• Wheels and tyres
• Steering system
• Environment sensing system
Each of these can typically be purchased as one subsystem. Each typically have mechanical and signal
interface to the remaining vehicle. Different vehicle manufacturers can have different strategies for
signal interface and how much the subsystems are allowed to be dependent of each other.
1.5.10.2 Vehicle function architecture

As the number of electronically controlled functions increase, an architecture for vehicle functionality,
or “Function Architecture”, becomes necessary to meet fast introduction of new functionality and to
manage different variety of vehicle configurations. A function architecture is a set of design rules for
how functions interact with each other (e.g. signalling between control functions). Also, the
decomposition in subsystems should be reflected in this architecture, see 1.5.10.1. An older expression
which is related to function architecture is cybernetics. Examples (from Vehicle Dynamics/Vehicle
motion function domain) of modern expressions which are related are Integrated Chassis Control
(GM), Integrated Vehicle Dynamics Control (Ford), Complete Vehicle Control (Volvo) and Vehicle
Dynamics Integrated Management (Toyota). There is no exact and generally well accepted definition of
such architecture. However, it becomes more and more essential, driven by increasing content of
electronic control in vehicles. One way to visualize a reference architecture is given in Figure 1-21 and
Figure 1-22.
43
INTRODUCTION
Physical connection
Vehicle control Information
driver sensor 1 Request
Function 1
Information
sensor 2
Fusion Function 2
Prediction
sensor 3
Arbitrator
Momentary & Capability
Request
Coordinator
Momentary & Capability
Actuated actuator 1 … … actuator N

environment
(road, obstacles, …) vehicle
Figure 1-21: Concepts of a vehicle motion function architecture. Arbitrators, Coordinators (or Allocator) and
Actuators are the most important architectural objects for vehicle dynamics (vehicle motion).
In order to be able to formulate design rules in reference architecture of functionality the following are
relevant questions:
• Which physical vehicle quantities should be represented on the interface between Sensors and
Actuators on one side and Vehicle Level Functionality on the other?
• Partitioning within a reference architecture of vehicle motion functionality could be realised as
shown in Figure 1-22. Different Layers/Domains are defined:
• Motion Support Device Layer: This includes the devices/actuators that can generate
vehicle motion. This layer is also consisting sensors which could include the capability and
status of each device e.g. max/min wheel torque.
• Vehicle Motion Layer: This includes the Energy management, powertrain coordination,
brake distribution, and vehicle stability such as ESC, ABS. This layer also estimates the
vehicle states e.g. 𝜔 . In addition, this layer would be able to give vehicle level
capability of max/min acceleration and their derivative.
• Traffic Situation Layer: Interpret the immediate surrounding traffic which the vehicle is in.
Automated driving assistance functions such as adaptive cruise control, collision
avoidance, and lane steer support are typical functions.
• Vehicle Environment: Includes surrounding sensors mounted on vehicle but also
communication with other vehicles (V2V) and infrastructure (V2I) and map information.
• Human Machine Interface: This includes the sensors/buttons which the driver uses to
request services from the vehicle’s embedded motion functionality.
• Formalisation of different types of:
• Blocks, e.g. Controller, Signal Fusion, Arbitrator, and Coordinator.
• Signals, e.g. Request, Status and Capability.
• Parameters used in Functional blocks. One can differ between Physical parameters (or
Model parameters) and Tuning parameters. Some parameters can be common across the
whole vehicle, which enables a kind of communication between blocks without normal
signals.
The functionality is then allocated to ECUs, and signals allocated to network communication. The
reference architecture can be used for reasoning where the allocation should be done. Which
functionality is sensitive for e.g. time delay and thus should be allocated in the same ECU. Detailed
control algorithm design is not stipulated by a reference architecture. Instead the reference
architecture assists how detailed control algorithms be managed in the complete problem of
44
INTRODUCTION
controlling the vehicle motion. Whether representation of solutions of Functional Safety (ISO 26262,
etc.) is represented in a reference architecture of functionality can vary.
Vehicles consisting of several units add special challenges, especially if the units are actuated. A heavy
truck trailer is always actuated with at least brakes.
1.5.10.3 Virtual vehicle architecture

Virtual vehicle architecture, VVA, here refers to how to organize complete vehicle models, which can
replace some real vehicle pre-series. VVA is a set of rules how a vehicle model should be modularized,
such as variable/signal interfaces, parameterization, format- and tool-chains. The vision is that a
whole project organisation, within an OEM and its suppliers, could deliver modules to virtual pre-
series and that these modules would fit together. The FMI format, see Section 1.5.7.6, is one example of
what could facilitate this.
Vehicle Traffic Situation Layer Human

Forrward Direction
Environment Machine
ACC Driver
V2V / V2I Interface
Traffic Interpretation
LaneSteering ACCbtn
Maps Interpretation
… Collision Dist btn
FwdSens Avoidaice Arbitration
Vehicle Capability Actual Req. speed / acceleration

SteWhl
AnSens
Vehicle Motion Layer
Driver APed
Vehicle Interpretation
Vehicle Motion
Motion BPed
Control
Estimator
Energy
Management Arbitration
Vehicle Req. e.g. Global Forces
Stability
Capability Coordination
Control
Estimator
Actual Actual Req. e.g. Wheel Torque

Device Capability
Motion Support Device Layer Physical connection to

⋯ FrntBrkAct RearBrkAct
brake only on passenger
EngAct SteAct cars, Trucks have
Vehicle TransmAct brake-by-wire.
Motion
Sens
requests
actual values
capabilities
physical connections
Figure 1-22: One example of reference architecture of vehicle motion functionality. Red arrows: Requests,
Blue arrow: Information, Black lines with dot-ends: physical connection.
1.6 Heavy trucks

The following section describes heavy trucks, mainly as compared to passenger vehicles.
1.6.1 General differences

Trucks are normally bought and owned by companies, not private persons. Each truck is bought for a
specialized transport task. Life, counted in covered distance, for trucks is typically 10 times passenger
cars. The life time cost of fuel is normally 5 times the vehicle cost, compared to passenger car where
these costs are about equal. The cost for driver salary is a part of mileage cost, typically same
magnitude as fuel cost. If investment cost for vehicle and repairs are distributed over travelled
distance, these are typically also of same magnitude. So, the cost for a transport typically comes from
one third fuel, one third driver salary and one third vehicle investment and repairs.
45
INTRODUCTION
1.6.2 Vehicle dynamics differences

A truck has 5..10 times less power installed per vehicle weight. Trucks have their centre of gravity
much higher, meaning that roll-over occurs at typically 4 m/s2 lateral acceleration, as compared to
around 10 m/s2 for passenger cars. Trucks have centre of gravity far behind mid-point between axles,
where passenger cars have it approximately symmetrical between the axles. Trucks are often driven
with more units after, see Figure 1-23. The weight of the load in a truck can be up to 2..4 times the
weight of the empty vehicle, while the maximum payload in passenger cars normally are significantly
lower than the empty car weight. Trucks often have many steered axles, while passenger cars normally
are only steered at front axle.
1.6.3 Definitions
In Figure 1-23, we can find the following units:
• Towing units: Tractor or Rigid (truck)
• Towed units: Full trailer, Semi-trailer, Centre-axle trailer, and (Converter) Dolly
• The couplings between the units can of 2 types:
o Fifth-wheel coupling (e.g. between Tractor and Semi-trailer). Designed to take
significant force in all 3 directions. Furthermore, it is designed to be roll-rigid, but free
in pitch and yaw. (A Turn-table is similar as fifth wheel, but might often have rolling
bearing instead of greased surface, which leads to less yaw friction and no pitch degree
of freedom. In converter dollies, one often sees a combination of both fifth wheel and
turntable.)
o Hitch coupling (e.g. between Rigid and Full trailer). Designed to take significant forces
in longitudinal and lateral directions, but only minor in vertical direction. Furthermore,
it is rotationally free in all 3 directions.
Figure 1-23: An overview over conventional and longer combinations. From (Kharrazi , 2012).
The 2 units “Dolly-Semitrailer” can also be replaced by a 1 unit “Full-trailer”.
1.7 Smaller vehicles

This section is about smaller vehicles, meaning bicycles, electric bicycles, motorcycles and car-like
vehicles for 1..2 persons. The last vehicle type refers to vehicles which are rare today, but there are
46
INTRODUCTION
reasons why they could become more common: Increasing focus on energy consumption and
congestion in cities can be partly solved with such small car-like vehicles, of which the Twizy in Figure
1-24 is one example. All vehicles in Figure 1-24 can be referred to as Urban Personal Vehicle (UPVs),
because they enable personalised transport in urban environments. The transport can be done with
low energy consumption per travelled person and distance, compared to today’s passenger cars. The
transport can be done with high level of flexibility and privacy for the travelling persons, compared to
today’s public transportation.
New solutions as in Figure 1-24 may require some categorization.
• Climate and user type: Sheltered or open.
• Transport and user type: Short travels (typically urban, 5-10 km, 50 km/h) or long travels
(typically inter-urban, 10-30 km, 100 km/h).
• Chassis concept:
o Narrow (e.g. normal bicycles and motorcycles) or wide (at least one axle with 2 wheels,
resulting in 3-4 wheels on the vehicle). Note that UPVs in both categories are typically still
less wide than passenger cars.
o Roll moment during cornering carried by suspension roll stiffness or roll moment during
cornering avoided by vehicle cambering. The first concept can be called “Roll-stiff vehicle”.
The second concept can be called “Cambering vehicle” or “Leaning vehicle”. 1-tracked are
always Cambering vehicles. 2-tracked are normally Roll-stiff, but there are examples of
Cambering such (see upper right in Figure 1-24). See Figure 1-25.
o (This compendium does only consider vehicles which are “Pitch-stiff”, i.e. such that can take
the pitch moment during acceleration and braking. Examples of vehicles not considered,
“Pitching vehicles”, are: one-wheeled vehicles as used at circuses and two-wheeled vehicles
with one axle, such as Segways.)
3 wheeled cambering Piaggio MP3
concept vehicle from BMW
Toyota
iRoad
Twizy from Renault
ZeeBee from Clean Motion Eco Electric Bicycle, Monark

Figure 1-24: Examples of Urban Personal Vehicles. Left: “Roll-stiff”. Right:
“Cambering“. From www.motorstown.com, www.cleanmotion.se, www.monarkexercise.se,
www.nycscootering.com.
Note that also Roll-stiff Vehicles camber while cornering, but outwards in curve and only slightly,
while Cambering Vehicles cambers inwards in curve and with a significant angle. Cambering Vehicles
is more intricate to understand when it comes to how wheel steering is used. In a Roll-stiff Vehicle, the
wheel suspension takes the roll-moment (maintains the roll equilibrium), which means that driver can
use wheel steering solely for making the vehicle steer (follow an intended path). In a Cambering
Vehicles, the driver has to use the wheel steering for both maintaining the roll equilibrium and
following the intended path. A model is given in Section 4.5.2.3. Figure 1-25 shows some possible
conceptual design of a cambering vehicle. It is not possible to do a partly roll-stiff and partly
47
INTRODUCTION
cambering vehicle, unless one uses suspension springs with negative stiffness, i.e. the suspension
would need to consume energy to tilt the vehicle inward in curve.
“Roll-stiff concept” … …as opposed to “Cambering concept”
Translate Lean sprung body Lean whole vehicle, Lean whole vehicle,
2 wheel axle single wheel axle
Figure 1-25: How Cambering vehicles avoid to take the roll moment while cornering. The vehicles are viewed
from rear and turns to the left. The dash-dotted arrows marks the resulting forces which are equal between
the 5 concepts.
Smaller vehicles might have significantly different vehicle dynamics in many ways:
• Influence of driver weight is larger than for other vehicle types. This especially goes for bicycles,
where driver weight can be typically 5 times larger than the vehicles own weight. Drivers not
only change the total inertia properties, but they can also actively use their weight and move it
during driving.
• Ratio between CoG height and wheel base is likely to be larger than passenger cars and trucks.
This gives a larger longitudinal load transfer during acceleration and braking. Some concepts
might even have the risk of “pitch-over”.
• For 2-tracked UPVs, the ratio between CoG height and track width is likely to be larger than
passenger cars, rather like trucks, because it is likely that one wants to keep swept area low for
UPVs.
o If Roll-stiff, this gives a larger lateral load transfer and roll angles during cornering. The risk
for roll-over is also likely to be larger than passenger cars.
o If Cambering, this opens for the risk for slide (“roll-over inwards in curve”, so called “low-
sider”), which is known from motorcycles.
• For future UPVs it is possible that the part costs need to be kept low, as compared to passenger
cars. This, and the fact that new inexperienced OEMs might show up on market, might lead to
low-cost and/or unproven solutions appear for the vehicle design. This might be a challenge for
driving experience and (driver and automated) active safety.
• Positive, for safety, is that future UPVs can have significantly lower top speed than passenger
cars. However, the acceleration performance up to this top speed might be high, e.g. due to
electric propulsion, which might cause new concerns for city traffic with surrounding.
1.8 Typical numerical data

The purpose of the tables in this section is to give approximate numerical values of parameters.
Vehicle parameters are often dependent on each other; changing one leads to that it is suitable to
change others. The parameters are given with the intention to be consistent with each other, for each
vehicle example. The tables balance between being generic and specific, which is difficult. Therefore,
please consider the table as very approximate.
1.8.1 For passenger vehicle

Table 1.2 gives typical numerical data for a medium sized passenger car.
48
INTRODUCTION
Table 1.3: Typical values of parameters, common for typical passenger cars and heavy trucks
Parameter Notation Typical Value Unit

Passenger car Heavy truck
Air density roh 1 kg/(m^3)
Earth gravity g 9.80665 m/(s*s)
Road friction, at dry asphalt 1.0
Road friction, at wet asphalt 0.6
mu N/N
Road friction, at snow 0.3
Road friction, at wet ice 0.1
Tyre lateral cornering stiffness, i.e. slip (N/rad)/N
CC 10..15 5..10
stiffness normalized with vertical load = (N/1)/N
Ratio of tyre longitudinal and lateral
CCx/Ccy 1.5..2 1
cornering stiffness
Table 1.4: Typical data for a passenger vehicle
Group/Type Parameter Notation Typical Unit Note / Typical / Range /

Value Relation to other
Vehicle Length, bumper to bumper 5.00 m
Longitudinal distance from 40-50% of wheel base:
Vehicle lf 1.3 m
CoG to front axle lf=0.55*L;
Vehicle Mass m 1700 kg
Radius of gyration is sligthly less
Moment of inertia for yaw
Vehicle Jzz 2900 kg*m*m (0.9) than half wheel base:
rotation
=m*(0.9*L/2)^2
Sum of 4 wheels with
Vehicle Unsprung mass mus 200 kg
suspensions
Vehicle Track width W 1.70 m
Vehicle Wheel base L 2.90 m
For calculation of air resistance.
Projected area in a
Vehicle Afront 2 m^2 Examples: Volvo XC90 has 2.78
transversal view
m^2, Volvo S60 has 2.27 m^2
For calculation of longitudinal
cd, clf, 0.31, 0.10,
Vehicle Aerodynamic coefficients 1 resistance and lift force over front
clr 0.07
and rear axle
Wheel and Tyre Rolling resistance 0.01 N/N
Wheel and Tyre Wheel radius 0.30 m
Wheel and Tyre Wheel rotational inertia 0.5kg*m*m For one wheel
Wheel and Tyre Tyre vertical stiffness ct 250 000 N/m For one wheel
Propulsion Engine inertia 0.5kg*m*m
Transmission ratio, highest rad/rad= Engine to wheel. In magnitude
Propulsion 5.00
gear Nm/Nm of 5.
Transmission ratio, lowest rad/rad= Engine to wheel. In magnitude
Propulsion 10.00
gear Nm/Nm of 10.
Brake proportioning, N/N or Often tuned so that braking both
Brake 70/30
front/rear Nm/Nm axles when braking 0.8*g
Steering Steering ratio 16 rad/rad Steering wheel to Road wheel
Vertical stiffness at wheel. Sum of 4
Suspension heave stiffness wheels. So that bounce frequency f
Suspension cs 100 000 N/m
(without roll) is between 1 and 2 Hz:
sqrt(c/m)=f*2*pi;
Suspension roll stiffness, only Vertical stiffness at wheel. Sum of
Suspension carb 14 000 N/m
from anti-roll-bars both axles.
Wheel rate. Sum of 4 wheels. Some
40..60% of critical damping:
Suspension Suspension damping ds 13 000 N/(m/s) d = (0.4..0.6)*2*sqrt(c*m);
2..3 times softer in compression
than rebound.
1.8.2 For heavy vehicle

Compared to passenger cars, trucks differ much more in size and configuration, see Figure 1-23. Also, a
certain individual vehicle also differs much more between empty and fully loaded. Globally, “Tractor
with Semitrailer” is the most common heavy vehicle, which is why it is used as example.
49
INTRODUCTION
load length=13.6
front overhang
=1.5 wheel base=8
overall height =
overall height
3.7 m
=4 m
forward rear rear
front offset =4 axle spread overhang
overhang overhang =1.5
=1.5 =1 =3.6
wheel base=5
vehicle overall length=16.5 m
Figure 1-26: Typical data for a heavy vehicle, exemplified with “Tractor with Semi-trailer”
Table 1.5: Typical data for a heavy vehicle, exemplified with “Tractor with Semi-trailer”
Typical Value Note / Typical / Range /

Group/Type Parameter Empty With Unit Relation to other
max
payload
Trackwidth, centre-to-centre for
Vehicle 2.25 m Outer tyre edge 2.55
single tyres
Projected area in a transversal For calculation of air
Vehicle 10.0 m^2
view resistance
c_d coefficient in aero dynamic
Vehicle Aerodynamic coefficient 0.4..0.6 1 resistance formula (normal truck
0.4, long combination truck 0.6)
Tractor CoG heigth 1.00 m

Tractor Mass (total), tractor 7500 kg
Tractor Moment of yaw inertia 22500 kg*m*m Around unit CoG
Tractor Unsprung mass 1700 kg Sum of all axles at unit
Semitrailer CoG heigth 1.00 2.00 m
Semitrailer Mass (total), semi-trailer 10000 32500 kg
Semitrailer Moment of yaw inertia 150000 500000 kg*m*m Around unit CoG
Semitrailer Unsprung mass 2400 kg Sum of all axles at unit
Wheel and Tyre Wheel radius 0.50 m 0.4-0.5
Wheel and Tyre Wheel rotational inertia 10.0 kg*m*m For one wheel, single tyre
Wheel and Tyre Rolling resistance 0.005 N/N Or less, 0.003
kNm/s=k
Propulsion Engine max power 370
W
Propulsion Engine inertia 5 kg*m*m
Propulsion Transmission ratio, highest gear 3 1=rad/rad Engine to wheel
=Nm/Nm
Propulsion Transmission ratio, lowest gear 30 Engine to wheel
Steering wheel to Road
Steering Steering ratio 20 rad/rad
wheel
Steering wheel to Road
Steering Steering wheel rotational inertia 0.04 kg*m^2
wheel
50
VEHICLE INTERACTIONS
2 VEHICLE INTERACTIONS
2.1 Introduction
The study of vehicle dynamics starts with the interfaces between the vehicle and its environment, see
Figure 2-1. The vehicle tyres are the primary force interface for all motion (in road plane, i.e.
acceleration, steering, braking) but also undesired disturbances (out-of- road plane, i.e. road
unevenness, road bumps, etc.). Additionally, the aerodynamic loads on the vehicle will create forces
that are often undesirable (wind resistance, side gusts, etc.) but can sometimes be exploited for better
contact with the road (down-force). An example of extreme interactions to the vehicle is the impact
forces from a crash. An interaction of another kind is the driver. One often divides the whole system
into 3 parts: vehicle, driver and environment, where environment is all remaining, as in Figure 2-1.
Driver and environment can sometimes be clustered as section 2.9 Driving and Vehicle Operation,
which includes all outside subject vehicle.
Driver
(via Driver) Traffic, Infrastructure
Driver/Rider,
User or
Customer Steering Wheel,
Pedals, Buttons,
Air, Weather Seat Accelerations,
Visual Instruments,
Sounds, …
Aerodynamic Relative Motion,
Collision Forces, (Object) vehicles, Other road users,
Air density, Weather, Forces, Vision, …
Information, … Traffic signs, Road markings,
Light/dark, …
Infrastructure, …
(Subject)
Freight
Vehicle Road surface construction, Covered
Weight with water/snow, Topography,
Payload, Curvature, Unevenness, …
Occupants, … and Location Tyre Forces
kg
Ground, Road
Figure 2-1: How the vehicle interacts with driver and environment.
Main focus of present chapter is on tyres, including derivation of models. Important aerodynamic
models are listed, without ambition on derivation. Driver models are discussed with some selected
example models.
2.1.1 References for this chapter

• Tyres: (Pacejka, 2005), (Rill, 2006), and (Michelin, 2003)
• Vehicle Aerodynamics: Reference (Barnard, 2010)
• Driver models: “Chapter 38 Driver Models in Automobile Dynamics Application” in Reference
(Ploechl, 2013)
2.2 Wheel & Tyre Terminology & Design

The tyre (or wheel) is best described in its own coordinate system, see Figure 2-2. The three forces
and moments are acting on the wheel from the ground. They are needed are recognized in any free
51
body diagrams of the vehicle. The other quantities, such as slip angle 𝛼 and camber angle 𝜀𝑤 , in Figure
2-2 are examples of variables which have influence on the forces and moments in the coming chapters.
perpendicular
to road plane
Plane perpendicular to road
𝑀
𝑀
𝑀
Figure 2-2: Tyre Coordinate System, forces and torques on tyre from ground. According to ISO.
2.2.1 Wheel angles

2.2.1.1 Steering angle
A wheel’s steering angle is the angle between the wheel plane and the longitudinal axis of the vehicle.
Steering angle can be defined for one wheel or one axle, using averaging of left and right wheel.
2.2.1.2 Camber angle

Camber angle (or sometimes Camber) is the angle between the wheel plane and the vertical, positive
when the top of the wheel leans outward. (This definition of sign is only applicable for axles with two
wheels. For instance, it is not applicable for motorcycles.) Camber generates lateral forces and gives
tyre wear. See also Section 2.4.6.3.
2.2.1.3 Caster angle

The angle in a side-view between the steering axis and the vertical axis is called Caster angle. It is
positive if the top of the steering axis is inclined backwards. Caster angle provides an additional
aligning torque, see 2.4.6.2 and Section 4.2.3.2, and changes the camber angle when the wheel is
steered.
2.2.1.4 Toe Angle

Toe angle (or toe-in) is defined for an axle with two wheels (not for a single wheel), as the difference
between right and left wheel’s steering angles. Toe angle is positive if front ends of the wheels are
pointing inward. Hence, toe can be called toe-in. Negative toe can be called toe-out. Toe angle
generates opposing lateral forces on each side.
The toe angles vary with the tyre forces, due to suspension linkage geometry and elasticity in
suspension bushings. This means that there is a static toe angle (when a vehicle parked in workshop,
with a certain load and not tyre forces in ground plane) has one value, while actual toe angle (in a
particular moment during a manoeuvre) has another value.
52
Zero toe gives low rolling resistance and low tire wear. Theoretically, toe-out on front axle and toe-in
on rear axle makes the vehicle most yaw stable (less over-steered). Tone-in on front axle makes
vehicle more yaw agile and improves on-centre steering feel. Normal design choice is toe-in on both
axles, and more on front axle.
2.2.1.5 Steering axis inclination

Steering axis inclination (or Kingpin inclination) is defined for a wheel with its suspension linkage (not
for an isolated wheel). The angle in a front view between the steering axis (or Kingpin) and the vertical
axis is called Kingpin inclination. See Section 4.2.3.2.
2.2.2 Tyre Design

The tyres of a vehicle have the following tasks:
o Carry the vertical load
o Generate longitudinal (brake, traction) and lateral (steering) forces
o Isolate vertical disturbances, for comfort, road grip and reduction of suspension fatigue
o Roll with minimum energy loss, tyre wear, particle emissions and noise emission
Before discussing the mechanics of tyre and road interactions, the physical structure of the wheel
assembly should be understood. Consisting of a steel rim and an inflated rubber toroid, pneumatic
tyres were invented and patented by Robert William Thomson in 1845 and are essentially the only
type of tyre found on motor vehicles today.
The physical construction of the tyre carcass affects the response of the tyre to different road loadings.
The carcass is a network of fabric and wire reinforcement that gives the tyre the mechanical strength.
The structure of the carcass can be divided into different types of tyres: Bias-ply, Bias-ply Belted, and
Radial-ply. Bias-ply tyres were the first types of pneumatic tyres to be used on motor vehicles. Radial
ply tyres followed in 1946 and became the standard for passenger car tyres. Figure 2-3 shows the
main features of these tyre types.
Note how the bias-ply constructions have textile structures oriented at an angle to the tyre centreline
along the x-z plane. This angle is referred to as the crown angle and is further illustrated in Figure 2-3.
Note the textile orientation for the bias-ply and radial tyres. Also note the difference in crown angles
between the two tyre constructions. This difference plays an important part in the rolling resistance
characteristics of the tyre which is Section 2.3.
The tyre components have been constructed to provide the best tyre performance for different loading
directions. A trade-off must be made between handling performance and comfort, between
acceleration and wear, as well as between rolling resistance and desired friction for generating forces
in ground plane. Some of the tyre components have an important role in vehicle performance. For
example, the radial tyre components are presented in more detail in Figure 2-3. The rubber
components and patterns incorporated in the tread are critical to the friction developed between the
tyre and road under all road conditions (wet, dry, snow, etc.). Friction is most relevant in longitudinal
and lateral vehicle dynamics. The belts define the circumferential strength of the tyre and thus braking
and acceleration performance. The sidewall and plies define the lateral strength of the tyre and thus
influence the lateral (cornering) performance of the vehicle. The sidewall as well as the inflation
pressure are also significant contributors to the vertical stiffness properties of the tyre and affect how
the tyre transmits road irregularities to the remainder of the vehicle. It becomes clear that a tyre that
has strong sidewalls will support cornering at the cost of vertical compliance – reducing the comfort in
the vehicle.
53
bias-ply:
belt
radial-ply (most common on cars):
sidewall
Figure 2-3: Left: Carcass Construction, (Wong, 2001). Left top: Bias-ply construction. Left bottom: Radial
construction. Right: Radial Tyre Structure, (Cooper Tire & Rubber Co., 2007).
marking: S/[mm] / 100*H/S construction Dr/[inch]

example, car: 165 / 65 R 14
example, truck: 315 / 80 R 22.5
Unladen radius R0 = (Dr*25.4+2*(H/S)*S)/2

for car example = (14*25.4+2*0.65*165)/2 = 285 mm
for truck example = (22.5*25.4+2*0.8*315)/2 = 538 mm
Figure 2-4: Tyre marking (radial tyre)

Figure 2-5 shows how a pneumatic tyre carries the vertical load. In a traditional bicycle or motorbike
wheel, the pre-tensioned spokes have a similar role as the air and rubber parts of a pneumatic tyre.
2.2.3 Tyre Model Architecture

This section describes how a tyre model can be instantiated once per wheel and integrated in a model
together with models of vehicle, driver, and environment. It is often difficult and seldom useful to
distinguish exactly which is the model of the wheel and the tyre, respectively. We then need to differ
between parameters and variables. The description assumes dynamic models used for simulation over
time domain. Within each simulation, parameters are constant and variables can vary. If design
optimization, as opposed to simulation, the design parameters also has a role of variables (i.e. they are
varied) in the optimization, but not in time domain.
54
Upper Sidewall
Air, inflated
Belt
Shaft
Lower Sidewall
(Front and Rear Sidewall are neglected here, but
could have been modelled as shear springs.)
Inflated, but not loaded Loaded
slightly increased push,
increased
Upper
pull
𝑭𝒛
pull
decreased pull
Sidewall,
but same
𝑠
push
pull
Belt,
Shaft Shaft Shaft
decreased
Lower These 3 all change
pull
pull
in direction to carry
Sidewall, 𝑠 .
The largest of these
Air, ≈ changes is in (lower)
𝑭𝒛 sidewall!
Figure 2-5: A pneumatic tyre is pre-tensioned by inflation pressure. This figure explains how the wheel and
tyre takes (vertical) load without (significant) increase of inflation pressure.
Operational parameters
Driver
(Wheel and Tyre)
Design Behavioural Physical
Parameters Rest of
Tyre response Wheel response
variables variables vehicle Environment
Wheel and Tyre model
Operational
variables
Figure 2-6: How wheel and tyre models can come into a model context.
2.2.3.1 (Wheel and Tyre) Design parameters

We limit ourselves to today’s traditional pneumatic tyres. Then the design is captured by:
• Carcass/Material: Rubber quality and plies arrangement.
• Tread/Grooves: Groove pattern, Groove depth, Tread depth, Spikes pattern (if spikes)
• Main dimensions: Outer radius, Width, Aspect ratio
• Installation parameters: Inflation pressure
2.2.3.2 Operational parameters

These are operating conditions which vary slowly, and in this description assumed to be constant
during one manoeuvre/driving cycle. These are:
• Road surface (dry/wet, asphalt/gravel/snow/ice, …)
• Road compliance and damping (hard/soft, …)
• Wear state of tyre
• Age of tyre
• Temperature
55
2.2.3.3 Operational variables

These are operating conditions which vary quickly, and in this description assumed to be variable in
time during one manoeuvre/driving cycle. These are:
• Tyre velocities ( 𝜔) or Tyre slip (𝑠 𝑠 )
• Vertical tyre force, ( )
• Camber angle
The wheel’s rotational velocity, 𝜔, can be contained in Wheel and Tyre model instead of inside Model of
rest of the vehicle, which would mean that 𝜔 will not be part of the Operational variables interface.
Tyre forces in road plane ( ) can be given instead of tyre slip. Another alternative is to give the
corresponding actuation, e.g. wheel shaft torque and wheel steering angle relative to wheel course
angle. With those setups, the response variables in Section 2.2.3.4.
Vertical tyre force can be modelled as arbitrarily varying in time or as an offset amplitude for different
frequencies where offset is from a mean value. The mean value would then be a parameter instead of a
variable. The latter alternative can be more efficient if simulating a longer driving cycle, where
following each road wave would be very computational inefficient.
2.2.3.4 Response variables

Response refers to response to Operational variables changes.
2.2.3.4.1 Physical response variables
The variables from tyre model to model of rest of the vehicle is essential the forces and moments, see
Figure 2-2:
• Longitudinal and lateral forces ( )
• Roll moment or Overturning moment (𝑀 )
• Rolling resistance moment (𝑀 )
• Spin moment or Aligning moment (𝑀 )
Other responses are:
• Wear rate [worn rubber mass or tread depth per time unit]
• Loaded radius (𝑅 )
2.2.3.4.2 Behavioural response variables
The Responses can be modelled as Behavioural variables:
• Slip characteristics (𝐶𝑠 𝐶𝑠 , peak friction coefficient, … or other parameters in physical or
empirical tyre slip model, as in Section 2.3.3.2)
• Rolling resistance coefficient and Loaded radius (𝑅 )
• Vertical stiffness and damping coefficients
• Wear coefficient
• Relaxation length, …
Generally speaking, one can select to not introduce Behavioural response variables but only use
Physical reponse variables. If Behavioural response variables are introduced, one should consider
them as special from DAE point of view, e.g. they can be difficult to differentiate. Also, they require an
agreed definition to be interpreted, which is the reason to why they are not proposed as interface
variables to the vehicle model.
56
2.3 Longitudinal Properties of Tyres

The longitudinal force in ground plane ( or 𝑜𝑢𝑛𝑑 ), depends on the design of the tyre, the ground
surface and the operational variables, such as velocities ( 𝜔) and vertical force ( ). This section
treats the main phenomena of tyre mechanics that influence the tyre-longitudinal force. The tyre can
be under braking, free rolling, pure rolling or traction.
The whole wheel, rim and tyre, can be seen as a transmission from rotational mechanical energy to
translational mechanical energy. The shaft is on the rotating side; where we find rotational speed, 𝜔,
and the torque, 𝑇. The torque 𝑇 is then the sum of torque on the propulsion shaft and torque on the
brake disk or drum. The wheel hub is on the translating side of the transmission; were we find
translating speed, , and force, ℎ𝑢 . See Figure 2-7. Normally, we approximate and denote
𝑜𝑢𝑛𝑑 = ℎ𝑢 = and 𝑜𝑢𝑛𝑑 = ℎ𝑢 = .
x-direction =
= vehicle longitudinal forward
vehicle, wheel
including torque source, excluding wheel 𝑭𝒉𝒖𝒃 𝒛
𝑭𝒉𝒖𝒃 𝒙 𝝎
𝒗𝒙
𝒗𝒙 𝑭𝒉𝒖𝒃 𝒙
𝑭𝒉𝒖𝒃 𝒛
𝑜𝑢𝑛𝑑
𝑅
𝑜𝑢𝑛𝑑
𝑜𝑢𝑛𝑑
𝑜𝑢𝑛𝑑
ground
Figure 2-7: A wheel as a transmission from rotational [𝜔; 𝑇] to translational [ 𝑥 ; ℎ𝑢 ].
2.3.1 Tyre Rolling and Radii

In fundamental engineering mechanics, the radius of an object that rolls without sliding is a
compatibility link between the translational velocity and angular velocity as shown in Figure 2-8 a).
This relationship does not hold when the tyre is deflected as in Figure 2-8 b). Not even the deflected
radius can be assumed to be a proportionality constant between angular and translational velocity,
since the tyre contact surface slides, or slips, versus the ground. Then, an even truer picture of a rolling
tyre looks like Figure 2-8 c), where the deformation at the leading edge also is drawn. This means that
is only ≈ 𝑅 ∙ 𝜔 and only for limited slip levels.
57
a) Ideal rolling: b) With vertical load c) With vertical load

and torque ≈𝑇≈ : and (significant) torque 𝑇 > :
𝑭𝒛 𝑭𝒛
w
vx vx 𝑭𝒙 vx
w w
𝑅
𝑅
𝑅0 = inflated unloaded radius 𝑅 = free radius when loaded 𝑅 = free radius when loaded and propelled
= 𝜔 𝑅0 𝑅 = lever when loaded 𝑅 = lever when loaded and propelled
𝑅 > 𝑅0 and 𝑅 < 𝑅 and
≈𝜔 𝑅 ≈𝜔 𝑅 𝜔 𝑅 𝜔 𝑅
Figure 2-8: Radius and speed relations of a rolling tyre 𝑅0 and 𝑅 do not have exactly the same numerical
values across a), b) and c).
There is a difference speed between tyre and the road surface. The ratio between this speed and a
reference speed is defined as the “tyre slip”. The reference speed can be the translational speed of the
tyre or the circumferential speed of the tyre depending on the application. For a driven wheel, the slip
is often defined as in Eq [2.1] and for braked wheels it is often defined as in Eq [2.2].
R∙𝜔
𝑠 = ; [2.1]
|R ∙ 𝜔|
R∙𝜔
𝑠 = ; [2.2]
| |
With the definition in Equation [2.1], the following transformation falls out:
1
= ∙ ; R∙𝜔 > ;
1 𝑠 𝑅
𝜔= [2.3]
1
= ∙ ; R∙𝜔 < ;
{ 1+𝑠 𝑅
Whether one should use 𝑅 ∙ 𝜔 or as reference speed (the denominators in Equations [2.1]..[2.2]), is
not obvious and is discussed in Section 2.3.32.4.
It is also not obvious which 𝑅 to use in Equations [2.1]..[2.2], e.g. 𝑅0 𝑅 𝑅 𝑅 or 𝑅 . However, this
compendium recommends the free radius (most general of these is 𝑅 ), rather than the loaded
radii (𝑅 or 𝑅 ) because the free radii are better average value of the radius around the tyre and the
tyre’s circumference is tangentially stiff so speed has to be same around the circumference.
Sometimes one defines the Rolling radius 𝑅𝑒 = ( ⁄𝜔)| 𝑇=0, i.e. a speed ratio with dimension length,
between translational and rotational speeds, measured when the wheel is undriven (𝑇 = ). This
radius can be used for relating vehicle longitudinal speed to wheel rotational speed sensors, e.g. for
speedometer or as reference speed for ABS and ESC algorithms.
Yet another approach is to use the radius ( ⁄𝜔)|𝐹𝑥 =0, i.e. the ratio when the wheel is pure rolling.
Using 𝑅 = ( ⁄𝜔)| 𝑇=0 or 𝑅 = ( ⁄𝜔)|𝐹𝑥 =0 in the slip definitions shifts (𝑠 ) curve and 𝑇(𝑠 ) curve,
see more in Section 2.3.3.3.
The variable 𝑠 is the longitudinal slip value, sometimes also denoted as 𝜅. When studying braking, one
sometimes uses the opposite sign definition, so that the numerical values of slip becomes positive.
58
2.3.2 Rolling Resistance of Tyres

2.3.2.1 Physical explanation of rolling resistance
The rolling resitance is difficult to model physically. In the following, some possible explanations are
given. It focuses tyres on hard ground and moderate speeds, both positive and negative. On soft
ground, there are mechanisms as in 2.3.2.3.4. At very high speeds, there are inertial impact mechanism
which causes energy loss. The explanation model below mainly takes on the challenge to explain why a
rolling resistance moment appears opposite to rolling direction, also when speed is close to zero.
The overall explanation of rolling resistance for pneumatic tyres on hard flat surfaces is that the
pressure distribution is off-set in rolling direction. Damping and friction, see Figure 2-9, is one reason
for this and it is not dependent on the longitudinal force. Figure 2-9 also shows another off-set effect,
which is directly influenced by the longitudinal force.
Pressure distribution components Resulting pressure and Equivalent forces
𝑭𝒉𝒛 𝑭𝒉𝒛 T
w T w
v v
𝑭𝒉𝒙 𝑭𝒉𝒙
Rl Fx Rl
Fx
Pressure from elastic
radial deformation e Fz
Pressure from radial deformation
speed (damping & friction) Pressure
Contact patch offset, due to shear
of tyre walls (drawn for Fx>0) equivalent
Figure 2-9: Normal force distribution on a tyre. The measure e is the force offset. In steady state, the forces in
hub and contact patch are the same: ℎ = and ℎ = .
The rolling resistance coefficient is almost the same for very low speeds; even when the wheel starts
rolling from zero speed, after gradually increasing the torque up to the rolling resistance moment. The
radial friction in Figure 2-9 can explain that, but there is also another explanation, see Figure 2-10. It
explains how the contact patch is moved ahead of wheel hub. The belt is tensionally stiff and takes a
short-cut along the corda, through the contact. This builds up shear stresses in the sidewall, 𝜏𝑠𝑤 . The
belt is flexible for bending, so belt radius is proportional to belt tension force, . This is because same
effect as for tension in rope around a cylinder: 𝑜𝑟 = 𝑝𝑟 𝑠𝑠𝑢𝑟 𝑟 𝑢𝑠 𝑤 𝑡ℎ;. In our case, the
pressure corresponds to the summed effects of inflation pressure 𝑝 𝑛 and radial stress in sidewalls
𝜎𝑠𝑤 . So, the radius becomes smaller than original radius in inlet and larger in outlet. Assumption of
constant contact length and geometric constraints from tensionally rigid belt requires that contact
patch is off-setted towards the inlet, so = si n(𝜔) | |;.
2.3.2.2 Mathematical representation of rolling resistance

The rolling resistance force is defined as the loss of longitudinal force on the vehicle body, as
compared to the longitudinal force which would have been transferred with an ideal wheel. The
Rolling Resistance Coefficient, , is the rolling resistance force divided by the normal force, .
Assuming force equilibria in longitudinal and vertical direction, ℎ𝑢 = 𝑜𝑢𝑛𝑑 = and ℎ𝑢 =
𝑜𝑢𝑛𝑑 = , see Figure 2-7.
𝑇
𝑅 [2.4]
=
59
1. wheel 2. wheel, except belt 2. belt

𝑟 𝑢𝑠 𝑇
w 0
𝑇
𝑇 𝑇
v
undeformed
Rl deformed 𝑠𝑤
𝜏𝑠𝑤
𝑠𝑤
3. belt in contact
4. wheel
𝑠𝑤 + 𝑠𝑤 +
= 0 + = 0
𝑟 𝑢𝑠 > 0
𝑟 𝑢𝑠 < 0 𝑠𝑤
Figure 2-10: 1. Wheel at low speed ( small and negative). 2. Free body diagrams with belt separate. 3. How
belt tension force change before and after contact. 4. Influence of radii variation on contact patch position.
denotes the longitudinal force on the wheel, 𝑇 denotes the applied torque and 𝑅 denotes the tyre
radius. For a free rolling tyre, where 𝑇 = , becomes simply ⁄ . One often see definitions of
which assumes free rolling tyre; but Eq [2.4] is more generally useful.
Regarding which radius to use, it should rather be 𝑅 than 𝑅0 , because 𝑅 represents the lever for the
longitudinal tyre force around the wheel hub.
A free body diagram of the forces on the wheel can be used to explain the rolling resistance. Consider
Figure 2-9 which represents a free rolling wheel under steady state conditions. The inertia of the
wheel is neglected.
Longitudinal and vertical force equilibria are already satisfied, due to assumptions above. However,
moment equilibrium around wheel hub requires:
𝑇
𝑇 ∙𝑅 ∙ = ⇒ = ∙ ; [2.5]
𝑅 𝑅
This result suggests that the force , which pushes the vehicle body forward, is the term 𝑇⁄𝑅 (arising
from the applied propulsion or brake torque T) minus the term ∙ ⁄𝑅 . The term can be seen as a
force 𝑜 and referred to as the rolling resistance force. The dimensionless ratio ⁄𝑅 expression is the
rolling resistance coefficient, :
= ; [2.6]
𝑅
Eq [2.6] is a definition of rolling resistance coefficient based assumed physical mechanisms internally
in the tyre with road contact. Rolling resistance coefficient can also be defined based on quantities
which are measurable externally, see Eq [2.4]. Sometimes one sees = ⁄ as a definition, but
that is not suitable since it assumes absence of torque.
60
It is important to refer to this phenomenon as rolling resistance as opposed to rolling friction. It is

not friction in the basic sense of friction, because ∙ except for the special case when un-
driven wheel (𝑇 = ). Figure 2-11 shows an un-driven and pure rolling.
Rolling resistance is a torque loss. Other torque losses, which can be included or not in tyre rolling
resistance, are:
• losses associated with friction in gear meshes,
• drag losses from oil in the transmission,
• wheel bearing (and bearing sealings) torque losses,
• drag from brake discs,
• drag losses from aerodynamic around the wheel, and
• uneven road in combination with suspension damping that dissipates energy.
These should, as rolling resistance, be subtracted from propulsion/brake torque when calculating 𝑇 in
Figure 2-11. However, sometimes they are included as part of the tyres rolling resistance coefficient,
which is somewhat misleading even if it works if done correctly. The wheel bearing torque loss have
two torque terms: one is proportional to vertical load on the wheel (adds typically . 3 to rolling
resistance coefficient), and the other is of constant magnitude but counter-directed to rotation speed.
The former term can be included in rolling resistance coefficient. The aerodynamic losses due to wheel
rotation are special since they vary with wheel rotational speed, meaning that they (for constant
vertical load) are relevant to include when studying the variation of rolling resistance coefficient with
vehicle speed. A summarizing comment is that one has to be careful with where to include different
torque losses, so that they are included once and only once.
Driven Undriven (T=0)
𝒆
Fx has to be: 𝑭𝒙 𝑭𝒉𝒛 ;
𝒍
w 𝑭𝒛 T
w 𝑭𝒛
v v
Fx
Fx
𝑅 1 |Fx|
Fx |Fx| 𝑅
𝑅
e Fz T Fz
Undriven (T=0) Pure rolling (Fx=0) e
Figure 2-11: Driven wheel with rolling resistance, special cases “Un-driven” and “Pure rolling”. Normally, the
force in hub and contact patch are the same: ℎ = .
One should note that torque losses as discussed above are not the only vehicle energy losses related to
the tyres. There are also speed losses: speed losses in the tyre slip, both longitudinal and lateral. These
cannot be modelled as rolling resistance.
2.3.2.3 Variation of rolling resistance

Several parameters will affect the rolling resistance moment (or Rolling resistance coefficient). Design
parameters, see Figure 2-6:
• Tyre material. Natural rubber often gives lower rolling resistance.
• Radial tyres have more flexible sides, giving lower rolling resistance also bias ply have a
greater crown angle causing more internal friction within the tyre during deflection.
• Geometry:
o Diameter. Large wheels often have lower coefficient of rolling resistance
o Width
o Groove depth
o Tread depth
61
• Higher inflation pressure gives lower rolling resistance on hard ground but higher rolling
resistance on soft ground (and vice versa), see Figure 2-12 and Figure 2-13.
Operational parameters, see Figure 2-6:
• High temperatures give low rolling resistance. Tyres need to roll approximately 30 km before
the rolling resistance drop to their lowest values.
• Road/ground, sometimes covered: Clean asphalt, Asphalt with water/leaves/sand/…,
Loose/hard gravel, Snow/ice. Soft ground or covered hard ground increases rolling resistance.
• Wear state. Worn tyres have lower rolling resistance than new ones (less rubber to deform).
Operational variables, see Figure 2-6:
• Vertical force.
• Speed. Rolling resistance increases with vehicle speed due to rubber hysteresis and air drag.
• Tyre loads (propulsion/braking and lateral forces)
2.3.2.3.1 Variation of tyre type
Trucks tyres have a much lower rolling resistance coefficient than passenger vehicle tyres,
approximately half or less. Tyres have developed in that way for trucks, because their fuel economy is
so critical.
2.3.2.3.2 Variation of vertical load
In a first approximation, the rolling resistance force is proportional to vertical force, i.e. RRC is
constant. But, generally, the RRC decreases slightly with vertical force. This is the main reason to why
multiple-axle commercial vehicles lift axles when they are driven with low payload.
2.3.2.3.3 Variation of speed
As an example, left part of Figure 2-12 shows the influence of tyre construction and speed on rolling
resistance. The sudden increase in rolling resistance at high speed is important to note since this can
lead to catastrophic failure in tyres. The source of this increase in rolling resistance is a high energy
standing wave that forms at the trailing edge of the tyre/road contact.
There are some empirical relationships derived for the tyre's rolling resistance. It is advisable to refer
to the tyre manufacturer's technical specifications when exact information is required. This type of
information is usually very confidential and not readily available. Some general relationships have
been developed, from (Wong, 2001):
Radial-ply passenger car tyres: = . 136 + . 4 ∙ 1 −6 ∙
Bias-ply passenger car tyres: = . 169 + .19 ∙ 1 −6 ∙
Radial-ply truck tyres: = . 6 + . 3 ∙ 1 −6 ∙
Bias-ply truck tyres: = . 7 + .45 ∙ 1 −6 ∙
As seen in Figure 2-12, a rule of thumb is that rolling resistance coefficient is constant up to around
100 km/h.
2.3.2.3.4 Variation of road surface
The road or ground can mainly vary in two ways: rigidness of road and slipperiness of road. It is not
solely a question of the road, because such rigidness should be judge relative to tyre inflation pressure
and such slipperiness should be judged relative to tyre tread pattern, including whether the tyre has
spikes or not.
Right part of Figure 2-12 shows that the rolling resistance changes a lot due to different road/ground
material and inflation pressures. As can be expected, a range of values exist depending on the specific
tyre and surface materials investigated. On hard ground, the rolling resistance decreases with
increased inflation pressure, which is in-line with the explanation model used above, since higher
pressure intuitively reduces the contact surface and hence reduces e. On soft ground the situation is
reversed, which requires a slightly different explanation model, see Figure 2-13. On soft ground, the
62
ground is deformed so that the wheel rolls in a “local uphill slope” with inclination angle . Intuitively,
a higher inflation pressure will lead to more deformation of the ground, leading to a steeper .
.05
Coefficient of rolling resistance, f

0.4
Coeff. of Rolling Resistance
.04
Sand
.03
0.2
.02 Bias Ply

Bias-Belted Mediumhard
Medium HardsoilSoil
Radial
Concrete
.01 0
0 25 50 75 100 125 150 175 0 20 40
Speed [km/h] Inflation Pressure [psi]
Figure 2-12 : Left: Rolling resistance coefficient variation with speed for different tyre types (Gillespie, 1992).
Right: Range of Coefficient of Rolling Resistance for different road/ground material
Wheel rolling on hard ground: Wheel rolling on soft ground:
w 𝑇 w 𝑇 w 𝑇
v horizontal v ⃗
j
𝑐
j
𝑐
⃗𝑐
Figure 2-13 : Rolling resistance explanation model for hard and soft ground.
2.3.2.3.5 Variation of longitudinal force, propulsion and braking
There is also a dependency on longitudinal force, see Figure 2-14. RRC increases with positive force
and decreases with negative force, due to that the shearing of the sidewall moves the contact forward
or rearwards. The wheel radius also decreases with magnitude of force. For negative forces, these two
effects have opposite influence, so the change is less for negative force, as seen in Figure 2-14.
Rolling resistance
coefficient, 𝑅𝑅𝐶 =
0.08
0.04
Normalised
Longitudinal
𝐹𝑥
-0.4 -0.2 0 +0.2 +0.4 Force, 0 = 𝐹
Figure 2-14 : Rolling resistance dependency of longitudinal tyre force. Inspired by (Wong, 2001)
63
2.3.3 Longitudinal force due to slip

The longitudinal force, , between the tyre and ground influences the vehicle propulsion and braking
performance. We can see it as a depending on the sliding between the tyre and the ground.
First, compare friction characteristics for a translating block of rubber with a rolling wheel of rubber.
Figure 2-2 shows the basic differences between classical dry-friction, or Coulomb friction, of such
sliding block and the basic performance of a rolling tyre. Experiments have shown that the relative
speed between the tyre and the road produces a frictional force that has an initial linear region that
builds to a peak value. After this peak is achieved, no further increase in the tangential (friction) force
is possible. There is not always a peak value, which is shown by the dashed curve in the figure. The
slope in the right diagram will be explained in Section 2.3.3.1, using the so called “tyre brush model”.
Sliding block (or wheel with ω=0) Rolling wheel of rubber

w
N
N vx
F
F
T
𝑅 𝜔
v F
F N N
N F (𝑁 = )
F N
F F
𝑠𝑡 𝑡 𝑁 𝑒 𝑘 𝑁
slide 𝑁
𝑑 𝑛 𝑁 𝑑 𝑛
stick
Dv=v slip=Dv/vref=(R*w-vx)/vref
slide
• A stick mode where Dv=0. • No stick mode (if w>0).

• In slide mode, the friction force, F, • The friction force, F, is depending on
is only depending on sign(Dv). relative sliding speed, Dv/vref.
Figure 2-15: Friction characteristics
2.3.3.1 Tyre brush model for longitudinal slip

The brush model is frequently used to explain how tyre develop forces in ground plane, see e.g.
References (Pacejka, 2005) and (Svendenius, 2007). The brush model is a physically based model that
uses shear stress and dry friction at a local level, i.e. for each contact point in the contact patch. Figure
2-16 shows the starting point for understanding the brush model. A first simple variant of the brush
model, uses the following assumptions:
o Sliding and shear stress only in longitudinal direction (as opposed to combined longitudinal
and lateral)
o Uniform and known pressure distribution over a constant and known contact length (as
opposed to using a contact mechanics based approach, which can calculate pressure
distribution and contact length.)
o Contact length is known (as opposed to function of vertical force).
o No difference between static and dynamic coefficient of friction
o Only studying the steady state conditions, as opposed to including the transition between
operating conditions. Steady state as opposed to stationary variations, which would arise for
driving on undulated road.
64
w*R
vx
bristle or block
Figure 2-16: Tyre ground contact for braked tyre. Origin to the “Brush model”. Picture from Michelin.
2.3.3.1.1 Uniform pressure distribution and known contact length
These assumptions lead to a model as drawn in Figure 2-17. The shear stress is the (tyre-) longitudinal
stress in the brushes.
The rubber bristles (rubber elements) are fixed to the wheel and have friction contact to the ground.
When going from Figure 2-17 to equations, we move from a view discretized in longitudinal and
lateral dimension to a continuous view. So, e.g., forces [𝑁] becomes stress [𝑁 ]. The shear stress of
the element develops as in Hooke’s law, see Equation [2.7].
𝜏 = 𝐺 ∙ 𝛾;
𝑤ℎ 𝑟 𝜏 = 𝑠ℎ 𝑟 𝑠𝑡𝑟 𝑠𝑠 𝐺 = 𝑠ℎ 𝑟 𝑜 𝑢𝑙𝑢𝑠 [2.7]
𝑛 𝛾 = 𝑠ℎ 𝑟 𝑜𝑟 𝑡 𝑜𝑛 𝑛 𝑙 ;
When a rubber element enters the contact patch, it lands un-deformed, i.e. with 𝛾 = . The further into
contact, along coordinate 𝜉, we follow the element, the more sheared will it become. Since the ground
end of the element sticks to ground, the increase becomes proportional to the speed difference ∆ and
the time 𝑡 it takes to reach the coordinate 𝜉:
∆ ∙ 𝑡(𝜉) (𝑅 ∙ 𝜔 ) ∙ 𝑡(𝜉)
𝛾= = ;
𝐻 𝐻 [2.8]
𝑤ℎ 𝑟 𝑡(𝜉) = 𝜉 ⁄ 𝑇 𝑛𝑠 𝑜 𝑡 𝑛 𝑇 𝑛𝑠 𝑜 𝑡 ≈ |𝑅 ∙ 𝜔|;
The velocity 𝑇 𝑛𝑠 𝑜 𝑡 is the velocity of which the brush bristles are transported through the contact.
Combining these equations yields the following, where the slip 𝑠 can be identified:
𝐺 𝑅∙𝜔 𝑅∙𝜔 𝐺
𝜏= ∙ ∙ 𝜉 = {𝑠 = ; } = ∙ 𝑠 ∙ 𝜉; [2.9]
𝐻 |𝑅 ∙ 𝜔| |𝑅 ∙ 𝜔| 𝐻
65
rolling direction
View from rear,
trailing end L=length of contact patch leading edge at one certain x:
of contact of contact
View from (=rubber tyre (=rubber
outlet) 𝑅 𝜔 inlet) W=width
side: x=L x=xc x=0
x
H=height
bristle road
(relative to wheel hub)
𝛾
Velocities of bristle
increasing deformation ends in stick zone:
Contact patch, 𝜔
𝑅 𝜔
view from above:
same deformation Drawn for 𝑅 𝜔 >
x
pressure 𝑅 𝜔
stick zone
x stick
slip
slip zone
“mode”
x shear stress 𝜏
(or shear deformation angle, 𝛾)
Figure 2-17: The physical model of simple brush model for longitudinal slip. Drawn for propelled tyre. The
bristles represent the rubber tread, so they don’t include elasticity in sidewall. Drawn for 𝑅 ∙ 𝜔 > .
With this definition of slip, 𝑠 , we automatically handle 𝜔 and both positive and 𝜔 and both
negative. The case when 𝜔 and have different signs will be handled below.
As long as friction limit is not reached (|𝜏| < 𝑝) within the whole contact, we can find the force as
this integral:
𝐿 𝐿
𝐺 𝐺 𝑊
= 𝑊 ∙ ∫𝜏 ∙ 𝜉 = 𝑊 ∙ ∫ ∙𝑠 ∙𝜉∙ 𝜉 = ∙𝑠 =
𝐻 𝐻
0 0 [2.10]
𝐺 𝑊
= {𝐶 = }=𝐶 ∙𝑠 ; 𝑜𝑟 𝑠 <
𝐻 𝐶
If friction limit is reached within the contact, i.e. at the break-away point 𝜉 = 𝜉𝑐 , we have to split the
𝐺 𝐶 𝑊 𝐿2
integral in two. The point 𝜉𝑐 is defined by 𝜏(𝜉𝑐 ) = 𝐻 𝑠 𝜉𝑐 = 𝑊 𝐿𝑥2 𝑠 𝜉𝑐 = 𝑝; ⇒ 𝜉𝑐 = 𝐶𝑥 𝑠𝑥
=
𝐹 𝐿
𝐶𝑥 𝑠𝑥
;. For 𝜉 > 𝜉𝑐 , the rubber element will slide with a constant 𝜏 = ∙ 𝑝.
𝐿 𝜉𝑐 𝐿
𝐶
= 𝑊 ∙ ∫𝜏 ∙ 𝜉 = 𝑊 ∙ ∫ ∙𝑠 ∙𝜉∙ 𝜉+𝑊∙ ∫ ∙𝑝∙ 𝜉 =
𝑊
0 0 𝜉𝑐
𝐶 𝜉𝑐
= ∙𝑠 ∙ + ∙𝑝∙𝑊∙( 𝜉𝑐 ) = {𝑝 = ; 𝑛 𝜉𝑐 = ;} = [2.11]
𝑊 𝐶 𝑠
= + ∙ ∙ (1 )= ∙ ∙ (1 ); 𝑜𝑟 𝑠 >
4 𝐶
⏟ 𝑠 ⏟ 𝐶 𝑠 4 𝐶 𝑠 𝐶
𝐹𝑥 𝑠𝑡𝑖𝑐𝑘 𝐹𝑥 𝑠𝑙𝑖𝑝
66
As seen in Eq [2.11], the force terms, 𝑠𝑡 𝑐𝑘 and 𝑠 , from each of stick and slip regions are
identified. These two terms are shown plotted in Figure 2-20.
The case when 𝜔 and have different signs leads to that 𝜉𝑐 = ;, since the bristles will deform in the
opposite direction. So, the whole contact has 𝜏 = ∙ 𝑝, which leads to:
𝐿 𝐿
= 𝑊 ∙∫𝜏 ∙ 𝜉 = 𝑊 ∙∫ 𝑝∙ 𝜉 =𝑊∙ 𝑝 ∙ = {𝑝 ∙ 𝑊 ∙ = }= ; [2.12]

0 0
Hence, the total expression for becomes as in [2.13], where a generalisation to cover also when 𝜔
and have different signs (𝜔 < ) is done. We also add subscript 𝑥 on 𝐺 and 𝐻, to prepare for a
corresponding model for lateral forces, in Section 2.4.
= si n(𝑠 ) ∙ ∙ 𝜔 <
∙ ∙
=𝐶 ∙𝑠 ; 𝑙𝑠 |𝑠 | ≤ ⇔| |≤
= ∙𝐶
∙ 1 [2.13]
= si n(𝑠 ) ∙ ∙ ∙ (1 ∙ ); 𝑙𝑠
{ 4 ∙ 𝐶 |𝑠 |
𝐺 ∙𝑊∙ 𝑅 𝜔
𝑤ℎ 𝑟 𝐶 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ] 𝑛 𝑠 = ;
∙𝐻 |𝑅 𝜔|
It is very important to reflect over which of the physical quantities that reasonably has to be modelled
as varying, see Section 2.2.3. This will of course depend on the driving manoeuvre studied, but here is
a typical situation: The slip 𝑠 and normal load are typical varying and defined by the vehicle model.
The quantities 𝐺 𝑊 𝐻 are often reasonably constant, so they can be parameters. However, it is
often not reasonable to assume that the contact length is constant. is rather a function of : ( );.
To assume constant (or constant 𝐶 ) is a common misuse of the brush model. Elaboration with
different ( ) is found in Section 2.3.3.1.3.
The shape of the (𝑠 ) curve becomes as shown in Figure 2-18, the curves “Reference contact length”
and “Constant contact length model”. Variations of and influence together, via .
The case when 𝜔 < is unusual and can only occur when |𝑠 | > 1. An example is when vehicle
moves rearward with = 1 𝑠, would be that the wheel spins forward, e.g. with 𝑅 𝜔 = 1 𝑠.
Then 𝑠 = + and = ∙ . Also, if increase to 𝑅 𝜔 = +∞ 𝑠, we get same = ∙ , but
with 𝑠 = +∞. One can also note that there is another 𝑅 𝜔, for same , which gives 𝑠 = + . This is
𝑅 𝜔 = 1⁄3 𝑠 and then = ∙ ∙ (1 ∙ ⁄(8 ∙ 𝐶 )), which is < . So, is uniquely
defined for any (𝜔 ), but has double solutions for some 𝑠 , when |𝑠 | > 1.
If we instead hold a certain forward vehicle speed, e.g. =1 𝑠, and study how varies with 𝜔, we
can identify three characteristic levels of 𝜔:
• Full rearward traction: 𝑅 𝜔 = ∞ ⁄𝑠. This gives 𝑠 = ∞, with different signs on 𝜔 and ,
and = ;
• Locked wheel: 𝑅 𝜔 = ⁄𝑠. This also gives 𝑠 = ∞ and = ;
• Pure rolling wheel: 𝑅 𝜔 = 1 ⁄𝑠. This gives 𝑠 = and = ;
• Full forward traction: 𝑅 𝜔 = +∞ ⁄𝑠. This gives 𝑠 = +1 and = ∙ ∙ (1
∙ ⁄(4 ∙ 𝐶 )) ≈ {𝑡𝑦𝑝 𝑙𝑙𝑦} ≈ ( .95. . .98) ∙ , which is < ;
The first case is achievable, with an electric motor braking but not with friction brakes, where 𝜔 and
have different signs. The last case shows that we cannot reach in the direction the vehicle
moves, since there will always be a small part on the inlet side of the contact where the shear stress
has not reached 𝑝. In practice, we can see it as ≈ , but when using the model mathematically, it
can be good to note such small phenomena.
67
Shear stress
distribution
above linear part
𝑠𝑦 𝑝𝑡𝑜𝑡 =1
Shear stress distribution

in linear part
Figure 2-18: Shape of force/slip relation derived with brush model with uniform pressure distribution and
different contact length model. Slope 𝐶 is the longitudinal slip stiffness.
2.3.3.1.2 Longitudinal tyre (slip) stiffness
In summary for many models (and tests!) the following is a good approximation for small longitudinal
slip, and certain normal load and certain friction coefficient:
=𝐶 ∙𝑠 [2.14]
For the brush model, or any other model which describes (𝑠 … ), one can define the
“Longitudinal tyre (slip) stiffness” 𝐶 , which have the unit 𝑁 = 𝑁⁄1 = 𝑁 (( 𝑠)⁄( 𝑠)). It is the
derivative of force with respect to slip. In many cases one means the derivative at sx=0:
𝜕
𝐶 =( (𝑠 … ))| [2.15]
𝜕𝑠
𝑠𝑥 =0
Note that 𝐶 is not a stiffness in the conventional sense, force/deformation. The tyre also has such a
conventional stiffness, defined by force and deformation of a not rotating wheel. Often, it is obvious
which stiffness is relevant, but to be unambiguous one can use the wording: “slip stiffness [N/1]” and
“deformation stiffness [N/m]”.
With the brush models with both pressure distributions, Eq [2.13] and Eq [2.19], we get the 𝐶 =
𝐺 ∙ 𝑊 ∙ ⁄( 𝐻). With 𝐺 = .5 [𝑀 𝑁⁄ ]; (typical shear modulus in rubber), 𝑊 = = .1. . .1 [ ];
(typical sizes of contact patch for passenger car) and 𝐻 = . 1 [ ]; (approximate tyre tread depth)
one gets ≈ 5[𝑘𝑁] < 𝐶 = 𝐺 ∙ 𝑊 ∙ ⁄( 𝐻) <≈ 4 [𝑘𝑁]; which is an interval that contains typically
measured stiffnesses for passenger car tyres. This indicates that the brush model models the essential
physical phenomena and that the sheared part (the bristles) is rather only the tread than the whole
elastic part sidewall and tread together.
The vertical load on the tyre affects the force generation, . An approximation is that is
proportional to . Then, one can define the Cornering Coefficient, 𝐶𝐶 :
68
𝐶
; ⇒ 𝐶 ≈ ⇒ 𝐶 = 𝐶𝐶 ; ⇒ 𝐶𝐶 = ; [2.16]
𝑠
With the brush models with both uniform pressure distribution we can connect and sizes of contact
patch with = 𝑝 𝑊 . Assuming approximately same contact pressure as inflation pressure, 𝑝 =
𝑝 𝑛 , we can express 𝐶 = 𝐺 ∙ 𝑊 ∙ ⁄( 𝐻) = 𝐺 ∙ ∙ ⁄( 𝑝 𝑛 𝐻). This gives 𝐶𝐶 =
𝐺 ∙ ⁄( 𝑝 𝑛 𝐻). With similar values as above, and typical passenger car tyre pressure 𝑝 𝑛 = 3
1 5 [𝑁⁄ ], we get 𝐶𝐶 ≈ 1 . With typical truck tyre pressure 𝑝 𝑛 = 9 1 5 [𝑁⁄ ], and slightly
larger and 𝐻, we get 𝐶𝐶 ≈ 5. These are both in the same magnitude and relative relation as one can
measure, see Section 1.8.
2.3.3.1.3 Influence of vertical load and friction in brush model
It is empirically found that longitudinal slip stiffness 𝐶 increases with vertical load . The increase is
almost proportional, but slightly degressive (𝜕 𝐶 ⁄𝜕 < ). The corresponding lateral slip stiffness
𝐶 has an even clearer degressive tendency, see Figure 2-31.
We will now try to explain why variation in gives a variation in slip stiffness. The simplest model for
lateral slip stiffness (Section 2.4.1.1) gives the same formula as for longitudinal: 𝐶 = 𝐶 = 𝐶 =
𝐺 ∙ 𝑊 ∙ ⁄( 𝐻) ;. The reasoning will therefore be generic for longitudinal both lateral. The
parameters 𝐺 𝑊 and 𝐻 are reasonably constant for varying , but not . So, variation of the contact
length with could explain why 𝐶 varies. One measurement of how contact length vary almost
linear is seen in Figure 2-31, but many measurements show a more degressive character. The contact
length will reasonably vary according to some deformation model.
Four deformation models will be compared:
• Constant contact length
• √ , which can be motivated by assuming Hertz’s contact theory for point contact. This is
3
implemented as = 𝑘𝑅 , where 𝑘𝑅 is a “material modulus per radius” with dimension

𝑜𝑟 ( 𝑟 𝑙 𝑛 𝑡ℎ). One can understand that length as a weighted average of wheel radius
and radius of the tyre toroid.
• √ , which can be motivated by assuming Hertz’s contact theory for line contact. This is
2
implemented as = 𝑘, where 𝑘 is a material modulus with dimension force/area.

• , which can be motivated by assuming that the tyre is a membrane with constant contact
pressure 𝑝 = inflation pressure 𝑝 (like a balloon).
Note that we use Eq [2.13] which implicitly means that we keep assuming a pressure which is constant
over the contact surface (but it varies with ). Using also the Hertz’s theory elliptic pressure
distribution over the contact surface would be more consistent, but we leave pressure distribution
uniform until Section 2.3.3.1.5. The results of the different deformation models are plotted in Figure
2-18 using a constant value of tyre parameters, but varying and .
Studying Figure 2-18, we find:
• Constant contact length model gives 𝐶( ) = 𝑜𝑛𝑠𝑡 𝑛𝑡;.
• √ gives 𝜕𝐶 ⁄𝜕 < ; (degressive).
3
• √ gives 𝐶( )
2
; (proportional).
• gives 𝜕𝐶 ⁄𝜕 > ; (progressive).
So, the deformation model “ √ ” can be one explanation of why slip stiffness is degressive with
3
vertical force.
Studying Figure 2-18, we find that increasing only increases the saturation level, while leaving the
slope at lower 𝑠 constant, or possibly slightly increased. Variation of involves the contact length
model. Hertz’s line contact simply scales the curve in force direction, but the other models can
69
motivate a slightly “degressive or progressive” scaling. Overall, it can be concluded that there are
arguments for two conceptual ways the tyre characteristics changes, see Figure 2-19.
In practice, it is relatively straight forward to vary the vertical force. But varying the friction coefficient,
for one certain tyre, cannot be done without changing the road surface. The local deformation and
friction model in the brush models above, 𝜏 = 𝐺 𝛾; and 𝜏 = 𝑝;, are well motivated on hard and dry
road, such as asphalt. When changing the road surface it is reasonable to add snow, water, ice, gravel,
etc., which can motivate totally different local models, e.g. 𝜏 = 𝑢𝑛 𝑡 𝑜𝑛(𝑅 𝜔 ), which might even
lead to that = (𝑅 𝜔 ); instead of simply = (𝑠 );.
Air
Asphalt temperature:
-15 to -10 C
⁄
Normalized force,
Snow
=
Reference 𝑠
Doubled vertical force (simplified) Ice
Doubled vertical force (tendency)
Doubled friction coefficient (simplified)
Doubled friction coefficient (tendency)
slip, 𝑠
Figure 2-19: How tyre characteristics typically vary due to varying vertical force and road friction. Left:
Theory. Right: Measurements with varying friction, i.e. varying surfaces, from PhD course by Ari Tuononen,
Aalto university, Finland, Saariselkä, Finland 2014-03-15..22.
2.3.3.1.4 Influence of different static and dynamic friction
A common model for friction is that coefficient of friction to remain sticking, 𝑠𝑡 𝑐𝑘 , is higher than the
coefficient of friction when slipping has started, 𝑠 . This is sometimes called “stiction”. If this is
implemented in the model, the brush model can explain why the overall (𝑠 ) often has a peak, as
indicated already in Figure 2-15. In the derivation in Eq [2.11], the differing between 𝑠𝑡 𝑐𝑘 and 𝑠 ,
𝐿 𝐹 𝐿
affects like this: 𝑠 =𝑊∙ 𝜉𝑐 𝑠 ∙ 𝑝 ∙ 𝜉 ; and 𝜉𝑐 = 𝑠𝑡𝑖𝑐𝑘
𝐶𝑥 𝑠𝑥
;.
= si n(𝑠 ) ∙ 𝑠 ∙ ; 𝜔 <
∙
𝑠𝑡 𝑐𝑘 𝑠𝑡 𝑐𝑘 ∙
=𝐶 ∙𝑠 ; 𝑙𝑠 |𝑠 | ≤ ⇔| |≤
= ∙𝐶
= si n(𝑠 ) ∙ ∙ (1 (1
𝑠𝑡 𝑐𝑘
)) ; 𝑙𝑠 [2.17]
𝑠 𝑠𝑡 𝑐𝑘
{ 𝐶 |𝑠 | 𝑠
𝐺 ∙𝑊∙ 𝑅 𝜔
𝑤ℎ 𝑟 𝐶 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ] 𝑛 𝑠 = ;
∙𝐻 |𝑅 𝜔|
This model only gives a peak if 𝑠𝑡 𝑐𝑘 > 𝑠 , see Figure 2-21. The model explains that a peak can
occur, but the shape of the curve does not correlate well with tyre measurements. The model can work
to explain some phenomena on vehicle level, but a better model can be found if assuming parabolic
pressure distribution, see Section 2.3.3.1.5.
2.3.3.1.5 Brush model with parabolic pressure distribution
Hertz’s contact theory for line contact motivate an elliptical pressure distribution. A parabolic
pressure distribution is an approximation of such elliptical and it gives an alternative brush model, as
compared to the one appearing from uniform pressure. The coefficients in the parabolic pressure
𝐿
function have to be chosen such that 0 𝑝(𝜉) 𝑊 𝜉 = and 𝑝( ) = 𝑝( ) = ;:
70
6∙ 𝜉 𝜉
𝑝(𝜉) = ∙ (1 ); [2.18]
𝑊
If we do the corresponding derivation as for the uniform pressure distribution, e.g. assuming a stick
and slip zones, the location of where slip starts, 𝜉𝑐 , becomes:
6∙ 𝜉𝑐 𝜉𝑐 𝐺 𝑊 𝑠
𝜏(𝜉𝑐 ) = (𝐺 𝐻) 𝑠 𝜉𝑐 = 𝑝(𝜉𝑐 ) = ∙ (1 ) ; ⇒ 𝜉𝑐 = ;
𝑊 6 𝐻
The slip where 𝜉𝑐 appears outside < 𝜉𝑐 < is when 𝜉𝑐 becomes < , which is when the whole
contact slips:
𝐺 𝑊 𝑠 𝐶 𝑠 3
𝜉𝑐 = = (1 ) < ;⇒ 𝑠 > ;
6 𝐻 3 𝐶
Total longitudinal force, , becomes:
= si n(𝑠 ) ∙ ∙ 𝜔 <
(𝐶 ∙ |𝑠 |) (𝐶 ∙ |𝑠 |) 3∙ ∙
= si n(𝑠 ) ∙ (𝐶 ∙ |𝑠 | + ); 𝑙𝑠 |𝑠 | < ⇔
= 3∙ ∙ 7∙( ) 𝐶
⇔| |≤ ∙
[2.19]
{= si n(𝑠 ) ∙ ∙ ; 𝑙𝑠
𝐺∙𝑊∙ 𝑅 𝜔
𝑤ℎ 𝑟 𝐶 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ] 𝑛 𝑠 = ;
∙𝐻 |𝑅 𝜔|
The shape of this curve becomes as shown in Figure 2-20. The uniform pressure distribution model is
drawn as reference. Note that the parabolic pressure distribution does not give any linear part, but the
derivative at 𝑠 = is same, 𝐶 = 𝐺 ∙ 𝑊 ∙ ⁄( 𝐻).
Figure 2-20: Shape of force/slip relation derived with brush model with parabolic pressure distribution and
different contact length models. Also, the force terms from stick- and slip-regions are shown.
We now introduce different 𝑠𝑡 𝑐𝑘 and 𝑠 . We also identify each force term, for stick and slip regions,
as follows:
71
𝜉𝑐 𝜉𝑐
𝐶 𝐶 𝑠
𝑠𝑡 𝑐𝑘 = 𝑊 ∙ ∫ 𝜏(𝜉) ∙ 𝜉 = 𝑊 ∙ ∫ ∙ 𝑠 ∙ 𝜉 ∙ 𝜉 = { 𝜉𝑐 = (1 );} = ⋯ =
𝑊 3 𝑠𝑡 𝑐𝑘
0 0
𝐶 𝑠
= 𝐶 ∙ (1 ) 𝑠 ;
3 𝑠𝑡 𝑐𝑘
𝐿 𝐿
6∙ 𝜉 𝜉
𝑝= ∙ (1 );
𝑊
𝑠 =𝑊∙ ∫ 𝑠 ∙𝑝∙ 𝜉 = 𝑠 ∙ 𝑊 ∙ ∫𝑝 ∙ 𝜉 = 𝐶 𝑠 =⋯=
𝜉𝑐 𝜉𝑐 𝜉 = (1 );
{ 𝑐 3 𝑠𝑡 𝑐𝑘 }
𝐶 𝑠 𝐶 𝑠
= 𝑠 ∙ ∙ (3 ) ( ) ;
3 𝑠𝑡 𝑐𝑘 3 𝑠𝑡 𝑐𝑘
3∙ 𝑠𝑡 𝑐𝑘 ∙
𝐵𝑜𝑡ℎ 𝑥𝑝𝑟 𝑠𝑠 𝑜𝑛𝑠 𝑙 𝑜𝑛𝑙𝑦 𝑜𝑟 𝑠 < ;
𝐶
The force terms, 𝑠𝑡 𝑐𝑘 and 𝑠 , from each of stick and slip regions are identified for 𝑠 <
3 ∙ 𝑠𝑡 𝑐𝑘 ∙ ⁄𝐶 . When slip is larger, 𝑠 > 3 ∙ 𝑠𝑡 𝑐𝑘 ∙ ⁄𝐶 , the whole contact slips, 𝑠𝑡 𝑐𝑘 = ; and
𝑠 = 𝑠 ;. If we sum to = 𝑠𝑡 𝑐𝑘 + 𝑠 we get:
si n(𝑠 ) =
= 𝑠 ∙ ; 𝜔 <
𝑠 (𝐶 ∙ |𝑠 |) 𝑠 (𝐶 ∙ |𝑠 |)
= 𝐶 ∙ |𝑠 | ( ) + (3 ) ;
𝑠𝑡 𝑐𝑘 3 ∙ 𝑠𝑡 𝑐𝑘 ∙ 𝑠𝑡 𝑐𝑘 7 ∙ ( 𝑠𝑡 𝑐𝑘 )
=
3 ∙ 𝑠𝑡 𝑐𝑘 ∙ [2.20]
𝑙𝑠 |𝑠 | <
𝐶
{= 𝑠 ∙ ; 𝑙𝑠
𝐺∙𝑊∙ 𝑅 𝜔
𝑤ℎ 𝑟 𝐶 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ] 𝑛 𝑠 = ;
∙𝐻 |𝑅 𝜔|
Typical values of friction coefficients that give good resemblance with measurements for the parabolic
pressure distribution are 𝑠𝑡 𝑐𝑘 ⁄ 𝑠 = 1.3. .1.8.
tyre_noStiction.Fx tyre_smallStiction.Fx tyre_criticalStiction.Fx tyre_largeStiction.Fx
tyre2_noStiction.Fx tyre2_smallStiction.Fx tyre2_criticalStiction.Fx tyre2_largeStict
9000 9000
8000 𝑠 = 1;
8000
𝑠𝑡 𝑐𝑘 = 3;
7000 7000
𝑠𝑡 𝑐𝑘 = 3; 𝑠𝑡 𝑐𝑘 = ;
6000 6000
𝑠𝑡 𝑐𝑘 = 1.5;
𝑠𝑡 𝑐𝑘 = ;
5000 5000
𝑠𝑡 𝑐𝑘 = 1;
4000 4000
𝑠𝑡 𝑐𝑘 = 1.5;
3000 𝑠𝑡 𝑐𝑘 = 1; 3000
2000 2000
1000 1000
0 0
0.0 0.2 0.4 0.6 0.8 0.0
1.0 0.2 0.4 0.6 0.8 1.0
sx sx
Figure 2-21: Brush model with uniform (left) and parabolic (right) pressure distribution. For different
𝑠𝑡 𝑐𝑘 𝑠 .
72
2.3.3.2 Empirical tyre models

The brush model is a physical model which explains the principles of how the forces are developed.
However, if a model is required which is numerically accurate to a certain tyre, one often uses a fitted
curve instead of Equation [2.13]. To limit the number of parameters to fit, one often uses a
mathematical curve approximation, using trigonometric and exponential formulas. These curve fit
models, it can be applied to any type of data with similar characteristics and can thus be used to
describe other loading behaviour of the tyre such as lateral stiffness and self-aligning torque.
Also empirical tyre models normally use the "simplifying collapse from 3 to 2 independent variables",
meaning that they give curve fits for 2 independent variables: [ ] = 𝑢𝑛 (𝑠 𝑠 ) =
𝑢𝑛 ((𝑅 𝜔 )⁄|𝑅 𝜔| ⁄|𝑅 𝜔|); rather than 3 independent variables: [ ]=
𝑢𝑛 (𝜔 ); This simplification is physically motivated and often used also by the empirical tyre
models. So, the "simplifying collapse from 3 to 2 independent variables" is actually a physically
motivated assumption, which I think that also the empirical models are based on.
2.3.3.2.1 Magic Formula Tyre Model
Probably the most well-known curve fit is called “Magic Formula” and it was proposed in (Bakker,
1987). This approach uses trigonometric functions to curve fit the experimental data. The curve fit has
the general form:
𝑜𝑟 = 𝑦(𝑥) = 𝐷 ∙ sin(𝐶 ∙ arctan(𝐵 ∙ 𝑥 𝐸 ∙ (𝐵 ∙ 𝑥 arctan(𝐵 ∙ 𝑥)))) ;

𝑌(𝑥) = 𝑦(𝑥) + 𝑆𝑉 ; [2.22]
𝑆𝑙 𝑝 = 𝑥 = 𝑋 + 𝑆𝐻 ;
Where B is a stiffness parameter, C is a shape parameter, D is a peak value parameter, and E is a

curvature parameter describing the curve. The variable x is the tyre slip value. The parameters 𝑆𝑉 and
𝑆𝐻 simply shifts the curve so that it passes through the origin, which might not be the case for
measurement data, since there can be errors in tyre radius and correction for rolling resistance. The
relationship between these parameters and the tyre slip/friction relation is shown in Figure 2-22.
Figure 2-22: Magic Formula Tyre Parameters, (Pacejka, 2005)

2.3.3.2.2 TMsimple and TMeasy Tyre Models
TMsimple and TMeasy are two other curve fit models. TMsimple is shown in Eq [2.23] (without
influence of vertical force) and both are shown in Figure 2-23. TMsimple is a simplified variant of
TMeasy. For example, in TMsimple, it is not possible to set the maximum force value to a specific slip.
Also, torques on the tyre are not described within TMsimple, just the tyre forces. TMeasy is described
in Ref (Hirschberg, et al., 2007).
73
(𝑠) = 𝑚 −|𝑠|⁄𝐴
sin(𝐵 (1 ) si n(𝑠)) ;
𝑤ℎ 𝑟 𝐵 = 𝜋 arcsin( ∞ ⁄ 𝑚 ) ; 𝑛 𝐴 = 𝑚 𝐵⁄arctan(𝐶) ; [2.23]
𝑤 𝑡ℎ 𝑚 ∞ 𝑛 𝐶 𝑠 𝑛 𝐸𝑞 [ 6]
2.3.3.2.3 More advanced models
There are numerous of more advanced tyre models, such as Swift and FTire. They mix physical and
curve fit parameters. FTire is almost a finite element model and demands very many parameters.
2.3.3.2.4 Very simple Tyre Models
There are many more models with different degree of curve fitting to experimental data. However, one
can often have use for very simple curve fits, such as:
• Linearized: = 𝐶 ∙ 𝑠
• Linearized and saturated: = 𝑠 𝑛(𝑠 ) ∙ 𝑛(𝐶 ∙ |𝑠 |; ∙ )
• Stiff: 𝑠 = ; (as if linear with 𝐶 ∞)
• Stiff and saturated: 𝑠𝑡 𝑘 𝑠 = ; 𝑙𝑠 = ; Discrete phases handled as: 𝑤ℎ 𝑛 𝑠 <
𝑡ℎ 𝑛 𝑠𝑡 𝑘 ← 𝑡𝑟𝑢 𝑙𝑠 𝑤ℎ 𝑛 > 𝑡ℎ 𝑛 𝑠𝑡 𝑘 ← 𝑙𝑠 ; (Approximately described.)
Tyre model TMeasy
𝐶
1
∞
Figure 2-23: Left: TMsimple (Lex, 2015). Right: TM-Easy Tyre Model, (Hirschberg, et al., 2007).
2.3.3.3 Tyre with both rolling resistance and slip

So far, the models in Section 2.3.32.4 has not included rolling resistance. As we regard rolling
resistance as a torque, not a force, it does not affect the (𝑠 ) curve. But the rolling resistance does
move the 𝑇(𝑠 ) curve vertical. The curve moves upwards if the wheel is rolling forward and
downwards if rolling rearwards, see Figure 2-24. The (𝑠 ) curve is normally the suitable view for
vehicle level studies, while the 𝑇(𝑠 ) curve is sometimes needed for involving a model of the
propulsion and brake systems. As mentioned in Section 2.3.1, one can select 𝑅 in slip definition so that
the = for 𝑠 = , as shown in Figure 2-24. This compendium claims this is the most appropriate if
one does not know 𝑅. As also mentioned in Section 2.3.1, one can alternatively select 𝑅 in slip
definition so that the 𝑇 = for 𝑠 = . This compendium claims that it is less appropriate and that
using 𝑇 = (𝑠 ) ∙ 𝑅 + si n(𝜔) ∙ 𝑅 ∙ is better.
74
+ +
+ 𝑇 𝑅 (if 𝜔 > )
+
𝑇 𝑅 (if 𝜔 < )
𝑅 𝜔
𝑠 =
𝑅 𝜔
Figure 2-24: Longitudinal tyre force ( ) and normalized wheel torque (𝑇 𝑅 ) including both rolling
resistance and slip. Slip defined so that curve (𝑠 ) passes through diagram origin, which means that
𝑇(𝑠 )⁄𝑅 does not.
2.3.3.4 Transients
Both the physical and empirical tyre models discussed above are based on the assumption of steady
state condition in the contact patch, meaning steady state deformation pattern and steady state sliding
speed distribution. Transients between different steady state conditions take some time and this
process is referred to as relaxation. Often in Vehicle Dynamics, the relaxation is such a quick process
that it can be assumed to take place instantaneously, i.e. the (𝑠 ) curve can be used as an algebraic
relation. But sometimes it is relevant to model the nature of the transient more carefully. Transients
are triggered by variations; variations in slip ( and/or 𝜔), vertical force ( ) and surface conditions
(e.g. expressed in varying ). Transients can also appear for constant conditions (constant 𝜔 )
if the tyre and its suspension constitute a system that comes into stick-slip oscillations.
Note that also start from stand-still and deceleration to stand-still are examples of strong transients in
slip, even if vehicle acceleration/deceleration is small. This is mathematically manifested by that slip
goes towards infinity when wheel rotational speed approaches zero. Modelling transients often solve
both normal transients in slip, vertical force and surface conditions as well as driving situations
involving/close to stand-still.
The physical phenomena to be modelled is elasticity; in tyre sidewalls and/or in contact patch.
2.3.3.4.1 Transients due to Elasticity of Sidewalls
Figure 2-25 shows a physical model which can model how the force change is delayed during quick
changes in slip. The model can also handle vehicle stand-still. This type of model is often called a rigid
ring tyre model, because the belt is modelled as a rigid ring. The ring is here massless, but can assumed
to have mass and rotational inertia. The longitudinal and vertical support are here rigid, but they can
be modelled as compliant. If no significant inertia in wheel hub, the driveshaft compliance will be
series coupled with the rotational compliance, 𝑠𝑤 . The torque 𝑇 𝑚 is the sum of torque from
propulsion system and brake system. Damper elements can be added beside the compliances.
The mathematical model becomes as follows, where 𝑆𝑆 denotes a steady state tyre model, e.g. from
Section 2.3.3.1 or 2.3.3.2:
̇ = 𝑠𝑤 ∙ (𝑅 𝜔 𝑚 𝑅 𝜔 𝑒 𝑡 );
𝑅 𝜔 𝑒𝑡 [2.24]
= 𝑆𝑆 (𝑠 … ); 𝑤ℎ 𝑟 𝑠 = ;
|𝑅 𝜔 𝑒 𝑡 |
75
Wheel rotational
elasticity of sidewalls Massless
ring/belt
Rim &
Rim &
wheel shaft Same dynamic wheel shaft
behaviour between
rim and road
Rigid longitudinal
and vertical supports
Road 𝑠𝑤
𝑅 𝜔 𝑒𝑡 𝑅 𝜔 𝑚
Brush model or similar:

= 𝑠 ;
𝑅𝜔 − Wheel rotational elasticity
𝑠 = 𝑅 𝜔 𝑙𝑡 𝑥 ; of sidewalls transformed to
𝑙𝑡
Note: This can be seen as a non-linear damper: tangential elasticity with
= 𝑅 𝜔 𝑒𝑡 ; stiffness 𝑠𝑤 𝑁 .
Figure 2-25: Tyre model including the rotational elasticity of tyre sidewalls.
Damping in parallel with the elasticity is often motivated also:
= 𝑠 + 𝑑;
̇ 𝑠 = 𝑠𝑤 ∙ (𝑅 𝜔 𝑚 𝑅 𝜔 𝑒 𝑡 );
𝑑 = 𝑠𝑤 ∙ (𝑅 𝜔 𝑚 𝑅 𝜔 𝑒 𝑡 );
[2.25]
𝑅 𝜔 𝑒𝑡
= 𝑆𝑆 (𝑠 … ); 𝑤ℎ 𝑟 𝑠 = ;
|𝑅 𝜔 𝑒 𝑡 |
If used in a system where 𝜔 𝑚 and are input variables to tyre, the force will become a state
variable. It is then not a problem that slip is undefined for 𝜔 = , because the explicit form of
equations will become as follows. Note that we simplify by only considering the case when 𝜔 𝑒 𝑡 >
. And the model validity is limited to … such that uniquely defines 𝑠 .
𝑠 ← 𝑆𝑆 ( … );
𝜔 𝑒𝑡 ← ; (𝑜𝑛𝑙𝑦 𝑙 𝑜𝑟 𝜔 𝑒𝑡 > )
𝑅 (1 𝑠 si n( )) [2.26]
̇ ← 𝑠𝑤 ∙ (𝑅 𝜔 𝑚 𝑅 𝜔 𝑒 𝑡 );
𝑤ℎ 𝑟 𝑆𝑆 𝑠 𝑡ℎ 𝑛 𝑟𝑠 𝑜 𝑢𝑛 𝑡 𝑜𝑛 𝑆𝑆 𝑠𝑢 ℎ 𝑡ℎ 𝑡 𝑠 = 𝑆𝑆 ( );
With 𝑆𝑆 =𝐶 𝑠 ;, and 𝜔 𝑚 > ; and > ;, we can simplify Eq [2.26]to Eq [2.27]:
𝑠𝑤 ∙
̇ = 𝑠𝑤 ∙𝑅 𝜔 𝑚 ; ⇒ {𝑠 = 𝑠𝑠 𝑙𝑙} ⇒
1 𝐶
𝐶 [2.27]
⇒ ̇ ≈ 𝑠𝑤 ∙𝑅 𝜔 𝑚 𝑠𝑤 ∙ ;
Eq [2.26] means that we read the function 𝑆𝑆 from force to slip 𝑠 . When knowing slip, we can
calculate the rotational speed 𝜔 𝑒 𝑡 . Then, the state derivative ̇ can be calculated, so that the state
can be updated in each time step.
76
With any reasonable tyre function 𝑆𝑆 , there is a maximum magnitude of force, | | = , above
which there is no slip 𝑠 that gives that . In most problems, one never ends up there in the
simulations, since when approaching | | = , the velocity 𝜔 𝑒 𝑡 changes quickly in the direction
that makes | | stays < . But, due to numerical errors in solutions, or if decreases stepwise,
one might end up there anyway for short time periods. In that case, it often gives physically acceptable
solutions on vehicle level, to simply saturate 𝑠 so that | | is saturated at a certain level, e.g. .95
. For the brush model with uniform pressure distribution, Eq [2.13], the inverted function becomes
as follows, including such saturation:
∙
= ; 𝑜𝑟 | | ≤ ;
𝐶
∙ si n( )
) = =4∙𝐶 ∙ ; 𝑜𝑟 | | < .95 ∙ ;
𝑠 = ( | |
1 ∙ [2.28]
∙ si n( )
= ∙ ; 𝑙𝑠
{ 4∙𝐶 1 .95
𝐺∙𝑊∙
𝑤ℎ 𝑟 𝐶 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ]
∙𝐻
2.3.3.4.2 Transients due to Relaxation in Contact patch modelled as filter

The elasticity in the side walls, modelled in Section 2.3.3.4.1, explains delay; e.g. when changing 𝜔 𝑚
stepwise, the force will not follow 𝑆𝑆 directly, but with a delay. Another phenomenon that causes
delay is that the bristles in the contact patch needs to adopt to a new deformation pattern and sliding
speed distribution. Following the brush model, the physically correct way would be to formulate the
equations as a partial differential equation (PDE), with derivatives with respect to both time and
position along the contact patch, 𝜉. A “quasi-physical” way to model this is to apply a 1st order time
delay of force:
̇ = (1 𝜏) ∙ ( 𝑆𝑆 (𝑠 …) );
where 𝑆𝑆 (𝑠 … ) is the force according to a steady state model, e.g.Eq
𝐿𝑟 𝐿 [2.29]
[2.13] and the time delay, 𝜏 = = |𝑅∙𝜔𝑟 | and is the relaxation length,
𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡 𝑟𝑖𝑚
which often is given as a fraction (≈ 5. .5 %) of tyre circumference.
Alternatively, one can also express the delay as a 1st order time delay of the slip, as follows:
= 𝑆𝑆 (𝑠 𝑑𝑒 𝑒𝑑 … );
𝑠̇ 𝑑𝑒 𝑒𝑑 = (1 𝜏) ∙ (𝑠 𝑠 𝑑𝑒 𝑒𝑑 );
[2.30]
where 𝑆𝑆 (𝑠 𝑑𝑒 𝑒𝑑 … ) is the force according to a steady state model, e.g.
Eq [2.13] and 𝜏 is as defined in Eq [2.29].
To delay force (Eq [2.29]) is more physical than delaying the slip (Eq [2.30]), since the nature of the
delay is motivated by elasticity and delay in an elasticity is modelled with force (or deformation) as
state variable. The delay in slip rather proposes that relative speed as state variable, which is not
physical in this context. However, the delayed force has the non-physical effect that sometimes can
become > in cases when wheel is off-loaded quickly, i.e. when ̇ is a large negative value. So, an
extension to = max( 𝑆𝑆 (𝑠 𝑑𝑒 𝑒𝑑 …) ) ; can be motivated.
It would make sense from physical point of view, if the relaxation length was approximately same
magnitude as the contact length, or possibly the length of the sticking zone. However, commonly given
size of relaxation length is 5. .5 % of tyre circumference, which is normally several times larger than
77
the contact length. This can be because one measures delay due to sidewall elasticity also, but then
interpreted as a relaxation length.
With 𝑆𝑆 =𝐶 𝑠 ;, and 𝜔 𝑚 > ; and > ;, we can simplify Eq [2.29] to Eq [2.31]:
𝐶 𝐶 𝑅 𝜔 𝑚
̇ = ∙𝑅 𝜔 𝑚 ∙ ∙ ; ⇒ {𝑠 = 𝑠𝑠 𝑙𝑙} ⇒
𝐶
[2.31]
𝐶 𝐶
⇒ ̇ ≈ ∙𝑅 𝜔 𝑚 ∙ ;
2.3.3.4.3 Relation between the two Transients models

The models in Section 2.3.3.4.1 and 2.3.3.4.2 describe two different phenomena which both are
present in the real world. Since it is difficult to distinguish between the two delay-creating parameters,
𝑠𝑤 and 𝐶 ⁄ , one often models only one phenomena. But then, one adjusts the numerical value of the
used parameter so that the model captures about the same delay as a real-world test. If 𝑠𝑤 = 𝐶 ⁄ ;,
the two models coincide approximately.
This compendium does not give any recommendation of which of the two models is best.
Note that, with unsaturated in Eq [2.29], one has to limit the integration of ̇ to | | ≤ ∙ .
2.4 Lateral force due to slip

After a vehicle starts moving, controlling the direction of travel becomes a high priority for the
operator. For wheeled vehicles, the primary mode to control travel direction is to change the
orientation of the tyre, i.e. to apply a steering angle. Tyres generate a lateral force when they are
oriented at an angle different to the direction of the vehicle motion. The tyre typically deforms as in
Figure 2-26.
a) yz plane, view from rear b) xy plane, view from above

rest of vehicle, left side wheel rest of vehicle,
left side wheel
view from rear
⃗𝑤 view from above
𝑭𝒔
𝑴𝒔 𝑭𝒔
𝑴𝒔
𝑭𝒔 𝑭𝒔
𝑴𝒔 𝑴𝒔
𝑭𝒚
𝑭𝒚
Figure 2-26: Deformation and forces of a Cornering Tyre, (Clark, 1971)

It is essential to distinguish between the steering angle and (lateral or side) slip angle of the tyre.
Lower right part of Figure 1-18 shows this difference. The steering angle, or d, is the angle between
vehicle longitudinal direction and tyre longitudinal direction. The tyre side slip angle, 𝛼 is the angle
between tyre longitudinal direction and the tyre translational velocity (=wheel hub velocity).
The relation between the lateral force of a tyre and the tyre side slip angle is typically as shown in
Figure 2-30. The behaviour of the curve is similar to that exhibited for longitudinal forces Figure 2-22
78
and Figure 2-23. It becomes even more similar if lateral slip angle is replaced by lateral wheel slip, 𝑠 ,
which is ≈ 𝛼 for pure and small lateral slip.
𝑠 =
𝑠 = ={ } = 𝑡 𝑛(𝛼) ; [2.32]
|𝑅 ∙ 𝜔| . . 𝑅 𝜔=
2.4.1 Tyre brush model for lateral slip

With corresponding simplification as in Section 2.3, we now use the brush model to also explain the
lateral properties.
2.4.1.1 Model with independent bristles

Figure 2-27 shows the model for pure lateral slip and should be compared to Figure 2-17. The
difference is that the model for lateral slip has the deformation of the bristles perpendicular to
drawing in the upper left view in the figure.
Each bristle in Figure 2-27 is thought of as a part of the tread in series with a part of the sidewall.
Further, the bristles are considered as independent of each other, which is debatable. However, it is
enough for a quantitative explanation of the brush model for lateral slip. The derivation of model
equations becomes similar as for the longitudinal model:
∆ ∙ 𝑡(𝜉) ∙ 𝑡(𝜉) 𝑡(𝜉) = 𝜉 ⁄ 𝑇 𝑛𝑠 𝑜 𝑡 ;
𝜏 =𝐺 ∙𝛾 =𝐺 ∙ =𝐺 ∙ ={ }=
𝐻 𝐻 𝑇 𝑛𝑠 𝑜 𝑡 ≈ |𝑅 ∙ 𝜔|;
∙𝜉 𝐺 𝐺 𝑊 𝐶
=𝐺 ∙ = ∙ ∙ 𝜉 = {𝐶 = ;𝑠 = ;} = ∙ 𝑠 ∙ 𝜉;
𝐻 |𝑅 ∙ 𝜔| 𝐻 |𝑅 ∙ 𝜔| 𝐻 |𝑅 ∙ 𝜔| 𝑊
Note that subscript 𝑦 has been introduced where we need to differ towards the longitudinal brush
model. Correspondingly, subscript 𝑥 should be used in longitudinal model. As for longitudinal model,
we have to express the force differently for when friction limit is not reached within the contact and
when it is reach at inlet and outlet.
∙ ∙
= 𝐶 ∙𝑠 ; |𝑠 | ≤ ⇔| |≤
∙𝐶
=
∙ 1
= si n(𝑠 ) ∙ ∙ ∙ (1 ∙ ); 𝑙𝑠
4 ∙ 𝐶 |𝑠 |
{ [2.33]
𝐺 ∙𝑊∙
𝑤ℎ 𝑟 𝐶 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ] 𝑛 𝑠 = ;
∙𝐻 |𝑅 ∙ 𝜔|
(Only valid for pure lateral slip, i.e. pure longitudinal rolling: = 𝑅 𝜔;)
79
rolling direction
L=length of contact patch view from rear,
at one certain x
tyre
𝑅 𝜔=
outlet inlet rim
x sidewall
x=L x=xc x=0 and tread
H=height
in series
road (relative
=𝑅 𝜔 to wheel hub)
Contact patch, W=width
view from above:
Velocities of bristle
line where upper ends
ends in stick zone:
of bristles are attached
line where lower ends of 𝑅 𝜔
bristles appear when sliding
Drawn for 𝑅 𝜔 =
x and >
pressure
x
stick
slip
“mode”
x
shear stress
(lateral, i.e. perpendicular
to top drawing)
Figure 2-27: The brush model’s physical model for lateral slip. The bristles are to be thought of as the tread
in series with sidewall lateral elasticity.
2.4.1.2 Model with dependent bristles, String model

Opposed to the assumption in Section 2.4.1.1, the lateral deformations of the bristles are dependent on
each other, especially since the tread is mounted on the belt and the belt is rather like a string. So, we
assume a certain deformation of the sidewall, expressed in 𝜀 𝑛 and 𝜀𝑜𝑢𝑡 for the belt=”string” in Figure
2-28. This gives a slightly different model compared to Section 2.4.1.1. Models with such belt
deformation are called “tyre string models”. The shape of the string is dependent on the sidewall
elasticity, e.g. tyre profile height, but also of the side force itself. So, the model is intrinsically implicit;
the string shape influences the side force and the side force influences the string shape.
The derivation of model equations becomes similar as for the longitudinal model. Here is an
intermediate result, an expression for the shear stress 𝜏 :
∆ ∙ 𝑡(𝜉) (𝑅 𝜔 tan(𝜀 𝑛 ) ) ∙ 𝑡(𝜉)
𝜏 =𝐺 𝑡 ∙𝛾 =𝐺 𝑡 ∙ =𝐺 𝑡 ∙ =
𝐻 𝑡 𝐻 𝑡
𝑡(𝜉) = 𝜉 ⁄ 𝑇 𝑛𝑠 𝑜 𝑡 ; (𝑅 𝜔 tan(𝜀 𝑛 ) )∙𝜉
={ }=𝐺 𝑡 ∙ =
𝑇 𝑛𝑠 𝑜 𝑡 ≈ |𝑅 ∙ 𝜔|; 𝐻 𝑡 |𝑅 ∙ 𝜔|
𝐺 𝑡 𝐺 𝑡 𝑊
= ∙ tan(𝜀 𝑛 ) si n(𝜔) ∙ 𝜉 = {𝐶 𝑡 = }=
𝐻 𝑡 |𝑅
⏟ ∙ 𝜔| 𝐻 𝑡
( 𝑠𝑦 )
𝐶 𝑡
= ∙ (𝑠
⏟ tan(𝜀 𝑛 ) si n(𝜔)) ∙ 𝜉;
𝑊
𝑠𝑦𝜀
80
Comparing to Section 2.4.1.1, we can note that subscript 𝑡𝑟 is added to underline that 𝐶 𝑡 and
𝐺 𝑡 𝐻 𝑡 now means only the tread, not including the sidewall. Sidewall elasticity is instead handled
with 𝜀 𝑛 . The variable 𝑠 𝜀 is to be seen as an auxiliary mathematical variable introduced only to make
the expressions more manageable; it will be eliminated later.
We have to express the force differently for when friction limit is not reached within the contact and
when it is reach at inlet and outlet.
∙
= 𝐶 𝑡 ∙𝑠 𝜀; | |≤
= ∙ 1
= si n(𝑠 𝜀) ∙ ∙ ∙ (1 ∙ ); 𝑙𝑠
4∙𝐶 𝑡 |𝑠 𝜀|
{
𝐺 𝑡 ∙𝑊∙
𝑤ℎ 𝑟 𝐶 𝑡 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ] 𝑛 𝑠 = 𝑛 𝑠 𝜀 =𝑠 tan(𝜀 𝑛 ) si n(𝜔) ;
∙𝐻 𝑡 |𝑅 ∙ 𝜔|
view from rear,

rolling direction at one certain x
rim
L=length of contact patch sidewalls
𝑅 𝜔 tan 𝜀 𝑛
tyre (relative to wheel hub)
outlet 𝑅 𝜔= belt
inlet
x 𝑅 𝜔 tan 𝜀 𝑛
x=L x=xc x=0 tread
H=height
road =𝑅 𝜔
Contact patch, view W=width

from above: rim
𝜀𝑜𝑢𝑡 𝜀𝑛
Belt deformation is
drawn simplified as
piecewise linear.
Velocities of bristle ends in stick zone:
𝑅 𝜔 tan 𝜀𝑜𝑢𝑡 𝑅 𝜔 𝑅 𝜔 tan 𝜀 𝑛
𝑅 𝜔
Drawn for 𝑅 𝜔 = and >
𝜉
pressure
𝜉 stick
slip
“mode”
𝜉 shear
stress
(lateral)
Figure 2-28: Tyre sidewall deformation and tread deformation with belt=”string” in between. The drawn
bristles are here assumed to represent only tread parts, while the sidewall is treated as an elastic structure
between rim and belt=“string”.
Now, we have to remember that this model is still implicit because 𝜀 𝑛 depends on 𝜏(𝜉). Introducing
simplest possible (linear) constitutive equation for this dependency as in Eq [2.34]:
81
∫𝜏 𝜉= = 𝐶𝑠𝑤 𝜀 𝑛 tan(𝜀 𝑛 ) ; [2.34]
Eq [2.34] makes it possible to derive the following explicit model:
𝐶𝑠𝑤 𝜀 𝑛 𝐶 𝑡 ∙
= ∙𝑠 ; | |≤
𝐶𝑠𝑤 𝜀 𝑛 + 𝐶 𝑡
= 𝐶𝑠𝑤 𝜀 |𝑠 | 𝐶𝑠𝑤 𝜀 𝑠 𝐶𝑠𝑤 𝜀

𝑛 𝑛 𝑛
1+ √(1 ) + 𝐶
∙ ∙ 𝑡
[2.35]
{= 𝑠 𝑛(𝑠 ) ∙ ∙ ; 𝑙𝑠
𝐺 𝑡 ∙𝑊∙
𝑤ℎ 𝑟 𝑠 = ; 𝑛 𝐶 𝑡 = ; 𝑛𝑠 𝑜𝑛: [ 𝑜𝑟 ]
|𝑅 ∙ 𝜔| ∙𝐻 𝑡
(Only valid for pure lateral slip, i.e. pure longitudinal rolling: = 𝑅 𝜔;)
It should be noted that the constitutive relation in Eq [2.34] only states how the string is angled. It is
still physically consistent to separately add, outside the tyre model, a constitutive equation for the
lateral translational deformation of the sidewall; something as = 𝑠𝑤 𝑡 𝑛𝑠 𝑡 𝑜𝑛 𝑡 𝑒 ;, where
𝑡 𝑒 would be the lateral deformation between wheel rim and contact patch. Such sidewall elasticity
would appear in series with the lateral tyre slip force model, in a similar way as the torsional sidewall
elasticity in Section Transients due to Elasticity of Sidewalls2.3.3.4.1 appeared in series with the
longitudinal tyre slip force model.
2.4.1.3 Comparison of the models

Eq [2.35] (dependant bristles) is to be compared to Eq [2.33] (independent bristles). Using 𝐶 𝑠𝑤 =
𝐶𝑠𝑤 𝜀 𝑛 𝐶 𝑡 ⁄(𝐶𝑠𝑤 𝜀 𝑛 + 𝐶 𝑡 ) ; the models are identical up to and have the same asymptot for
𝑠 ∞. See Figure 2-29.
Figure 2-29: Comparison of model with independent bristles and dependent bristles (String model).
The influence of vertical load , discussed already in Section 2.3.3.1.2, is better explained with
dependent bristles. Assumes that only the tread stiffness 𝐶 𝑡 (and not 𝐶𝑠𝑤 𝜀 𝑛 ) varies with contact
length, and that this variation is proportional, 𝐶 𝑡 ( ) ;, as we found in Section 2.3.3.1.2. This
indicates a degressive characteristics of 𝐶 ( ), which is also observed in measurements.
Notes, for both models:
82
• The tyre lateral slip is the sliding speed in lateral direction, divided by the same “transport
speed” as for longitudinal slip, i.e. the longitudinal transport speed.
• There is a minus sign appearing, because of sign conventions.
• The tyre lateral slip definition in Eq [2.33] and Eq [2.35] is to be compared to lateral wheel
slip 𝑡 𝑛(𝛼) = ⁄ , mentioned in context of Figure 2-26:
o “Lateral wheel slip”= 𝑠 𝑤ℎ𝑒𝑒 = 𝑤ℎ𝑒𝑒 ⁄ 𝑤ℎ𝑒𝑒 = arctan(𝛼), which is defined by
how wheel hub moves over ground and independent of wheel rotational speed.
o “Lateral tyre slip” = 𝑠 𝑡 𝑒 = ⁄|𝑅 ∙ 𝜔|, valid as slip in the constitutive relation =
(𝑠 𝑡 𝑒 ).
o If no longitudinal tyre slip, i.e. if 𝑅 𝜔 = , we have 𝑠 𝑤ℎ𝑒𝑒 = 𝑠 𝑡 𝑒 . Then 𝑠 𝑤ℎ𝑒𝑒 can
be used in the constitutive relation.
• For a linearization, the most correct way is that lateral force is 𝑠 , not α. Often one
find α as starting point in the literature, but this compendium uses 𝑠 .
(In (Pacejka, 2005), pp184-185, there is a note that slip is more appropriate than angle for
large side slip, which supports 𝑠 .)
(A difference is how one linearize a vehicle model in and 𝜔 . A non-steered axle modelled
with α, needs to be approximated with “𝛼 = 𝑟 𝑡 𝑛( ⁄ )≈ ⁄ ” to make the
vehicle model linear, see derivation of Eq [4.55]. However, with the model 𝑠 ; makes the
vehicle model linear without further approximations. For a steered axle it is less obvious, but it
does not help the linearization to use α.)
The model with dependent bristles is probably more correct. Anyway, we will use the other most in
this compendium, since it is much easier to combine with the longitudinal model (Eq [2.13]) to model
combined (longitudinal and lateral) slip.
2.4.2 Lateral tyre (slip) stiffness

In summary for many models (and tests!) the following is a good approximation for small lateral slip
(and negligible longitudinal slip and constant normal load):
= 𝐶 ∙𝑠 ; or = 𝐶𝛼 ∙ 𝛼; [2.36]
For the brush model, or any other model which describes (𝑠 … ) or (α … ), one can
define the “Lateral tyre slip stiffness” or “Tyre Cornering Stiffness”, 𝐶 or 𝐶𝛼 , which have the unit N/1
or N/rad. It is the derivative of force with respect to slip or slip angle. Reference (ISO8855) defines the
cornering stiffness as 𝐶𝛼 for slip angle 𝛼 = ;:
𝜕 𝜕
𝐶𝛼 = ( )| = 𝐶 = ( )| [2.37]
𝜕𝛼 𝛼=0 𝜕𝑠
𝑠𝑦 =0
When using only small slip, it does not matter if the cornering stiffness is defined as the slope in an
versus 𝛼 diagram or versus 𝑠 = 𝑡 𝑛(𝛼) diagram. Therefore, the notation for cornering stiffness
varies between 𝐶𝛼 and 𝐶 . Cornering stiffness has the unit 𝑁 which can be interpreted as 𝑁 = 𝑁⁄1 =
𝑁 (( 𝑠)⁄( 𝑠)) or 𝑁 = 𝑁⁄𝑟 .
The longitudinal tyre slip stiffness, 𝐶 , is normally larger than the lateral tyre slip stiffness, 𝐶 , which
can be explained with that the tyre is less stiff in lateral direction. Since it is the same rubber one could
argue that both 𝐺 and 𝐻 should be the same, but both due to longitudinal grooves in the tread and due
to lateral deformable sidewall, it is motivated to introduce different subscripts: (𝐺 ⁄𝐻 ) > (𝐺 ⁄𝐻 ) .
One could elaborate with different friction coefficients and , but in this compendium it is claimed
that friction is well modelled as isotropic. More about this in Section 2.5
83
2.4.3 Empirical tyre models

The cornering tyre forces initially exhibit a linear relation with the slip angle. A non-linear region is
then exhibited up to a maximum value. In Figure 2-30, the maximum slip angle is only 16 degrees (or
𝑠 = tan(16 ) = . 9) and one can expect that the tyre forces will drop as the slip angle
approaches 90 degrees. The general form of the lateral force versus slip angle curve is also suitable for
the Magic Formula, TM-Easy, or similar curve fitting approach when sufficient test data is available.
2.4.4 Influence of vertical load

As discussed for longitudinal slip and the rolling resistance behaviour of tyres, the vertical load on the
tyre affects the force generation. The general behaviour of the tyre’s cornering performance as the
vertical load changes is presented in Figure 2-30. These figures show that the cornering stiffness is
influenced by vertical load. One term is used to describe the dependence of cornering stiffness on
vertical load: cornering coefficient, and is defined as:
𝐶𝛼 𝐶
𝐶𝐶𝛼 = = 𝐶𝐶 = ; [2.38]
In total, we can approximate = ( 𝑠 ) ≈ 𝐶 𝑠 = 𝐶𝐶 𝑠 ; and then we have linearized in

two ways: w.r.t. and 𝑠 . A slightly better linearization is ≈ 𝐶𝐶0 ( 0 ) 𝑠 ;, where subscript
zero is the tyre’s nominal load. The influence of vertical tyre force on the generated lateral forces is an
important aspect of vehicle performance in cornering and will be further discussed in 4.
Adhesion Limit
Adhesion Limit
|Cornering Force| [kN]
|Slip Angle| [deg]
0 4 8 12 16
Figure 2-30: Left: Influence of tyre design. Right: Influence of inflation pressure, (Gillespie, 1992).
Contact lenght [mm]
Lateral, 𝐶
Figure 2-31: Measurements for varying vertical force. Left: Slip stiffnesses. Right: Contact length.
84
Figure 2-32: Example of cornering stiffness versus vertical load for a truck tyre 295/80R22.5.
Figure 2-33: Cornering stiffness versus vertical load for some passenger car tyres. From flat track tests.
2.4.5 Relaxation in contact patch

As for longitudinal, there is a delay in how fast the steady state conditions can be reached in contact
patch, which is sometimes important to consider. A similar model, as for relaxation in longitudinal
direction Section 2.3.3.4.2, is to add a first order delay of the force:
̇ = 𝐴 ∙ ( (𝑠 …) );
where (𝑠 … ) is the force according to a steady state model and 𝐴 =
𝑇𝑟𝑎𝑛𝑠𝑝𝑜𝑟𝑡 |𝑅∙𝜔| | 𝑥|
[2.39]
𝐿𝑟
= 𝐿𝑟
or 𝐴 = 𝐿𝑟
and is the relaxation length, which is a
fraction (≈ 5. .5 %) of tyre circumference.
85
2.4.6 Other lateral effects

The deformations of the tyre during cornering are quite complex when compared to the case of pure
longitudinal motion, see Figure 2-26. Hence, there are more effects than simply a lateral force. Some of
these will be discussed in the following.
2.4.6.1 Overturning moment

The contact patch is deflected laterally from the centre of the carcass. This creates an overturning
moment 𝑀 due to the offset position of the normal force. The lateral force also contributes to the
overturning moment.
2.4.6.2 Tyre aligning moment

(This section has strong connection with Section 4.2.3.2 Steering system forces.)
In the lowest diagram in Figure 2-27, one can see that shear stress is concentrated to the outlet side of
the contact patch for small slip angles. So, the equivalent lateral force acts behind the centre of wheel
rotation for small slip angles. As seen in Figure 2-26 b) it acts at a position 𝑡 behind the wheel’s y axis.
The distance 𝑡 is called the pneumatic trail, see also Figure 4-5, and the resulting yawing moment
around a vertical axis through centre of the contact patch will be = 𝑡 , which is often called the
aligning moment, 𝑀 . The aligning moment is named after that is normally acts to steer the wheel to
zero side slip.
If the tyre is on a steered axle, the aligning moment influence on steering wheel torque is important.
When finding that influence, the moment around the steering axis intersection with ground is the
important moment = (𝑡 + ), which can be called the steering moment. The distance is the
mechanical trail, which is built into the suspension linkage design. One typically designs the
suspension so that > , which makes the whole steering moment act in the same direction as the
aliging moment. If driver takes his hands from steering wheel in a curve, and/or if steering power
assistance is lost, the steering will tend to steer in the direction of body motion above the steered axle,
which is normally relatively smooth and safe.
steering axis
shear stress, 𝜏
intersection
with ground 𝑡
contact
patch
𝑡 𝑡 +
tyre
steering moment peaks at lower
slip than lateral tyre force.
lateral tyre force = Fy (for a tyre with peak)
total trail = caster trail + pneumatic trail = + 𝑡
steering moment = +𝑡
side slip angle
Figure 2-34: General Response of Steering torque to Side slip angle. Tyre aligning moment = ∙ 𝑡 is one
part of the steering moment= ∙ (𝑡 + ).
86
Figure 2-34 shows the combined response of lateral force and slip angle. It is interesting to note that
the steering torque reaches a peak before the maximum lateral force capacity of the tyre is reached. It
can be used by drivers to find, via steering wheel torque, a suitable steering angle which gives a large
lateral force but still does not pass the peak in lateral force. The reason why pneumatic trail can
become slightly negative is because pressure centre is in front of wheel centre, see Figure 2-9.
2.4.6.2.1 Aligning Torque Model
A model for (yawing) aligning moment around a vertical axle through centre of contact point, 𝑀 , will
now be derived. Any model for lateral shear stress can be used, but we will here only use the uniform
pressure distribution and independent bristles in Section 2.4.1.1. A corresponding expression as Eq
[2.33] is derived, but for 𝑀 instead of .
𝑀 = 𝑊 ∙ ∫ 𝜏 ∙ (𝜉 ) 𝜉=⋯
0
= Cy ∙𝑠 ; 𝑜𝑟 < ⇔𝑠 <
6 𝐶
= [2.40]
∙
= (1 ); 𝑙𝑠
{ 8∙𝐶 ∙𝑠 3∙𝐶 ∙𝑠
𝐺 𝑊
𝑤ℎ 𝑟 𝐶 = ;
𝐻
The curve of Eq [2.40] is plotted in Figure 2-35.

The lateral force and the aligning torque can be used to calculate the steering forces. If also the
steering assistance is known, the steering wheel torque can be calculated. It can be noted that the
model does not include the moment from steering rotation itself, i.e. the torque counteracting ̇ .
Small and increasing ,

and constant non-zero
lever
Large , but lever

approaches zero
Figure 2-35: Aligning moment (𝑀 ) around contact patch center for uniform pressure distribution.
2.4.6.3 Camber force

Camber force (also called Camber thrust) is the lateral force caused by the cambering of a wheel. One
explanation model is shown in Figure 2-36. It is that Camber force is generated when a point on the
outer surface of a leaning (cambered) and rotating tyre, that would normally follow a path that is
elliptical when projected onto the ground, is forced to follow a straight path due to friction with the
ground. This deviation towards the direction of the lean causes a deformation in the tyre tread and
carcass that is transmitted to the vehicle as a force in the direction of the lean. Another explanation
87
model is that the tyre “climbs” sideways as the inlet edge is directed, because there is more stick and
less slip in the inlet edge as compared to the outlet edge; see brush model.
Camber thrust is approximately linearly proportional to camber angle for small angles: Camber
thrust= 𝑐 = 𝐶𝛾 ∙ 𝛾. The camber stiffness, 𝐶𝛾 , is typically 5-10 % of the cornering stiffness and
opposite sign. The tyre lateral forces due to lateral slip and due to camber are superimposable for
small lateral slip and small camber angle. The approximate Eq [2.36], can then be developed to:
= 𝑆 𝑑𝑒𝑆 + 𝐶 𝑚 𝑒 = 𝐶 ∙ 𝑠 + 𝐶𝛾 ∙ 𝛾; or = 𝐶𝛼 ∙ 𝛼 + 𝐶𝛾 ∙ 𝛾 [2.41]
There is also a rotation perpendicular to the ground due to the camber. The total rotation vector of the
wheel is directed along the wheel rotation axis. The component of this vector that is perpendicular to
ground creates friction moment which steers the wheel towards the side it tilts; on a cambering
vehicle like a bicycle, the wheel is steered “into the fall”, which counteracts falling.
2.5 Combined Longitudinal and Lateral Slip

The operation of motor vehicles often involves a combination of turning, braking, and steering actions.
Hence, the performance of the tyre under combined (tyre) slip is important.
If the tyre has isotropic adhesion properties in the lateral and longitudinal direction, one can assume
that the maximum force magnitude is determined by the maximum resultant friction force, ∙ .
𝐹 𝐹
= + ≤( ∙ ) ⇒ ( 𝑥) + ( 𝑦) ≤ [2.42]
𝐹 𝐹
Equation [2.42] can be plotted as a circle, called the “Friction Circle”. Since the lateral and longitudinal
properties are not really isotropic (due to carcass deflection, tread patterns, camber, etc) the shape
may be better described as a “Friction Ellipse” or simply “Friction limit”.
𝛾
View from rear
or front:
A point on tyre surface travels basically along the

edge of an ellipse, when the tyre has a camber
View from above: angle…
𝜔
inlet edge …but the ellipse is flattened in the contact patch
(if friction makes the rubber stick and wheel hub
𝑐 𝑐 is moving in pure longitudinal direction).
This is the path the point would take if not

outlet edge flattened. Consequently, to flatten the ellipse, it
requires a force on the tyre, as arrow “camber
thrust” shows.
Figure 2-36: Camber thrust
88
When not cornering, the tyre forces are 1

described by a position between -1 maximum turning left
(braking) and +1 (acceleration) along the without propulsion and
Y-axis. Note that the scales of both axes are braking
0.5 turning left while
normalized to the maximum value for braking
friction. partial propulsion
Fy/(mu*Fz)
0
The “actuation” of the wheel means the max propulsion
propulsion, braking and steering (and
sometimes suspension control) of the -0.5
wheel. An ideal actuation allows all
conditions within the boundaries of the
friction circle to be achieved anytime -1
during a vehicle manoeuvre. An example of -1 -0.5 0 0.5 1
limitation in actuation is a wheel on a non- Fx/(mu*Fz)
steered rear axle. They cannot access any of
the lateral parts of the circle; unless the Figure 2-37: Friction Circle with some examples of
utilization. View from above, forces on tyre.
vehicle slides laterally.
At the boundary of the friction circle, tyres become more sensitive to changes in slip. It is therefore
extra important to model the direction of the force in relation to shear deformation and relative slip
motion in the tyre contact patch. Here, isotropic shear and friction properties are assumed:
𝑙 𝑡 𝑟 𝑙 𝑠𝑙 𝑛 𝑠𝑝
= = ; [2.43]
𝑙𝑜𝑛 𝑡𝑢 𝑛 𝑙 𝑠𝑙 𝑛 𝑠𝑝 𝑅∙𝜔
Significant testing and modelling work have been undertaken to quantify the relationship between
tyre slip (𝑠 and 𝑠 , or 𝑠 and 𝛼) and tyre forces ( and ). As an example, Figure 2-38 presents one
method to present the maximum lateral force for different longitudinal slips. Note that each line
represents a constant slip angle and that the maximum lateral force occurs for a slip angle between 10
and 18 degrees. Also the maximum value for longitudinal slip does not occur at zero, but for a slight
braking condition (around 0.05), which can be motivated using slip definition in Eq [2.1], but not from
Eq [2.2].
Figure 2-38: Combined Longitudinal and Lateral Slip
89
Figure 2-39: Left: Effects of combined longitudinal and lateral utilization of friction. Right: Tyre force
diagrams with iso-curves for longitudinal and lateral slip. From (Andreasson, 2007). 𝜅 is longitudinal slip
and is ⁄ .
2.5.1 Isotropic combined slip, any 𝑭 = 𝒇(𝒔);

A simple but physically consistent combined slip model can be expressed using “effective slip, s”, as
shown in Equation [2.44].
+𝑠 𝑅 ∙𝜔
[ ]=[ 𝑠 ] =[ 𝑤 ] ;
𝑠
√(𝑅𝑤 ∙ 𝜔 ) +( )
𝑠 = 𝑠 = 𝑠 +𝑠 ;
𝑅𝑤 ∙ 𝜔
𝑠 = ; 𝑛 𝑠 = ; 𝑛 = |𝑅𝑤 ∙ 𝜔|;
𝑇 𝑛𝑠 𝑜 𝑡 𝑇 𝑛𝑠 𝑜 𝑡
𝑇 𝑛𝑠 𝑜 𝑡 [2.44]
= ∙∙ (𝑠 );
. . (𝑠 ) = {𝐸𝑞 [ 13] 𝑜𝑟 [ 19]]};
𝐶0
𝑜𝑟 (𝑠 ) = min ( ∙ 𝑠 ; 1) ; 𝐶0 ≈ {𝑡𝑦𝑝 𝑙𝑙𝑦} ≈ 5. .15 [non-dimensional];
{
The model becomes singular if √(𝑅𝑤 ∙ 𝜔 ) +( ) ;. This can be avoided for some such
𝐶0
conditions using that (𝑠) ≈ 𝑙 𝑛 𝑟 for small 𝑠 . Inserting (𝑠) = min ( ∙ 𝑠 ; 1) ; in Eq [2.44] gives:
𝑅𝑤 ∙ 𝜔 𝐶0 +𝑠
=[ ] =[ 𝑠 ] 𝐶0 ;
|𝑅𝑤 ∙ 𝜔|
… √(𝑅𝑤 ∙ 𝜔 ) +( ) < 𝐶0 |𝑅𝑤 ∙ 𝜔|

[ ]= [2.45]
𝑅𝑤 ∙ 𝜔 ∙
=[ ] ; … 𝑙𝑠
√(𝑅𝑤 ∙ 𝜔 ) +( )
{
2.5.2 Isotropic brush model for combined slip

The brush model can be applied also for combined slip. If doing that, with uniform pressure
distribution and same (𝐺 𝐻) = (𝐺 𝐻) ⇔ 𝐶 = 𝐶 = 𝐶 (isotropic stiffness), the expression for
in Eq [2.44] will be as follows.
90
∙ ∙
=𝐶 ∙ 𝑠; … s≤ ⇔| |≤
∙𝐶 𝐺∙𝑊∙
= 𝑤ℎ 𝑟 𝐶 =
∙ 1 ∙𝐻
= ∙ ∙ (1 ∙ ); … 𝑙𝑠
{ 4∙𝐶 s
Or, equivalent: [2.46]
1 ∙
= 𝐶0 ∙ ∙ ∙ 𝑠; … s≤ ⇔| |≤
∙ 𝐶0 𝐶
= 𝑤ℎ 𝑟 𝐶0 = ;
1 1
= ∙ ∙ (1 ∙ ); … 𝑙𝑠
{ 4 ∙ 𝐶0 s
2.5.3 Anisotropic brush model, combined slip

The longitudinal model (Eq [2.13]) and the lateral model with independent bristles (Section 2.4.1.1)
will now be used with the anisotropy 𝐶 𝐶 ;, motivated by that sidewall contributes more to lateral
bristle elasticity than to longitudinal bristle elasticity. With significantly different tread grooves in
longitudinal and lateral direction, one could motivate also difference in friction coefficient , but here
we assume same in both directions. The derivation of model equations becomes similar as for the
pure longitudinal and lateral models. For the stick zone:
𝐶 𝐶
𝜏 = ∙ 𝑠 ∙ 𝜉; 𝑛 𝜏 = ∙ 𝑠 ∙ 𝜉;
𝑊 𝑊
If no slip in contact:
𝐿 𝐿 𝐿
𝐶 𝐶 𝐶
= 𝑊 ∙ ∫𝜏 ∙ 𝜉 = 𝑊 ∙ ∫ ∙𝑠 ∙𝜉∙ 𝜉 = ∙ 𝑠 ∙ ∫𝜉 ∙ 𝜉 = ∙𝑠 ∙ =𝐶 ∙𝑠 ;
𝑊
0 0 0
𝐿 𝐿 𝐿
𝐶 𝐶 𝐶
= 𝑊 ∙ ∫𝜏 ∙ 𝜉 = 𝑊 ∙ ∫ ∙𝑠 ∙𝜉∙ 𝜉 = ∙𝑠 ∙∫𝜉 ∙ 𝜉 = ∙𝑠 ∙ =𝐶 ∙𝑠 ;
𝑊
0 0 0
Now, if there is a break-away point (𝜉𝑐 ) where slip starts, we can find it from 𝜏(𝜉𝑐 ) = 𝑝; ⇒ 𝜏 +
𝜏 = 𝑝 ;. Introducing an auxiliary parameter, 𝑘 = 𝐶 ⁄𝐶 ;, and an auxiliary variable, 𝑠𝑘 =
√(𝑘 ∙ 𝑠 ) + 𝑠 ;, gives the following break-away point:
𝜏(𝜉𝑐 ) = 𝑝; ⇒ 𝜏 + 𝜏 = 𝑝 ; ⇒ ⋯ ⇒ 𝜉𝑐 = ;
𝐶 𝑠𝑘
The forces, when < 𝜉𝑐 < and 𝜔 > becomes:
𝐿 𝜉𝑐 𝐿
𝐶
= 𝑊 ∙ ∫𝜏 ∙ 𝜉 = 𝑊 ∙ ∫ ∙𝑠 ∙𝜉 ∙ 𝜉 +𝑊 ∙ ∫𝜏 𝑠 𝜉=
𝑊
0 0 𝜉𝑐
𝜉𝑐 𝐿
𝐶 𝑠
=𝑊∙∫ ∙𝑠 ∙𝜉∙ 𝜉+𝑊∙ ∫ ∙𝑝 𝜉=⋯=
𝑊
0 𝜉𝑐 √𝑠 +𝑠
𝑠𝑥 𝐹 𝑘𝑠
= ∙ 𝑠
(1 𝐶𝑦 𝑠𝑘
(1 𝑠𝑘
)) ; and
91
𝐿 𝜉𝑐 𝐿
𝐶 𝑠
= 𝑊 ∙ ∫𝜏 ∙ 𝜉 =… = 𝑊 ∙ ∫ ∙𝑠 ∙𝜉∙ 𝜉+𝑊∙ ∫ ∙ 𝜉=⋯=
𝑊 𝑊 𝑠
0 0 𝜉𝑐
𝑠 1 s
= ∙ (1 (1 )) ;
𝑠 𝐶 𝑠𝑘 𝑠𝑘
rolling direction view from rear,

showing rubber element
L=length of contact patch on one certain x
tyre 𝑅 𝜔 W=width
outlet inlet
x
H=height x=L x=xc x=0
road (relative
(relative
to wheel hub)
to wheel hub)
Contact patch, Velocities of bristle ends
view from above: in stick zone:
𝑅 𝜔
force on bristle
𝐻 𝑠 ⁄𝐺
𝑠 𝐺 𝑠 ⁄𝐻 Drawn for 𝑅 𝜔 >
𝑠 𝑠 𝐺 𝑠 ⁄𝐻 and >
𝐻 𝑠 ⁄𝐺 𝑠
x
pressure
x stick
slip
“mode”
x 𝜏 𝑠 longitudinal
Continuous 𝜏 across break-away
𝜏 𝑐 shear stress point (𝜏 𝑠 + 𝜏 𝑠 =
x 𝜏 𝑠 lateral
= 𝜏 𝑐 + 𝜏 𝑐 ), but redistribution
𝜏 𝑐 towards sliding direction in slip zone.
shear stress
Figure 2-40: Physical model for deriving brush model for combined slip.
Arranging for all combinations of ( 𝜔), we find Eq [2.47]. Figure 2-41 shows results from the
model. It can be observed that is independent of 𝑠 and is independent of 𝑠 for ≤ ∙ ⁄𝑠 .
This is a reasonable consequence of that no sliding occurs so that forces are purely defined by the
elasticity, not the friction. At 𝑠 ≈ . 1. . . 4, we see that increases with utilization of | | at some
areas. This is a redistribution of force from longitudinal to lateral, due to that the tyre stiffer in
longitudinal than lateral.
92
𝑘 𝑠
+ ∙
𝑠𝑘
= 𝑠 ; 𝜔 <
∙
[ 𝑠𝑘 ]
+𝐶 𝑠 ∙ ∙
=[ 𝐶 𝑠 ]; 𝑙𝑠 𝑠𝑘 ≤ ⇔ =√ + ≤
[ ]= ∙𝐶
𝑠 1 𝑘 𝑠
+ ∙ (1 (1 ))
𝑠 𝐶 𝑠𝑘 𝑠𝑘 [2.47]
= ; 𝑙𝑠
𝑠 1 s
∙ (1 (1 ))
{ [ 𝑠 𝐶 𝑠𝑘 𝑠𝑘 ]
𝑅 𝜔
𝑤ℎ 𝑟 𝑠 = ; 𝑠 = ; 𝑠 = √𝑠 +𝑠 ;
|𝑅 ∙ 𝜔| |𝑅 ∙ 𝜔|
𝐶
𝑛 𝑘= ; 𝑠𝑘 = √(𝑘 𝑠 ) + 𝑠 ;
𝐶
2.5.4 An approximate combined slip model

A combined model for cases when one knows without involving 𝑠 is shown in Equation [2.48]. It
can be if we know the wheel torque, e.g. by prescribed propulsion or braking. The equation is not
completely physically motivated, but works relatively well. One can consider it as “a mathematical
scaling, inspired by the friction circle”.
≈ √1 ( ) ∙ 0; 𝑤ℎ 𝑟 0 = |𝑠 [2.48]
∙ 𝑥 =0
A complement case would be if the lateral force is known without involving 𝑠 . This is not usual, since
steering actuation often in steering angle, but one could imagine situations where steering torque is
actuated. Then, the corresponding version of the model becomes:
𝐹𝑦
≈ √1 ( ) ∙ 0; 𝑤ℎ 𝑟 0 = |𝑠𝑦 =0 [2.49]
∙𝐹
If this concept of scaling is applied on the slip stiffnesses instead, one gets:
√( 𝐹 )2 −𝐹𝑥2 √( 𝐹 )2 −𝐹𝑦2
𝐶 ( )=( 𝐹
)∙𝐶 ; 𝑛 𝐶 ( )=( 𝐹
) ∙𝐶 ; [2.50]
93
Figure 2-41:Tyre force diagrams with iso-curves for longitudinal and lateral slip. Results from Eq [2.47].
2.5.5 Transients and Relaxation

The models above for combined slip assume steady state conditions. For separate longitudinal and
lateral slip, there is a delay in how fast the steady state conditions can be reached, which is sometimes
important to consider. A similar model, as for relaxation in longitudinal direction Section 2.3.3.4.2, is to
add a first order delay of the force:
[ ̇ ; ̇ ] = [𝐴 ∙ ( (𝑠 𝑠 …) ); 𝐴 ∙ ( (𝑠 𝑠 …) )];
where [ ; ] are the forces according to steady state models and [𝐴 ; 𝐴 ] are [2.51]
relaxation lengths, as defined in Eq [2.29] and Eq [2.39].
Most reasoning in Section 2.3.3.4 is applicable also for combined slip relaxation.
94
2.6 Vertical Properties of Tyres

The most important vertical property of a tyre is probably the stiffness. It mainly influences the
vertical dynamics, see 5. For normal operation, the vertical force of the tyre can be assumed to vary
linearly with vertical deflection. If comparing a tyre with different pressures, the stiffness increases
approximately linear with pressure. See Figure 2-42.
60000
60
0
0
Figure 2-42: Vertical properties of a truck tyre.
Figure 2-43: Different tyre models which will filter road irregularities differently. Picture from Peter
Zegelaar, Ford Aachen.
(Rhyne, 2005)
2.7 Tyre wear

There are many other aspects of tyres, for instance the wear. Wear models are often based around the
Archard’s (or Reye’s) wear hypothesis: worn material is proportional to work done by friction, i.e.
friction force times sliding distance. Wear rate (worn material per time) is therefore friction force
times sliding speed. Different approaches to apply this to tyres and expanding to temperature
dependency etc. is found for instance in Reference (Grosch, et al., 1961). A generalization of 𝑊 𝑟𝑅 𝑡
[in mass/s or mm tread depth/s], for one certain tyre at certain temperature, becomes as follows:
95
𝑊 𝑟𝑅 𝑡 𝑟 𝑡 𝑜𝑛 𝑜𝑟 ∙ 𝑆𝑙 𝑛 𝑆𝑝 ⇒𝑊 𝑟𝑅 𝑡 = 𝑘𝑤 ∙ ∙∆ ≈
≈𝑘∙ ∙ √(𝑅 𝜔 ) + ≈ 𝑘𝑤 ∙ (𝐶 ∙ 𝑠) ∙ √(𝑅 𝜔 ) + ≈
≈ 𝑘𝑤 ∙ (𝐶 ∙ 𝑠) ∙ (𝑠 ∙ 𝑇 𝑛𝑠 𝑜 𝑡 ) ⇒𝑊 𝑟𝑅 𝑡 ≈ 𝑘𝑤 ∙𝐶∙𝑠 ∙ 𝑥; [2.53]
where =√ + ; 𝑠 = √𝑠 +𝑠 ; 𝑇 𝑛𝑠 𝑜 𝑡 defined as in Eq [2.8] and 𝑘𝑤

is a constant for a certain tyre with a certain temperature, rolling on a certain road
surface, which characterises the wear averaged over the contact patch.
2.8 Vehicle Aerodynamics

The flow of air around the vehicle body produces different external forces and moments on the vehicle.
Considerable research and engineering has been carried out, and the fluid mechanics will not be
covered in this course. However practical first order models for aerodynamics forces have been
established and are presented here.
2.8.1 Longitudinal wind velocity

The most relevant force of interest in this course is the resistance force to forward motion,
aerodynamic drag , which is proportional to the square of the longitudinal component of the
wind speed relative to the vehicle, 𝑒 . For aerodynamic loads resisting forward motion of the
vehicle, the Equation [2.55] can be used. The parameters 𝑑 𝜌 𝑛 𝐴 𝑜𝑛𝑡 represent the drag
coefficient, the air density and a reference area, respectively. The reference area is the area of the
vehicle projected on a vehicle transversal plane.
= ∙ 𝑑 ∙𝜌∙𝐴 𝑜𝑛𝑡 ∙ 𝑒 ; [2.55]
Typical values of drag coefficients (cd) for cars can be found from sources such as: (Robert Bosch
GmbH, 2004), (Barnard, 2010) and (Hucho, 1998). These coefficients are derived from coast down
tests, wind tunnel tests or CFD (Computational Fluid Dynamics) calculations. For light vehicles (cars,
SUVs, pick-up trucks) the air resistance can often be neglected for city traffic up to 30-50 km/h, but not
at highway speeds.
Since a car structure moving through the air is not unlike an aircraft wing, there are also aerodynamic
lift forces. Considering the forward motion only, a lift force and pitch moment will develop. This affects
the normal loads on front and rear axle, and consequently the tyre to road grip. Hence, it affects the
lateral stability.
= ∙ ∙𝜌∙𝐴 𝑜𝑛𝑡 ∙ 𝑒 ;
[2.56]
𝑀 = ∙ 𝑚∙𝜌∙𝐴 𝑜𝑛𝑡 ∙ 𝑐∙ 𝑒 ;
The coefficient represents the lift characteristics of the vehicle. The lift force is here assumed
to act through the centre of gravity. The 𝑚 (pitch moment coefficient) and 𝑐 (characteristic length,
usually wheelbase) must be reported together.
With identical effect on the (rigid) vehicle body one can replace the lift force in centre of gravity, ,
and the pitch moment, 𝑀 , with lift forces at two different longitudinal positions, typically over each
axle:
96
= ∙ ∙𝜌∙𝐴 𝑜𝑛𝑡 ∙ 𝑒 ;
[2.57]
= ∙ ∙𝜌∙𝐴 𝑜𝑛𝑡 ∙ 𝑒 ;
The coefficients and are lift coefficient over front and rear axle, respectively. Relations between
Equation [2.56] and Equation [2.57] are: = + ; and 𝑚 = ( ) ∙ 𝑐 ;.
A third way of representing identical effect on the (rigid) vehicle body is to move the longitudinal force
to a height ℎ = ℎ 𝑀 ⁄ above ground and remove the pitch moment, 𝑀 . The 3 ways
are visualised in Figure 2-44.
ℎ
Figure 2-44: Three force equivalent ways to describe longitudinal wind aerodynamic force play..
2.8.2 Lateral wind velocity

When the wind comes from the side, there can be direct influences on the vehicle lateral dynamics.
Especially sensitive are large but light vehicles (such as buses or vehicles with unloaded trailers). The
problem can be emphasized by sudden winds (e.g. on bridges or exiting a forested area). Besides direct
effects on the vehicle lateral motion, side-winds can also disturb the driver through disturbances in the
steering wheel feel.
Similar expressions to the longitudinal loads are derived for lateral forces and from side-winds.
= ∙ 𝑠 ∙𝜌∙𝐴∙ 𝑒 ;
[2.58]
𝑀 = 𝑚∙𝜌∙𝐴∙ 𝑐∙ 𝑒 ;
The speed 𝑒 is the lateral component of the vehicle velocity relative to the wind. Note that A and Lc
may now have other interpretations and values than in Equations [2.55]-[2.57].
2.9 Driving and Vehicle Operation

The driver drives and experience the vehicle in the short time scale through pedals, steering wheel and
seat. But also the choice of load and choice of route is important on a longer time scale.
2.9.1 Mission, Road and Traffic

This section could be much longer, but is kept very short. It should remind reader about that the load
of goods/passengers can very well vary as well as how it is positioned and secured on the vehicle. Also,
the actual choice of road influences through topography and curvature. Choice of road and time of day
influences traffic, i.e. other road users. See more in Section 3.3.1.
2.9.2 Driver
To study how different vehicle designs work in a vehicle operation a driver model is needed. In its
easiest form, a driver model can be steering angle 𝑠𝑡𝑤 ;. Another extreme interpretation of what
can be called a driver model is an implicit/inverse statement, like “driver will push accelerator pedal
97
so that speed [ ⁄𝑠] during the manouvre”, which leads to that accelerator pedal position
becomes an output, as opposed to input, to the vehicle model. Beyond those very simple driver models,
there is often need for a driver model which react on vehicle states in relation to an environment or
traffic. In this section, driver models are primarily thought of as models of the driver of the subject
vehicle, but when modelling surrounding traffic carefully, each object vehicle can also use a driver
model.
The driver interacts with the vehicle mainly through steering wheel, accelerator pedal and brake
pedal. In addition to these, there are clutch pedal, gear stick/gear selector, and various buttons, etc.,
see Figure 2-45, but we focus here on the first 3 mentioned.
• Accelerator Pedal Position, Other:
• Brake Pedal Force, • Clutch Pedal and Gear
• Steering Wheel Angle Stick or Gear Selector
• Parking Brake
• Direction indicators
• HMI (buttons, lamps,
• Accelerator Pedal Force, text, sounds, …)
• Brake Pedal Position, • …
• Steering Wheel Torque,
• Seat motion,
• View of environment relative to vehicle
Figure 2-45: Interface of driver, and some commonly assumed causality.

Driver’s control of vehicle dynamics, or vehicle motion including position, can be discussed in
longitudinal (mainly pedals) and lateral (mainly steering wheel).
Driver reacts on several stimuli, such as motion (mainly through seat), sounds, and optical. Among
motion, it is primarily the accelerations (and their time derivative, jerk) that is sensed by the driver,
but also rotational velocity in yaw can be sensed by human. Among the optical there is looming
(optical expansion of an object in the driver’s field of view [ 𝑠]) is often used as a cause for how
driver uses the pedals. The optical flow (the pattern of apparent motion of objects, surfaces, and edges
in the driver’s field of view) is often used as a cause for highest comfortable speed and yaw rate.
Driver models are here discussed as models of the “human driver” for use in vehicle verification
simulations. However, driver models can also be understood as models of “virtual driver”, and then
they are actually implemented as algorithms in the vehicle product, e.g. as prediction algorithms or
automated driving controllers. In the first context, it is often important to vary the driver model (at
least its parameters, maybe even its equations) for robustness, as mentioned in Figure 1-2. In the
latter context, the driver model of the subject vehicle is rather varied for optimization/satisficing, see
Figure 1-2 again.
An important aspect of driver modelling is how the user (driver or occupant) experiences the vehicle.
This is often referred to as subjective evaluation, but for some cases one can establish methods to
objectively calculate a measure of how good or bad the experience is. The measure can sometimes be a
physical quantity but often it has to be a rating or grading without unit. Examples are “driveability
[rating 0-10, high is good]”, “steering effort [deg/s, low is good]” and “ride comfort [m/s2, low is
good]”.
2.9.2.1 Different driver modelling

As in all modelling, it is important to distinguish between different model use when modelling or
selecting a model.
Driver modelling for “verification of vehicle functions” and for implementation in “driving
automation functions” have large similarities in that they should react on the vehicle’s environment,
but one can also see differences. Driver models for verification of vehicle functions should be as
human-like as possible. They should also judge feedback to driver, such as assessing steering effort.
Driver models for use in driving automation functions should also be human-like to facilitate
cooperation between human and automated driving, such as hand-over/take-back or simultaneous
98
control. However, there are also reasons to not mimic all aspects from a human driver, such as the
human’s inability to watch in several directions simultaneously.
A categorization of modelling concept is whether the model uses relations that reflect the biological
processes human’s perception, cognition and neuro-muscular or relations reflecting the vehicle
motion. The first concept (exemplified in 2.9.2.3.3) would rather use angle to obstacle as opposed to
distance to obstacle, since humans rather see angles than distances. The latter concept (exemplified in
2.9.2.2.1) assumes that driver has adapted to the specific vehicle and (subconsciously) operates the
vehicle in a good way; a kind of inverse model thinking. Overall, both concepts can reflect
approximately the same driving, but they are differently parameterized; typically, in biological
parameters and vehicle parameters, respectively.
2.9.2.2 Driver Longitudinal or Pedals Operation

Each pedal has both pedal position and pedal force as communication channels to the driver.
In broad terms, the accelerator pedal position is interpreted by a control system as a certain requested
engine torque, which means a certain wheel shaft torque on the propelled axle/axles. In some
advanced solutions, one could think of using accelerator pedal force as a feedback channel from
vehicle to driver.
In broad terms, the brake pedal force gives a certain friction brake torque application on each axle. In
some advanced solutions, one could think of changing brake pedal position versus force characteristics
as a feedback channel from vehicle to driver. But already with traditional (hydraulic) brake systems
for passenger cars, the brake pedal gives a feedback to driver in that ABS interventions are felt.
For electric and hybrid vehicles, the brake pedal can typically influence the propulsion system as well,
in that at low levels of pedal braking, the electric propulsion motors give negative wheel shaft torque.
This is for regenerating energy. Regenerative braking can even be engaged at low levels of accelerator
pedal position. The propulsion and brake systems are more described in 3. LONGITUDINAL DYNAMIC.
A model of how the driver operates the pedals is called a longitudinal driver model. But different
manoeuvres call for different driver models. Examples of manoeuvres/situations are:
• Full acceleration or deceleration, including driver delays
• Keep constant speed
• Keep distance to lead vehicle
• When a cut-in happens in front of object vehicle: Expand to comfortable/safe gap
• Approach slower object vehicle from behind: Reduce to comfortable/safe gap
• Longitudinal speed control during overtaking
• Longitudinal speed control to stand-still
2.9.2.2.1 Example of longitudinal driver model
An example for approaching a slower vehicle from behind is defined through simple kinematic
relations and some assumptions, see Figure 2-46 which indicates how to calculate the required
(constant) acceleration to reach a certain safe gap. The safe gap can be defined by “gap needed to
exactly avoid collision if braking as object vehicle but with a certain reaction time”. In traffic schools,
one is often taught that safe time gap 𝑆𝑇𝐺 is 𝑆𝑇𝐺 = 3 [𝑠], as a rule of thumb. In Eq [2.61], it is shown
how such a model is derived, subscripts: s=subject, o=object, 0=present, p=predicted. The requested
acceleration 𝑒𝑞 just have to be converted to pedal force or position.
99
Prediction, of object and subject:

𝑠𝑜 = 𝑠𝑜0 + 𝑜 𝑡 ; (object vehicle assumed to not accelerate)
𝑠 = 𝑠0 + 𝑒𝑞 𝑡 ;
𝑠𝑠 = 𝑠𝑠0 + 𝑠0 𝑡 + 𝑒𝑞 𝑡 ⁄ ;
Requirement on prediction at 𝑡 :
𝑠𝑜 𝑠𝑠 = (𝑠𝑜 𝑠𝑠 )𝑠 𝑒 = 𝑜 𝑆 𝑇 𝐺 𝑝= 𝑜 𝑆𝑇𝐺;
= [2.59]
𝑠 𝑜;
Solve for 𝑒𝑞 and 𝑡 gives the driver model on explicit form:

( 𝑜 𝑠0 )
𝑒𝑞 = ;
(𝑠𝑜0 𝑠𝑠0 𝑆𝑇𝐺 𝑜)
(𝑠𝑠0 𝑠𝑜0 + 𝑆𝑇𝐺 𝑜)
𝑡 = ;
𝑜 𝑠0
Use only if 𝑡 > , else “prediction” happens in passed time.
Examples of applying the driver model in Eq [2.61] is shown in Figure 2-46, where also a general PVA-
time-diagrams (PVA=position, velocity, acceleration) is shown. PVA-time-diagrams are often useful
when studying vehicle operations with other moving objects, see Figure 2-46. Different scenarios can
be tested, and driver models can be derived using simple formulas, such as constant speed and
constant acceleration.
predicted collision (if
𝑙𝑜𝑛 𝑡𝑢 𝑛 𝑙 no speed changes)
predicted collision (if
𝒐𝒔𝒊𝒕𝒊𝒐𝒏 𝑠
certain deceleration
of object vehicle)
known (sensed/estimated) unknown (predicted, several scenarios)
safe speed adaptation

prediction (if object keeps
safe time gap speed and subject decelerates
with constant ̇ )
prediction time 𝑡 for “safe speed adaptation”
TTC=Time To Collision
eTTC=enhanced TTC
𝑝𝑟 𝑠 𝑛𝑡 𝑡
𝑡
𝑙𝑜𝑛 𝑡𝑢 𝑛 𝑙
𝒗𝒆𝒍𝒐𝒄𝒊𝒕𝒚
subject vehicle
object vehicle
𝑡
𝑙𝑜𝑛 𝑡𝑢 𝑛 𝑙
𝒂𝒄𝒄𝒆𝒍𝒆 𝒂𝒕𝒊𝒐𝒏
𝑡
𝑒𝑞
Figure 2-46: Left: PVA-time-diagrams (PVA=position, velocity, acceleration) for studying longitudinal
interaction with an object vehicle ahead. Right: PVA diagram for the driver model in Eq [2.61] using
𝑆 𝑇 𝐺 𝑝 = [𝑠] and some randomly selected 𝑠𝑠0 and 0 .
100
2.9.2.3 Driver Lateral or Steering Wheel Operation

2.9.2.3.1 Causality of steering wheel angle and torque
The driver interacts with steering wheel through two channels, steering wheel angle and steering
wheel torque. The angles of the road wheels are functions of steering wheel angle, almost independent
of forces and torques in the steering system. Further, the steering wheel torque is a function of the
lateral force on the front axle, scaled down with assistance from the power steering or steering servo,
but almost independent of the angles.
A frequently used model is that the driver decides a steering wheel angle and expects a certain
steering wheel torque, 𝑇𝑠𝑤 , feedback. Only in extreme situations, the causality is modelled the other
way, e.g. studying what happens if driver takes hands-off, i.e. steering wheel torque =0.
A way to mix angle and torque prescribing driver steering models is to let the steering wheel inertia,
𝑠𝑡𝑒𝑒 , be common for both driver and vehicle. This is done in (Tagesson, o.a., 2017), from where also
Figure 2-47 is borrowed.
Figure 2-47: Driver steering model, where the steering wheel inertia, 𝑠𝑡𝑒𝑒 , is common for driver and vehicle
model. From (Tagesson, o.a., 2017)
2.9.2.3.2 Steering wheel angle driver models

How driver operates the steering wheel can also be called a lateral driver model. But different
manoeuvres call for different driver models. Examples of manoeuvres/situations are:
• Going into curve from straight and vice versa
• Turning in intersection
• Avoidance manoeuvre
• Stabilizing vehicle when sudden lose yaw stability, e.g. in curve
• Compensating road unevenness which cause lateral disturbance
• Lateral speed control during overtaking
2.9.2.3.3 Example of steering angle driver model
An example of a driver model is the Salvucci and Gray model in Figure 2-48 and Eq [2.61]. It is
motivated from humans biological cognition. The steering wheel angle is denoted 𝑠𝑡𝑤 .
101
nearpoint
𝑊
𝑛
farpoint
𝑊 is some width
𝑊 (such as vehicle,
road or lane width)
Figure 2-48: Salvucci and Gray model for how driver steers.
̇
𝑠𝑡𝑤 =𝑘 ̇ + 𝑘𝑛 ̇ 𝑛 + 𝑘𝐼 𝑛; ⇔
𝑡 [2.61]
⇔ 𝑠𝑡𝑤 (𝑡) = 𝑠𝑡𝑤 (𝑡0 ) + 𝑘 (𝑡) + 𝑘𝑛 𝑛 (𝑡) + 𝑘𝐼 (𝑡 . . 𝑡)
𝑡0 𝑛 0
𝑡;
The 3 𝑘 parameters that have to be tuned to certain driver, vehicle and operation. Additionally, the
distance to aim-points needs to be tuned, such as if there should be extra margins on top of 𝑊 and
whether 𝑊 is vehicle, road or lane width. Further on, all these assumptions can change during a
simulation, e.g. as driving on a windy country road and suddenly a meeting vehicle shows up might
make the far point jump. So, even if the model might look simple, it is not obvious how to tune and use
it. Hence, there are variants of it which has only one aim-point. We will refer to the reduced model,
with only one aim-point:
𝑡
̇
𝑠𝑡𝑤 =𝑘 ̇ + 𝑘𝐼 ; ⇔ 𝑠𝑡𝑤 (𝑡) = 𝑠𝑡𝑤 (𝑡0 ) + 𝑘 (𝑡) + 𝑘𝐼 (𝑡 . . 𝑡) 𝑡; [2.62]
𝑡0 𝑛 0
This kind of model is fed with variables from the vehicle and environment states. A drawback is that
the dynamic vehicle behaviour is built into the 𝑘 parameters, meaning that if vehicle dynamics is
changed, e.g. adding a trailer or reducing road friction, the 𝑘 parameters reasonable have to be re-
tuned. If identifying the 𝑘 parameters as physical parameters, using a dynamic vehicle model, one can
reduce the need for such re-tuning.
Above driver model can be categorized as having “look ahead”. Simpler driver model can be based on
only where the vehicle is at present time, such as present lateral position in lane. More advanced
driver models use a dynamic model for prediction. The vehicle dynamics model is especially needed in
predicting models.
102
LONGITUDINAL DYNAMIC
3 LONGITUDINAL
DYNAMICS
3.1 Introduction
The primary purpose of a vehicle is transportation, which requires longitudinal dynamics. The chapter
is organised with one group of functions in each section as follows:
• 3.2 Steady State Function
• 3.3 Functions over (long)
• 3.4 Functions in (short) events
• 3.5 Control functions
Most of the functions in “3.5 Control functions”, but not all, could be parts of ”3.4 Functions in shorter
events”. However, they are collected in one own section, since they are special in that they partly rely
on software algorithms.
There are a lot of propulsion related functions, originating from the attribute Driving dynamics.
Examples of such, not covered in this compendium are:
• Off-road accessibility: Ability to pass obstacles of different kind, such as uneven ground,
extreme up- and down-hills, mud depth, snow depth, etc.
• Shift quality: Quick and smooth automatic/automated gear shifts
• Shunt & shuffle: Absence from oscillation for quick pedal apply, especially accelerator pedal.

• “Chapter 23. Driveline” in Reference (Ploechl, 2013).
• “Chapter 24. Brake System Dynamics” in Reference (Ploechl, 2013).
• “Chapter 27 Basics of Longitudinal and Lateral Vehicle Dynamics” in Reference (Ploechl, 2013)
• “Chapter 6: Adaptive Cruise Control” in Reference (Rajamani, 2012)
3.2 Steady State Functions

Functions as top speed and grade-ability are relevant without defining a certain time period. For such
functions, it is suitable to observe the vehicle in steady state, i.e. independent of time. Those functions
are therefore called steady state functions, in this compendium.
The main subsystems that influences here are the propulsion system, see Section 3.2.1, and the
(Friction) Brake system, see Section 3.4.3.
3.2.1 Propulsion System

A generalised propulsion system is shown in Figure 3-12, along with a specific example of a
conventional one. There are 1+2=3 control degrees of freedom marked for the generalized one (e.g.
engine power, transmission ratio and storage power) while there is only 1+1=2 for the conventional.
Note that the approach in Figure 3-1 is one-dimensional: we consider neither the differential between
left and right wheel on the driven axle nor distribution between axles. Instead, we sum up the torques
at all wheels and assume same rotational speed.
103
Generalized propulsion system:

Transportation
Driver
task
Acceleration
Request (1)
Propulsion control other actuation systems, e.g. brake

(1) (2)
Prime T,w T,w Wheels Road /
Transmission (incl. road Body
mover (as many as
contact)
𝑠 … Traffic
driven wheels)
P (position along
road, grade, …)
Energy storage (e.g. battery

in electric or hybrid vehicle)
Conventional propulsion system:

Driver
e-throttle control other systems, e.g. brake
(1)
T, w T, w
(on one
driven axle)
Body
Prime mover
(ICE, internal Wheels
combustion
engine)
Transmission (Clutch &
Gearbox & Final gear)
Figure 3-1: Propulsion system
It is often suitable to model propulsion systems “connecting in nodes”, see Section 1.5.2.7.1. Notation
and sign conventions then becomes as in Figure 3-2.
input signal output signal • Nodes only on connections (shafts).
• A node has a positive direction, from left to
Component Component Component
right, in which 𝜔 and are positive.
A Node 1 B Node 2 C • A system model with natural causality has to
have, for each component, equally many
equations as connected shafts, plus number
Component
𝑇 Node N 𝑇 of output signals.
D In example: Components [A, B, C, D] has to
Node 3 𝜔 have [1, 3+1, 1, 1] equations.
Figure 3-2: Propose notation and sign conventions for Propulsion system models.
3.2.1.1 Prime movers

The conversion of stored energy to power occurs in the prime mover, see Figure 3-1. Details of the
conversion processes and transmission of power to the tyres are not covered in this compendium.
Some basic background is still necessary to describe the longitudinal performance of the vehicle. The
main information that is required is a description of the torque applied to the wheels over time and/or
as a function of speed. Sketches of how the maximum torque varies with speed for different prime
movers (internal combustion engine (ICE), electric motor or similar) are shown in Figure 3-3. The
torque speed characteristics vary dramatically between electric and internal combustion engines. Also,
gasoline and diesel engines characteristics vary.
The curve for electric motors in Figure 3-3 shows that the main speciality, compared to ICEs, is that
their operation range is nearly symmetrical for negative speeds and torques. However, the curve
104
should be taken as very approximate, since electric motors can work at higher torque for short periods
of time. The strong time duration dependency makes electric motors very different to ICEs from a
vehicle dynamics point of view. Other properties that makes them special are quick and accurate
response, well known actual torque and that it is much more realistic to divide them into several
smaller motors, which can operate on different wheels/axles.
Consumes power 𝑇 Generates power
𝜔
𝑇 =𝑇 𝜔
𝜔
Generates power Consumes power
ICE Electric Motor Human (on bicycle)

𝑇 𝑇
𝑇 𝜔
[𝑊 𝑘 ]
𝜔 1
𝜔
𝑢𝑟 𝑡 𝑜𝑛 [𝑠]
5
1 1 1 1 1 1
Figure 3-3: Torque Characteristics of 3 Prime Movers

A bicycle uses the human body as prime mover. It reminds of an electric motor in that it has torque
from zero speed. The duration dependency is high, with maximum power ≈ 1 W/kg up to a minute
duration and ≈ W/kg for an hour duration.
3.2.1.1.1 Efficiency and consumption
The efficiency (𝜂) is the output power per input power. The specific consumption can be the inverted
value, but often the input power is measured in fuel rate [ 𝑠] or [𝑙 𝑡𝑟 𝑠], which makes specific
consumption inversely proportional to efficiency and not the exact inverse value.
𝑜𝑢𝑡
𝐸 𝑛 𝑦=𝜂= ;
𝑛
𝑢 𝑙𝑅 𝑡 𝑛 𝑢 𝑙𝑅 𝑡
𝑆𝑝 𝐶𝑜𝑛𝑠𝑢 𝑝𝑡 𝑜𝑛 = 𝑆𝑝 𝐶𝑜𝑛𝑠𝑢 𝑝𝑡 𝑜𝑛 = = ; [3.1]
𝑜𝑢𝑡 𝑜𝑢𝑡 𝑜𝑢𝑡
𝑤ℎ 𝑟 𝑢 𝑙𝑅 𝑡 𝑠 𝑛 [𝑘 𝑠] 𝑜𝑟 [𝑙 𝑡𝑟 𝑠]
𝑛 = 𝑠𝑝 𝑛 𝑟 𝑦 𝑜𝑛𝑡 𝑛𝑡 𝑛 [ ⁄𝑘 ] 𝑜𝑟 [ ⁄𝑙 𝑡𝑟 ]
The efficiency is dependent on the operating point in the speed vs torque diagram, or map, for the
prime mover. An example of a specific consumption map for an ICE is given in Figure 3-4. Maps with
similar function can be found for other types of prime movers, such as the efficiency map for an
electric motor, see Figure 3-5.
Figure 3-4 and Figure 3-5 also show that the efficiencies can be transformed to the traction diagram.
The maps for different gears partly overlap each other, which show that an operating point of the
vehicle can be reached using different gears. The most fuel or energy efficient way to select gear is to
select the gear which gives the lowest specific fuel consumption, or highest efficiency. Such a gear
selection principle is one way of avoiding specifying the gear selection as a function over time in the
driving cycle. For vehicles with automatic transmission, that principle can be programmed into the
105
control algorithms for the transmission. However, the gear selection is often a trade-off with
acceleration reserve, see Section 3.3.4.5, which argues for lower gear. Assuming very tightly stepped
transmission, or CVT, gives that a high efficiency can be kept down to around 0.3 of maximum power.
With gear ratio for lowest

Lowest specific
specific consumption
Efficiency
consumption
for that power
force, F
Propulsion
0 = 𝑚
gear 1 0𝑐 𝑡 1
( 0𝑐 𝑡 typically between 0.1 and 0.5.)
gear 2
0
0 Speed, v
Figure 3-4: Left: Fuel consumption map. Curves with constant specific fuel consumption [ ⁄(𝑘𝑊 ℎ)],
which is 1⁄efficiency. Middle: Specific fuel consumption curves transformed to Traction Diagram, for
different gears. Right: How efficiency with efficiency-optimal gear ratio drops when < 0 𝑐 𝑡 𝑚 .
Propulsion
force, F
gear 1
gear 2
0
0 Speed, v
Figure 3-5: Left: Efficiency map for a typical brushless DC motor, from (Boerboom, 2012). Elliptic curves
show where efficiency is constant. Right: The efficiency curves can also be transformed in Traction Diagram,
for a given gear.
3.2.1.2 Transmissions
In some contexts, “transmission” means the 1-dimensional transmission of rotational mechanical
power from an input shaft to one output shaft. Such is called “Main transmission”, Section 3.2.1.2.1. In
other contexts, “transmission” means the system that distributes the energy to/from an energy buffer
and to/from multiple axles and/or wheels. Such is called “Distribution transmission”, Section 3.2.1.2.2.
3.2.1.2.1 Main transmissions
Main transmission can be either stepped transmissions or continuously variable transmissions.
Among stepped transmissions, there are manual an automatic. Among automatic, there are those with
power transmission interruption during shifting and other with powershifting, see Section 3.2.1.3 .
106
Clutches and torque converters can also be part of models of main transmissions, see Sections 3.2.1.4
and 3.2.1.5. A stepped transmission, can be modelled e.g. as:
𝑇𝑜𝑢𝑡 =𝑟 ∙ 𝑇 𝑛 ∙ 𝜂𝐶𝑜 𝑀𝑒𝑠ℎ𝑒𝑠 Δ𝑇0 ∙ si n(𝜔𝑜𝑢𝑡 ) ;

𝜔𝑛
𝜔𝑜𝑢𝑡 = ; [3.2]
𝑟
𝑤ℎ 𝑟 𝑟 = 𝑟 𝑟 ⋯ 𝑟 ; 𝑟 ;
Δ𝑇0 is the “parasitic” or “load independent” losses, arising from oil, sealings and bearings. Eq [3.2] is
not valid for neutral gear, because then there is no speed equation, but instead two torque equations:
𝑇𝑜𝑢𝑡 = Δ𝑇0 𝑜𝑢𝑡 ∙ si n(𝜔𝑜𝑢𝑡 ) ; and 𝑇 𝑛 =Δ𝑇0 𝑛 ∙ si n(𝜔 𝑛 ) ;.
For any 1-dimensional transmission of rotational mechanical power between two rotating shafts, the
total efficiency, 𝜂𝑡𝑜𝑡 = 𝑜𝑢𝑡 ⁄ 𝑛 = 𝑇𝑜𝑢𝑡 ∙ 𝜔𝑜𝑢𝑡 ⁄(𝑇 𝑛 ∙ 𝜔 𝑛 ) ;, is depending on operating condition. If
assuming a nominal ratio, 𝑟, the total efficiency can be decomposed in 𝜂𝑡𝑜𝑡 = 𝜂 𝑇 ∙ 𝜂𝜔 ;, where 𝜂 𝑇 =
𝑇𝑜𝑢𝑡 ⁄(𝑟 ∙ 𝑇 𝑛 ) ; and 𝜂𝜔 = 𝜔𝑜𝑢𝑡 ⁄(𝜔 𝑛 ⁄𝑟) ;.
3.2.1.2.2 Distribution transmissions
The distribution to energy buffer and multiple axles and/or wheels can basically be done in two ways:
• Distribute in certain fractions of (rotational) speed. A (rotationally) rigid shaft between left
and right wheel is one example of this. We find this in special vehicles, such as go-carts, and in
more common vehicles when a differential lock is engaged. There are 3 shafts in such an axle:
input to the axle and two outputs (to left and right wheel). The equations will be:
𝜔 𝑛 = 𝜔 𝑒 𝑡;
𝜔 𝑛 = 𝜔 ℎ𝑡 ; [3.3]
𝑇𝑛 = 𝑇𝑒 𝑡 + 𝑇 ℎ𝑡 ;
• Distribute in certain fraction of torque. This requires some type of planetary gear arrangement.
A conventional (open) differential gear is one example of this, where the equations will be:
𝜔𝑒 𝑡+𝜔 ℎ𝑡
𝜔𝑛 = ;
𝑇 𝑒 𝑡 = 𝑇 ℎ𝑡 ; [3.4]
𝑇𝑛 = 𝑇𝑒 𝑡 + 𝑇 ℎ𝑡 ;
Generally speaking, the open differential is rather straight-forward to use in most vehicle dynamics
manoeuvres: The speeds are given by vehicle motion (e.g. curve-outer wheel runs faster than curve-
inner wheel, defined by vehicle yaw rate and track width). The torques are defined by the differential,
as half of the propulsion torque at each side.
Also, a locked differential, it is generally more complex to model and understand in a vehicle
manoeuvre. Here, the wheels are forced to have same rotational speed and, in a curve, that involves
the tyre longitudinal slip characteristics. The solution involves more equations with shared variables.
So, open/locked differential is the basic concept choice. But there are additions to those: One can build
in friction clutches which are either operated automatically with mechanical wedges or similar or
operated by control functions. One can also build in electric motors which moves torque from one
wheel to the other. However, the compendium does not intend to go further into these designs.
Part of the distribution transmissions are also shafts. If oscillations is to be studied, these has to be
modelled with energy storing components:
• Rotating inertias or Flywheels ( 𝜔̇ = 𝑇 𝑛 𝑇𝑜𝑢𝑡 ;) and
• Elasticities, compliances or springs: (𝑇̇ = ∙ (𝜔 𝑛 𝜔𝑜𝑢𝑡 )).
107
3.2.1.3 Powershifting transmissions

The most common powershifting transmission is using planetary gears and torque converter, see
Figure 3-14. During 2000-2010, developments in mechatronics has enabled to leave out the torque
converter and the planetary gears resulting in what often meant with powershifting transmissions, see
concept in Figure 3-6. In practice, one can often manage with 2 clutches, and instead select different
paths through gear wheels with synchronisers. An advanced design of powershifting transmission for
a hybrid propulsion system is seen in Figure 3-7. A dummy sequence of shifting is simulated in Figure
3-8.
Interpret pedals.
Request engine torque Decides how to shift. Engages
and when to shift gear. and disengages the clutches
C3
e in C2
𝑒
out Driven wheel
Vehicle body
Prime
mover C1 w x
Figure 3-6: Conceptual design of powershifting transmission with 3 gears.
89
82
85
44
41
37
Electric
motor
m 5 2
𝑚
9
C1
4
3
ICE 8
e 1 C0 C2 6 B1 x
𝑒 𝑅𝑤
7
Figure 3-7: Hybrid propulsion system with powershifting, designed using planetary gears. Upper left: Design.
Upper right: Gear/Clutch schedule. Lower half: Dynamic model.
108
C0.c C0.Tleft wm Tm
400 1500
200
1000
0
500
-200
0 10 20 30 40 50 60 70
0
C1.c C1.Tleft
400 -500
0 10 20 30 40 50 60 70
200
0 we Te
-200
0 10 20 30 40 50 60 70 400
C2.c C2.Tleft 0
200 0 10 20 30 40 50 60 70
0 vx Fx_kN
20
0 10 20 30 40 50 60 70
15
B1.c B1.Tleft
10
400
200 5
0 0
-200 -5
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
Figure 3-8: Simulation of the transmission in Figure 3-7 (with approximate clutch model from Eq [3.6]).
Example sequence of shifts: Simply shift each 10th second, in order as in table in Figure 3-7.
3.2.1.4 Clutches and Brakes

In this content, clutches include brakes as a special cases, characterised by that one of the clutch halves
is not rotating. Clutches with dry friction characteristics are often difficult to model in dynamic
systems since they introduce “discrete dynamics” by sticking and slipping. The general mathematical
model of a controlled clutch is given in Eq[3.5]. Here, the interface variables are 𝑇 𝑒 𝑡 𝑇 ℎ𝑡 𝜔 𝑒 𝑡
𝜔 ℎ𝑡 and . The capability in[𝑁 ] is an input signal. It tells the magnitude of torque when the
clutch is slipping. In the model below, a discrete state variable, 𝑠𝑙 𝑝 = 1 𝑜𝑟 + 1, is introduced to
model the discrete dynamics.
𝑇𝑒 𝑡 = 𝑇 ℎ𝑡 ;
= 𝑇𝑒 𝑡 + ; 𝑠𝑙 𝑝 = 1
= {= 𝜔 𝑒 𝑡 𝜔 ℎ𝑡 ; 𝑠𝑙 𝑝 =
= 𝑇𝑒 𝑡 ; 𝑠𝑙 𝑝 = +1
[3.5]
𝑤ℎ 𝑛 (𝑠𝑙 𝑝 1 𝐴𝑁𝐷 𝑤ℎ 𝑛 𝜔 𝑒 𝑡 > 𝜔 ℎ𝑡 ): ℎ 𝑛 𝑠𝑙 𝑝 𝑡𝑜 ;
𝑤ℎ 𝑛 (𝑠𝑙 𝑝 𝐴𝑁𝐷 𝑤ℎ 𝑛 𝑇 𝑒 𝑡 < ): ℎ 𝑛 𝑠𝑙 𝑝 𝑡𝑜 1;
𝑤ℎ 𝑛 (𝑠𝑙 𝑝 𝐴𝑁𝐷 𝑤ℎ 𝑛 𝑇 𝑒 𝑡 > + ): ℎ 𝑛 𝑠𝑙 𝑝 𝑡𝑜 + 1;
𝑤ℎ 𝑛 (𝑠𝑙 𝑝 +1 𝐴𝑁𝐷 𝑤ℎ 𝑛 𝜔 𝑒 𝑡 < 𝜔 ℎ𝑡 ): ℎ 𝑛 𝑠𝑙 𝑝 𝑡𝑜 ;
The clutch model is to be connected between surrounding components, which have shaft ends as
connectors, see examples in Figure 3-9. The implementation can be almost as Eq [3.5]. But slightly
different implementations is generally needed for different surroundings; energy dissipating or
generating components (dampers or power sources), kinetic energy storing components (flywheels)
or potential energy storing components (elasticities). The challenge is to handle that the set of state
variables can change between the different discrete states. Modelling format and tools have to allow
discrete state events and resetting of states when events occurs. For automatic transmissions, where
there are several clutches involved, the implementation of the ideal model in Eq [3.5] can be very
demanding.
109
Connected to control algorithm, which
calculates “clutch torque capacity”, .
…
Further components,
ending with 𝑇(𝜔) … J J … Further components, ending
with wheels, connected to … …
prime mover model. vehicle body, typically 1D,
2D or 3D motion.
flywheel clutch flywheel elasticity clutch elasticity
…
…
… J … … …
damper clutch damper
(e.g. 𝑇(𝜔) prime (e.g. (𝑠 )
flywheel clutch elasticity mover model) tyre model)
Figure 3-9: Examples of differently modelled surroundings of a clutch.
𝑐𝑢 𝑐̇ 𝑢 //Clutch:
if abs(slip) == 1 then der(T) = sign(slip)*der(cclu);
else w1 = wx; end if;
𝑐𝑢 𝑡
𝑠 if pre(slip) == +1 and w1 < wx then slip = 0;
𝜔 𝜔 elseif pre(slip) == -1 and w1 > wx then slip = 0;
𝑇 𝑇 elseif pre(slip) == 0 and T > +cclu then slip = +1;
elseif pre(slip) == 0 and T < -cclu then slip = -1;
𝑇 𝜔 𝑇 else slip = pre(slip);
𝜔 𝜔 end if;
clutch elasticity when slip == -1 or slip == +1 then

reinit(T, slip*cclu);
end when;
//Spring:
der(T) = cspr*(wx-w2);
//Clutch and Spring:
der(T) = if T > cclu and w1 > w2 then +der(cclu)
elseif T < -cclu and w1 < w2 then -der(cclu) else cspr *(w1-w2);
Figure 3-10: Model of clutch and spring in series, modelled so that speed is input from both sides.
3.2.1.4.1 Implementation of clutches a stiff dampers
A way around these problems is to use an approximate clutch model as in Eq [3.8]. The advantage is
that it can handle any surrounding without leading to changing set of state variables. The drawback is
that it has a trade-off between modelling the intended physics (Eq [3.8]) and the computational
efficiency in simulation; the more ∞, the closer to the intended model we come, but the simulation
time will increase towards infinity. The approximation can be seen as putting a stiff damper in direct
series with an ideal clutch. There are also other approximations, which could be seen as putting a stiff
spring with stiffness in series with an ideal clutch. This will work if clutch is surrounded by inertias:
𝑇𝑒 𝑡 =𝑇 ℎ𝑡 ; 𝑛 𝜔 𝑒 =𝜔𝑒 𝑡 𝜔 ℎ𝑡 ;
𝑇𝑒 𝑡
si n(𝜔 𝑒 )=
1 𝜀 |𝜔 𝑒 | 𝜀 |𝜔 𝑒 | |𝜔 𝑒 |
= min ( ; 1+ ( 1)) ; <1
𝜀 𝜔𝑛𝑜𝑚 1 𝜀 𝜔𝑛𝑜𝑚 𝜔𝑛𝑜𝑚 [3.6]
=
1 𝜀 |𝜔 𝑒 |
= max ( ; ); 𝑙𝑠
𝜀 𝜔 1 𝜀 𝜔𝑛𝑜𝑚
1 + 1 𝜀 ( 𝑛𝑜𝑚 1)
{ |𝜔 𝑒 |
Figure 3-12 shows an example that the ideal and approximate models can give similar results with
respect to torques and speeds. About computational efficiency, the ideal needs around 5 𝑠 time step
with Euler forward integration, while the approximate needs 100 times smaller time step. If better
agreement than in Figure 3-12 is needed, 𝜀 needs to be reduced, which slows down the approximation
even more. Note also that, for energy dissipation, the approximate model of course calculates a higher
energy dissipation, since it assumes the clutch has to slip to transfer torque.
110
∞
(Note: “Semi-invert diagram axes”, i.e. that
value varies linearly below 1, but 1/value
varies linearly above 1.)
ℎ𝑡
ℎ𝑡
𝑇
𝑇
=
=
𝑡
𝑡
𝑇𝑒
𝑇𝑒
∞
∞ ∞
𝜔𝑒 𝜔𝑒
𝜔𝑛𝑜𝑚 𝜔𝑛𝑜𝑚
Figure 3-11: Example of approximation of clutch model. Strategies: 𝜔𝑛𝑜𝑚 = typical slip speed of clutch, 𝜀 ≪ 1.
Dashed curve shows before approximation.
There are clutch models also in the standard Modelica (see Section 1.5.7.5) library, see Figure 3-13.
Note that the library is built such that the “small inertia” is needed, which forces down computational
efficiency during clutch slip.
Brakes, one-way clutches and backlashes often causes similar difficulties and can be modelled
similarly as clutches.
approxClutch.Aped approxClutch.c0
approxClutch.Aped
2 approxClutch.c0
2 𝐴 𝑙 𝑟 𝑡𝑜𝑟 𝑝 𝑙
1
1 𝐶𝑙𝑢𝑡 ℎ 𝑛 𝑛𝑡
0
0 0 2 4 6 8 10 12 14 16
0 2 4 6 8 10 12 14 16
approxClutch.w eng approxClutch.w gin idealClutch.w eng idealClutch.w gin
approxClutch.w
600eng approxClutch.w gin idealClutch.w eng idealClutch.w gin
600
550 𝜔𝑒𝑛 [𝑟 𝑠]
𝑛𝑒
550
500
500
450
450 𝜔 [𝑟 𝑠]
𝑒 𝑜 𝑛
400
400
350
350
300
300
250
250
200
200 0 2 4 6 8 10 12 14 16
0 2 4 6 8 10 12 14 16
Modelled in two ways: approxClutch.vx idealClutch.vx
approxClutch.vx idealClutch.vx
• Approximation, 𝜀 = . 5 20
[ 𝑠]
• Ideal 20
0
0 0 2 4 6 8 10 12 14 16
Figure 3-12: Example

0
with
2
clutch,
4
modelled
6 8
in two
10
ways.
12 14 16
idealGear
engine engine_flywheel small_inertia body
spring damper
tau ratio=7.5 c=1000e3

clutch d=500
J=0.5 J=0.001 J=1500*0.3*0.3
(Inputs to "eninge" and "clutch" governed by equations)

Figure 3-13: Similar example, but built with standard Modelica library “Modelica.Mechanics.Rotational”.
3.2.1.5 Hydrodynamic torque converters

Hydrodynamic torque converters serve almost same purpose as a clutch, but it is much less complex to
model as a member of a dynamic system. Such converters have a pump in input side and turbine on
output side. They can operate with substantial slip, and when slip, there is a torque amplification,
111
which leads to that a vehicle with converter have typically good acceleration performance and
driveability also without the corresponding lowest gear needed on same vehicle without converter.
The following model gives the steady state characteristics of a hydrodynamic torque converter. Steady
state characteristics are often enough, but combined with that hydrodynamic torque converters often
are possible to lock-up with a clutch, mounted in parallel to the impellers.
𝑇𝑜𝑢𝑡 𝑇𝑛 𝜔𝑜𝑢𝑡
= (𝜈); = 𝜆(𝜈); 𝑤ℎ 𝑟 𝑛 𝜆 𝑟 𝑢𝑛 𝑡 𝑜𝑛𝑠 𝑛 𝜈 = ;
𝑇𝑛 𝜔𝑛 𝜔𝑛
𝑇𝑜𝑢𝑡 𝜔𝑜𝑢𝑡
But if locked-up: 𝑇𝑖𝑛
= 1; 𝑛 𝜔𝑖𝑛
= 1; [3.7]
The moment capacity, 𝜆 is often given on a dimensionless form

𝜆 = 𝑇 𝑛 ⁄(𝜔 𝑛 𝑙𝐷 𝑛𝑠 𝑡𝑦 𝑢𝑡 𝑟𝑅 𝑢𝑠 5 ).
𝑇𝑜𝑢𝑡 𝑇𝑛
= 𝜆=
𝑇𝑛 𝜔𝑛
stepped 1
e in out gearbox
𝑒
(with planetary
lock-up gears) 1 1
clutch
𝜈 = 𝜔𝑜𝑢𝑡 ⁄𝜔 𝑛 𝜈 = 𝜔𝑜𝑢𝑡 ⁄𝜔 𝑛
Figure 3-14: Left: Traditional automatic transmission. Right: Conceptual curves of the torque converter.
3.2.1.6 Energy storages

Fuel tank and battery are two examples of energy storages. An “energy buffer” often refers to an
energy storage that can not only be emptied (during propulsion), but also refilled by regenerating
energy from the vehicle during deceleration. A fuel tank is an energy storage, but not an energy buffer.
Also, a battery which can only be charged from the grid, and not from regenerating deceleration
energy, is not an energy buffer.
Energy buffers in vehicles are today often electro-chemical batteries. However, other designs are
possible, such as flywheels and hydrostatic accumulators. A simple model of a buffer is as follows:
=( 𝑛 𝑜𝑢𝑡 ) ∙
𝜂𝑐ℎ 𝑒; 𝑜𝑟 𝑛 > 𝑜𝑢𝑡 ;
𝐸̇ = {
=( 𝑛 𝑜𝑢𝑡 ) 𝜂𝑢𝑠𝑒 ; 𝑙𝑠 ; [3.8]
𝑤ℎ 𝑟 𝑛 = 𝑇 𝑛 ∙ 𝜔 𝑛; 𝑛 𝑜𝑢𝑡 = 𝑇𝑜𝑢𝑡 ∙ 𝜔𝑜𝑢𝑡 ;
Including how the buffer is connected, one more equation can be found: Typically, 𝜔 𝑛 = 𝜔𝑜𝑢𝑡 ; or
𝑇 𝑛 = 𝑇𝑜𝑢𝑡 ;.
The model uses stored energy, 𝐸. For batteries, one often uses state of charge, 𝑆𝑜𝐶, instead.
Conceptually, 𝑆𝑜𝐶 = 𝐸 𝐸𝑚 ;, where 𝐸𝑚 is a nominal maximum charge level.
A first approximation of the efficiencies, can be 𝜂𝑐ℎ 𝑒= 𝜂𝑢𝑠𝑒 = 𝑜𝑛𝑠𝑡 𝑛𝑡 < 1, but typically the
efficiency is dependent of many things, such as 𝑛 𝑜𝑢𝑡 . The model above does not consider any
leakage when buffer is “resting”: 𝑛 = 𝑜𝑢𝑡 .
3.2.2 Traction diagram

The force generated in the prime mover is transmitted through a mechanical transmission to the
wheel which then generates the propulsive forces in the contact patch between tyre and road. In an
electric in-wheel motor, the transmission can be as simple as a single-step gear. In a conventional
vehicle, it is a stepped transmission with several gear ratios (i.e. a gearbox). Then, the drivetrain can
be drawn as in Figure 3-15. The torque and rotational speed of the engine is transformed into force
112
and velocity curves via the mechanical drivetrain and driven wheel. The result is a Traction diagram.
The transformation follows the following formula, if losses are neglected:
𝑇 𝜔
=𝑟 𝑡 𝑜∙ ; 𝑛 = 𝑊ℎ 𝑙𝑅 𝑢𝑠 ∙ ; [3.9]
𝑊ℎ 𝑙𝑅 𝑢𝑠 𝑟 𝑡𝑜
prime mover driven wheel 𝒂𝒕𝒊𝒐

𝑭= ∙
𝒂𝒕𝒊𝒐 ∙ 𝒂𝒅𝒊𝒖𝒔
𝒂𝒅𝒊𝒖𝒔
𝝎 𝝎⁄ 𝒂𝒕𝒊𝒐 𝒗= ∙𝝎
𝒂𝒕𝒊𝒐
trans-
(vehicle) body
mission 𝒗
𝝎
𝒂𝒕𝒊𝒐 ∙
𝑭
𝝎⁄ 𝒂𝒕𝒊𝒐
prime mover
characteristics:
Traction Diagram
𝒂𝒕𝒊𝒐
Multiply torque by 𝑭
𝝎 Multiply rotational speed by 𝒗
𝒂𝒕𝒊𝒐
Figure 3-15: Construction of Traction Diagram, i.e. diagram for transmission of torque to longitudinal force
on vehicle.
A traction diagram for a truck is given in Figure 3-16, which also shows that there will be one curve for
each gear.
lowest gear
highest gear
Figure 3-16: Example of engine map and corresponding traction diagram map from a truck. (D13C540 is an
I6 diesel engine of 12.8-litre and 540 hp for heavy trucks.)
Losses in transmission can be included by loss models for transmission, such as:
𝑇
= 𝜂𝑇 ∙ 𝑟 𝑡 𝑜 ∙ ; 𝑤ℎ 𝑟 𝜂 𝑇 ≤ 1;
𝑟 𝑢𝑠
𝜔
= 𝜂𝜔 ∙ 𝑟 𝑢𝑠 ∙ ; 𝑤ℎ 𝑟 𝜂𝜔 ≤ 1; [3.10]
𝑟 𝑡𝑜
𝑒ℎ 𝑐 𝑒 ∙
𝑤ℎ 𝑟 𝜂 𝑇 ∙ 𝜂𝜔 = 𝜂𝑡𝑜𝑡 = = ≤1
𝑒𝑛 𝑛𝑒 𝑇∙𝜔
113
This will basically move the curves in the first quadrant downwards due to ηT < 1 and to the left due
to 𝜂𝜔 < 1. Tyre rolling friction is a torque loss mechanism, which on its own yields 𝜂𝜔 = 1 and ηT < 1.
Tyre longitudinal slip is a speed loss mechanism, which on its own yields 𝜂𝜔 < 1 and ηT = 1. See
Section 2.3.
A traction diagram is a kind of one degree of freedom graphical model. The traction diagram is
normally on complete vehicle level, so that the force axis is the sum of forces from all wheels, which
can include more than one propulsion system and also brakes.
3.2.3 Driving Resistance

From studying Figure 3-16 one could extrapolate that a very low transmission ratio, i.e., a very high
gear, we would enable almost infinite speed. We know that this is not possible. Something obviously
limits the top speed: the limit comes from “driving resistance”. As presented in Chapter 2, there are
tyre rolling-resistance forces and aerodynamic driving-resistance forces. In Chapter 2 we learnt that
rolling resistance primarily causes a torque, not a force. However, it causes a force on a none-driven
(and non-braked) wheel. One can select to put the rolling resistance as something that reduces the
propulsion curve or something that increases the resistance curve. To show both appearances, we
represent it as a torque (which reduces the propulsion curve) on driven axle and a force (which adds
to the resistance curve) on non-driven axles. This is the best if we want to draw a line for friction limit
for spinning wheels, as in Figure 3-22.
There are additional resistance force when going uphill: the grade or gravitational load on the vehicle.
This resistance can also be negative, i.e. act as propulsion down-hill. The formula for driving resistance
becomes as follows, using Equation [2.55]. Notation f means rolling resistance coefficient.
1
𝑒𝑠 = 𝑜 + ∙ ∙ sin( )+ ∙ 𝑑 ∙𝜌∙𝐴∙( 𝑤 𝑛𝑑 ) ;
𝑜 = ∑ 𝑤ℎ𝑒𝑒 ; [3.11]
𝑢𝑛𝑑 𝑒𝑛 𝑤ℎ𝑒𝑒 𝑠
𝑙𝑙 𝑤ℎ 𝑙𝑠 𝑟 𝑢𝑛 𝑟 𝑛: 𝑜 = ∙ ∙ ∙ cos( );
Superimposing the external loads on the Traction Diagram allows us to identify when the “demands”
can be met by the “supply”. As seen in Figure 3-17, the combination of external resistance loads and
the internally generated drive forces can be presented on the same figure. The resulting intersection of
supply (prime mover) and demand (driving resistance) identifies the top speed of the vehicle. These
results hold only for steady state (no acceleration) conditions.
Traction diagram
Supplied force (from propulsion system):
longitudinal force
𝑭𝒙 = 𝑇 𝑑 𝑒𝑛 𝑅𝑤 ;
Available force for acceleration: 𝑭𝒙 𝑭 𝒆𝒔
(Total) Driving resistance:
𝑭 𝒆𝒔 = 𝑑𝑒 + 𝑜 + ; (downhill positive,
air resistance: after ISO8855)
= .5 𝜌 𝐴 𝑑 ;
rolling resistance: Note: The rolling resistance of driven wheels is
included as a torque loss in the transformation from
𝑜 = 𝑢𝑛𝑑 𝑒𝑛 ; the prime mover to the traction force on the tyre .
𝒗𝒙 Picture is general and do not assume which axles
topspeed road grade resistance: are driven and undriven, so and 𝑜 are drawn
𝑑𝑒 = sin ; outside vehicle.
Figure 3-17: Traction diagram. Head wind speed, 𝑤 𝑛 𝑥, is assumed to be zero. See also Figure 3-22.
As seen in Figure 3-17, the acceleration can be identified as a vertical measure in the traction diagram,
divided by the mass. However, one should be careful when using the traction diagram for more than
114
steady state driving. We will come back to acceleration performance later, after introducing the two
effects “Load transfer” and “Rotating inertia effect”.
3.2.3.1 Losses due to lateral tyre slip

(Section 3.2.3.1 might require some studying of Chapter 4 for full understanding.)
There are more driving-resistance effects than covered in Equation [3.11]. One example is that none-
Ackermann steering geometry (toe or parallel steering on an axle, or two non-steered axles).
Another effect, which appears also for Ackermann steering geometry, is that power is lost due to
lateral axle slip. Now, we use the same simple model as in Figure 4-25, but additionally use =
⁄𝜔 ; and define power losses 𝑜𝑠𝑠 as sliding velocity counterdirected to force:
𝑜𝑠𝑠 = = 𝑠 𝑠 ;.
We also define a Cornering Resistance Coefficient, 𝐶𝑅𝐶 :
𝑜𝑠𝑠 ⁄ 𝑙 1 𝑙𝑟 1
𝐶𝑅𝐶 = = (( ) +( ) );
𝐶𝑟 𝐶 [3.12]
𝐶𝑅𝐶 is such that the extra propulsion force needed is = 𝐶𝑅𝐶
During a transport operation, the cornering in each time instant is typically described by two
variables, e.g. ( 𝑅 ), but only the combined scalar measure = ⁄𝑅 influences 𝐶𝑅𝐶 . Hence, we
can plot the following graph:
Figure 3-18: Left: Cornering Resistance Coefficient. Right: Required steering angle.
Vehicle data: = 15 𝑘 ; = 3; 𝑙 = 1. 5; 𝐶 = 6 [𝑘𝑁 𝑠]; 𝐶 = 8 [𝑘𝑁 1].
Notes:
• The model used above is not advanced enough to differ between which axle is driven. For such
purpose, one would need e.g. the model in Figure 4-22.
• Normal driving is often below 2 or 3 ⁄𝑠 , so the coefficient typically stays below 0.01.
However, the influence on energy consumption, during such “maximum normal” negotiation of
corners, is still of the same magnitude as rolling resistance coefficient.
• For ideally tracking axles, see Section 4.2.5, 𝐶 ∞ and 𝐶 ∞, which gives that 𝐶𝑅𝐶 and
consequently no power loss and no required propulsion force. Therefore, high cornering
stiffness is fuel efficient when cornering.
• When driving extreme cornering, such as driving as fast as possible in a circle on a test-track,
one will experience that the top speed is much lower than driving straight ahead. That is NOT
explained by [3.12]. Instead, an explanation is that tyres are saturated which leads to
3.2.4 Functions after start

Figure 3-19 shows how the functions can be found in a traction diagram.
115
3.2.4.1 Top speed *

Function definition: Top speed is the maximum longitudinal forward speed the vehicle can reach and maintain
on level and rigid ground without head-wind.
Top speed is the speed where the sum of all driving resistance terms is equal to the available
propulsion forces.
3.2.4.2 Grade-ability *
Function definition: Grade-ability is the maximum grade that a vehicle is capable to maintain the forward
motion on an uphill road at a certain constant speed, at a certain road friction level and with a certain load. (from
Reference (Kati, 2013))
For vehicles with high installed propulsion power per weight, the road friction can be limiting, but this
is not visualised in Figure 3-19. Since the speeds are higher than for start-ability, the air resistance
cannot be neglected.
Road slope of this curve Road slope of this curve (Grade-ability here limited by
( 𝑟 ) is the Grade- ( 𝑟 ) is the Grade-
ability for speed ability for speed propulsion, not road friction)
force, Fx
Longitudinal
𝑒𝑠 for
varying
road grades
Road slope of this curve
( 𝑟 ) is the Down-
grade holding capability
for speed
0
0 Top Speed,
speed
maximum engine brake
Figure 3-19: How Top speed, Grade-ability and Down grade holding capability is read-out from Traction
diagram.
3.2.4.3 Down-grade holding capability *

Function definition: Down-grade holding capability is defined as the maximum down-grade in which the
vehicle with certain weight is able to maintain a certain speed without using friction brakes.
The function is typically of interest for heavy trucks and certain typical certain weight and certain
speed is payload corresponding to maximum allowed weight and 80 km/h downhill.
The function is defined assuming there are clear friction brakes and other brakes, where the other
brakes are typically engine brake and retarders. For newer vehicle concepts having electric
propulsion, also regenerative braking via reversed electric propulsion motors can be discussed to be
allowed. However, because sometimes the energy storage will be full so that regenerative braking
cannot be applied. Also, a small energy storage will have a limited downhill length, which might call for
also prescribing a certain downhill distance.
3.2.5 Starting with slipping clutch

As seen in previous traction diagrams, there is no available positive propulsion force at zero speed.
This means that the diagram can still not explain how we can start a vehicle from stand-still.
The concepts in Figure 3-15 were used to create the force-velocity diagram in
116
Figure 3-20. It shows the smooth curve of a Continuous Variable ratio Transmission (CVT) in
comparison to the stepped transmission. The CVT is the ideal situation for the engine since it can
always let the engine work at a maximum power or minimum fuel consumption (minimum for the
momentarily required power). If the CVT has unlimited high ratio, it can actually have a non-zero
propulsion force at zero vehicle speed. Without losses, this force would be infinite, but in reality, it is
limited, but still positive, so the vehicle can start from stand-still.
A stepped transmission, as well as a CVT with limited ratio range, instead needs a clutch to enable
starting from vehicle stand-still. This is shown in Figure 3-20. The highest force level on each curve can
be reached at all lower vehicle speeds, because the clutch can slip. It requires the clutch to be engaged
carefully to the torque level just below the maximum the engine can produce. In traditional automatic
transmissions, the slipping clutch is replaced with a hydrodynamic torque converter, to enable start
from stand-still. Longitudinal
force, 𝑭𝒙
ICE on continuously variable ratio

(can keep power at maximum ICE power, 𝑇 = 𝑚 )
ICE on lowest gear (highest ratio)
slipping clutch ICE on highest gear (lowest ratio)
fully engaged clutch

(start from v=0 is not possible)
0
0 Speed, 𝒗𝒙
Figure 3-20: Force/Speed Curves for a Multiple Gear Transmission and for CVT.
3.2.6 Steady state vertical force distribution over

axles
The vehicle performance discussed previously does not rely on knowing the distribution of (vertical)
load between the axles. To be able to introduce limitations due to road friction, this distribution must
be known. Hence, we set up the free-body diagram in Figure 3-21.
m*g
Figure 3-21: Free Body Diagram for steady state vehicle. With ISO coordinate system, the road gradient is
positive when downhill. (Rolling resistance force on non-driven axles is included in .)
From the free-body diagram we can set up the equilibrium equations as follows and derive the formula
for load on front and rear axle:
117
Moment equilibrium, around rear contact with ground:

∙ ∙ (𝑙 ∙ cos( ) + ℎ ∙ sin( )) ∙ℎ ∙ (𝑙 + 𝑙 ) = ; ⇒
𝑙 ∙ cos( ) + ℎ ∙ sin( ) ℎ
⇒ = ∙ ∙ ∙
𝑙 +𝑙 𝑙 +𝑙
[3.13]
Moment equilibrium, around front contact with ground:
∙ (𝑙 + 𝑙 ) ∙ ∙ (𝑙 ∙ cos( ) ℎ ∙ sin( )) ∙ℎ = ;⇒
𝑙 ∙ cos( ) ℎ ∙ sin( ) ℎ
⇒ = ∙ ∙ + ∙
𝑙 +𝑙 𝑙 +𝑙
For most vehicles and reasonable gradients, one can neglect ℎ ∙ sin( ) since it is ≪ |𝑙 ∙ cos( )| ≈
|𝑙 ∙ cos( )|.
3.2.7 Friction limit

With a high-powered propulsion system, there is a limitation to how much the vehicle can be
propelled, due to the road friction limit. It is the normal load and coefficient of friction, which limits
this. For a vehicle which is driven only on one axle, it is only the normal load on the driven axle,
𝑑 𝑒𝑛 , that is the limiting factor:
= min( 𝑃 𝑜 𝑆 𝑠𝑡 ; ∙ 𝑑 𝑒𝑛 ) [3.14]
One realises, from Figure 2-11, that the rolling resistance on the driven axle works as a torque loss and
that the road friction limitation will be limiting 𝑇𝑒𝑛 ∙ 𝑟 𝑡 𝑜 ∙ 𝑑 𝑒𝑛 rather than limiting 𝑇𝑒𝑛 ∙
𝑟 𝑡 𝑜. Expressed using the rolling resistance coefficient, 𝑜 , gives:
𝑇𝑒𝑛 ∙ 𝑟 𝑡 𝑜
= min( 𝑃 𝑜 𝑆 𝑠𝑡 ; ∙ 𝑑 𝑒𝑛 ) = min ( 𝑜 ∙ 𝑑 𝑒𝑛 ; ∙ 𝑑 𝑒𝑛 ) [3.15]
𝑟 𝑢𝑠
This is shown in the traction diagram in Figure 3-22, where it should also be noted that the rolling
resistance curve only consists of the rolling resistance on the non-driven axles. See also Figure 2-11.
𝑇𝑒𝑛 𝑟 𝑡𝑜
force, Fx
Longitudinal
𝑑 𝑒𝑛
𝑃 𝑜 𝑆 𝑠𝑡 = =
𝑟 𝑢𝑠
𝑇𝑒𝑛 𝑟 𝑡𝑜
= 𝑜 𝑑 𝑒𝑛
𝑟 𝑢𝑠
Friction limit: 𝑑 𝑒𝑛
𝑒𝑠
Maximum 𝑒𝑠 = .5 𝜌 𝑑 𝐴 ;
available 𝑒𝑠 𝑜 = 𝑜 𝑢𝑛𝑑 𝑒𝑛 ;
𝑒𝑠 𝑠 𝑜 𝑒 = sin ;
0
0 Speed, 𝒗
Figure 3-22: Traction diagram with Road Friction limitation and Driving Resistance curves.
118
3.2.8 Start functions

3.2.8.1 Start-ability *
Function definition: Start-ability is the maximum grade that a vehicle is capable to start in and maintain the
forward motion at a certain road friction level and a certain load. (Reference (Kati, 2013))
Figure 3-23 shows how we find the start-ability in the traction diagram. There are two phenomena
that can limit start-ability: propulsion system or road friction. Also, in each case, we can theoretically
reach somewhat higher start-ability by allowing clutch or tyre to slip. However, in practice the start-
ability has to require “forward motion without significant slip in clutch or tyre”, because there will be a
lot of wear and heat in the slipping clutch or tyre. Hence, the lower curves in Figure 3-23 are used. The
reduction is however very small, since the resistance curves does not change very much in this speed
interval (the resistance curves in the figure have exaggerated slope; the air resistance can typically be
neglected for start-ability).
However, the energy loss (heat, wear) in clutch and tyre should be limited also during the starting
sequence. This can limit the start-ability more severely than the slope of the resistance curves, but it
cannot be shown in the traction diagram, since it is limited by energy losses in clutch or tyre, which is a
time integral of 𝑇 ∙ 𝜔𝑐 𝑢𝑡𝑐ℎ and 𝑇𝑤ℎ𝑒𝑒 ∙ 𝜔𝑤ℎ𝑒𝑒 . There can also be quite other limitations of start-ability,
such as deliberately limited engine torque at low gears, to save drive shafts.
Start-ability limited by propulsion system Start-ability limited by road friction

Road slope of this curve
force, Fx
force, Fx
Longitudinal
Longitudinal
is the start-ability with

spinning tyres.
Friction limit Road slope of
𝑒𝑠for this curve is the
varying start-ability with
rolling tyres. 𝑒𝑠 for
road slope varying
Road slope of this curve is road slope
the start-ability (ending
with slipping clutch). Friction
limit
lowest gear lowest gear
Road slope of this curve is the start-
ability (ending with engaged clutch).
0 0
0 Speed, 𝒗 0 Speed, 𝒗
Figure 3-23: How Start-ability is read-out from Traction diagram.
3.2.8.2 Towing capacity *

Function definition: Towing capacity is the maximum vehicle-external longitudinal force the vehicle can have
on its body and start and maintain a certain forward speed at a certain road friction and a certain up-hill gradient.
The driving situation for defining towing capacity is similar to the one for defining grade-ability.
Towing capacity describes how much load the vehicle can tow, in Figure 3-24, on a certain up-hill
gradient. Since towing a load is more relevant as part of a longer transport mission, it is normally also
for a particular constant speed, typically in range 80 to 100 km/h. Since the speed is that large, the air
resistance may not be neglected. It is also important consider air resistance of the trailer and that axle
loads can change, which changes both friction limitation and rolling resistance. A free-body diagram is
shown in Figure 3-24. It is noticeable, that there can also be an additional air resistance of the trailer
which will influence in a test of Towing capacity.
For pure off-road vehicles and agriculture tractors, the term “towing” can mean the maximum pulling
force at very low forward speeds at level ground. This is related but different to the above described
towing capacity for road vehicles.
119
mt*g
m*g
Figure 3-24: Towing Loads. The towing vehicle is front axle driven.
3.3 Functions over (long) cycles

Functions as fuel or energy consumption and emissions are relevant only over longer periods of
driving, typically some minutes to hours of driving. A collective name for this kind of driving can be
cycles. There are different ways of defining such cycles.
3.3.1 Description formats of Vehicle Operation

This section is about how to describe “all except the vehicle”, e.g. road, traffic, driver and payload. The
overall idea is to model the vehicle operation as independent of the vehicle, so that different vehicles, or
different designs or configurations of a certain vehicle, can be compared in a fair way.
3.3.1.1 (Traditional) Driving cycles

One way to model vehicle operation is a so-called driving cycle; where the relevant variables are pre-
scribed as function of time. At least on defines speed as a function of time. Examples of commonly used
driving cycles are given in Figure 3-25 and Figure 3-26. In addition, it can also be relevant to give road
inclination as function of time. Engine temperature and selected gear as functions of time may also be
defined. For hybrid vehicles, the possibility to regenerate energy via electric machines is limited in
curves, so curvature radius can also be prescribed as function of time. For heavy vehicles, the weight of
transported goods can be another important measure to prescribe.
80 km/h
Figure 3-25: New European Driving Cycle (NEDC). From (Boerboom, 2012)
120
80 km/h
20 min
80 km/h
10 min
Figure 3-26: Top: FTP cycle from http://www.epa.gov/oms/regs/ld-hwy/ftp-rev/ftp-tech.pdf. Bottom: HFTP

cycle from http://www.epa.gov/nvfel/methods/hwfetdds.gif.
FTP and HFTP are examples of cycles derived from logging actual driving, mainly used in North
America. NEDC is an example of a “synthetically compiled” cycle, mainly used in Europe. Worldwide
harmonized Light duty driving Test Cycle (WLTC) is a work with intention to be used world-wide, see
Figure 3-27. WLTC exists in different variants for differently powered vehicle [power/weight].
Figure 3-27: WLTC cycle from http://www.unece.org.
121
3.3.1.2 Driving pattern

A driving cycle can be condensed into a Driving pattern, i.e. a 2-dimensional function of speed and
acceleration, as shown in Figure 3-28. Note that the chronological order is no longer represented in
such representation; it is not a dynamic model. Figure 3-28 shows simply a scatter plot of time-
sampled combinations of speed and acceleration. Using the same diagram axis, such information can
also be shown as durations (in seconds or fractions of total time or frequency). A Driving pattern can
only be combined with a steady state model of the vehicle, such as “fuel consumption=function(speed,
acceleration)”, as opposed to a dynamic model of the vehicle. The Driving pattern itself includes the
driver, so no driver model is needed.
Driving patterns can use more than 2 dimensions, such as [speed, acceleration, road gradient]. In
principle, they can also use less than 2 dimensions, maybe only [speed]. The (steady state) vehicle
model has to reflect the corresponding dimensions.
Figure 3-28: FTP cycle converted to a Driving pattern, i.e. a distribution of operating duration in speed and
acceleration domain. From http://www.epa.gov/oms/regs/ld-hwy/ftp-rev/ftp-tech.pdf.
3.3.1.3 Transport task/Operating cycle

Desired speed, road inclination, weight of transported goods etc. are defined in position rather than
time. For stops along the route, the stop duration or departure time has to be separately defined. This
is a more advanced and realistic description of the vehicle usage. It can be extended to include
operation at stand-still, e.g. idle and loading/unloading cargo.
Simulation of such vehicle usage require some kind of driver model. A consequence is that different
driver models will give different results, e.g., different fuel consumption due to different driver
preferred acceleration. Hence, the driver model itself can be seen as a part of the vehicle usage
definition.
3.3.2 Rotating inertia effects

In Figure 3-17 it was shown that the acceleration cannot be found directly as a force difference
(distance between curves) divided by vehicle mass. This is because the Traction Diagram does not
contain any dynamics, and dynamics are more complicated than simply accelerating the vehicle mass.
The phenomenon that occurs is referred to here as “rotating inertia effect”. The rotating part of the
propulsion system, e.g. engine and wheels, must be synchronically accelerated with the vehicle mass.
122
This “steals” some of the power from the propulsion system. This affects the required propulsion force
when following accelerations in a driving cycle.
Consider a wheel rolling which is ideally rolling (no slip), with a free-body diagram and notations as in
Figure 3-29. Setting up 2 equilibrium equations and 1 compatibility equation gives:
T
w ∙ ̇= ;
vx ∙ 𝜔̇ = 𝑇 ∙ 𝑅; [3.16]
= 𝑅 ∙ 𝜔 ⇒ ̇ = 𝑅 ∙ 𝜔̇ ;
̇ m,J
F
Figure 3-29: Rolling wheel
Assume that we know torque T and want to 𝑇 𝑇

compare this with the situation without ∙ ̇= ; 𝑛 = ; [3.17]
rotational inertia. We get Eq [3.17. 𝑅 𝑅
Then we use the compatibility equation to 𝑇 1 𝑇

𝑘∙ ∙ ̇= ; 𝑛 =
∙ ;
eliminate 𝜔 in Equation [3.16]. We get Eq 𝑅 𝑘 𝑅 [3.18]
[3.18]. + ⁄𝑅 + 𝑡 𝑛𝑠
𝑤ℎ 𝑟 𝑘= = ;
Thus, we can see the effect of the rotating inertia as making the mass a factor k larger and making the
reaction force correspondingly lower. We call the factor k the “rotational inertia coefficient”.
In a vehicle propulsion system, there are rotational inertias mainly at two places:
• Rotating inertias before transmission, i.e. elements rotating with same speed as engine: 𝑒
• Rotating inertias after transmission, i.e. elements rotating with same speed as the wheel: 𝑤
The effective mass, 𝑘 ∙ , will be dependent on the ∙𝑟 𝑡 𝑜

; [3.19]
𝑤 𝑒
gear ratio as well: 𝑘∙ = + +
𝑅 𝑅
Typically for a passenger car, k=1.4 in the first gear and k=1.1 in the highest gear. So, in the first gear,
approximately one third of the engine torque is required to rotationally accelerate the driveline, and
only two thirds is available for accelerating the vehicle!
When the clutch is slipping, there is no constraint between engine speed and vehicle. The term with 𝑒
disappears from Equation [3.19]. If the wheel spins, both rotational terms disappear.
We can now learn how to determine acceleration from the Traction Diagram, see Figure 3-30.
123
force, Fx
Propulsion • 𝑘 = rotating inertia coefficient on gear 1,
without engine intertia and without
wheel inertia, since slipping tyre, i.e.
Gear 1 𝑘 =1
Friction limit
Resistance • 𝑘 = rotating inertia coefficient on gear 1,
• 𝑘 = rotating inertia coefficient on gear 2,

without engine intertia, since slipping
Gear 2
clutch
𝑘
𝑘
• 𝑘 = rotating inertia coefficient on gear 2

Speed, vx
𝑘
𝑘
Figure 3-30: Acceleration in Traction Diagram. Rotating inertia effects are shown assuming that the engine
is run on its maximum curve and the gear (or slipping clutch) for highest acceleration is selected.
3.3.3 Four quadrant traction diagram

When the driving cycle shows a deceleration, which is larger than can be achieved with resistance
force, we need to brake with a combination of engine braking and friction brakes. If only friction
braking is used, it can be with engaged or disengaged clutch. That influences the rotating inertia
coefficient by using or not using the 𝑒 term in Equation [3.19], respectively. The traction diagram can
be extended to also cover engine braking and friction braking. However, the friction brake system is
seldom limiting factor for how negative the longitudinal force can be. But the road friction is, see
Figure 3-31.
Propulsion force, Fx
Braking friction limit, +
Gear 1
Propulsion friction limit,
𝑚
𝑑 𝑒𝑛 ≈
Gear 2
Speed, vx
This linetype marks upper and lower

limits between where operating points
can be found using propulsion system
Propulsion friction limit, and friction brake system, including
𝑚 slipping clutch and spinning driven axle.
𝑑 𝑒𝑛 ≈
Reverse gear
Braking friction limit, -
Figure 3-31: Traction Diagram in 4 quadrants. One of two axles is assumed to be driven, which limits
propulsion to approximately half of braking friction limit. Up-hill slope is assumed, which is seen as
asymmetric resistance.
124
3.3.4 Functions over cycles

3.3.4.1 Energy consumption *
Function definition: Energy (or Fuel) Consumption is the amount of energy [ ] (or fuel [𝑘 ] or [𝑙 𝑡𝑟 ])
consumed by the vehicle per performed transportation amount. Transportation amount can e.g. be measured in km,
𝑘 𝑝 𝑟𝑠𝑜𝑛 𝑘 𝑡𝑜𝑛 𝑘 or 𝑘 . The vehicle operation has to be defined, e.g. with a certain
driving cycle (speed as function of time or position), including road gradient, cargo load, road surface conditions, etc.
The consumption arises in the prime mover, see Figure 3-4 and Figure 3-5.
How to predict the consumption for a vehicle during a certain driving cycle is rather straight-forward
using what has been presented earlier in this chapter. Since a driving cycle is a prediction of how the
vehicle is moving, it actually stipulates the acceleration of a mass, which calls for an “inverse dynamic
analysis”. In such one assumes that the driving cycle is met exactly, which means that both required
wheel speed and required wheel torque can be calculated for each time instant. Then, via a propulsion
and brake system model, the corresponding fuel consumption in the engine can also be found for each
time instant. A summary of such an inverse dynamic algorithm for prediction of fuel or energy
consumption is given in Equation [3.20].
For each time step in the driving cycle:

• Calculate operating point for vehicle (speed and acceleration) from
driving cycle. Acceleration is found as slope of v(t) curve. Other
quantities, such as road slope, also needs to be identified;
• Select gear (and clutch state, tyre spin, friction brake state, etc) to
obtain this operating point. Select also friction brake, especially for
operating points which can be reached using only friction brake. If the
vehicle has an energy buffer, regenerative braking is also an option;
• Calculate required actuation from propulsion system on the driven wheel,
i.e. rotational speed and shaft torque; [3.20]
• Calculate backwards through propulsion system, from wheel to prime
motor. It gives the operating point for prime mover (rotational speed
and torque);
• Read prime mover consumption [in kg/s or W=J/s, not specific
consumption, not efficiency] from prime mover consumption map;
• Sum up consumption [in kg or J] with earlier time steps, e.g. using the
Euler forward integration method: AccumulatedConsumption =
AccumulatedConsumption + Consumption*TimeStepLength;
end;
The final accumulated consumption [in kg or J] is often divided by the total covered distance in the
driving cycle, which gives a value in kg/km or J/km. If the fuel is liquid, it is also convenient to divide
by fuel density, to give a value in litre/(100*km). It can also be seen as a measure in , which is the
area of the “fuel pipe” which the vehicle “consumes” on the way.
3.3.4.1.1 Forward and backward simulation
We should note that the calculation scheme in Equation [3.20] does not always guarantee a solution.
An obvious example is if the driving cycle prescribes such high accelerations at such high speeds that
the propulsion system is not enough, i.e. we end up outside maximum torque curve in engine diagram.
This is often the case with “inverse dynamic analysis”, i.e. when acceleration of inertial bodies is
prescribed and the required force is calculated. An alternative is to do a dynamic analysis, which
means that a driver model calculates the pedals in order to follow the driving cycle speed
approximately, but not exactly. Inverse dynamic analysis is often more computational efficient, but
limits what can phenomena that can be modelled in the propulsion and brake system. The
computational benefit is especially large if state variables can be omitted, which is often the case but
not always. Inverse dynamics and dynamic simulations are sometimes referred to as backward and
forward simulation, respectively; see Reference (Wipke, o.a., 1999).
125
Driving cycles are used for legislation and rating for passenger cars. For commercial vehicles, the
legislation is done for the engine alone, and not for the whole vehicle. Rating for commercial vehicles is
under development, see http://ec.europa.eu/clima/events/docs/0096/vecto_en.pdf.
3.3.4.2 Emissions *
Function definition: As Energy consumption, but amount of certain substance instead of amount of energy.
There are emission maps where different emission substances (NOx, HC, etc.) per time or per
produced energy can be read out for a given speed and torque. This is conceptually the same as
reading out specific fuel consumption or efficiency from maps like in Figure 3-4 and Figure 3-5. A
resulting value can be found in mass of the emitted substance per driven distance.
Noise is also sometimes referred to as emissions. It is not relevant to integrate noise over the time for
the driving cycle, but maximum or mean values can have relevance. Noise emissions are very
peripheral to vehicle dynamics.
3.3.4.3 Tyre wear *

Function definition: Tyre wear is the worn out tyre tread depth on a vehicle per performed transportation
amount. Transportation amount can be measured as for Energy consumption. Tyre wear as a vehicle function has
to consider all tyres on the vehicle, e.g. as maximum over the wheels (assuming that all tyres are changed when one
is worn out) or average (assuming that single tyres are exchanged when worn out).
There are models for tyre wear (e.g. outputting “worn tread depth per time”), see Equation [2.53]. For
a certain driving cycle, we can integrate the 𝑊 𝑟𝑅 𝑡 [in mass/s or mm tread depth/s], over time
similar to energy consumption rate, which becomes worn material [in mass or mm tread depth]. The
wear rate per wheel is a function of the total slip, so it can include both longitudinal slip (propulsion
and brake) and lateral slip (from cornering and toe angles).
Generally, the worn material will be different for different axles, or wheels, so a tyre change strategy
might be necessary to assume to transform the worn material on several axles into one cost. The cost
will depend on whether one renews all tyres on the vehicle at once or if one change per axle. The tyre
wear is a cost which typically sums up with energy cost and cost of transport time (e.g. driver salary,
for commercial vehicles).
3.3.4.4 Range *
Function definition: Range [km] is the inverted value of Energy consumption [kg/km, litre/km or J/km], and
multiplied with fuel tank size [kg or litre] or energy storage size [J].
The range is how far the vehicle can be driven without refilling the energy storage, i.e. filling up fuel
tank or charging the batteries from the grid. This is in principle dependent on how the vehicle is used,
so the driving cycle influences the range. In principle, the same prediction method as for energy
consumption and substance emissions can be used. In the case of predicting range, you have to
integrate speed to distance, so that you in will know the travelled distance.
3.3.4.5 Acceleration reserve *

Function definition: Acceleration reserve is the additional acceleration the vehicle will achieve within a certain
time (typically 0.1..1 s) without manual gear-shifting by pressing accelerator pedal fully, when driving in a certain
speed on level ground without head-wind. For vehicles with automatic transmissions or CVTs the certain time set
can allow automatic gear-shift (or ratio-change) or not. The reserve can also be measured in propulsion force.
In general terms, the lowest consumption is found in high gears. However, the vehicle will then tend to
have a very small reserve in acceleration. It will, in practice, make the vehicle less comfortable and less
safe to drive in real traffic, because one will have to change to a lower gear to achieve a certain higher
acceleration. The gear shift gives a time delay.
Figure 3-32 shows one way of defining a momentary acceleration reserve. The reserve becomes
generally larger the lower gear one selects.
126
A characteristic of electric propulsion systems is that an electric motor can be run at higher torque for
a short time than stationary, see
Figure 3-3. On the other hand, the stationary acceleration reserve is less gear dependent, since an
electric motor can work at certain power levels in large portions of its operating range.
One can calculate the acceleration reserve at each time instant over a driving cycle. However,
integration of acceleration reserve, as we did with fuel, emissions and wear, makes less sense. Instead,
a mean value of acceleration reserve tells something about the vehicle’s driveability. Minimum or
maximum values can also be useful measures.
Acceleration reserve was above described as limited by gear shift strategy. Other factors can be
limiting, such as energy buffer state of charge for parallel hybrid vehicles or how much overload an
electric machine can take short term, see right part of Figure 3-32.
Vehicle with conventional propulsion system Vehicle with electric propulsion system
Longitudinal
Longitudinal
force, 𝑭𝒙
force, 𝑭𝒙
Mass∙Acceleration
Mass∙Acceleration gear 1 Reserve on gear 1
Reserve on gear 1 Mass∙Acceleration
gear 1 Mass∙Acceleration Reserve on gear 2
Reserve on gear 2
Short term
Acceleration
gear 2 Reserve
gear 2
0 0
0 Speed, 𝒗 0 Speed, 𝒗
Figure 3-32: Acceleration reserves for different gears. Large dots mark assumed operating points, each with
its acceleration reserve shown.
3.3.5 Load Transfer with rigid suspension

Longitudinal load transfer redistributes vertical force from one axle to the other. The off-loaded axle
can limit the traction and braking. This is because the propulsion and brake systems are normally
designed such that axle torques cannot always be ideally distributed.
For functions over longer events it is often reasonable to consider the suspension as rigid. We start
with the free-body diagram in Figure 3-33, which includes acceleration, .
x
y
Figure 3-33: Free Body Diagram for accelerating vehicle. Rolling resistance in and .
Note that the free-body diagram and the following derivation is very similar to the derivation of
Equation [3.13], but we now include the fictive force ∙ .
127
Moment equilibrium, around rear contact with ground:

∙ + ∙ ∙ (𝑙 ∙ 𝑜𝑠( ) + ℎ ∙ 𝑠 𝑛( )) ∙ℎ ∙ ∙ ℎ = ;⇒
𝑙 ∙ 𝑜𝑠( ) + ℎ ∙ 𝑠 𝑛( ) ℎ ℎ
⇒ = ∙( ∙ ∙ ) ∙
[3.21]
Moment equilibrium, around front contact with ground:
+ ∙ ∙ ∙ (𝑙 ∙ 𝑜𝑠( ) ℎ ∙ 𝑠 𝑛( )) ∙ℎ ∙ ∙ ℎ = ;⇒
𝑙 ∙ 𝑜𝑠( ) ℎ ∙ 𝑠 𝑛( ) ℎ ℎ
⇒ = ∙( ∙ + ∙ )+ ∙
These equations confirm what we know from experience, the front axle is off-loaded under
acceleration with the load shifting to the rear axle. The opposite occurs under braking.
The load shift has an effect on the tyre’s grip. If one considers the combined slip conditions of the tyre
(presented in Chapter 2), a locked braking wheel limits the amount of lateral tyre forces. The same is
true for a spinning wheel. This is an important problem for braking as the rear wheels become off-
loaded. This can cause locking of the rear wheels if the brake pressures are not adjusted appropriately.
See more in Section 3.4.5.
3.3.5.1 Varying road pitch

The model in 3.4.5..3.4.7 assumes flat but not level road, i.e. is constant. An example where
varies is when passing a crest or a hollow, or meeting uphill or meeting downhill, see Figure 3-34. If
negotiating a curve at the same time as a crest, a vehicle can loose vertical force under tyres so that
lateral grip is affected.
crest hollow
s
𝑦
= = s
s
• The variable s is the distance along the

road.
• Road gradient versus inertial coordinate
system, ≈ , is a function of s
𝑑
• where = ; is a function of s.
𝑑𝑠
• Vertical acceleration in inertial coordinate
system, ̈ ≈ ;,
̈
𝑑2
• where = ; is a function of s.
𝑑𝑠 2
Figure 3-34: Free Body Diagram for driving over arbitrary vertical road profile.
Moment equilibrium, around rear and front wheel contact with ground gives:
″
𝑙 ∙ cos( ) + h ∙ sin( ) ℎ ℎ
= ∙ (( + )∙ ∙ ) ∙ ;
L L
[3.22]
″
𝑙 ∙ cos( ) h ∙ sin( ) ℎ ℎ
= ∙ (( + )∙ + ∙ )+ ∙ ;
𝑑 𝑑 𝑑2
Assuming that we have the road as (𝑠), then = arctan (𝑑𝑠 ) ≈ 𝑑𝑠
; and ″
= 𝑑𝑠2 ;.
128
Note that this model is assuming that vertical variations of road are larger than wheel base and track
width and same on left and right side of the road/vehicle. Else the variation would be called road
unevenness, which will be more treated in Chapter 5.
If models with body vertical and pitch motion and suspension springs, such as in Sections 3.4.6 and
3.4.7 it is often suitable to express the vertical fictive force, ̈ with 𝜔 instead of ″
.
The fictive force downwards will then be 𝜅 = 𝜔 instead. This can be understood
from basic geometry, ″ ≈ 𝜅 , where 𝜅 is the road pitch curvature [1⁄𝑙 𝑛 𝑡ℎ], see Figure 3-34.
3.3.6 Acceleration
Acceleration performance like, typically, 0-100 km/h over 5..10 s, will be addressed in this section.
These accelerations are relatively steady state (vehicle pitch and heave is relatively constant), so the
suspension compliance is not considered.
Accelerations will also be covered in Section 3.4, as being shorter events. The vehicle pitch and heave
varies more and consequently, the suspension compliance becomes important to model. This
modelling is also more suited for braking, which typically involve suspension more than propulsion.
3.3.6.1 Acceleration performance *

Function definition: Acceleration performance is the time needed to, with fully applied accelerator pedal,
increase speed from a certain speed to another certain higher speed, at certain road friction on level ground without
head-wind and certain load.
3.3.6.2 Solution using integration over time

A front-wheel-drive passenger car with a stepped gearbox should accelerate from 0 to 100 km/h. A
Matlab code is given in Equation [3.23], which simulates the acceleration uphill from stand-still, using
simple numerical integration. The code calculates the possible acceleration in each of the gears, and
one mode with slipping clutch. In each time step it selects that which gives the highest acceleration.
The numerical data and results are not given in the code, but some diagrams are shown in Figure 3-35.
The code is not fully documented, only using equations so far presented in this compendium.
dt=0.1; t_vec=[0:dt:10]; vx_vec(1)=0;
for i=1:length(t_vec)
vx=vx_vec(i);
Fres=m*g*sin(p)+froll*m*g*cos(p)+0.5*roh*A*cd*vx*vx;
%if gear 1 (clutch engaged)

ratio=ratios(1);
we=vx*ratio/radius;
Te=interp1(Engine_w,Engine_T,we);
Fx=Te*ratio/radius;
ax=(Fx-Fres)/(m+(Jw+Je*(ratio^2))/(radius^2));
Ffz=m*(g*lr/L-ax*h/L);
if Fx>mu*Ffz
Fx=mu*Ffz;
ax=(Fx-Fres)/m;
end
ax1=ax;
[3.23]
ratio=ratios(2);
… then similar as for gear 1
ax2=ax;

ratio=ratios(3);
… then similar as for gears 1 and 2
ax3=ax;
%if clutch slipping on gear 2

ratio=ratios(2);
wc=vx*ratio/radius; %speed of output side of clutch
Te=max(Engine_T);
we=Engine_w(find(Engine_T>=Te)); %engine runs on speed where max torque
Fx=Te*ratio/radius;
129
ax=(Fx-Fres)/(m+Jw/(radius^2));
Ffz=m*(g*lr/L-ax*h/L);
if Fx>mu*Ffz
Fx=mu*Ffz;
ax=(Fx-Fres)/m;
end
if wc>we %if vehicle side (wc) runs too fast, we cannot slip on clutch
ax=-inf;
end
ax0=ax;
[ax,gear_vec(i)]=max([ax0,ax1,ax2,ax3]); vx_vec(i+1)=vx+ax*dt;
end
Phenomena that are missing in this model example are:

• Gear shifts are assumed to take place instantly, without any duration
• The option to use slipping clutch on 1st and 3rd gear is not included in model
• The tyre slip is only considered as a limitation at a strict force level, but the partial slip is not
considered for simplifcation. The code line “we=vx*ratio/radius;” is hence not fully
correct. Including the slip, the engine would run at somewhat higher speeds, leading to that it
would lose its torque earlier, leading to worse acceleration performance.
• Load transfer is assumed to take place instantly quick; delays due to Suspension compliance, as
described in Section 3.3, are not included.
Engine 4
Traction diagram
x 10
2
Torque[Nm] gear 1
160
Power[kW] 1.8 gear 2
gear 3
140 1.6
resistance
slip limit
120 1.4
Longitudinal force [N]
max achievable
1.2
100
1
80
0.8
60
0.6
40
0.4
20
0.2
0 0
0 100 200 300 400 500 600 700 0 5 10 15 20 25 30 35 40
rotational speed [rad/s] speed [m/s]
Simulation result
14
12
10
vx [m/s]
6
gear (0=clutch slips on gear 2)
0
0 1 2 3 4 5 6 7 8 9 10
t [s]
Figure 3-35: Example of simulation of acceleration, using the code in Equation [3.23].
3.3.6.3 Solution using integration over speed

An alternative way to find the relation between and 𝑡 is to separate the differential equation:
130
∙
∙ = ∙ = ( ) 𝑒𝑠 ( ); ⇒ = 𝑡⇒
𝑡 ( ) 𝑒𝑠 ( )
𝑡 [3.24]
∙ ∙
⇒∫ =∫ 𝑡⇒ 𝑡= ∫ ;
( ) 𝑒𝑠 ( ) ( ) 𝑒𝑠 ( )
0 0 0
Now, the time is calculated by means of integration over speed, as opposed to integration over time. If
simple mathematic functions are used to describe ( ) and 𝑒𝑠 ( ) the solution can be mathematically
explicit, but the previous integration over time is more general and works for more advanced models.
3.4 Functions in (short) events

This section targets models and methods to define and predict functions in a certain and shorter time
frame, typically 0.5 to 5 seconds. It can be both acceleration and deceleration. (Friction) Brake system
and phenomena as load transfer then becomes important, why these are presented early. But first,
some typical driving manoeuvres are presented.
3.4.1 Typical test manoeuvres

When applying the longitudinal actuator systems (propulsion system and brake system) there are a
couple of different situations which are typical to consider:
• Straight line maximum braking from, typically 100 km/h to stand-still for passenger cars.
• Braking in curve with significant lateral acceleration, see References (ISO, 2006) and (ISO,
2011).
• Straight line acceleration, typically 0 to 100 km/h and 80-100 km/h.
• Accelerating in curve with significant lateral acceleration.
For these four main situations, one can also vary other, typically:
• At high road friction and at low friction, often called “hi-mu” and “lo-mu”
• At different road friction left and right, often called “split-mu”
• At sudden changes in road friction, called “step-mu”
• At high speed, typically 200 km/h, for checking stability
• At different up-hill/down-hill gradients
• At different road banking (slope left to right)
A propelled or braked wheel or axle develops a longitudinal force, , counter-acting the rotation. is
limited by the road friction: | |𝑚 = ∙ ;. (not | |𝑚 = ∙ ∙ , see Figure 2-11).
Braking Coefficient = ⁄ , is a property defined for an axle or a single wheel. It can be seen as the
utilized friction coefficient, 𝑢𝑡 = brake force/normal load. The 𝑢𝑡 should not be mixed up with
(available) friction coefficient, ; the relation between them is 𝑢𝑡 ≤ .
3.4.2 Deceleration performance

There are some different functions that measures braking performance or deceleration performance.
3.4.2.1 Braking efficiency *

Function definition: Braking Efficiency is the ratio of vehicle deceleration and the best brake-utilized axle (or
wheel), while a certain application level of the brake pedal at a certain speed straight ahead, at certain road friction
on level ground without head-wind and certain load at a certain position in the vehicle.
In equation form, the Braking Efficiency becomes = ( ⁄ )⁄( max 𝑢𝑡 ) ;. If Braking
∀ 𝑒𝑠
Efficiency=1=100%, the distribution of braking is optimal.
131
3.4.2.2 Braking Distance *

Function definition: Braking Distance is the distance travelled during braking with fully applied brake pedal
from a certain speed straight ahead to another certain lower speed, at certain road friction on level ground without
head-wind and certain load at a certain position in the vehicle.
For passenger cars one typically brakes fully from 100 km/h and then the braking distance is typically
around 40 m (average deceleration 9.65 m/s2). For a truck it is typically longer, 51..55 m (7.5..7 m/s2).
3.4.2.3 Stopping Distance *

Function definition: Stopping Distance is the distance travelled from that an obstacle becomes visible to driver
have taken the vehicle to stand-still. Certain conditions, as for Braking Distance, have to be specified, but also a
certain traffic scenario and a certain driver to be well defined.
Stopping Distance is the braking distance + the “thinking/reaction distance”, which depends on the
speed and the reaction time. The reaction time of a driver is typically 0.5..2 seconds.
3.4.3 Brake System

Generally speaking, there are several systems that can brake a vehicle:
• Service brake system (brake pedal and ABS/ESC controller, which together applies brake pads
to brake discs/drums)
• Parking brake (lever/button that applied brake pads to brake discs/drums, normally on rear
axle on cars but all axles on heavy vehicles)
• Engine braking (ICE operates at “engine brake” as marked in Figure 3-3)
• Electric machines (machines can be used symmetrical, i.e. both for positive and negative
torques, see Figure 3-3)
• Retarders, or auxiliary brakes, for heavy vehicles. They normally use hydraulic or Eddy current
to dissipate engine, as opposed to dry-friction. So, they cannot brake at low speeds or stand-
still.
• Large steering angles will decelerate the vehicle, see Section 3.2.3.1.
This section is about Friction brakes, meaning Service brakes and Parking brake. In vehicle dynamics
perspective, these have the following special characteristics:
• Friction brakes are almost unlimited in force for a limited time since they can lock the wheels
for most driving situations and road friction (ICE and electric motors are often limited by their
maximum power, since it is often smaller than available road friction.) If, however, the friction
brakes are used for a long time the brake lining will start to fade, e.g. lose friction coefficient
due to high temperature (oxidation and melting of pad/lining material).
• Friction brakes can only give torque in opposite direction to wheel rotation. (Electric motors
can brake so much that wheel spins rearwards.)
• Friction brakes can hold the vehicle at exact standstill. (If using electric machines for holding
stand-still in a slope, a closed loop control would be necessary, resulting in that vehicle “floats”
a little.)
The basic design of a passenger car brake system is a hydraulic system is show in Figure 3-36. Here,
the brake pedal pushes a piston, which causes a hydraulic pressure (pressure = pedal force/piston
area). The hydraulic pressure is then connected to brake callipers at each wheel, so that a piston at
each wheel pushes a brake pad towards a brake disc ( 𝑠 𝑜𝑟 = 𝑝𝑟 𝑠𝑠𝑢𝑟 𝑝 𝑠𝑡𝑜𝑛 𝑟 ). The
brake torque on each wheel is then simply: 𝑡𝑜𝑟𝑞𝑢 = 𝑛𝑢 𝑏 𝑟 𝑜 𝑟 𝑡 𝑜𝑛 𝑠𝑢𝑟 𝑠
𝑠 𝑜 𝑛𝑡 𝑜 𝑟 𝑡 𝑜𝑛 𝑠 𝑟 𝑢𝑠 𝑠 𝑜𝑟 . (Normally, there are 2 friction surfaces, since
double-acting brake calipers.) By selecting different piston area and disc radii at front and rear, there
is a basic hydro mechanical brake distribution ratio between front and rear axle. There are normally
two circuits for redundancy.
132
Brake systems for heavy trucks are generally based on pneumatics, as opposed to hydraulics, see
Figure 3-38. In reference (Tagesson, 2017), there is a good descriptive chapter about brake systems
for heavy vehicles.
Driver pushes Master cylinder

Front Front
Slave cylinder Front
on pedal
Hydraulics
II HI
Pistons with
brake pad
Front Front
Brake
disc
Rear
Hydraulic pump for

ABS/ESC and hydraulic X
valves are not drawn. (common) LL HH
Figure 3-36: Layouts of a hydraulically applied brake system, which is conventional on passenger cars.
ABS intervention 1.
2.
Master Cylinder
Inlet Valves
ESC intervention
(Pedal braking. One wheel 3. Wheel Cylinders/Callipers (No pedal braking. One wheel
brake controlled=released) 4. Brake Discs brake controlled=applied)
5. Outlet Valves
6. Tank/Low Pressure
Accumulator
1. 7. Pump 1.
8. Check Valve
9. 9. Isolation Valve/Traction 9.
Control Valve
8. 8.
7. 7.
2. 5. 5.
5. 2.
2. 5.
6. 6.
2.
3. 3. 3. 3.
controlled = controlled =
= released 4. … 4.
= applied 4. … 4.
(several wheels, only 2 drawn) (several wheels, only 2 drawn)
Figure 3-37: Concept of hydraulically applied brake system for ABS and ESC functions.
3.4.4 Pedal functions

3.4.4.1 Pedal Response *
Function definition: Accelerator pedal response is how vehicle acceleration varies with accelerator pedal
position, for a certain vehicle speed and possibly certain gear, on level ground without pressing the brake pedal.
Function definition: Brake pedal response is how vehicle deceleration varies with brake pedal force, for a
certain vehicle speed, on level ground without pressing the accelerator pedal.
133
These functions, together with the functions in Section 3.4.4.2, enable the driver to operate the vehicle
longitudinally with precision and in an intuitive and consequent way. The requirements based on
above function definitions, are typically that the translation of pedal position (or force) to vehicle
acceleration (or deceleration) should be consistent, progressive and oscillation-free.
For accelerator pedal steps, there should be enough acceleration, but also absence of “shunt and
shuffle” (driveline oscillations). When accelerator pedal is suddenly lifted off, there shall be certain
deceleration levels, depending on vehicle speed and gear selected.
Figure 3-38: Pneumatically applied brake system for heavy vehicles. Electronics Brake System, EBS, from
Volvo GTT, Mats Sabelström.
3.4.4.2 Pedal Feel *

Function definition: Accelerator pedal feel is the pedal’s force response to pedal position.
Function definition: Brake pedal feel is the pedal’s position response to pedal force.
These functions, together with the functions in Section 3.4.4, enable the driver to operate the vehicle
longitudinally with precision and in an intuitive and consequent way.
3.4.5 Brake proportioning

Consider a vehicle where either the front axle or rear axle is braked to locking (𝜔 = ). Figure 3-39
shows the vehicle’s response to a small disturbance and argues for that:
• A vehicle with locked front wheels tends to be stable. However, steering ability is lost, so
vehicle continues straight, incapable of curving its path.
• A vehicle with locked rear wheels tends to be unstable. It turns around and ends up sliding
with the rear first.
• In most driving situations, it is preferred that the front wheels lock first. But it is of course also
important that both axles are used as much as possible during braking, to improve braking
efficiency. Hence, there are trade-offs when designing the brake torque distribution, see
Section 3.4.5.
• The basic function of a brake system is that brake pressure (hydraulic on passenger cars and
pneumatic on trucks) is activated so that it applies brake pads towards brake discs or drums.
134
In a first approximation, the pressure is distributed with a certain fraction on each axle. For
passenger cars this is typically 60..70% of axle torque front. In heavy trucks, the proportioning
varies a lot, e.g. 90% for a solo tractor and 30% for heavy off-road construction rigid truck. The
intention is to utilize road friction in proportion to the normal load, but not brake too much
rear to avoid yaw instability, see Figure 3-39.
Front locked (Rear rolling) Rear locked (Front rolling)
4 4
3 3
> ≈ >
𝜔̇ > 2 𝜔̇ > 2
Δ = 𝜔 > Δ = 𝜔 >
1 Δ𝜔 > 1 Δ𝜔 >
3 4 3 4
> ≈
1. Disturb with an small impact yaw moment, so Δ𝜔 > , but Δ = .

2. The disturbance gives (requires) the fictive force 𝜔 > and 1-4. <as for case ”Front locked”>
fictive moment 𝜔̇ > .
3. It also gives lateral sliding > and . 51. So, dominates over .
4. Lateral axle force is counter-directed sliding. A rolling tyre is much 62. Compared to “Front locked”, there is less
laterally stiffer 𝑁⁄ ⁄𝑠 than a locked. counter-acting yaw moment from the force
55. So, dominates over . couple and 𝜔 , so less stable.
66. The yaw moment from the force couple and (Or, express 𝜔̇ through equilibrium
𝜔 counter-acts Δ𝜔 , which is stabilizing. around front axle:
(Or, express 𝜔̇ through equilibrium around rear axle: 𝜔̇ + 𝜔 𝑙 = ; ⇒
𝑑𝜔̇ 𝑚 𝑥 𝑟 𝑚 𝑥
𝜔̇ 𝜔 𝑙 = ; ⇒ = < ; ⇒ 𝑆𝑡 𝑏𝑙 ) ⇒
𝑑𝜔̇
= > ; ⇒ 𝑛𝑠𝑡 𝑏𝑙 )
𝑑𝜔 − 𝑑𝜔
Figure 3-39: Locking of one axles. The numbers 1..6 give an intuitive causal explanation to why locked front
axle tends to be stable and locked rear tends to be unstable. (See more in Section 4.4.2.5 and 4.4.2.6).
Same reasoning works for propulsion, if “locked“ is replaced by “spin”, meaning large positive
longitudinal tyre slip. Spin at front makes vehicle more yaw stable than spin rear.
If neglecting air resistance and road grade in Eq [3.21], the vertical axle loads can be calculated as
function of deceleration (= ). An ideal brake distribution would be if each axle always utilize same
𝐹 𝑥 𝐹𝑟𝑥 𝐴𝑠𝑠𝑢 𝐹 𝑥 𝐹 𝐹 𝑥
fraction of available friction: = ⇒ { =
∙𝐹 𝑟 ∙𝐹𝑟 ;} ⇒ 𝐹𝑟𝑥
= 𝐹𝑟
⇒ 𝐹𝑟𝑥
=
𝑙 ℎ
𝑚∙( ∙ 𝑟 − 𝑥 ∙ ) ∙ 𝑟 − 𝑥 ∙ℎ
𝑙
𝐿 𝐿
ℎ
= ; Combining with + = ∙ ; gives the optimal and :
∙ + 𝑥 ∙ℎ
𝑚∙( ∙ +
L 𝑥 ∙L)𝐿
𝑙 ℎ 𝑙 ℎ
= ∙ ∙( ); 𝑛 = ∙ ∙( + );
[3.25]
𝑚 𝑚
or, if eliminating : +( ℎ
𝑟
) +( ℎ
+ )= ;
Eq [3.25] is plotted for variation in centre of gravity height and longitudinal position in Figure 3-40.
The proportioning is done by selecting pressure areas for brake calipers, so the base proportioning
will be a straight line, marked as “Hydrostatic brake proportioning”. For passenger cars, one typically
designs this so that front axle locks first for friction below 0.8 for lightest vehicle load and worst
variant. For heavier braking than .8 , or higher (or front-biased) centre of gravity, rear axle will
lock first, if only designing with hydrostatic proportioning.
135
To avoid rear axle lock up, one restricts the brake pressure to the rear axle. This is done by pressure
limiting valve, brake pads with pressure dependent friction coefficient or Electronic Brake Distribution
(EBD). In principle, it bends down the straight line as shown in Figure 3-40. With pressure dependent
values one gets a piece-wise linear curve, while pressure dependent friction coefficient gives a
continuously curved curve. EBD is an active control using same mechatronic actuation as ABS. EBD is
the design used in today’s passenger cars, since it comes with ABS, which is now a legal requirement
on most markets.
8000
7000
6000
5000
𝑁
Fxr⁄[N]
4000
3000
2000 lf=1.25; lr=1.5; h=0.4;

lf=1.25; lr=1.5; h=0.5; (higher load)
1000
lf=1.15; lr=1.6; h=0.4; (front-biased load)
0
0 5000 10000 15000
⁄ [N]
Fxf 𝑁
Figure 3-40: Brake Proportioning diagram. The curved curves mark optimal distribution for some variation
in position of centre of gravity.
On heavy vehicles with EBS (Electronic Brake System) and vertical axle load sensing, the brake
pressure for each axle can be tailored. For modest braking (corresponding to deceleration ≤ ⁄𝑠 )
all axles are braked with same brake torque, to equal the brake pad wear which is importance for
vehicle maintenance. When braking more, the brake pressure is distributed more in proportion to
each axle’s vertical load.
Ideal curve
ECE regulation
limits to this region
[data from BMW 320i E46]
Figure 3-41: Brake Proportioning. From (Boerboom, 2012). If looking carefully, the “HydroStatic” curve is
weakly degressive, which is thanks to brake pads with pressure dependent friction coefficient.
136
3.4.6 Steady state heave and pitch, trivial linkage

Additional to that the axle vertical loads change due to acceleration , there are also change in out-of-
road-plane motion (heave and pitch). In the following section, we study constant acceleration, e.g.
when mild braking for a long time. We propose the steady state model in Figure 3-42. The model
differs between the “unsprung mass” (wheels and the part of the suspension that does not heave) and
the “sprung mass” or “body” (parts that heaves and pitches as one rigid body).
There is no damping included in model, because their forces would be zero, since there is no
displacement velocity, due to the steady-state assumption. As constitutive equations for the
compliances (springs) we assume that displacements are measured from a static condition and that
the compliances are linear. The road is assumed to be smooth, i.e. = = .
= 0+ ∙( ) 𝑛 = 0 + ∙( )
[3.26]
𝑤ℎ 𝑟 0+ 0 = ∙ 𝑛 0∙𝑙 0∙𝑙 =
Suspension model with “trivial suspension”

ax
z
Fair m*ax
zf x zr
py
m*g
hair
h
Ffx Frx
zfr=0 zrr=0 Ffz Frz
lf lr
L Quasi steady-state assumed, so that
z,x,py,zf,zr, are displacements from longitudinal acceleration (ax) may
a static stand-still position. be non-zero, but vertical and pitch
zfr=zrr=0 means that road is smooth. acceleration are zero.
Figure 3-42: Model for steady state heave and pitch due to longitudinal wheel forces.
We see already in free-body diagram that and always act together, so we rename + =
𝑤 , where w refers to wheel. This and equilibrium give:
∙ + 𝑤 = ;
∙ = ; [3.27]
∙𝑙 ∙𝑙 𝑤 ∙ℎ ∙ (ℎ ℎ) = ;
Compatibility, to introduce body displacements, and 𝑝 , gives:
= 𝑙 ∙ ; 𝑛 = +𝑙 ∙ ; [3.28]
Combining constitutive relations, equilibrium, compatibility and renaming + = 𝑤, where w

refers to wheel), gives, as Matlab script:
137
clear, syms zf zr Ffz Frz Ffz0 Frz0 ax z py
sol=solve( ...
'Ffz=Ffz0-cf*zf', ...
'Frz=Frz0-cr*zr', ...
'Ffz0+Frz0=m*g', ...
'Ffz0*lf-Frz0*lr=0', ...
'-Fair-m*ax+Fxw=0', ... [3.29]
'm*g-Ffz-Frz=0', ...
'Frz*lr-Ffz*lf-Fxw*h-Fair*(hair-h)=0', ...
'zf=z-lf*py', ...
'zr=z+lr*py', ...
zf, zr, Ffz, Frz, Ffz0, Frz0, ax, z, py);
The solution from Matlab script in Equation [3.29] becomes:
𝑤
= ;
∙𝑙 ∙𝑙
= ∙( 𝑤 ∙ℎ+ ∙ (ℎ ℎ)) [3.30]
∙ ∙
+
𝑝 = ∙( 𝑤 ∙ℎ+ ∙ (ℎ ℎ))
∙ ∙
In agreement with intuition and experience the body dives (positive pitch) when braking (negative
𝑤 ). Further, the body centre of gravity is lowered (negative z) when braking and weaker suspension
front than rear ( ∙ 𝑙 < ∙ 𝑙 ), which is normally the chosen design for cars.
The air resistance force is brought into the equation. It can be noted that for a certain deceleration,
there will be different heave and pitch depending on how much of the decelerating force that comes
from air resistance and from longitudinal wheel forces. But, as already noted, heave and pitch does not
depend on how wheel longitudinal force is distributed between the axles.
3.4.6.1 Effective stiffness

A very simplified suspension is assumed in Figure 3-42. The stiffnesses and are the effective
stiffnesses at each axle, also called wheel spring rate if measured per wheel. The physical spring may
have another stiffness, but its effect on vertical force is captured in the effective stiffness. An example
of how the effective stiffness is found from a real suspension design is given in Figure 3-43.
Note that the factor ( ⁄𝑏) is not the only difference between effective and physical stiffness, but the
effective can also include compliance from other parts than just the spring, such as bushings and tyre.
There will also be a need for a corresponding effective damping coefficient, see Section 3.4.7.2, or
wheel (damping) rate. How forces in road plane is transferred is not well captured in this model, see
Figure 3-44. Compared to suspension models later in compendium, we can call it a “trivial linkage”
model. In Section 3.4.7, the suspension linkage is better modelled, which allows validity for transients
braking and acceleration.
Suspension design is further discussed at these places in this compendium: Section 3.4.6, Section
4.3.9.4 and Section 5.2.
138
cp*zp
Fzr
physical spring zp
with stiffness cp
cp zrr
zp
a
Fzr
b
Fzr
Moment equilibrium of arm:
Effective suspension design cp*zp*a=Fzr*b
 Fzr=cp*zp*a/b;
virtual spring
with stiffness cr Compatibility: zp/a=zrr/b;
Equivalence in stiffness:
cr Fzr=cr*zrr;
zrr Together:
 cp*(zrr*a/b)*a/b=cr*zrr 
Fzr  cr=cp*(a/b)2;
Figure 3-43: From physical suspension design to effective stiffness. Upper left: From
http://www.procarcare.com/icarumba/resourcecenter/encyclopedia/icar_resourcecenter_encyclopedia_suspsteer3.asp.
interpretation
including forces in
road plane
Figure 3-44: Interpretation of effective stiffness model, as a “trivial linkage” suspension model.
3.4.7 Load Transfer with suspension compliance

and linkage
In previous model of load transfer, see Section 3.3.5, we assumed nothing about the transients of
transfer (vertical wheel) loads between front and rear axle. So, if we study longer events when the
wheel force is applied and then kept constant for a longer time (1-5 s), it is often a good enough model.
But if the wheel forces vary more, we need to capture the transients better. Then it is important to
consider that the linkage can transfer some of the wheel longitudinal forces. When studying the
transients, it is also relevant to consider the damping.
139
Another reason for doing a better model than in Section 3.3.5 can be that one is interested in the
displacements, heave and pitch, which are not covered in the model in Section 3.3.5.
There are basically two modelling ways to include the suspension in the load transfer: through a pitch
centre or through a pivot point for each axle, see Figure 3-45.
Model with pitch center Model with Axle Pivot Points

vz
wx vx
h-hPC
h-hPC
PC
lPCf lr lf
lPCf
PC=Pitch Centre
Figure 3-45: Models for including suspension effects in longitudinal load transfer
3.4.7.1 Load Transfer model with Pitch Centre

This model will not be deeply presented in this compendium. It has drawbacks in that it has only one
suspended degree of freedom. Also, it does not take the distribution of longitudinal wheel forces
between the axles into account. These short comings is not very important for studying dive and squat,
but they are essential when studying rapid individual wheel torque changes in time frames of 0.1 s,
such as studying ABS or traction control. So, since the model with axle pivot points is more generally
useful and not much more computational demanding (and probably easier intuitively), that model is
prioritized in this compendium.
3.4.7.2 Load Transfer model with Axle Pivot Points

Behold the free-body diagram in Figure 3-46. The road is assumed to be flat, = . The force
play in the rear axle is shown in more detail. and are reaction forces in the pivot point. 𝑠 is the
force in the elasticity, i.e. where potential spring energy is stored. The torque 𝑇𝑠 is the shaft torque, i.e.
from the propulsion system. Both torque from propulsion and brake system contribute to . But
torque from friction brake 𝑇 𝑒 𝑐 is not visible in free-body diagram, unless decomposed in suspension
and wheel, as in the right-most part of Figure 3-46. This is because the friction braking appears as
internal torque (or, depending on the brake design, probably forces) between brake pad and brake
calliper. Any part of the longitudinal wheel force that is applied via reaction torque to unsprung parts
will not add to shaft torque, such as an electric motor mounted on unsprung parts or propulsion via
longitudinal propeller shaft to a final gear (as usual for rigid propelled axles). The term 𝜔 in is
often negelctable, except for during high speed, i.e. large . The term can be explained as the other
centripetal accelerations (giving centrifugal (fictive) forces) in Section 4.4.2.2.
We assume that displacements are measured from the forces 𝑠 0 and 𝑠 0, respectively, and that the
compliances are linear. The total constitutive equations become:
𝑠 = 𝑠 0 + ∙( ) 𝑛 𝑠 = 𝑠 0 + ∙( ) [3.31]
Now, there are two ways of representing the dynamics in spring-mass systems: Either as second order
differential equations in position or first order differential equations in velocity and force. We select
the latter, because it is easier to select suitable initial values. Then we need the differentiated versions
of the compliances constitutive equations:
140
̇
𝑠 = + ∙( ̇ ̇ )= ∙ 𝑛 ̇ =
𝑠 + ∙( ̇ ̇ )= ∙ [3.32]
= ̇ 𝜔 ; Decomposed in smaller
Rear axle: free body diagrams, to
𝑠 + 𝑑
explain 𝑇𝑠 and 𝑇 𝑒 𝑐 .
Suspension:
𝑠 + 𝑑
ℎ 𝑇𝑠
Wheel:
= = 𝑇𝑠
𝑙 𝑙
𝑥 are displacements from a static stand-still position. = = means flat road.
Figure 3-46: Free-body diagram for model with Axle Pivot Points. (The model assumes drive shafts from
propulsion system mounted on body. If rigid shaft with longitudinal propeller shaft, 𝑇𝑠 = .)
The damper forces are denoted 𝑑 and 𝑑 . They will appear in the equilibrium equations quite
similar to 𝑠 and 𝑠 . Note that the damping coefficients, and , are the effective ones, i.e. the ones
defined at the wheel contact point with ground, as opposed to the physical ones defined for the actual
physical damper. C.f. effective stiffness in Section 3.4.6.
𝑑 = ∙( ̇ ̇ ) = ∙ 𝑛 𝑑 = ∙( ̇ ̇ )= ∙ [3.33]
Now, 3 equilibria for whole vehicle and one moment equilibria around pivot point for each axle gives:
∙ + + = ; (𝑤 𝑡ℎ = ̇ ;)
∙ ∙ + + = ; (𝑤 𝑡ℎ = ̇ 𝜔 ;)
∙ 𝜔̇ + ∙𝑙 ∙𝑙 ( + )∙ℎ = ; [3.34]
( 𝑠 𝑑 )∙ ∙ + 𝑇𝑠 = ;
( 𝑠 + 𝑑 )∙ ∙ + 𝑇𝑠 = ;
It can be noted that a “trivial suspension” (wheel kinematics simply vertically supported), the same
equations is valid if we let ∞ and ∞. It is then no difference if the vehicle is actuated with a
via shaft torque 𝑇𝑠 or via reaction to unsprung parts, 𝑇 𝑒 𝑐 .
Compatibility, to connect to body displacements, and , gives:
= 𝑙 ∙ ; 𝑛 = +𝑙 ∙ ;
= 𝑙 ∙𝜔 ; 𝑛 = +𝑙 ∙𝜔 ; [3.35]
̇= ; 𝑛 ̇ =𝜔 ;
By combining constitutive relations, equilibrium and compatibility we can find explicit function so:
• 𝑆𝑡 𝑡 𝐷 𝑟 𝑡 𝑠 = Explicit orm unction(𝑆𝑡 𝑡 𝑠 𝐼𝑛𝑝𝑢𝑡𝑠);
• 𝑆𝑡 𝑡 𝑠 = [ 𝜔 𝑠 𝑠 ];
• 𝑆𝑡 𝑡 𝐷 𝑟 𝑡 𝑠 = [ ̇ ̇ 𝜔̇ ̇𝑠 ̇
𝑠 ̇ ̇ ];
• 𝐼𝑛𝑝𝑢𝑡𝑠 = [ 𝑇𝑠 𝑇𝑠 ];
The ExplicitFormFunction can be integrated with well-established methods for numerical ode. Such
simulation of is shown in Section 3.4.9.1.
141
3.4.7.3 Steady state heave and pitch with non-trivial linkage

If we study long term steady state for the model described in Section 3.4.7.2 we will find a model
comparable with the model in Section 3.4.6. So, Equations [3.31] to [3.35] are combined. We also
neglect air resistance force for clarity of equations. Equation [3.31] becomes, in Matlab format:
clear, syms zf zr Ffz Frz Fsf0 Fsr0 ax Fsf Fsr z py
sol=solve( ...
'Fsf=Fsf0-cf*zf', 'Fsr=Fsr0-cr*zr', ...
'Fsf0+Fsr0=m*g', 'Fsf0*lf-Fsr0*lr=0', ...
'-m*ax+Ffx+Frx=0', ...
'-m*(0+0)-m*g+Ffz+Frz=0', ...
'-J*0+Frz*lr-Ffz*lf-(Ffx+Frx)*h=0', ...
'(Frz-Fsr-0 )*gr-Frx*er+Tsr=0', '(Fsf+0-Ffz)*gf-Ffx*ef+Tsf=0', ...
'zf=z-lf*py', 'zr=z+lr*py', ...
zf, zr, Ffz, Frz, Fsf0, Fsr0, ax, Fsf, Fsr, z, py);
%results:
[3.36]
% ax = (Ffx + Frx)/m
% z = -(Tsr*cf*gf*lf^2 - Tsf*cr*gr*lr^2 + Tsr*cf*gf*lf*lr - Tsf*cr*gr*lf*lr

- Frx*cf*er*gf*lf^2 + Ffx*cr*ef*gr*lr^2 + Ffx*cf*gf*gr*h*lf +
Frx*cf*gf*gr*h*lf - Ffx*cr*gf*gr*h*lr - Frx*cr*gf*gr*h*lr -
Frx*cf*er*gf*lf*lr + Ffx*cr*ef*gr*lf*lr)/(cf*cr*gf*gr*L^2)
% py = -(Tsr*cf*gf*lf + Tsr*cf*gf*lr + Tsf*cr*gr*lf + Tsf*cr*gr*lr +

Ffx*cf*gf*gr*h + Frx*cf*gf*gr*h + Ffx*cr*gf*gr*h + Frx*cr*gf*gr*h -
Frx*cf*er*gf*lf - Ffx*cr*ef*gr*lf - Frx*cf*er*gf*lr -
Ffx*cr*ef*gr*lr)/(cf*cr*gf*gr*L^2)
The solution should be compared with corresponding solution in Equation [3.30]. One can see that
is exactly the same. Then, a general reflection is that the displacement, z and py, in Equation [3.36]
follows a complex formula, but that they are dependent on how the 𝑤 = + is applied: both
dependent on distribution between axles and dependent on how much of the axle forces ( and )
that are actuated with shaft torques (𝑇𝑠 and 𝑇𝑠 , respectively). In Figure 3-49, dashed lines show the
solutions from Equation [3.30].
3.4.7.4 Examples of real suspension designs

In Figure 3-46, some kind of “trailing arm” is drawn both for front and rear axle. For rear axle, that is a
realistic design even if other designs are equally common. However, for front axle a so-called
McPherson suspension is much more common, see Figure 3-47. Figure 3-48 shows a suspension for a
heavy vehicle.
gf
ef
90 deg
ef
gf
Figure 3-47: Example of typical front axle suspension, and how pivot point is found. The example shows a
McPherson suspension. From Gunnar Olsson, LeanNova.
Additional phenomena
It is relevant to point out the following, which are not modelled in this compendium:
142
• Stiffness and damping may be dependent of wheel (vertical) displacement and wheel
steering angle. One way of inserting this in the model is to make the coefficients varying with
spring force, which is a measure of how much compressed the suspension is. Here, non-
linearity within spring working range, as well as bump stops, can be modelled. Also, position
of pivot points (or pitch and roll centres) can be dependent of wheel displacement steering
angle.
• Dampers are often deformation direction dependent, i.e. different damping coefficients are
suitable to use for compression and rebound. Typical is 2..4 times softer (smaller d [N/(m/s)])
in compression than in rebound.
Non reactive Level sensor Reactive

Air spring
V-stay
Stabilizer bar
Bump stop
Air spring member

V-stay Spring link
bracket
U-bolts
Anti-roll bar link Spring link bearing
Shock absorber
Reaction rod Spring hanger bracket
Reaction rod bracket Anti-roll bar
𝑠 𝑠
=0
𝑇𝑠 𝑇𝑠
=∞
Figure 3-48: Axle suspensions/installations for double rear axle heavy vehicles.
3.4.8 Pitch functions at quick force changes

3.4.8.1 Dive at braking *
Function definition: Dive at braking is pitch angle of the vehicle body when applying a step in deceleration
request to a certain level. Either peak pitch or quasi-steady state pitch angle can be addressed.
Now, study the suspension at front axle in Figure 3-46. When the axle is braked, will be negative
and push the axle rearwards, i.e. in under the body. The front of the vehicle will then be lifted as in pole
jumping. This means that this design counter-acts the (transient) dive of the front. (Only the transient
dive will be reduced, while the dive after a longer time of kept braking is dependent only on the
stiffnesses according to Equation [3.30].) The design concept for front axle suspension to place the
pivot point behind axle and above ground is therefore called “anti-dive”.
If the braking is applied without shaft torque 𝑇𝑠 , a good measure of the Anti-dive mechanism is .
This is the normal way for braking, since both the action and reaction torque acts on the axle. For in-
board brakes, or braking via propulsion shaft, the reaction torque is not taken within the axle, but the
reaction torque is taken by the vehicle body. The action torque 𝑇𝑠 then appears in the equilibrium
143
equation for the axle, as shown in Equation [3.34]. If we neglect the wheel rotational dynamics for a
while, we can insert 𝑇𝑠 = ∙ 𝑅𝑤 in the equation with 𝑇𝑠 in Equation [3.34]:
( 𝑠 + 𝑑 )∙ ∙ + 𝑇𝑠 = ; 𝑤 𝑡ℎ 𝑇𝑠 = ∙ 𝑅𝑤 ; ⇒
⇒( 𝑠 + 𝑑 )∙ ∙ + ∙ 𝑅𝑤 = ; ⇒ [3.37]
⇒( 𝑠 + 𝑑 )∙ ∙( 𝑅𝑤 ) = ;
We can then see that a good measure of the Anti-dive mechanism is ( 𝑅𝑤 ) instead.
3.4.8.2 Squat at propulsion *

Function definition: Squat at propulsion is pitch angle of the vehicle body when applying a step in acceleration
request to a certain level. Either peak pitch or quasi-steady state pitch angle can be addressed.
Now, study the suspension at rear axle in Figure 3-46. When the axle is propelled, will push the
axle in under the body. This means that this design reduces the rear from squatting (transiently). The
design concept for rear axle suspension to place the pivot point ahead of axle and above ground is
therefore called “anti-squat”.
3.4.8.3 Anti-dive and Anti-squat designs

With Anti-dive front and Anti-squat rear, we avoid front lowering at braking and rear lowering at
acceleration, respectively. But how will the designs influence the parallel tendencies: that rear tend to
lift at braking and front then to lift at propulsion? Well, they will luckily counteract also these: Braking
at rear axle will stretch the rear axle rearwards and upwards relative to the body. When propelling the
front axle, the propulsion force will stretch the front axle forwards and upwards relative to the body.
(If one brakes at one axle and propels at the other, the reasoning is not valid. This mode may seem
irrelevant, but could be desired for a hybrid vehicle with ICE on front axle and electric motor on rear
axle, if one would like to charge batteries “via the road”. This is an example of that novel designs may
rise new issues.)
In summary: Anti-dive and anti-squat refer to the front diving when braking and the rear squatting
when acceleration. Anti-dive and anti-squat can be measured in fractions: Anti-dive for = or =
( 𝑅𝑤 ) and Anti-squat= or = ( 𝑅𝑤 ) . Normal values are typically 0.05..0.15.
3.4.9 Acceleration and Deceleration

Acceleration performance like, typically, 0-100 km/h over 5..10 s, was addressed in Section 3.3.6. In
present Section we address the similar functionality but include larger transients, such as when wheel
longituidinal wheel force is changed more rapidly, typically changing ± ⁄ during 0.2-0.5 s.
3.4.9.1 Deceleration performance *

Function definition: See Section 3.4.2.2.
Deceleration performance can now be predicted, including the suspension mechanisms. It is a very
important function, and every decimetre counts when measuring braking distance in standard tests
like braking from 100 to 0 km/h. The active control of the brake torques (ABS function) is then very
important, and this is so fast dynamics that the suspension mechanisms of Anti-lift and Anti-dive
influences. The position of the load in the vehicle will influence, since it influences the load transfer.
We will now set up a mathematical model, see Equation [3.38], which shows how the normal forces
change during a braking event. It is based on the physical model in Figure 3-46. Note that we cannot
neglect the term 𝜔 as we did in Section 3.4.7.3, since the manoeuvre cannot be seen as quasi-
steady state. Driving resistance contributes normally with a large part of the deceleration, but we will
neglect this for simplicity, just to show how the suspension mechanism works. The equations in the
model are presented in the dynamic modelling standardized format “Modelica”, and are hence more or
less identical to Equation [3.32] to [3.35].
144
Ffx = if 1 < time and time < 3 then -0.4*m*g else 0;
Frx = if 3 < time and time < 7 then -0.4*m*g else 0;
Tsf/Rw = 0;
Tsr/Rw = if 5 < time and time < 7 then -0.4*m*g else 0;
// Motion equations:
der(z) = vz;
der(py) = wy;
// Constitutive equations for the springs:

der(Fsf) = -cf*vzf;
der(Fsr) = -cr*vzr;
// Constitutive equations for the dampers:

Fdf = -df*vzf; [3.38]
Fdr = -dr*vzr;
//(Dynamic) Equilibrium equations:

-m*der(vx) + Ffx + Frx = 0;
-m*(der(vz)-wy*vx) - m*g + Ffz + Frz = 0;
-Jy*der(wy) + Frz*lr - Ffz*lf - (Ffx + Frx)*h = 0;
(Frz - Fsr - Fdr)*gr - Frx*er + Tsr = 0;
(Fsf + Fdf - Ffz)*gf - Ffx*ef + Tsf = 0;
//Compatibility:
zf = z - lf*py;
zr = z + lr*py;
vfz = vz - lf*wy;
vrz = vz + lr*wy;
The simulation results are shown in Figure 3-49. It shows a constant deceleration, but it is changed
how the decelerating force is generated. At time=3 s, there is a shift from braking solely on front axle
to solely on rear axle. The braking is, so far, only done with friction brakes, i.e. generating torque by
taking reaction torque in the axle itself. At time=5 s, there is a shift from braking with friction brakes to
braking with shaft torque. It should be noted that if we shift axle or shift way to take reaction torque,
gives transients even if the deceleration remains constant.
One can also see, at time=1 s, that the normal load under the braked axle first changes in a step. This is
the effect of the Anti-dive geometry. Similar happens when braking at rear axle, due to the Anti-squat
geometry. Since brake performance is much about controlling the pressure rapidly, the transients are
relevant and the plots should make it credible that it is a control challenge to reach a high braking
efficiency.
3.4.9.2 Acceleration performance *

Function definition: See Section 3.3.6.1.
The model presented in Equation [3.38] can also be used to predict acceleration performance in a
more accurate way compared to Section 3.3.6. Especially, the more accurate model is needed when
propelling or braking on the limit of tyre to road adhesion, since the normal load of each tyre then is
essential. It is a challenge to control the propulsion and brake wheel torques to utilize the varying
normal loads under each axle.
145
w ithCentrifugalForce.vx
50
0
0 1 2 3 4 5 6 7 8 9
w ithCentrifugalForce.z [m] w ithCentrifugalForce.py [rad] w ithCentrifugalFo
0.03
Different displacements, both

0.02
transiently and steady state
0.01
py= [𝑟 ]
0.00 [ ]
[ ]
-0.01
[ ]
-0.02
0 1 2 3 4 5 6 7 8 9
w ithCentrifugalForce.Fzf w ithCentrifugalForce.Fzr w ithoutCentrifugalForce
1.4E4
1.2E4
1.0E4
8.0E3
Same steady state vertical
6.0E3
forces, but transiently different
4.0E3
0 1 2 3 4 5 6 7 8 9
Braking on front axle Braking on rear axle Braking on rear axle
with friction brakes with friction brakes with shaft torque
Figure 3-49: Deceleration sequence with constant vehicle deceleration, but changing between different ways
of actuation. (With the centripetal term 𝜔 (solid) and without (dashed). Dotted shows without anti-
dive/-squat geometry, i.e. = = ∞.)
3.5 Control functions

Some control functions will be described. First, some general aspects on control are given.
3.5.1 Longitudinal Control

Some of the most important sensors available in production vehicles and used mainly for longitudinal
control are listed below. (Sensors for instrumented vehicles for testing can be many more.)
• Wheel Speed Sensors, WSS. For vehicle control design, one can often assume that “sensor-
close” software also can supply information about longitudinal vehicle speed.
146
• Vehicle body inertial sensors. There is generally a yaw rate gyro and a lateral accelerometer
available, but sometimes also a longitudinal accelerometer. The longitudinal accelerometer is
useful for longitudinal control and longitudinal velocity estimation.
• Pedal sensors. Accelerator pedal normally has a position sensor and brake pedal force can be
sensed via brake system main pressure sensor. Heavy vehicles often have both a brake pedal
position and brake pressure sensors.
• High specification modern vehicles have environment sensors (camera, radar, GPS with
electronic map, etc.) that can give information (relative distance and speed, etc.) about objects
ahead of subject vehicle. It can be both fixed objects (road edges, curves, hills, …) and moving
objects (other road users, animals, …).
• Information about what actuation that is actually applied at each time instant is available, but
it should be underlined that the confidence in that information often is questionable.
Information about axle propulsion torque is generally present, but normally relies on imprecise
models of the whole combustion process and torque transmission, based on injected amount of
fuel and gear stick position. (Electric motors can typically give better confidence in estimation,
especially if motor is close to the wheel without too much transmission in between.) Wheel
individual friction brake torque is available, but normally rely on imprecise models of the brake
systems hydraulic/pneumatic valves and disc friction coefficient, based on brake main cylinder
pressure.
• Information about what actuation levels that are possible upon request (availability or
capability) is generally not so common. It is difficult to agree of general definitions of such
information, because different functions have so different needs, e.g. variations in accepted time
delay for actuation.
• Sometime wheel/axle forces can be sensed. One case is when pneumatic suspension. More
extreme variants are under development, such as sensors in the wheel bearings which can sense
forces (3 forces and roll and yaw moment) (SKF) and sensors in drive shafts (ABB).
3.5.2 Longitudinal Control Functions

3.5.2.1 Pedal driving *
Function definition: See Section 3.4.4.
These functions, Pedal functions
Pedal Response *, are often not seen as comparable with other control functions, but they become
more and more relevant to define as such, since both accelerator and brake pedals tend to be
electronically controlled, and hence they become increasingly tuneable. Also, more and more
functions, such as those below in section 3.5.2, will have to be arbitrated with the pedals.
In modern passenger vehicles, Accelerator pedal is normally electronically controlled but the Brake
pedal is basically mechanical. In modern heavy commercial vehicles, both functions are electronically
controlled.
The functions in Section “3.4.4.2 Pedal Feel *” are normally not actively controlled, but in there are
concept studies with active pedals, where also the pedal feel can be actively changed to give feedback
to driver.
3.5.2.2 Cruise Control and Adaptive Cruise Control (CC, ACC) *

Function definition: Cruise Control, CC, controls the vehicle’s longitudinal speed. Driver can activate the
function and decide desired speed.
Function definition: Adaptive Cruise Control, ACC, is an addition to CC. ACC controls the vehicle’s time
gap to a lead vehicle. Driver can activate the function and decide desired gap. When there is no lead vehicle, CC
controls the vehicle’s speed.
147
The purpose of CC is to keep the vehicle at a driver selected longitudinal speed, while driver not
pushes the accelerator pedal. The actuator used is the propulsion system. In heavy vehicles also the
braking system (both retarders and service brakes) is used to maintain or regulate the vehicle speed.
ACC is an addition to CC. The purpose of ACC is to keep a safe distance to the lead vehicle (vehicle
ahead of subject vehicle). ACC uses also friction brake system as actuator, but normally limited to a
deceleration of . .4 𝑠 .
The safe distance which ACC aims for, is often expressed as a time gap, driver adjustable in the range
. .3 𝑠. A model behind this is that the time gap is the driver reaction time and the subject vehicle can
decelerate as much as the object vehicle. More advanced models can allow smaller time gap in certain
situations. This is desired because it reduces the risk that other vehicles cut in between subject and
leading object vehicle. Such models can take into account, e.g., acceleration of the object vehicle, pedal
operation of subject vehicle, road gradient, road curvature, road friction and deceleration capabilities
of subject and object vehicles.
CC is normally only working down to 3 .. 4 𝑘 ℎ. ACC can have same limitation, but with automatic
transmission, good forward looking environment sensors, brake actuators and speed sensing, ACC can
be allowed all the way down to stand-still.
3.5.2.2.1 Topography adapted CC
A development of CC has varying set speed which is optimized for a predicted road topography about a
minute ahead. Such products are on the market, e.g. Volvo iSee and Scania Active Prediction.
3.5.2.3 Anti-lock Braking System, ABS *

Function definition: Anti-lock Braking System, ABS, prohibits driver to lock the wheels while braking.
The wheel brake torques requested are limited by ABS in a way that each individual wheel’s longitudinal slip
stays above a certain (negative) value. An extended definition of ABS also includes vehicle deceleration requested
by other functions than pedal braking, such as AEB. ABS only uses friction brake as actuator.
The purpose of ABS is to avoid losing vehicle brake force due to that the tyre force curve drops at high
slips AND to leave some friction for steering and cornering, see Sections 2.3, 0 and 2.5. ABS is a wheel
slip closed loop control, active when driver brakes via brake pedal. It keeps the slip above a certain
value, typically -15..-20 %. ABS uses the friction brakes as actuator.
Each wheel is controlled individually, but all wheels’ speed sensors contribute to calculation of vehicle
longitudinal speed, for calculation of actual slip. In the ABS function, it may be included how slip are
distributed between the wheels, such as normally the front axle is controlled to a slip closer to locking
than the rear axle. Also, a sub-function called “select-low” which makes the wheel closest to locking
decide the pressure also for the other wheel at the same axle. Select-low is typically used at rear axles.
Figure 3-50: ABS control. Principle and control sequence
148
3.5.2.4 Electronic Brake Distribution, EBD *

Function definition: Electronic Brake Distribution prohibits driver to over-brake the rear axle while
braking. An extended definition of EBD also includes vehicle deceleration requested by other functions than pedal
braking, such as AEB. EBD only uses friction brake as actuator.
With a fix proportioning between front and rear axle braking, there is a risk to over-brake rear axle
when friction is very high, since rear axle is unloaded so much then. Before electronic control was
available, it was solved by hydraulic valves, which limited the brake pressure to rear axle when pedal
force became too high. In today’s cars, where electronic brake control is present thanks to legislation
of ABS, the software base function EBD fulfils this need. In heavy vehicles, pneumatic valves are used
that limits the brake pressure in relation to the rear axle load (deflection of mechanical spring
suspension or air pressure in air suspension).
"DataLog_Braking.xls","ABS straight"
0.05
-0.05
-0.1
sxfl/[1]
-0.15 sxfr/[1]
sxrl/[1]
-0.2 sxrr/[1]
-0.25
1 1.5 2 2.5 3 3.5 4 4.5 5

"DataLog_Braking.xls","ABS straight"
140
120
100
80
60
pfl/[bar]
40 pfr/[bar]
prl/[bar]
20 prr/[bar]
vx/[km/h]
0
1 1.5 2 2.5 3 3.5 4 4.5
time [s] 5
Figure 3-51: ABS control, Data log from passenger car test.
There are other side functions enabled by having ABS on-board. Such are “select low”, which means
that the brake pressure to both wheels on an axel is limited by the one with lowest pressure allowed
from ABS. So, if one wheel comes into ABS control, the other gets the same pressure. This is most
relevant on rear axle (to reduce risk of losing side grip) but one tries to eliminate the need of it totally,
because it reduces the brake efficiency when braking in curve or on different road friction left/right.
It is often difficult to define strict border between functions that is a part of ABS and which is part of
EBD, which is why sometimes one say ABS/EBD as a combined function.
3.5.2.5 Traction Control, TC *

Function definition: Traction Control prohibits driver to spin the driven axle(s) in positive direction while
accelerating. An extended definition of TC also includes vehicle acceleration requested by other functions than pedal
braking, such as CC. TC uses both friction brakes and propulsion system as actuators.
149
The purpose of Traction Control is to maximise traction AND to leave some friction for lateral forces
for steering and cornering. Traction control is similar to ABS, but for keeping slip below a certain
value, typically +(15..20)%.
Traction control can use different ways to control slip, using different actuators. One way is to reduce
engine torque, which reduces slip on both wheels on an axle if driven via differential. Another way is to
apply friction brakes, which can be done on each wheel individually. Vehicles with propulsion on
several axles can also redistribute propulsion from one axle to other axles, when the first tends to slip.
Vehicles with transversal differential clutch or differential lock can redistribute between left and right
wheel on one axle.
3.5.2.6 Engine Drag Torque Control, EDC *

Function definition: Engine Drag Torque Control prohibits over-braking of the driven axle(s) while
engine-braking. EDC uses both friction brakes and propulsion system as actuators.
The purpose of Engine Drag Torque Control is as the purpose of ABS, but the targeted driving situation
is when engine braking at low road friction, when engine drag torque otherwise can force the wheels
to slip too much negative. Similarly to ABS, it keeps the slip above a certain negative value. However, it
does it by increasing the engine torque from negative (drag torque) to zero (or above zero for a short
period of time).
3.5.2.7 Automatic Emergency Brake, AEB *

Function definition: Automatic Emergency Brake decelerates vehicle without driver having to use brake
pedal when probability for forward collision is predicted as high.
The purpose of AEB is to eliminate or mitigate collisions where subject vehicle collides with a lead
vehicle. AEB uses friction brake system as actuator, up to full brake which would be typically 10 m/s2.
An AEB system is often limited by that it cannot be designed to trigger too early, because driver would
be disturbed or it could actually cause accidents. Therefore, in many situations, AEB will rather
mitigate than avoid collisions.
Conceptually, an AEB algorithm can be assumed to know physical quantities as marked in Figure 3-52.
The quantity time-to-collision, TTC, can then be defined as 𝑇𝑇𝐶 = 𝑥𝑜 ⁄( 𝑠 𝑜 ), which means the
time within a collision will appear if no velocities changes. Subscript 𝑠 and 𝑜 means subject and object
vehicle, respectively. Also, one can define enhanced time-to-collision measure, eTTC, which takes into
account the present accelerations of the lead vehicle. TTC and eTTC can also be drawn in PVA-time-
diagram, see Figure 2-46. TTC shall not be mixed up with “time gap” (TG), which is the time when
subject vehicle will reach the present position of the lead vehicle, 𝑇𝐺 = 𝑥𝑜 ⁄ 𝑠 ;.
AEB function shall, continuously, decide whether or not to trigger AEB braking. AEB shall intervene by
braking when driver can be assumed to collide without intervention. If no other information, this can
be predicted as when driver can NOT avoid by normal driving. Avoidance manoeuvres that have to be
considered are (normal) deceleration and (normal) lateral avoidance to the left and to the right. What
to assume as normal driving is a question of tuning; here the following limits are used | ̇𝑠 | < 𝑠 𝑛 =
. . 4 ⁄𝑠 and | ̇𝑠 | < 𝑠 𝑛 = . . 6 ⁄𝑠 . A physical model based algorithm can start from the
following simple models.
• Normal deceleration ( ̇𝑠 = 𝑠 𝑛 = 4 ⁄𝑠 ) leads to collision if:
min(𝑥𝑜 (𝑡)) < ⇒ min(𝑥𝑜 + 𝑜 ∙ 𝑡 ( 𝑠 ∙ 𝑡 + 𝑠 𝑛 ∙ 𝑡 ⁄ )) < ⇒
𝑡>0 𝑡>0
⇒ (𝑥𝑜 + 𝑜 ∙𝑡 ( 𝑠 ∙𝑡+ 𝑠 𝑛 ∙ 𝑡 ⁄ ))|𝑡=𝑣𝑜𝑥−𝑣𝑠𝑥 < ⇒
𝑎𝑠𝑥𝑛
𝑜 𝑠𝑥 − 𝑜𝑥 𝒗𝒔𝒙 −𝒗𝒐𝒙
⇒ 𝑥𝑜 ∙(− 𝑠𝑥𝑛 )
∙( 𝑠 𝑜 ) < ⇒ < ∙(− 𝑠𝑥𝑛 )
⇒ 𝑪< ;
𝑠𝑥 − 𝑜𝑥 𝟖
150
• Normal avoidance to the left ( ̇𝑠 = 𝑠 𝑛 =6 ⁄𝑠 ) leads to collision if:
𝑤 𝑡 2 𝑤
(𝑦𝑜 (𝑡) + )| < ⇒ (𝑦𝑜 𝑠 𝑛∙ + )| 𝑥𝑜 < ⇒
𝑜 =0 𝑡=
𝑣𝑠𝑥 −𝑣𝑜𝑥
𝑜 𝑜𝑙 +𝑤 0.6+ .8
⇒ <√ ∙ = { . .} = √ ≈ .4 𝑠 ⇒ 𝑪 < 𝟎. 𝟒 𝒔;
𝑠𝑥 − 𝑜𝑥 𝑠𝑦𝑛
• Normal avoidance to the right ( ̇𝑠 = 𝑠 𝑛 = 6 ⁄𝑠 ) leads to collision if:

𝑜 − 𝑜𝑟 −𝑤⁄
⋯ ⇒ <√ ∙ =⋯ ⇒ 𝑪 < 𝟎. 𝟒 𝒔;
𝑠𝑥 − 𝑜𝑥 𝑠𝑦𝑛
• Assuming that the AEB intervention decelerates the vehicle with ̇𝑠 = 𝑠 𝐴𝐸𝐵 = 8 ⁄𝑠 , a
forward collision can be avoided if AEB intervenes AND if:
𝑜 𝒗𝒔𝒙 −𝒗𝒐𝒙 − 𝒗 −𝒗
⋯⇒ −
> 𝟐∙(−𝒂 )
= 𝑠𝑥 6 𝑜𝑥 ⇒ 𝑪 > 𝒔𝒙𝟏𝟔 𝒐𝒙 ;
𝑠𝑥 𝑜𝑥 𝒔𝒙𝑨𝑬𝑩
Figure 3-52 shows a diagram where different condition areas are marked. The sectioned area shows
where AEB will be triggered, using above rules. The smaller of the sectioned areas shows where it also
will be possible to trigger AEB so timely that a collision is actually avoided; with the assumed numbers,
this is for speeds up to 6.4 m/s≈23 km/h.
𝑥𝑜
(or lead vehicle)
= 𝑇𝑇𝐶
object vehicle
𝑥𝑜 .4 𝑠
Collision if normal
lateral avoidance
𝑦𝑜 𝑚
( 𝑠 = 6 2)
𝑦𝑜 𝑠
3. 𝑠 6.4 𝑠 𝑜
AEB triggered AND avoids collision

(or own vehicle)
𝑤⁄ 𝑤⁄
subject vehicle
AEB triggered AND mitigates (not avoids) collision

“Mild AEB” or FCW might be triggered, since too late
𝑠
for normal lateral avoidance
“Mild Automatic Lane Change” or FCW might be triggered,
since too late for normal longitudinal deceleration
Figure 3-52: Left: Quantities known for an AEB algorithm in the subject vehicle, assuming “symmetrically
behind” (𝑦𝑜 = 𝑦𝑜 ). Right: Model based decision of triggering AEB and effectiveness of AEB if triggered.
The reasoning above is very simplified. Additional information can improve effectiveness of AEB, such
as knowing if a lateral avoidance on one side of object vehicle is blocked, the acceleration of the object
vehicle, pedal operation of subject vehicle, road gradient, road curvature and road friction is. The
vehicle dynamics model used is simply a point mass with predicted constant velocity and certain
assumed acceleration capability, which of course can be extended a lot; both with taking actual
accelerations into account and more advanced vehicle dynamics models. The decision to intervene is
dependent of many pieces of information and simple models; vehicle dynamics and driver behaviour
(in both subject and object vehicles) as well as road characteristics. AEB function has to be designed
together with other similar functions, such as ACC and Forward Collision Warning (FCW).
AEB is on market and legal requirement for both passenger vehicles and heavy vehicles.
Related functions are, e.g. extra force assistance in brake pedal when driver steps quickly onto brake.
Another related function is automatic braking triggered by a first impact and intended to mitigate or
avoid secondary accident events, starts to appear at market, see Reference (Yang, 2013). In semantic
151
meaning, this could be seen as AEB, but they are normally not referred to as AEB; AEB normally refers
to functions that use environment sensors (forward directed radar, camera, etc.).
When designing and evaluating AEB, it is important to also know about the function Forward Collision
Warning, FCW. FCW is a function that warns the driver via visual and/or audio signals when a forward
collision is predicted. FCW is typically triggered earlier than AEB.
3.5.3 Longitudinal Motion Functionality in a

reference architecture
All control functions controls have to cooperate and they have to be transferable between platforms
and vehicle variants. It is very complex to take all functions into consideration, but with a scope
limited to the longitudinal Motion functionality
Figure 3-53 can be drawn as a solution within the reference architecture.
By using a reference architecture, it can be illustrated that Adaptive Cruise control and cruise control
can be seen as part of Traffic Situation Layer (ACC=CC if no vehicle ahead). The Traffic Situation Layer
has the purpose and scope to understand the ego vehicle’s surrounding traffic by looking at e.g.
Forward Sensors. The forward-looking sensor is in this case part of Vehicle Environment sensors.
Vehicle Motion and Coordination Layer would include the arbitration of Driver’s Acceleration and
Brake pedal input and Traffic functionality, see Figure 3-53. In addition, Vehicle Motion and
Coordination Layer would perform the powertrain coordination and brake distribution. The
coordinated requests are then sent to Motion Support Device Layer.
The Human Machine Interface would include the services available for the driver to activate or
request, E.g. ACC activation to Traffic Situation or Deceleration by pressing the brake pedal.
Vehicle Traffic Situation Layer Human

Forrward Direction
Enivormnent Machine
Dist Dist. Ctrl (ACC) Interface
AEB Speed Ctrl (CC) ACCbtn
Maps
… Dist btn
FwdSens
Arb(Min)
APed
Vehicle Motion and Coordination Layer Arb(sum) APedInterp
Speed Arb (prio BPed, BPed
BPedInterp
else Max)
Energy Managament
Powertrain Coord ABS &
TC ESC RSC
EBD
Arb(min)
Arb(min) Arb
Coord
Coord
Arb (switching) Arb (switching)
Motion Support Device Layer Legend

Vehicle Border of Vehicle
EngAct GboxAct RearBrkAct Level Functionality
Motion FrntBrkAct
Sens
SW connection (signals,
here 2 scalar signals)
HW connection
Sensors & Actuators
End Customer Function
Arbitration
Coordination
Figure 3-53: Functional architecture for conventional front axle driven passenger car. Mainly longitudinal
functions (plus ESC, RSC) are shown, e.g. no steering. Cf Figure 1-22.
If a reference architecture is used, it can assist function developers from OEM’s Electrical, Powertrain,
and Chassis departments and suppliers to have a common view of how vehicle’s embedded motion
functionality is intended to be partitioned and to understand how different functions relate and
interact with each other and what responsibilities they have.
152
LATERAL DYNAMICS
4 LATERAL DYNAMICS
4.1 Introduction
The lateral motion of a vehicle is needed to follow the roads’ curves and select route in intersections as
well as to laterally avoid obstacles that appear. The vehicle needs to be steerable. With some
simplification, one can say that lateral dynamics is about how steerable the vehicle is for different
given longitudinal speeds. Vehicle steering is studied mainly through the vehicle degrees of freedom:
yaw rotation and lateral translation.
A vehicle can be steered in different ways:
• Applying steering angles on, at least one, road wheel. Normally both of front wheels are
steered with approximately same angle.
• Applying longitudinal forces on road wheels; directly by unsymmetrical between left and right
side of vehicle, e.g. one sided braking, or indirectly by deliberately use up much friction
longitudinally on one axle in a curve, so that that axle loses lateral force.
• Articulated steering, where the axles are fixed mounted on the vehicle body, but the vehicle
itself can bend.
The chapter is organised with one group of functions in each section as follows:
• 4.2 Low speed manoeuvers
• 4.3 Steady state cornering at high speed
• 4.4 Stationary oscillating steering
• 4.5 Transient handling
• 4.6 Lateral Control Functions
Most of the functions in “4.6 Lateral Control Functions”, but not all, could be parts of ”4.5 Transient
handling”. However, they are collected in one own section, since they are special in that they partly
rely on software algorithms.
The lateral dynamics of vehicles is often experienced as the most challenging for the new automotive
engineer. Longitudinal dynamics is essentially motion in one plane and rectilinear. Vertical dynamics
may be 3 dimensional, but normally the displacements are small and in this compendium the vertical
dynamics is mainly studied in one plane as rectilinear. However, lateral dynamics involves motion in
the vehicle coordinate system which introduces curvilinear motion since the coordinate system is
rotating as the vehicle yaws.
The turning manoeuvres of vehicles encompass two concepts. Handling is the driver’s perception of
the vehicle’s response to the steering input. Cornering is usually used to describe the physical
response (open-loop) of the vehicle independent of how it influences the driver.

• “Chapter 25 Steering System” in Reference (Ploechl, 2013).
• “Chapter 27 Basics of Longitudinal and Lateral Vehicle Dynamics” in Reference (Ploechl, 2013).
• “Chapter 8: Electronic Stability Control” in Reference (Rajamani, 2012)
4.2 Low speed manoeuvers

This section is about operating vehicles in low speeds, including stand-still and reverse. Specific for
low speed is that inertial effects can be neglected, i.e. one can assume that left hand side in motion
equation (𝑀 𝑠𝑠 ∙ 𝑡 𝑟 𝑙𝐴 𝑙 𝑟 𝑡 𝑜𝑛 = 𝑠𝑢 𝑜 𝑜𝑟 𝑠) is zero.
153
LATERAL DYNAMICS
In low speed, one often needs to find the path with orientation and understand the steering system
and how tyres can be modelled to track ideally. This, and the resulting one-track model for low speeds,
is described in Sections 4.2.1, …,4.2.6.
4.2.1 Path with orientation

The path and path with orientation was introduced y
in Section 1.5. The path, in global coordinate 𝜔
system, is related to vehicle speeds, in vehicle fix
coordinates, as given in Figure 4-1 and Equation
[4.1].
Knowing ( (𝑡) (𝑡) 𝜔 (𝑡)), we can determine
“path with orientation” (𝑥(𝑡) 𝑦(𝑡) (𝑡)), by time x
integration of the right hand side of the equation.
Hence, the positions are typically “state variables” Figure 4-1: Model for connecting “path with
in lateral dynamics models. orientation” to speeds in vehicle coordinate
system.
𝑥̇ cos( ) sin( )
[ ]=[ ]∙[ ];
𝑦̇ sin( ) cos( ) [4.1]
̇ =𝜔 ;
It should be noted that in some problems, typically manoeuvring at low speed, the real time scale is of
less interest. Then, the problem can be treated as time independent, e.g. by introducing a coordinate, s,
along the path, as in Equation [4.2].
𝑥 = ∙ cos( ) ∙ sin( );
𝑠̇ 𝑠̇
𝑦 = ∙ cos( ) + ∙ sin( ) ;
𝑠̇ 𝑠̇ [4.2]
𝜔
= ;
𝑠̇
𝑤ℎ 𝑟 𝑝𝑟 𝑛𝑜𝑡 𝑠 𝑟 𝑛𝑡 𝑡 𝑜𝑛 𝑤 𝑡ℎ 𝑟 𝑠𝑝 𝑡 𝑡𝑜 𝑠
Here, 𝑠̇ can be thought of like an arbitrary time scale, with which all speeds are scaled. One can
typically chose 𝑠̇ = 1 [ ⁄𝑠]. However, in this compendium we will keep notation t and the dot notation
for derivative.
4.2.2 Vehicle and wheel orientations

For steered wheels, there are often reason to translate forces and velocities between vehicle
coordinate system and wheel coordinate system, see Figure 4-2 and Equation [4.3].
Transformation from wheel coordinates to vehicle coordinates:

cos( ) sin( ) 𝑤 cos( ) sin( ) 𝑤
[ ]=[ ]∙[ ]; 𝑛 [ ]=[ ]∙[ ];
sin( ) cos( ) 𝑤 sin( ) cos( ) 𝑤
[4.3]
Transformation from vehicle coordinates to wheel coordinates:
𝑤 cos( ) sin( ) 𝑤 cos( ) sin( )
[ ]=[ ]∙[ ]; 𝑛 [ ]=[ ]∙[ ];
𝑤 sin( ) cos( ) 𝑤 sin( ) cos( )
154
LATERAL DYNAMICS
Velocity of body, Components in vehicle Components in wheel

above this wheel coordinates coordinates
𝒗𝒙𝒚 𝒗𝒙𝒚 𝒗𝒙𝒗 𝒗𝒙𝒚 𝒗𝒙𝒘

𝒗𝒚𝒗 𝒗𝒚𝒘
Force in road plane, Components in vehicle Components in wheel

acting on this wheel coordinates coordinates
𝑭𝒙𝒚 𝑭𝒙𝒚 𝑭𝒙𝒗 𝑭𝒙𝒚 𝑭𝒙𝒘
𝑭𝒚𝒗 𝑭𝒚𝒘
Figure 4-2: Transformation between forces and velocities in vehicle coordinate

system and wheel coordinate system.
4.2.3 Steering System

The steering system is here referred to the link between steering wheel and the road wheel’s steering,
on the steered axle. It is normally the front axle that is steered. Driver’s interaction is two-folded, both
steering wheel angle and torque, which is introduced in Section 2.9. In present section, we will focus
on how wheel steering angles are distributed between the wheels.
4.2.3.1 Chassis steering geometry

The most basic intuitive relation between the wheels steering angles is probably that all wheels
rotation axes always intersect in one point. This is called Ackermann geometry and is shown in Figure
4-3. The condition for having Ackermann geometry is, for the front axle steered vehicle that:
1 𝑅 𝑤⁄
= ;
tan( ) 1 1 𝑤
⇒ = + ; [4.4]
1 𝑅 + 𝑤⁄ tan( 𝑜 ) tan( )
= ;
tan( 𝑜 ) }
The alternative to Ackermann steering geometry is parallel steering geometry, which is simply that
= 𝑜 . Note that Ackermann geometry is defined for a vehicle, while parallel steering is defined for
an axle. This means that, for a vehicle with 2 axles, each axle can be parallel steered, which means that
the vehicle is non-Ackermann steered. However, the vehicle can still be seen as Ackermann steered
with respect to mean steering angles at each axle.
For low-speed, Ackermann gives best manoeuvrability and lowest tyre wear. For high-speed, Parallel
is better in both aspects. This is because vehicles generally corner with a drift outwards in curves,
which means that the instantaneous centre is further away than Ackermann geometry assumes, i.e.
more towards optimal for parallel. Hence the chosen geometry is normally somewhere between
Ackermann and parallell.
155
LATERAL DYNAMICS
Practical arrangement to design the steering geometry is shown in Figure 4-7. The design of linkage
will also make the transmission from steering wheel angle to road wheel steering angle non-linear.
This can lead to different degrees of Ackerman steering for small and large steering wheel angles.
di do 1 turn centre
a Ackermann error, front.
b Ackermann error, rear.
1 tan(d o )  1 tan(d i )  w L
L
Common intersection of all
wheels’ axes of rotation
Rr w
Figure 4-3: Ackermann steering geometry. Left: One axle steered. Right: Both axles steered and including
“Ackermann errors”. From (ISO 8855).
In traditional steering systems, the steering wheel angle has a monotonically increasing function of the
steering angle of the two front axle road wheels. This relation is approximately linear with a typical
ratio of 15..17 for passenger cars. For trucks the steering ratio is typically 18..22. In some advanced
solutions, steering on other axles is also influenced (multiple-axle steering, often rear axle steering).
There are also solutions for dynamically adding steering angle through a planetary gear and electric
angle-controlled motor on the steering shaft, so called Active Front Steering (AFS).
In reference (Tagesson, 2017), there is a good descriptive chapter about steering systems for heavy
vehicles.
4.2.3.2 Steering system forces

(This section has large connection with Section 2.4.6.2 Tyre aligning moment.)
The steering wheel torque, 𝑇𝑠𝑤 , should basically be a function of the tyre/road forces, mainly the
wheel-lateral forces. This gives the driver a haptic feedback of what state the vehicle is in. The
torque/force transmission involves a servo actuator, which helps the driver to turn the steering
system, typically that assists the steering wheel torque with a factor varying between 1 and 10, but
less for small 𝑇𝑠𝑤 (highway driving) than large 𝑇𝑠𝑤 (parking), see Figure 4-4. Here, the variation in
assistance is assumed to be hydraulic and follows a so-called boost curve. At 𝑇𝑠𝑤 = , the assistance is
≈0.45/0.55≈1 and for 𝑇𝑠𝑤 = 4 Nm, it is ≈0.9/0.1≈10.
Rack
Force Assisted
Assistance
UnAssisted
SteWhlTq
Figure 4-4: Left: Boost Curve with different working areas depending on the driving envelope. Middle:
Torque distribution between manual torque, FM, and assisting torque, FA, depending on applied steering
wheel torque. From Reference (Rösth, 2007). Right: Unassisted and assisted steering wheel torque.
156
LATERAL DYNAMICS
For vehicle dynamics, one important effect of a steered axle, is that the lateral force on the axle tries to
align the steering in the direction that the body (over the steered axle) moves, i.e. towards a zero tyre
side slip. This is designed in via the sign of the caster trail, see Figure 4-5. Also, asymmetry in
longitudinal tyre forces (wheel shaft torques and/or brake torques) affects the steering wheel torque.
This is analysed in the following.
Steering-axis inclination angle, Wheel 1 with

KPI Caster (friction)
angle, CA brake brake
View from View from
torque, Tb:
rear: right: (Tb<0 if w>0) Tb
bearing
w
forward
Fx1=Tb/R
Castor offset at
wheel center
(drawn positive)
Wheel 1 with
Fx1 shaft torque, Ts:
Castor offset at ground, c, R

(drawn positive)
Steering axis offset
at ground, s Ts
(drawn positive)
(sometimes called Scrub radius) Fy1
Ts*sin(KPI)
Fy1 = resulting
lateral force
Pneumatic trail, t Fx1=Ts/R
(positive if “behind”) lateral force
distribution
Figure 4-5: Steering geometry and offsets.

On a steered axle, there is a Steering axis (or Kingpin axis) around which the wheel is rotated when it is
steered. The steering axis intersects with the ground plane at a point which normally has an offset
from the contact patch centre, both longitudinally and laterally. The offsets are called Castor offset at
ground and Steering axis offset at ground, respectively, see Figure 4-5. (Steering axis offset at ground is
sometimes called Scrub radius, but that is also used for the resulting distance of both offsets, so it is an
ambiguous name.) This figure also defines Normal steering axis offset at wheel centre . We will use the
3 latter measures to explain why steering is affected by differences in shaft torque left/right,
differences in brake torque left/right, and lateral wheel forces. The steering axis distance from wheel
centre in side view, is called Castor offset at wheel centre .
Often, the actual forces between tyre and ground are not in the exact centre of the nominal contact
patch, which also creates “effective” version of caster offset and scrub radius. For instance, the
Pneumatic trail adds to the Castor offset at ground due to the lateral force distribution being
longitudinally offset from nominal contact point; so that the lever becomes not only , but + 𝑡. The
moment, 𝑇𝑠𝑡𝑒𝑒 , on steering system (around Steering axis, turning towards increasing steering angle) is
affected by tyre forces on a steered axle as in Equation [4.5].
Castor offset at ground is, on passenger vehicles, 15-20 mm (at motorcycles approximately 100 mm).
On rear wheel driven passenger vehicles it can typically be 5 mm, due to higher Castor Angle which
gives a beneficial higher Camber angle gain at cornering. On a front wheel driven passenger vehicle a
non-zero caster offset is not chosen due to drive axle lateral displacement. Caster offset gives a self-
aligning steering moment, which generally improves the steering feel.
157
LATERAL DYNAMICS
(Notations in Equation [4.5] are defined in Figure 4-5.)
(𝑇 + 𝑇𝑠 ) 𝑇𝑠
𝑇𝑠𝑡𝑒𝑒 = ∙ 𝑠 cos(𝐾 𝐼) + 𝑅 sin(𝐾 𝐼) +
𝑅 𝑅
(𝑇 + 𝑇𝑠 ) 𝑇𝑠
+ 𝑠 cos(𝐾 𝐼) + 𝑅 sin(𝐾 𝐼) +
𝑅 𝑅 [4.5]
+( 1) ( ( +𝑡 )+ ( + 𝑡 )) cos(𝐶𝐴) =
𝑇 𝑇 𝑇𝑠 𝑇𝑠
= 𝑠 cos(𝐾 𝐼) + 𝑘 ( ( +𝑡 )+ ( + 𝑡 )) cos(𝐶𝐴) ;
𝑅 𝑅
The equation shows that difference in both brake torque and shaft torque affects steering and so does
the sum of lateral forces. For reducing torque steer and disturbances from one-sided longitudinal
forces due to road irregularities, kingpin offset, scrub radius different road friction should be as small
as possible, but it is limited by geometrical conflicts between brake disc, bearing, damper, etc.
Positive scrub radius contributes to self-centring, thanks to lifting the car body, see below. Negative
scrub radius compensates for split- braking, or failure in one of the brake circuits. Hence, the scrub is
a balance between these two objectives. Scrub radius is often slightly negative on modern passenger
cars. Scrub radius is often positive on trucks, maybe 10 cm, due to packaging.
The geometry in Figure 4-6 shows one part of the lifting effect. This is that if steering angle is changed
from zero, it lifts the vehicle slightly which requires a steering torque. One can see this as a kind of
“return spring effect”.
Figure 4-6 shows how KPI and scrub radius causes the vehicle body to lift a distance 𝑠 = 𝑠 ∙
cos(𝐾 𝐼) ∙ (1 cos(| 𝑠𝑡𝑒𝑒 |)) ∙ sin(𝐾 𝐼). This will require a work 𝑇𝑠𝑡𝑒𝑒 ∙ 𝑠𝑡𝑒𝑒 = ∙ 𝑠 . This leads
to an 𝑇𝑠𝑡𝑒𝑒 (additional to Eq [4.5]) as follows:
∙𝑠
( 𝑡 𝑜𝑛 𝑙) 𝑇𝑠𝑡𝑒𝑒 = ∑ =
= 𝑒 𝑡 𝑛𝑑 ℎ𝑡 𝑠𝑡𝑒𝑒
[4.6]
𝑠 ∙ cos(𝐾 𝐼) ∙ (1 cos(| 𝑠𝑡𝑒𝑒 |)) ∙ sin(𝐾 𝐼)
= ∙ 𝑒 ;
𝑠𝑡𝑒𝑒
It should be noted that Eq [4.6] is not complete with respect to all “returning effects”. There are also
effects from Castor angle and Caster trail as well as that the tyre has a width and radius. However, in
total, these give rise to a returning steering torque which is depending on the steering angle.
It can also be noted that steering effort for low speed or stand-still is largely influenced by whether
brakes are applied or not, due to the magnitude (not sign) of scrub radius.
Contact point Contact point

𝑇𝑠𝑡𝑒𝑒 tyre/ground for 𝒔𝒕𝒆𝒆 𝟎
sin 𝐾 𝐼
for 𝒔𝒕𝒆𝒆 = 𝟎
View from 𝑠𝑡𝑒𝑒
rear: s
KPI
=𝑠
𝑠
Fx1
Steering axis offset 𝑠𝑡𝑒𝑒

at ground, s
Figure 4-6: Lift effect due to steering angle and positive Steering axis offset at ground.
158
LATERAL DYNAMICS
4.2.4 One-track models

When studying the vehicle cornering response, it is useful to combine the effects of all tyres on the axle
into one virtual tyre. This assumption, referred to as the one-track model (or single-track model or
bicycle model), facilitates understanding but can also capture most important phenomena. A one-track
model of a two-axle vehicle is shown in Figure 4-8. One-track model for truck with trailer is
exemplified in Figure 4-9.
Ackermann
steering
(trapezoidal 𝑠𝑤
geometry)
(Erasmus Darwin 1758,
Rudolph Ackermann
1810.)
ℎ𝑡
𝑒 𝑡
= 𝑢𝑛 𝑝𝑜𝑠 𝑡 𝑜𝑛
𝑒 𝑡 𝑐𝑘 =
= 𝑢𝑛 𝑟 𝑐𝑘 ∙ 𝑠𝑤 ;
ℎ𝑡 = 𝑢𝑛 𝑝𝑜𝑠 𝑡 𝑜𝑛 𝑐𝑘 =
= 𝑢𝑛 𝑟 𝑐𝑘 ∙ 𝑠𝑤 ;
Figure 4-7: Example of Trapezoidal Steering. Left: Conceptual use of steering arms. Right: More exact design,
common today. From Gunnar Olsson, LeanNova.
di df do
lf
L
lr
Common
intersection of all
wheels’ axes of
rotation Rr
Figure 4-8: Collapsing to one-track model.
159
LATERAL DYNAMICS
Figure 4-9: Upper: Two-track model of 3-axle truck with 2-axle trailer. Lower: One-track model of a 3-axle
truck with a 2+5 axle trailer. Note: SAE coordinate system used. (Kharrazi , 2012).
Phenomena which one-track models not capture are, e.g.:
• Deviations from Ackerman geometry within an axle.
• Varying axle cornering stiffness due to lateral load shift (4.3.5.2) and axle propulsion/braking
(4.3.5.6).
• Added yaw moment due to left/right-asymmetric wheel torque, such as ESC interventions.
4.2.5 Ideally tracking wheels and axles

In Section 2.4, lateral slip models for tyres were introduced. It is the constitutive equation = 𝐶 ∙
𝑠 . If the tyre force, , is small compared to cornering stiffness, 𝐶 , one would expect side slip, 𝑠 , to be
very small. If this assumption is taken to its extreme, one get something we can call “ideal tracking” for
a tyre or for an axle, which is that side slip is zero and side force can be any (finite) value. For low
speed, this is often a good enough model of the tyre. The validity of this assumption is limited by when
lateral force becomes large with respect to available lateral force, ∙ . Hence, the speed border
between low speed and high speed is lowered on low road friction.
Models using ideally tracking wheels are sometimes referred to as “kinematic models”. However, a
strictly kinematic model does not contain the forces at all, why validity cannot be checked.
The assumption about ideally tracking wheel or axle is a constitutive assumption, although that the
equation (tyre lateral speed=tyre side slip=0) does not couple force and speed, but only stipulates
speed. If forces should be calculated, it has to be done using the equilibrium.
4.2.6 One-track model for low speeds, with

Ackermann geometry
Low speed manoeuvres are characterised by that the inertial forces are neglected, i.e. ∙ = . If the
geometry is according to Ackermann, it is reasonable to assume ideally tracking axles. We can model
160
LATERAL DYNAMICS
this as a constitutional relation with infinite cornering stiffness: Eq [2.36] ⇒ = 𝐶 𝑠 ∞ 𝑠 ;⇒

𝑠 = ; ⇒ 𝑤 = ;.
4.2.6.1 Geometric model (without forces)

The geometric effects of this is that the intersection point of the wheels rotational axes coincides with
the instantaneous centre of vehicle rotation in ground plane. This directly relates steering angles and
path radius to each other. For the model in Figure 4-8 this connection becomes:
tan( ) = ;
𝑅 }⇒ = arctan ∙ si n(𝑅) ≈ ∙ si n(𝑅) ≈ ;
𝑅 [4.7]
𝑅 =𝑅 +𝑙 ; 𝑙 ) √𝑅
√𝑅 𝑙
(
𝑤ℎ 𝑟 𝑅 > 𝑛𝑠 𝑡ℎ 𝑡 𝑛𝑠𝑡 𝑛𝑡 𝑛 𝑜𝑢𝑠 𝑛𝑡𝑟 𝑜 𝑟𝑜𝑡 𝑡 𝑜𝑛 𝑠 𝑙 𝑡 𝑜 𝑡ℎ ℎ 𝑙
4.2.6.2 Model with forces

The model in Eq [4.7] predicts a motion without involving forces, or actually assuming forces are zero.
To get a more complete model, where more variables can be extracted, we can set up the model in
Figure 4-10.
speeds: 𝛼 = arctan 𝑠 𝛽 = arctan forces:
vy 𝑤
Ffxv ⃗
⃗
wz
Ffyv
vry=0
vfyv
Fry
vx Frx df
vrx vfxv
⃗ ⃗
lr lf lr lf
L L
Figure 4-10: One-track model with ideally tracking axles. Lower view of front wheel shows conversion
between wheel and vehicle coordinate systems.
The “physical model” in Figure 4-10 gives the following “mathematical model”:
Equilibrium (longitudinal, lateral and yaw-rotational):

= + ;
= + ;
= ∙𝑙 ∙𝑙 ;
Transformation between vehicle and wheel coordinate systems:
= 𝑤 ∙ cos( ) 𝑤 ∙ sin( );
= 𝑤 ∙ sin( ) + 𝑤 ∙ cos( );
[4.8]
= 𝑤 ∙ cos( ) 𝑤 ∙ sin( );
= 𝑤 ∙ sin( ) + 𝑤 ∙ cos( );
Compatibility between CoG and axles:
= ; 𝑛 = +𝑙 ∙𝜔 ;
= ; 𝑛 = 𝑙𝑟 ∙ 𝜔 ;
Ideal tracking (Constitutive relation, but without connection to forces):
𝑤 = ; 𝑛 = ;
161
LATERAL DYNAMICS
Path with orientation (compatibility):

𝑥̇ = ∙ cos( ) ∙ sin( );
𝑦̇ = ∙ cos( ) + ∙ sin( );
̇ = 𝜔;
Controls (driver or actuation):
= (35 ∙ 𝜋⁄18 ) ∙ sin( .5 ∙ ∙ 𝜋 ∙ 𝑡) ; 𝑡 < 4.5;
={
= 35 ∙ 𝜋 18 ; 𝑙𝑠
Constitution (Rear axle undriven, which gives drag from roll resistance):
= 1 ;
The “Compatibility” in Eq [4.8] neglects the influence of steering axis offsets at ground, see Section
4.2.3.2. The terms neglected are of the type 𝑡 𝑟 𝑙 𝑠 𝑡 ∙ ̇ ; in the equation for and
𝑜𝑛 𝑡𝑢 𝑛 𝑙 ̇
𝑠 𝑡 ∙ in the equation for . This is generally well motivated for normal road
vehicles, except for quick steering when vehicle is stand-still.
Equation [4.8] is written in Modelica format in Equation [4.9]. Comments are marked with //. The
subscript v and w refers to vehicle coordinate system and wheel coordinate system, respectively. The
actual assumption about ideal tracking lies in that 𝑤 = = . Global coordinates from Figure 4-1
is also used.
//Equilibrium:
0 = Ffxv + Frx;
0 = Ffyv + Fry;
0 = Ffyv*lf - Fry*lr;
//Ideal tracking (Constitutive relation, but without connection to forces):

vfyw = 0; vry = 0;
//Compatibility:
vfxv = vx; vfyv = vy + lf*wz;
vrx = vx; vry = vy - lr*wz;
//Transformation between vehicle and wheel coordinate systems:

Ffxv = Ffxw*cos(df) - Ffyw*sin(df); [4.9]
Ffyv = Ffxw*sin(df) + Ffyw*cos(df);
vfxv = vfxw*cos(df) - vfyw*sin(df);
vfyv = vfxw*sin(df) + vfyw*cos(df);
//Path with orientation:

der(x) = vx*cos(pz) - vy*sin(pz);
der(y) = vy*cos(pz) + vx*sin(pz);
der(pz) = wz;
// Prescription of actuation:
df = if time < 4.5 then (35*pi/180)*sin(0.5*2*pi*time) else 35*pi/180;
//Rear axle undriven, which gives drag from roll resistance:
Frx = -100;
The longitudinal speed is a parameter, = 1 𝑘 ℎ. A simulation result from the model is shown in
Figure 4-11. It shows the assumed steering angle function of time, which is an input. It also shows the
resulting path, 𝑦(𝑥). The variables 𝑥 𝑦 𝑝 are the “state variables” of this simulation.
162
LATERAL DYNAMICS
df [rad] y [m]
35 deg
x[m]
time [s]
Figure 4-11: Simulation results of one-track model with ideal tracking.
The variables 𝑥 𝑦 𝑝 are the only “state variables” of this simulation. If not including the path model
(Equation [4.1]), the model would actually not be a differential equation problem at all, just an
algebraic system of equations. That system of equations could be solved isolated for any value of
steering angle without knowledge of time history. These aspects are the same for the steady state
model in section 4.3.2.
A driving resistance of 100 N is assumed on the rear axle ( = 1 ;). This is to exemplify that
forces do not need to be zero, even if forces normally are not so interesting for low speed manoeuvres.
Anyway, one should note that the speeds and forces are weakly coupled, since there are no inertial
forces modelled. The modelling of forces is more important for the case where we don’t have
Ackerman geometry. For a two-axle one-track model we always have Ackermann geometry, because
there is always an intersection point between the front and rear wheels rotational axes.
4.2.7 Low speed functions

4.2.7.1 Turning circle *
Function definition: Turning diameter is the diameter of the smallest possible circular path obtained steady
state at low speed, measured to a certain point at the vehicle. The certain point can be either outer-most point on
wheel (Kerb Turning diameter) or outer-most point on body (Wall Turning diameter).
The end of the simulation in Figure 4-11 is made with constant steering angle. If we assume that it is
the maximum steering angle, the circle actually shows the turning circle (diameter) for centre of
gravity. If we add the path for the outermost wheels, we get the kerb turning circle diameter, see
Figure 4-12. If we add the path for the outermost point at the vehicle body we get the wall turning
circle diameter, also shown in Figure 4-12. The outermost point at the vehicle body is normally the
front outer corners of the vehicle body, i.e. the outer front bumper end.
If we have a vehicle with Ackermann geometry, it is tempting to model without involving forces, using
Eq [4.7]. But the forces from rolling resistance do influence, see Figure 4-10. When introducing rolling
resistance on the un-driven axle, the force equilibrium is obtained by counter-directed forces on the
two axles. Due to the steering angle, a lateral force on the front wheel is required, which gives a lateral
tyre slip, 𝛼 . This changes the motion compared to Figure 4-8. Road grade resistance influences in
same way and superimposes on rolling resistance.
163
LATERAL DYNAMICS
kerb turning diameter
wall turning diameter

y [m]
• CoG path diameter=8.54 m

• front right wheel path diameter=11.19 m
• real left wheel path diameter=6.29 m
x[m]
Figure 4-12: Adding paths for wheels and body points, on top of result in Figure 4-11.
Rear axle propulsion: Front axle propulsion:

𝑤 = 𝑜
= 𝐶 𝛼
⃗
𝛼
⃗ 𝛼
⃗ ⃗
= 𝑜 = = 𝐶 𝛼
𝑤 = 𝑜
𝑜 𝑤
Figure 4-13: Smaller turning circle diameter for front axle propulsion, as compared to rear axle propulsion
due to rolling resistance on the un-driven axle.
For heavy combination vehicles, these effects can be significant, adding also rolling resistance from
many towed units and non-Ackermann effects of several non-steered axles on some units. Then, the
function to turn can be quantified by required road friction on steered and driven axle, as opposed to
radius at maximum steering angle.
A turning circle can also be defined for high speed, but it is then more common to set requirement on
curvature gain. Curvature gain is not the curvature at maximum steering angle, but the curvature per
steering angle, see Section 4.2.9.
4.2.7.2 Swept path width *

Function definition: Swept path width is the distance between the outermost and innermost trajectories of
wheels (or body edges). The trajectories are then from a certain turning or lane change manoeuvre at a certain speed.
For manoeuvrability, there is a function which is complementary to turning radius diameter. It is
“Swept path width” (SPW), see Figure 4-12. It can occur as a kerb and wall version as well as turning
circle radius can. It is the distance between the outermost and innermost wheel/point on the vehicle.
The SPW should be as small as possible for improving manoeuvrability.
“Swept path width” is often used for low vehicle speeds. For higher speeds, it is more common to talk
about off-tracking, see next section.
164
LATERAL DYNAMICS
4.2.7.3 Off-tracking *
Function definition: Off-tracking is the distance between the outermost and innermost trajectories of centre
points of the axles. The trajectories are then from a certain turning or lane change manoeuvre at a certain speed.
Another measure of manoeuvrability is “Off-tracking”, see Figure 4-14. It is like swept path width, but
for the centre point of the axles. It is also used for higher speeds, and then the rear axle often tracks on
a larger radius than front axle.
Off-tracking is most relevant for vehicles with

several units, such as truck with trailer. Off-tracking
for driving several rounds in a circle (“Circle off- L
tracking”) is well defined since it is a steady state
(with respect to articulation angle states), but df
seldom the most relevant. Often, it is a more relevant
to set requirements on a “Manoeuvre Off-tracking”,
e.g. off-tracking when driving from straight, via curve
with certain outer radius, to a new straight in a
certain angle from the first straight, e.g. 90 degrees
and 12.5 m is common. A way to predict off-tracking
for this is to simulate with time integration. The
states in such simulation are the path coordinates
with orientation (𝑥 𝑦 ) and articulation angles
( …).
Figure 4-14: Adding one-track model, on top
of result in Figure 4-11.
From geometry in Figure 4-14 on can find an expression for off-tracking at low speed:
∆(𝑅 )=𝑅 𝑅 =𝑅 √𝑅 ;
[4.10]
∆( )=𝑅 𝑅 = ⁄sin( ) ⁄tan( ) ;
4.2.7.4 Steering effort *

Function definition: Steering effort at low speed is the steering wheel torque needed to turn the steering
wheel a certain angle at a certain angular speed at vehicle stand-still on high road friction.
Function definition: Steering effort at high speed is the steering wheel torque (or subjectively assessed effort)
needed to perform a certain avoidance manoeuvre at high road friction.
At low or zero vehicle speed, it is often difficult to reach a low steering wheel torque, due to:
• Caster offset in Figure 4-5 gives the wheel a side slip when steering and hence a tyre lateral
force is developed. Tyre lateral forces times caster offset increases the steering wheel torque.
• Additionally, there is a spin moment in the contact patch, 𝑀𝑍𝑇 in Figure 2-2. It does not
influence very much, except for at very low vehicle speed, which is why quantitative models for
𝑀𝑍𝑇 are not presented in this compendium.
A critical test for steering effort at low speed is to steer a parked vehicle with a certain high steering
wheel rotational speed, typically some hundred 𝑠. The steering wheel torque is then required to
stay under a certain design target value, normally a couple of Nm. The torque needed will be
dependent on lateral force, spin moment and steering geometry (which is not steering speed
dependent) and dependent on the capability of the power steering system (which is dependent on
steering, due to delays in the steering assistance actuator). A failure in this test is called “catch-up”,
referring to that driver catches up with the power steering system. It can be felt as a soft stop and
measured as a step in steering wheel torque.
165
LATERAL DYNAMICS
At higher vehicle speeds, the steering effort is normally less of a problem since unless really high
steering wheel rate. Hence, steering wheel torque in avoidance manoeuvres in e.g. 70 km/h can be a
relevant requirement. In these situations, the subjective assessment of steering effort can also be the
measure. Then, steering effort is probably assessed based on both steering wheel rate and steering
wheel torque.
4.2.8 One-track model for low speeds, with non-

Ackerman geometry
The model in Figure 4-10 can rather easily be extended to more axles, more units and two-track
model, if Ackermann geometry. However, if not Ackermann geometry, one has to do a structural
change in the model, because ideal tracking is no longer consistent with geometry. One example is a
two-track model of a two-axle vehicle which has parallel steering on one axle. Another example is a
one-track model of a truck with three axles, whereof the two rear are non-steered, see Figure 4-15.
We will go through model changes needed in the latter example. In order to compare the models as
closely as possible, we simply split the rear axle into two rear axles, in the example in Section 4.2.6.
The physical model becomes as in Figure 4-15. The measures appear in Figure 4-17, and you see that it
is not a truck, but a very unconventional passenger size car with two rear axles.
speeds: vy vfxv ⃗ forces: Ffxv ⃗

vr2y vr1y wz Fr2y Fr1y
vfyv
Ffyv
vx df Fr2x Fr1x df
vr2x vr1x
⃗ ⃗
lr lf
L
b b
Figure 4-15: Top: Rigid Truck with 3 axles, whereof only the first is steered.
Bottom: One-track model. Not Ackermann geometry, due to un-steered rear axles.
The changes we have to do in the model appear as underlined in Equation [4.11]. There has to be
double variables for , denoted 1 and 2 respectively. Also, we cannot use 𝑤 = =
anymore, but instead we have to introduce a lateral tyre force model, as described in Section 2.4.
//Equilibrium:
0 = Ffxv + Fr1x + Fr2x;
0 = Ffyv + Fr1y + Fr2y;
0 = Ffyv*lf - Fr1y*(lr - b) - Fr2y*(lr + b);
//Constitutive relation, i.e.

Lateral tyre force model (instead of Ideal tracking):
Ffyw = -Cf*sfy;
Fr1y = -Cr1*sr1y;
Fr2y = -Cr2*sr2y; [4.11]
sfy = vfyw/vfxw;
sr1y = vr1y/vr1x;
sr2y = vr2y/vr2x;
//Compatibility:
vfxv = vx;
vfyv = vy + lf*wz;
vr1x = vx;
166
LATERAL DYNAMICS
vr2x = vx;
vr1y = vy - (lr - b)*wz;
vr2y = vy - (lr + b)*wz;

Ffxv = Ffxw*cos(df) - Ffyw*sin(df);

der(pz) = wz;
// Prescription of steering angle:

//Rear axles undriven, which gives drag from roll resistance:
Fr1x = -100/2;
Fr2x = -100/2;
The new result is shown in Figure 4-16, which should be compared to Figure 4-11. The radius of the
final path radius increases a little. If we read out more carefully, we can draw the different locations of
the instantaneous centre for both cases. This is shown, in scale, in Figure 4-17.
df [deg] y [m]
35 deg
vx=2.778m/s
vy=1.036m/s
wz=0.652rad/s
x[m]
time [s]
Figure 4-16: Simulation results of one-track model with ideal tracking. Non-Ackermann geometry due to
two non-steered rear axles.
167
LATERAL DYNAMICS
We could tune the steering angle required to

reach exactly the same path radius as for the • instantaneous centre for
2-axle reference vehicle. Then, we would have 3-axle vehicle
to steer a little more than the 35 degrees used, • instantaneous centre for
and we could find a new instantaneous centre, 2-axle reference vehicle
and we could identify a so called Equivalent
wheelbase. This leads us to a definition: The
equivalent wheel base of a multi-axle vehicle is
the wheel base of a conventional two-axle
vehicle which would exhibit the same turning
behaviour as exhibited by the multi-axle
vehicle, given same steering angle and similar b=1 b=1
axle cornering stiffnesses. df
Section 4.2.8 shows that the lateral tyre force
model, which is a constitutional relation, can
sometimes be needed also at low speeds. lr=1.5 lf=1.3
However, it is always needed in next section
about high speed cornering, Section 4.2.9. Figure 4-17: Instantaneous centre with a 3-axle vehicle,
with the corresponding 2-axle vehicle as reference.
4.2.9 One-track model for low speeds, with trailer

When the vehicles has an articulation point, also the low speed case (disregarding inertias) has
transients in the sense that the articulation angles changes transiently. Consider the case of
instantaneous step steer. For a vehicle without articulation point, a steady state is reached directly,
since inertia is not considered. But for an articulated vehicle it takes some travelled distance (not time,
since it can be studied independent of speed and time) before a steady state articulation angle is
achieved.
So, for articulated vehicles, a scalar requirement on turning radius is not so relevant as for two axle
vehicles. But, the function “4.2.7.2 Swept path width *” and “4.2.7.3 Off-tracking *” is, if road geometry
is reflected, e.g. through outer radius and total angle of turning.
4.3 Steady state cornering at high speed

Steady state cornering refers to that all time derivatives of vehicle speeds ( 𝜔 ) are zero. The
physical understanding is then that the vehicle drives on a circle with constant yaw rate, see Figure
4-18. Alternatively, this can be described as driving with constant tangential speed ( ), on a constant
path radius (𝑅) with and with a constant side slip angle (𝛽).
4.3.1 Steady state driving manoeuvres

When testing steady state function, one usually runs on a so called “skid-pad” which appears on most
test tracks, see Figure 4-8. It is a flat circular surface with typically 100 m diameter and some
concentric circles marked. A general note is that tests in real vehicles are often needed to be
performed in simulation also, and normally earlier in the product development process.
Typical steady state tests are:
• Constant path radius. Driven for different longitudinal speeds.
• Constant longitudinal speed. Driven for different path radii.
• Constant steering wheel angle. Then increase accelerator pedal (or apply brake pedal) gently.
(If too quick, the test would fall under transient handling instead.)
168
LATERAL DYNAMICS
𝛽 Steady state cornering can be defined by either 3

or 2 quantities, depending on assumptions:
• For a general vehicle it is fully defined only
by all 3 quantities, e.g. 𝜔 or 𝛽𝑅.
• For a certain vehicle it is often considered as
fully defined by 2 inputs from driver:
accelerator pedal and steering wheel. (Then,
one have neglected other inputs from driver,
such as brake pedal and parking brake. One
have also neglected other possible automatic
actuation, such as all wheel drive distribution
and ESC-braking one wheel.)
Figure 4-18: Steady state cornering. (𝛽 will be negative for larger , i.e. vehicle will point inwards.)
Relevant standards for these test manoeuvres are: References (ISO 4138) and (ISO 14792).
Vehicle Handling
Dynamics track
Area
Skid pad
High speed
track
Figure 4-19: An example of test track and some parts with special relevance to Vehicle Dynamics. The
example is Hällered Proving Ground, Volvo Car Corporation.
169
LATERAL DYNAMICS
=Vehicle Dynamics Area
Figure 4-20: An example of test track. The example is AstaZero (Active Safety Test Arena), SP Technical
Research Institute of Sweden and Chalmers University of Technology.
High
speed
circle
Skid pad
Low
friction
Vehicle Dynamics Area
strips
Hill
strips
Figure 4-21: An example of test track. The example is CASTER’s (virtual) test track. Used for CASTER’s
driving simulator at Chalmers University of Technology.
4.3.2 Steady state one-track model

In steady state we have neither inertial effects from changing the total vehicle speed ( = √ +
is constant) nor from changing the yaw rate (𝑤 is constant). However, the inertial “centrifugal” effect
of the vehicle must be modelled. The related acceleration is the centripetal acceleration, 𝑐 = 𝑅 ∙
𝜔 = 𝑅=𝜔 ∙ .
170
LATERAL DYNAMICS
A vehicle model for this is sketched in Figure 4-22. The model is a development of the model for low-
speed in Figure 4-10 and Equation [4.9], with the following changes:
• Longitudinal and lateral accelerations are changed from zero to components of centripetal
acceleration, ac, as follows (see Figure 4-22):
o = 𝑐 sin(𝛽) = 𝜔 sin(𝛽) = 𝜔 ;
o = + 𝑐 cos(𝛽) = +𝜔 cos(𝛽) = +𝜔 ;
• The constitutive assumptions for the axles are changed from ideal tracking to a (linear)
relation between lateral force and lateral slip. The relations should capture the slip
characteristics for the tyres, see Section 2.4, but they can also capture steering system
compliance, side force steering, and roll steering (see Section 4.3.5.3). The total mathematical
relations can anyway be written as:
o 𝑤 = 𝐶 ∙ 𝑠 ; where 𝑠 = 𝑤⁄ 𝑤;
o = 𝐶 ∙ 𝑠 ; where 𝑠 = ⁄ ;
speeds: vy forces:
vfxv ⃗ m*ay Ffxv ⃗
vry wz
Ffyv
vfyv
Fry
vx df Frx m*ax df
vrx
lr lf lr lf
L L
⃗ ⃗
vehicle path
centre
-ax
ac=wz*v
b
-ax=ac*sin(b); ay
+ay=ac*cos(b); vx=v*cos(b) v
vy=v*sin(b)
b
Figure 4-22: One-track model. Dashed forces and moment are fictive forces.
The model in Figure 4-22 is documented in mathematical form in Equation [4.12] (in Modelica
format). The subscript v and w refers to vehicle coordinate system and wheel coordinate system,
respectively.
//Equilibrium:
m*ax = Ffxv + Frx; //Air and grade resistance neglected
m*ay = Ffyv + Fry;
J*0 = Ffyv*lf - Fry*lr; // der(wz)=0
-ax = wz*vy;
+ay = wz*vx;
//Constitutive relation, i.e. Lateral tyre force model: [4.12]

Ffyw = -Cf*sfy;
Fry = -Cr*sry;
sfy = vfyw/vfxw;
sry = vry/vrx;
//Compatibility:
171
LATERAL DYNAMICS
vfxv = vx;
vfyv = vy + lf*wz;
vrx = vx;
vry = vy - lr*wz;


der(pz) = wz;

// Rear axle undriven, which gives drag from roll resistance:
Frx = -100; // =10000; //
//Ffxw=0;
The longitudinal speed is a parameter, = 1 𝑘 ℎ. A simulation result from the model is shown in
Figure 4-23. It shows a certain input steering angle function of time. It also shows the resulting path,
y(x). A driving resistance of 100 N is assumed on the rear axle (Frx=100;). This is to exemplify that
longitudinal forces do not need to be zero, even if longitudinal forces normally are not so interesting
for steady state high speed manoeuvres.
The variables 𝑥 𝑦 and 𝑝 = are the only “state variables” of this simulation. If not including the path
model (Equation [4.1]), the model would actually not be a differential equation problem at all, just an
algebraic system of equations. That system of equations could be solved isolated for any value of
steering angle without knowledge of time history. These aspects are the same for the low speed model
in section 4.2.6. Note that steering start to the left, but vehicle path starts bending to the right. This is a
model inaccuracy, namely that the steady state model assumes that steady state is found directly, and
turning left has the steady state to the right, due to centrufugal force.
Now, the validity of a model always has to be questioned. There are many modelling assumptions
which could be checked, but in the following we only check the assumption =𝜔 ; instead of the
more correct = ̇ +𝜔 ;, which we will learn in Section “4.4.2 Transient one-track model”.
Comparison of the terms gives | ̇ |𝑚 ≈ |𝜔 |𝑚 ≈ 1 ⁄𝑠 , so |𝜔 | is large and this
jeopardizes the validity. Large |𝜔 | happens during < 𝑡 <≈ 𝑠, so the model is not very valid
there. But, at 𝑡 >≈ 𝑠, the model is valid, at least in this aspect, since there | ̇ | ≈ ≪ |𝜔 |. So,
the model is not very valid during the initial sinusoidal steering. This shows that a steady state models
should not be thrusted outside steady state conditions.
Equation [4.12] is a complete model suitable for simulation, but it does not facilitate understanding
very well. We will reformulate it assuming small (i.e. cos( ) = 1, sin( ) = , and = ).
Eliminate slip, all forces that are not wheel longitudinal, and all velocities that are not CoG velocities:
𝜔 ( +( +𝑙 𝜔 ) )=
= 𝑤 ( +( +𝑙 𝜔 ) )+𝐶 ( +𝑙 𝜔) + ( +( +𝑙 𝜔 ) );
𝜔 ( +( +𝑙 𝜔 ) )=
𝑦− 𝑟 𝜔
= 𝐶 ( +( +𝑙 𝜔 )) + 𝑤 𝐶 ( +( +𝑙 𝜔 ) ); [4.13]
𝑥
𝑙 𝜔
𝐶 ( +( +𝑙 𝜔 )) 𝑙 𝑤 𝑙 =𝐶 𝑙 ( +( +𝑙 𝜔) );
Equation [4.13] is a complete model, which we can see as a dynamic system without state variables.
• Actuation: Steering and wheel torque on each axle: 𝑤 .
• Motion quantities: 𝜔
172
LATERAL DYNAMICS
df [rad] y [m]
5 deg = + =
= 8.37 𝑠
Ffyw
Ffxw
Frx-100 Nm
vx = 100 km/h=27.78 m/s
wz = 0.3995 rad/s
vy = -5.761 m/s
x[m]
time [s]
Figure 4-23: Simulation results of steady state one-track model. The vehicle sketched in the path plot is not
in scale, but correctly oriented.
For example, we can choose 𝑤 𝑥 and use the 3 equations to calculate 𝑦 𝜔 . In steady state
manoeuvres, one can often disregard the longitudinal equilibrium, which means to skip 1st equation
and the unknown :
𝐶 𝐶 + (𝐶 𝑙 + 𝐶 𝑙 ) 𝑤
𝜔 = ;
𝐶 𝐶 + (𝐶 𝑙 𝐶 𝑙 )
[4.14]
𝐶 (𝐶 𝑙 𝑙 ) + (𝐶 𝑙 𝑙 ) 𝑤
= ;
𝐶 𝐶 + (𝐶 𝑙 𝐶 𝑙 )
4.3.2.1 Relation 𝜹𝒇 𝒗𝒙 and

Solving the first equation in Equation [4.14] yields:
𝐶 ∙𝐶 ∙ + (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙ ∙ 𝜔
= ≈ ∙
𝐶 ∙ 𝐶 ∙ + (𝐶 ∙ 𝑙 + 𝐶 ∙ 𝑙 ) ∙ 𝑤
𝐶 ∙𝑙 𝐶 ∙𝑙 𝜔
≈ { 𝑠𝑠𝑢 : 𝑤 ≈ }≈( + ∙ ∙ )∙ =
𝐶 ∙𝐶 ∙
𝐶 ∙𝑙 𝐶 ∙𝑙 𝑙 𝑙 𝜔 [4.15]
={ 𝑛 : 𝐾𝑢 = = }= ∙ + 𝐾𝑢 ∙ ∙ ≈
𝐶 ∙𝐶 ∙ 𝐶 ∙ 𝐶 ∙
𝜔
= {𝑢𝑠 𝑠𝑡 𝑦 𝑠𝑡 𝑡 : =𝜔 ∙ }= ∙ + 𝐾𝑢 ∙ ∙ 𝜔 ∙ ≈
𝜔 1 ∙
≈ { 𝑠𝑠𝑢 𝑠 𝑙𝑙 ⇒ ≈ } ≈ + 𝐾𝑢 ∙ ;
𝑅 𝑅 𝑅
The coefficient 𝐾𝑢 is the understeer gradient and it will be explained in next section.
173
LATERAL DYNAMICS
4.3.2.2 Relation 𝜹𝒇 𝒗𝒙 and 𝜷

Solving the second equation in Equation [4.14] yields:
𝐶 ∙𝐶 ∙ + (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙ ∙
= ∙ ≈
𝐶 ∙ (𝐶 ∙ 𝑙𝑙 ∙ ∙ ) + (𝐶 𝑙 𝑙 )∙ 𝑤
𝐶 ∙𝐶 ∙ (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙ ∙
≈ {𝑢𝑠 : 𝑤 = }≈ ∙ ≈
𝐶 ∙𝐶 ∙𝑙 ∙ 𝐶 ∙𝑙 ∙ ∙
[4.16]
→ ∙ ;
𝑥 0 𝑙
⇒ 𝐶 ∙𝑙 𝐶 ∙𝑙
→ ∙ = 𝐾𝑢 ∙ 𝐶 ∙ ∙ ;
{ 𝑥 ∞ 𝐶 ∙𝑙 𝑙
𝑦
We can see that there is a speed dependent relation between steering angle and side slip, . The side
𝑥
𝑦
slip can also be expressed as a side slip angle, 𝛽 = arctan ( ). Since normally 𝐾𝑢 > , the side slip
𝑥
changes sign, when increasing speed from zero to sufficient high enough. This should feel intuitively
correct, if agreeing on the conceptually different side slip angles at low and high speed, as shown in
Figure 4-24. We will come back to this equation in context of Figure 4-34.
path of
low speed front axle
high speed
𝜷>𝟎 path of 𝜷 𝜷
rear axle 𝜷<𝟎
v v
vehicle
vehicle path
path centre
centre
path of
front axle
path of
rear axle
Figure 4-24: Body Slip Angle for Low and High Speed Steady State Curves
4.3.2.3 Relation 𝒗𝒙 and 𝜷

If we approximate 𝑤 = and use both equations in Equation [4.14] to eliminate we get:
𝐶 𝑙 𝑙 𝑙
=𝑙 = = ≈ 𝑅 tan(𝛽) ; ⇒
𝐶 𝐶 𝜔 𝜔
[4.17]
𝑙 1
⇒ = tan(𝛽) = (𝑙 ) ;
𝐶 𝑅
4.3.2.4 Simpler derivation of same model

A simpler way to reach almost same expression as first expression in Eq [4.14] is given in Figure 4-25.
Here, the simplifications are introduced earlier, already in physical model, which means e.g. that the
influence of 𝑤 is not identified.
The validity of the steady state model described in this section, section 4.3.2, is of course limited by if
the manoeuvre is transient which would mean that steady state is not reached. But it is also limited by
174
LATERAL DYNAMICS
if the assumption of linear tyre characteristics, = 𝐶 ∙ 𝑠 ;, is violated. Therefore, one should check if
some axle (or wheel) is calculated to use more than some certain fraction of available friction, =
√ + > 𝑟 𝑡 𝑜𝑛 ∙ ∙ ;. Referring to the tyre brush model with uniform pressure distribution
one can argue for using 0.5 as this limiting fraction, since the tyre force is linear with slip up to this
value.
Physical model: Mathematical model:

• Path radius >> the vehicle. Then, all forces Equilibrium:
(and centripetal acceleration) are
approximately co-directed. ≈ + ; ≈ 𝑙 𝑙 ;
• Small tyre and vehicle side slip. Then, 𝑅
angle=sin(angle)=tan(angle). Constitution:
(Angles are not drawn small, which is the reason = 𝐶 𝑠 ; = 𝐶 𝑠 ;
why the forces not appear co-linear in figure.)
Compatibility:
𝑦 𝑦+ 𝜔
2 +𝛼 =𝛽 ; 𝛽 ≈ = ;
𝑥 𝑥 𝑥
𝑅 𝑟𝑦 𝑦− 𝑟 𝜔
𝛼 =𝛽 ≈ = ; ⇒
𝑥 𝑥
𝛽 𝑥
𝜔 ≈ ; 𝛼 ≈𝑠 ; 𝛼 ≈𝑠 ;
𝑅
𝛼
𝛽 ⇒ +𝛼 𝛼 ≈ ;
𝛽 𝑅
𝜔 Eliminate 𝛼 𝛼 𝛽 𝛽 𝜔 yields:
𝑙 𝑙 ≈ + 𝐾𝑢 ;
𝑅 𝑅
𝐶 𝑙 𝐶 𝑙
𝐾𝑢 = ;
𝐶 𝐶
Figure 4-25: Simpler derivation final step in Equation [4.15].
4.3.3 Under-, Neutral- and Over-steering *

4.3.3.1 Understeering as built-in measure
Function definition: Understeering (gradient) is the additional steering angle needed per increase of lateral
force (or lateral acceleration) when driving in high speed steady state cornering on level ground and high road
friction. Additional refers to low speed. The gradient is defined at certain high speed steady state cornering
conditions, including straight-line driving. Steering angle can be either road wheel angle or steering wheel angle.
𝐾𝑢 in Equation [4.15] is called “understeer gradient” and has hence the unit rad/N or 1/N. Sometimes
one can see slightly other definitions of what to include in definition of understeer gradient, which
have different units, see 𝐾𝑢 and 𝐾𝑢 in Equation [4.18]. (When defined as 𝐾𝑢 , one can sometimes see
the unit “rad/g” used, which present compendium recommended to not use.)
175
LATERAL DYNAMICS
𝐶 ∙𝑙 𝐶 ∙𝑙 ∙ ∙
= + ∙ =
𝑅 𝐶 ∙𝐶 ∙ ∙𝑅
𝐶 ∙𝑙 𝐶 ∙𝑙
= {𝑢𝑠 : 𝐾𝑢 = ∙ ∙ [1 𝑜𝑟 𝑟 ]} = + 𝐾𝑢 ∙ ;
𝐶 ∙𝐶 ∙ 𝑅 ∙𝑅
[4.18]
𝐶 ∙𝑙 𝐶 ∙𝑙 ∙
= + ∙ =
𝑅 𝐶 ∙𝐶 ∙ 𝑅
𝐶 ∙𝑙 𝐶 ∙𝑙 1 𝑟
= {𝑢𝑠 : 𝐾𝑢 = ∙ [ 𝑜𝑟 ]} = + 𝐾𝑢 ∙ ;
𝐶 ∙𝐶 ∙ ⁄𝑠 ⁄𝑠 𝑅 𝑅
The first term in the final expression in Equation [4.15], L/R, can be seen as a reference steering angle.
This is referred to as either “low speed steering angle” or “Ackermann steering angle”. The
understanding of this steering angle can be expanded to cover more general vehicles: It is the angle
required for low speed turning. For a general vehicle, e.g. with other steering geometry than
Ackermann, the angle cannot be calculated as simple as L/R, but with some specific derived formula,
or it can be measured on a real vehicle. It is common to use the subscript A. It can be defined for front
axle road wheel steering angle, 𝐴 . It can also be defined for steering wheel angle, 𝑠𝑤𝐴 .
The understeering gradient, 𝐾𝑢 , is normally positive, which means that most vehicles require more
steering angle for a given curve, the higher the speed is. Depending on the sign of 𝐾𝑢 a vehicle is said to
be oversteered (if 𝐾𝑢 < ), understeered (if 𝐾𝑢 > ) and neutral steered ( 𝐾𝑢 = ). In practice, all
vehicles are designed as understeered, because over steered vehicle would become unstable very
easily.
The understeering gradient 𝐾𝑢 can be understood as how much additionally to the Ackermann
steering angle one has to steer, in relation to centrifugal force, 𝑐 :
∙ 𝐴 ∆
= 𝐴 + 𝐾𝑢 ∙ = 𝐴 + 𝐾𝑢 ∙ 𝑐; ⇒ 𝐾𝑢 = = ;
𝑅 𝑐 𝑐
Or, using the steering ratio, = 𝑠𝑤 ⁄𝑟𝑠𝑡𝑒 : [4.19]

𝑟𝑠𝑡𝑒 ∙ ∙ 𝑠𝑤 𝑠𝑤𝐴 ∆ 𝑠𝑤
𝑠𝑤 = 𝑠𝑤𝐴 + 𝐾𝑢 ∙ = 𝐴 + 𝐾𝑢 ∙ 𝑟𝑠𝑡𝑒 ∙ 𝑐; ⇒ 𝐾𝑢 = = ;
𝑅 𝑟𝑠𝑡𝑒 ∙ 𝑐 𝑟𝑠𝑡𝑒 ∙ 𝑐
4.3.3.2 Understeering as varying with steady state lateral

acceleration
So far, the understeering gradient is presented as a fix built-in vehicle parameter. There is nothing that
says that a real vehicle behaves linear, so in order to get a well-defined value of 𝐾𝑢 , the ∆ and the 𝑐
should be small when using Eq [4.20]. However, if we accept that 𝐾𝑢 can change with 𝑐 , 𝐾𝑢 can be
defined as a differential quantity:
𝐾𝑢 can also be understood as how much the additional steering angle, ∆ , has to increase per
increased centrifugal force, 𝑐 , or per lateral acceleration, :
𝜕(∆ ) 𝜕 𝜕∆ 𝜕(∆ ) 𝜕 𝜕
𝐾𝑢 = = ( 𝐴) = ; 𝑜𝑟 𝐾𝑢 = = ( 𝐴) = ; [4.20]
𝜕 𝑐 𝜕 𝑐 𝜕 𝑐 𝜕 𝜕 𝜕
Equation [4.20] shows the understeering gradient as a (mathematical) function, rather than a scalar
parameter. But it is still fix and built-in in the vehicle. If assessing understeering for a lateral forces up
to near road friction limit, Equation [4.20] is more relevant than Equation [4.15], because it reflects
that understeering gradient changes.
176
LATERAL DYNAMICS
4.3.3.3 Understeering as a varying quantity during a transient

manoeuvre
A third understanding of the word understeering is quite different and less strictly defined. It is to see
the understeering as a variable during a transient manoeuvre. For instance, a vehicle can be said to
understeer if tyre side slip is larger on front axle than on rear axle, | | > | |, and over-steer if
opposite, | | > | |. This way of defining understeering and oversteering is not built in, but varies
over time through a (transient) manoeuvre. E.g., when braking in a curve a vehicle loses grip on rear
axle due to temporary load transfer from rear to front. Then the rear axles can slide outwards
significantly and the vehicle can be referred to as over-steering at this time instant, although the built-
in understeering gradient is >0. This “instantaneous” under-/over-steering can be approximately
found from log data with this simple approximation:
∙𝜔 ∙
𝑛𝑒𝑢𝑡 = ≈{ ≈𝑅∙𝜔 }≈ ≈{ ≈ ∙𝜔 }≈ ; [4.21]
𝑅
If the actual vehicle has | | < | 𝑛𝑒𝑢𝑡 | the vehicle oversteers, and vice versa. This is often very
practical since it only requires simply logged data, and . Note that when and 𝑛𝑒𝑢𝑡 have
different signs, neither understeer or oversteers is suitable as classification, but it can sometimes be
called “counter-steer”.
150
100
50
-50
-100 dfront /[10*deg]

2
ay /[10*m/s ]
-150
1 1.5 2 2.5 3 3.5 4 vx4.5
/[km/h] 5
t wz /[deg/s]
beta /[10*deg]
betaRear /[10*deg]
10
df /[deg]
dneutral /[deg]
5 US /[boolean]
OS /[boolean]
-5
-10
1 1.5 2 2.5 3 3.5 4 4.5 5
t
Figure 4-26: Upper: Log data from passenger car with ESC system in a double lane change. Lower: 𝑛𝑒𝑢𝑡
can be used to approximately assess “instantaneous” under-/over-steering (US/OS).
A second look at Equation [4.15] tells us that we have to assume absence of propulsion and braking on
front axle, 𝑤 = , to get the relatively simple final expression. When propulsion on front axle
( 𝑤 > ), the required steering angle, , will be smaller; the front propulsion pulls in the front end
of the vehicle. When braking on front axle ( 𝑤 < ), the required steering angle, , will be larger;
the front braking hinders the front end to turn in. To keep constant, which is required within
177
LATERAL DYNAMICS
definition of steady state, one have to propel the vehicle because there will always be some driving
resistance to overcome. Driving fast on a small radius is a situation where the driving resistance from
tyre lateral forces becomes significant, which is a part of driving resistance which was only briefly
mentioned in Section 3.2.
4.3.3.4 Neutral steering point

An alternative measure to understeering coefficient is the longitudinal position of the neutral steering
point. The point is defined for lateral force disturbance during steady state straight-ahead driving, as
opposed to steady state cornering without lateral force disturbance. The point is where a vehicle-
external lateral force, such as wind or impact, can be applied on the vehicle without causing a yaw
velocity (𝜔 = ), i.e. only causing lateral velocity ( ). From this definition, we can derive a
formula for calculating the position of the neutral steering point, see Figure 4-27.

• Steady state ( ̇ = ̇ = 𝜔̇ = ) Equilibrium:
• Straight ahead driving (𝜔 = ) = + + 𝑒;
• No steering = 𝑙 𝑙 𝑒 𝑙𝑠 ;
• Small tyre and vehicle side slip. Then, Constitution:
angle=sin(angle)=tan(angle). = 𝐶 𝑠 ;
= 𝐶 𝑠 ;
Compatibility:
Fey
Fry Ffy 𝑠 =𝑠 = ;
𝑙𝑠
Eliminate 𝑠 𝑠 𝑒 yields:
b 𝐶 𝑙 𝐶 𝑙
𝑙𝑠 = ;
𝐶 +𝐶
𝐶𝑟 𝑟 −𝐶
𝜔 = Identify understeering gradient, 𝐾𝑢 = :
𝐶 𝐶𝑟 𝐿
𝑙 𝑙 𝐶 ∙ 𝐶𝑟
L 𝑙𝑠 = 𝐾𝑢 ;
𝐶 + 𝐶𝑟
Figure 4-27: Model for definition and calculation of neutral steering point.
The result is 𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙𝐶 𝐶 ∙𝑙 𝐶 ∙𝑙
condensed in Eq [4.22]. 𝑙𝑠 = = 𝐾𝑢 ∙ ; 𝑤ℎ 𝑟 𝐾𝑢 = ; [4.22]
𝐶 +𝐶 𝐶 +𝐶 𝐶 ∙𝐶 ∙
We can see that the understeer gradient from steady state cornering model appears also in the
formula for neutral steering point position, 𝑙𝑠 . Since 𝐶 𝐶 𝑛 are positive, the neutral steering point
is behind of CoG for understeered (two-axle) vehicles, and in front of CoG for oversteered (two-axle)
vehicles. This is why 𝑙𝑠 and 𝐾𝑢 can be said to be alternative measures for the same vehicle
function/character, the yaw balance.
4.3.4 Normalized required steering angle

A fundamental property of the vehicle is what steering angle that is required to negotiate a certain
curvature (=1/path radius = 1 𝑅 ). This value can vary with longitudinal speed and it can be
normalized with wheel base, . From Equation [4.15], we can conclude:
178
LATERAL DYNAMICS
∙𝑅 ∙
𝑵𝒐 𝒎𝒂𝒍𝒊𝒛𝒆𝒅 𝒆𝒒𝒖𝒊 𝒆𝒅 𝒔𝒕𝒆𝒆 𝒊𝒏𝒈 𝒂𝒏𝒈𝒍𝒆 = = 1 + 𝐾𝑢 ∙ ; [4.23]
The normalized required steering angle is plotted for different understeering gradients Figure 4-28. It
is the same as the inverted and normalized curvature gain, see 4.3.6.2.
2.5
Ku = 2.525e-6 [1/N]
Understeered
2
required steering, df*R/L [rad]
1.5
Neutral steered
Ku = 0e-6 [1/N]
1
Oversteered
0.5 Ku = -1.794e-6 [1/N]
Characteristic speed (for Critical speed (for

understeered vehicle) oversteered vehicle)
0
0 5 10 15 20 25 30 35 40 45 50
vx [m/s]
Figure 4-28: Normalized steering angle ( ∙ 𝑅 ⁄ ) for Steady State Cornering
4.3.4.1 Critical and Characteristic speed *

Function definition: Critical speed is the speed above which the vehicle becomes unstable in the sense that the
yaw rate grows significantly for a small disturbance in, e.g., steering angle.
Function definition: Characteristic speed is the speed at which the vehicle requires twice as high steering angle
for a certain path radius as required at low speed (Figure 4-28). (Alternative definitions: The speed at which the
yaw rate gain reaches maximum (Figure 4-31). The speed at which the lateral acceleration gain per longitudinal
speed reaches its highest value. (Figure 4-33).)
We can identify that zero steering angle is required for the over-steered vehicle at 28 m/s. This is the
so called Critical Speed, which is the speed where the vehicle becomes unstable. It should be noted
here, that there are stable conditions also above critical speed, but one has to steer in the opposite
direction then. Normal vehicles are built understeered, which is why a Critical speed is more of a
theoretical definition. However, if studying (quasi-steady state) situations where the rear axle is
heavily braked, the cornering stiffness rear is reduced, and a critical speed can be relevant.
For understeered vehicles, we can instead read out another measure, the Characteristic Speed. The
understanding of Characteristic Speed is, so far just that required steering increases to over twice
what is needed for low speed at the same path radius. A better feeling for Characteristic Speed is easier
suggested in a later section, see Section 4.3.6.3.
From Equation [4.15], we can find a formula for critical and characteristic speeds:
∙ 𝑐 𝑡 𝐶 ∙𝐶 ∙
= + 𝐾𝑢 ∙ = ⇒ 𝑐 𝑡 =√ =√ ;
𝑅 𝑅 𝐾𝑢 ∙ (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙
[4.24]
∙ 𝑐ℎ 𝐶 ∙𝐶 ∙
= + 𝐾𝑢 ∙ = ∙ ⇒ 𝑐ℎ =√ =√ ;
𝑅 𝑅 𝑅 𝐾𝑢 ∙ (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙
179
LATERAL DYNAMICS
4.3.5 How to design for understeer gradient

The values of cornering stiffness for each axle are design parameters to influence understeer gradient.
However, cornering stiffness is an abstract design parameter, in the sense that one cannot put it on a
drawing. Instead, cornering stiffness is a combined effect from various, more concrete, design
parameters. Such more concrete design parameters are briefly presented in the following.
4.3.5.1 Tyre design, inflation pressure and number of tyres

The cornering stiffness of each tyre is an obvious parameter which influences the axle stiffness. The
cornering stiffness of an axle is influence by the sum of cornering stiffness for all tyres. There are
normally two tyres per axle, but there are also vehicles with one tyre (e.g. bicycles) or 4 (typically 2
double mounted on each side in heavy trucks).
Tyre design influences, which is geometrical dimensions and material selection. Inflation pressure is in
this context very close to a design parameter.
In a first approximation, tyre cornering stiffness is approximately proportional to vertical load: 𝐶 =
𝑘 ∙ . For a vehicle with equally many and same tyres front and rear, this means that it will be
neutral steered. This is because, in steady state cornering, vertical loads are distributed over the axles
in the same relation as lateral loads. Using definition of understeer gradient:
𝑙 𝑙 𝑙𝑜𝑛 𝑡𝑢 𝑛 𝑙
𝐾𝑢 𝑙 𝑙
= = = {𝑙𝑜 𝑡𝑟 𝑛𝑠 𝑟} ⇒
∙ 𝐶 ∙ 𝐶 ∙ 𝑘 ∙ ∙ 𝑘 ∙ ∙
∙ ∙ℎ
∙ ∙𝑙 ∙ ∙𝑙
⇒ 𝐾𝑢 = =
𝑙 ℎ 𝑙 ℎ [4.25]
𝑘 ∙( ∙ ∙ ∙ ∙ )∙ 𝑘 ∙( ∙ ∙ + ∙ ∙ )∙
=
𝑙 ⁄ 𝑙 ⁄
= ={ 𝑛 }= ;
𝑙 ℎ 𝑙 ℎ
𝑘 ∙( ∙ ) 𝑘 ∙( + ∙ ) 𝑘 =𝑘
Longitudinal load transfer (influence of in the equations) show that braking increases over-steering
tendency. It is actually so, that the critical speed 𝑐 𝑡 = √ ⁄( 𝐾𝑢 ∙ ) (see Equation [4.24]) can
come down to quite reachable levels when braking hard; i.e. hard braking at high speed may cause
instability. This is especially so for front biased CoG location. See Figure 4-29, inspired by Reference
(Drenth, 1993).
However, the cornering stiffness varies degressively, e.g. 𝐶 = 𝑘 ∙ 𝑘𝑑∙ . This is further
studied in Reference (Drenth, 1993).
If taking the degressiveness of tyre cornering stiffness into account, the weight distribution plays a
role also without longitudinal load transfer; front biased weight distribution gives under-steered
vehicles and vice versa. Also, the number of wheel per axle influence stronger; single wheel front (or
double-mounted rear) gives under-steered vehicles and vice versa.
It should be noted that if the longitudinal acceleration is due to wheel torques, as opposed to road
grade or aerodynamic forces, the tyre combined slip effects influences the curves in Figure 4-29.
Basically, as the axle that takes most longitudinal force gets reduced cornering stiffness.
180
LATERAL DYNAMICS
Cf/Ffz=kf=13.4831 and Cr/Frz=kr=13.4831 Solid=vxCrit, Dashed=vxChar

200
0.1 180
160
0.05 140
Ku2 [1 or rad]
120
vx [km/h]
0 100
80
-0.05 60
lf/L=0.25 40
lf/L=0.5 20
-0.1
lf/L=0.75
0
-10 -5 0 5 10 -1 -0.5 0 0.5 1
ax [m/(s*s)] ax [g]
Figure 4-29: Left: Under-steering gradient as function of longitudinal acceleration, ax, and static load
distribution, lf/L. Right: Critical and characteristic velocity as function of acceleration and load distribution.
4.3.5.2 Roll stiffness distribution between axles

During cornering, the vertical load is shifted towards the outer wheels. Depending on the roll stiffness
of each axle, the axles take differently much of this lateral load transfer. This also influences the yaw
balance. The more roll stiff an axle is, the more of the lateral load shift it takes. Tyre cornering stiffness
varies degressively with vertical load. Together, this means that increasing the roll stiffness on the
front axle, leads to less front cornering stiffness, see Figure 2-30, and consequently more understeered
vehicle. Increasing roll stiffness on rear axle makes the vehicle more oversteered. The total roll
stiffness of the vehicle does not influence the understeering gradient. Normally one makes the front
axle more roll stiff than the rear axle. This means that vehicle becomes more and more understeered
for increased lateral acceleration, e.g. more steering angle is needed to maintain a certain path radius if
speed increases. One can change the roll stiffness of an axle by changing roll centre height, wheel
stiffness rate and anti-roll bar stiffness.
3.7 1 5𝑁 𝑟
3.1 1 5𝑁 𝑟
Both wheels, if no load transfer.

.1 1 5 Axle cornering stiffness ≈
𝑁 𝑟 ≈ 3.1 = 6. 1 5 𝑁 𝑟
Cornering
stiffness
Left and right wheel, if load transfer ± 5 𝑘𝑁.

Axle cornering stiffness ≈ .1 + 3.7 = 5.8 1 5𝑁 𝑟
± 5 𝑘𝑁 ± 5 𝑘𝑁
Figure 4-30: The wheels cornering stiffness (𝜕⁄𝜕𝛼 ( )|𝑠 ) changes degressively with vertical load. The
𝑦 =0
axle cornering stiffness therefore decreases with increased load transfer.
4.3.5.3 Steering and suspension compliances

4.3.5.3.1 Steering system compliance
A compliant steering system will reduce the effective cornering stiffness on the steered axle. With
normal front axle steering, a compliant steering means a more understeered vehicle.
181
LATERAL DYNAMICS
The size of the steering system compliance depends on how much of the steering system one takes
into account. When studying steering response for human driver, it would be all the system up to the
steering wheel. Then, also the steering assistance reduces the axle’s effective cornering compliance.
When studying response for an automated steering where a virtual driver (automated steering)
actuates the steering, it would typically be only the parts up to the steering actuator, e.g. the steering
rack. Finally, the reduction of axle cornering stiffness becomes lesser, the closer to tyre grip loss, since
the caster offset decreases, see Figure 2-34.
4.3.5.3.2 Side-force steer gradient
Side-force steer gradient, 𝑆𝐹𝑆 , is defined for an axle and it is how much the wheels on an axle steers
[deg] negatively per lateral force [N]. Negative is chosen since the normal case is that pivot point is
ahead of wheel, so that wheel steers negative for a positive force. Also, a non-steered axle steers due to
side force steering, which depends on the compliance of the suspension bushings.
It can be modelled as an extra compliance, with the constitutional equations: = 𝑆𝐹𝑆 ∙ Δ ; and
= 𝑆𝐹𝑆 ∙ Δ ;, where the Δ marks additional steering angle due to the lateral force. These extra
compliances come into play as series connected with the tyre cornering compliances. If we update
Equation [4.15] with side force steering it becomes:
∙
= + 𝐾𝑢 ∙ ;
𝑅 𝑅
𝑙 𝑙 𝐶 𝑡𝑜𝑡 ∙𝑙 𝐶 𝑡𝑜𝑡 ∙𝑙
𝑤ℎ 𝑟 𝐾𝑢 = = ;
𝐶 𝑡𝑜𝑡 ∙ 𝐶 𝑡𝑜𝑡 ∙ 𝐶 𝑡𝑜𝑡 ∙𝐶 𝑡𝑜𝑡 ∙ [4.26]
1 1
𝑤ℎ 𝑟 𝐶 𝑡𝑜𝑡 = 𝑛 𝐶 𝑡𝑜𝑡 = ;
1 1 1 1
𝐶 + 𝑆𝐹𝑆 𝐶 +
𝑆𝐹𝑆
For vehicles with largely varying vertical axle load (such as heavy trucks), one has to consider that the
contribution from tyre to axle cornering stiffness is rather proportional to vertical axle load, while the
contribution from side-force steering comes from suspension elasticities and is rather constant. So,
utilizing side-force steering makes the vehicles lateral manoeuvrability inconsistent with vertical load.
4.3.5.3.3 Roll steer gradient
Roll steer gradient, 𝑅𝑆 , is defined for an axle and it is how much the wheels on an axle steers [deg] per
vehicle roll angle [deg]. Also, a non-steered axle steers due to roll-steering, which depends on the
suspension linkage geometry. The added steering angle can hence be expressed: Δ = 𝑅𝑆 ∙ ;
If one assumes steady state cornering, the vehicle roll angle depends on the vehicle’s roll gradient, 𝑜
as: = 𝑜 ∙ ;.For steady state, it is also possible to express = ∙ ∙( 𝑙 )⁄ ;. In total:
Δ = 𝑅𝑆 ∙ 𝑜 ∙ ∙ ⁄( ∙ ( 𝑙 )) = 𝑅𝑆 𝐹 ∙ ;. The parameter 𝑅𝑆 𝐹 comes into the
equations in exactly the same way as side force steering gradient, 𝑆𝐹𝑆 , so:
∙
= + 𝐾𝑢 ∙ ;
𝑅 𝑅
𝑙 𝑙 𝐶 𝑡𝑜𝑡 ∙𝑙 𝐶 𝑡𝑜𝑡 ∙𝑙
𝑤ℎ 𝑟 𝐾𝑢 = = ;
𝐶 𝑡𝑜𝑡 ∙ 𝐶 𝑡𝑜𝑡 ∙ 𝐶 𝑡𝑜𝑡 ∙𝐶 𝑡𝑜𝑡 ∙ [4.27]
1 1
𝑤ℎ 𝑟 𝐶 𝑡𝑜𝑡 = 𝑛 𝐶 𝑡𝑜𝑡 = ;
1 1 1 1
𝐶 + 𝑅𝑆 𝐹 𝐶 + 𝑅𝑆 𝐹
182
LATERAL DYNAMICS
4.3.5.3.4 Quantified combined effect

Side force steering and roll-steering are very similar. Roll-steering requires that sprung body changes
roll angle which takes significantly longer time; typically roll eigen-frequency is 1..2 Hz. Side force
steering does not require a roll angle change, so side-force steering has much less time delay. Roll-
steering also comes into play for one-sided road unevenness, i.e. also without cornering and without
body roll.
The combined effect from steering system (section 4.3.5.3.1), side force steering (section 4.3.5.3.2) and
roll-steering (section 4.3.5.3.3) often represent a significant part of the front axle cornering
compliance, e.g. reduces compliance with 20..40% compared to tyre cornering compliance only,
Reference (Wedlin, o.a., 1992). Most of this is due to steering system compliance.
On rear axles on passenger cars, the influence is typically less and in opposite direction, e.g. increase
1..5% compared to tyre cornering compliance only.
Rear axles on heavy vehicles are typically designed without significant side-force steering. However,
the frame compliance can cause a relevant amount of side-force oversteering due to that the whole
frame steers curve-outwards rear and curve-inwards front. The frame compliance is especially
influencive on a tractor with fifth wheel ahead of rear axle(s); clearly larger than sideforce compliance
on a single axle and on a rigid truck about the same, depending on the body-build on the frame.
Patents exist for making the rear axle suspension on heavy vehicle’s trailers sideforce steering so that
axle becomes less cornering compliant, reducing compliance with typically 1/3 compared to tyre
cornering compliance only. This increases yaw stability, which is very much same concept as using
side-force understeering rear at a two-axle vehicle. Using this concept can lead to very yaw stable
vehicles. The drawback is reduced yaw agility. If really exaggerated, it can take the rear axle to
effectively negative cornering stiffness, which makes vehicle unstable.
4.3.5.4 Camber steer

Negative camber (wheel top leaning inwards) increases the cornering stiffness. One explanation to this
is that curve outer wheel gets more vertical load than the curve inner wheel. Hence, the inwards
directed camber force from outer wheel dominates over outwards directed camber force from the
inner wheel. Negative camber is often used at rear axle at passenger cars. Drawback with non-zero
camber is tyre wear.
4.3.5.5 Toe angle

Toe has some, but limited effect on an axle’s cornering stiffness. Non-zero toe increases tyre wear. Toe-
angle: When rolling ahead, tyre side forces pre-tension bushes.
If toe (=toe-in) is positive there are tyre-lateral forces on each tyre already when driving straight
𝑡𝑜𝑒
ahead, even if left and right cancel out each other: = (𝐶 𝑒 𝑡 𝐶 ℎ𝑡 ) ∙ = ;. Then, if the axle
takes a side force, the vertical loads of the wheels are shifted between left and right wheel, which also
changes the tyre cornering stiffnesses. The outer wheel will get more cornering stiffness. Due to
positive toe, it will also have the largest steering angle. So, the axle will generate larger lateral force
than with zero toe. For steady-state cornering vehicle models, this effect comes in as an increased axle
cornering stiffness, i.e. a linear effect.
4.3.5.6 Wheel Torque effects

Wheel torque give tyre longitudinal force, directed as the wheel is pointing. If the wheel is steered, the
wheel longitudinal forces can influence the yaw balance, see also 𝑤 in Equation [4.15].
Unsymmetrical wheel torques (left/right) will give a direct yaw moment in the yaw equilibrium in
Equation [4.12]. The actuated yaw moment around CoG is then of the magnitude of wheel longitudinal
wheel force times half the track width. ESC and Torque vectoring interventions have such effects.
183
LATERAL DYNAMICS
High longitudinal utilization of friction on an axle leads to that lateral grip is reduced on that axle. The
changed yaw moment, compared to what one would have without using friction longitudinally, can be
called an indirect yaw moment. The actuated change in yaw moment around CoG is then of the
magnitude of change in wheel lateral wheel force times half the wheel base. It influences the yaw
balance. That is the reason why a front axle driven vehicle may be more understeered than a rear axle
driven one. On the other hand, the wheel-longitudinal propulsion force on the front axle does also help
the turn-in, which acts towards less understeering.
4.3.5.7 Transient vehicle motion effects on yaw balance

The effects presented here are not so relevant for steady state understeering coefficient. However,
they affect the yaw balance in a more general sense, why it is relevant to list them in this section.
• Longitudinal load transfer changes normal forces. E.g. strong deceleration by wheel forces
helps against under-steering, since front axle gets more normal load. This effect has some
delay. Also, it vanishes after the transient.
(This effect can be compared with the effect described in Section 4.3.5.2, which is caused by
tyre cornering stiffness varying degressively with vertical load, while the longitudinal load
transfer effect can be explained solely with the proportional variation.)
• Change of longitudinal speed helps later in manoeuvre. E.g. deceleration early in a manoeuvre
makes the vehicle easier to manoeuvre later in the manoeuvre. It is the effect of the term 𝜔
that is reduced.
4.3.5.8 Some other design aspects

High cornering stiffness is generally desired for controllability.
Longer wheel base (with unchanged yaw inertia and unchanged steering ratio) improves the transient
manoeuvrability, because the lateral forces have larger levers to generate yaw moment with.
4.3.6 Steady state cornering gains *

Function definition: Steady state cornering gains are the amplification from steering angle to certain vehicle
response measures for steady state cornering at a certain longitudinal speed.
From Equation [4.15], we can derive some interesting ratios. We put steering angle in the
denominator, so that we get a gain, in the sense that the ratio describes how much of something we get
“per steering angle”. If we assume 𝑤 = , we get Equation [4.28].
𝒀𝒂𝒘 𝒂𝒕𝒆 𝒈𝒂𝒊𝒏 =

𝜔 𝜔 1 ⁄𝑅 𝒗𝒙
= {𝑢𝑠 : ≈ }≈ = ;
𝑅 ∙ 𝑳 + 𝑲𝒖 ∙ 𝒎 ∙ 𝒗𝒙 𝟐
𝑅 + 𝐾𝑢 ∙ 𝑅
𝑪𝒖 𝒗𝒂𝒕𝒖 𝒆 𝒈𝒂𝒊𝒏 =
κ 1⁄𝑅 1⁄𝑅 𝟏
= = = = ; [4.28]
∙ 𝑳 + 𝑲𝒖 ∙ 𝒎 ∙ 𝒗𝒙 𝟐
𝑅 + 𝐾𝑢 ∙ 𝑅
𝑳𝒂𝒕𝒆 𝒂𝒍 𝒂𝒄𝒄𝒆𝒍𝒆 𝒂𝒕𝒊𝒐𝒏 𝒈𝒂𝒊𝒏 =
𝑢𝑠 :
=𝜔 ∙ ; 𝑅 𝒗𝒙 𝟐
= ={ 𝜔 1 }≈ = ;
𝑛 ≈ ; ∙ 𝑳 + 𝑲𝒖 ∙ 𝒎 ∙ 𝒗𝒙 𝟐
𝑅 𝑅 + 𝐾𝑢 ∙ 𝑅
Yaw rate gain is also derived for 𝑤 , and then we get Equation [4.29].
184
LATERAL DYNAMICS
𝑾𝒊𝒕𝒉 𝑭𝒇𝒙𝒘 𝒕𝒂𝒌𝒆𝒏 𝒊𝒏𝒕𝒐 𝒂𝒄𝒄𝒐𝒖𝒏𝒕: 𝒀𝒂𝒘 𝒂𝒕𝒆 𝒈𝒂𝒊𝒏 =

𝜔 𝐶 ∙ 𝐶 ∙ + (𝐶 ∙ 𝑙 + 𝐶 ∙ 𝑙 ) ∙ 𝑤 [4.29]
= = ∙
𝐶 ∙𝐶 ∙ (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙ ∙
4.3.6.1 Yaw Rate gain

30
Oversteered
25
Ku = -1.794e-6 [1/N]
yaw rate gain, wz/df [(rad/s)/rad]
20
Neutral steered
Ku = 0e-6 [1/N]
Characteristic
15
speed (for
understeered
vehicle)
10
Understeered
Ku = 2.525e-6 [1/N]
5
Critical speed (for
oversteered vehicle)
0
0 5 10 15 20 25 30 35 40 45 50
vx [m/s]
Figure 4-31: Yaw rate gain (𝜔 ) for Steady State Cornering. Each “cluster of 3 curves”: Mid curve 𝑤 =
. Upper 𝑤 =+ . 5∙ . Lower 𝑤 = . 5∙ .
The yaw rate gain gives us a way to understand Characteristic Speed. Normally one would expect the
yaw rate to increase if one increases the speed along a circular path. However, the vehicle will also
increase its path radius when speed is increased. At the Characteristic Speed, the increase in radius
cancel out the effect of increased speed, so that yaw rate in total decrease with increased speed. One
can find the characteristic speed as the speed where one senses or measures the highest value of yaw
rate increase for a given steering angle step. The curves for 𝑤 = + . 5∙ and 𝑤 = . 5∙
in Figure 4-31 are generated using the first stages in Eq [4.15], before assuming 𝑤 ≈ .
4.3.6.2 Curvature gain

3
Oversteered
Ku = -1.794e-6 [1/N]
2.5
curvature gain, (1/R)/df [1/(m*rad)]
1.5
Neutral steered
0.5 Ku = 0e-6 [1/N]
Ku = 2.525e-6 [1/N] Understeered

0
0 5 10 15 20 25 30 35 40 45 50
vx [m/s]
⁄𝑅
Figure 4-32: Curvature gain ( ) for Steady State Cornering
𝛿
185
LATERAL DYNAMICS
If driving on a constant path radius, and slowly increase speed from zero, an understeered vehicle will
require more and more steering angle (“steer-in”), to stay at the same path radius. For an over-steered
vehicle one has to steer less (“open up steering”) when increasing the speed.
4.3.6.3 Lateral Acceleration gain

Figure 4-34 shows the lateral acceleration gain as function of vehicle speed. The characteristics speed
is once again identified in this diagram, and now as the speed when lateral acceleration gain per
longitudinal speed (( ⁄ )⁄ ) reaches its highest value. This is an alternative definition of
characteristic speed, c.f. Section 4.3.4.1.
500
Oversteered
450 Ku = -1.794e-6 [1/N] Neutral steered
lateral acceleration gain, ay/df [(m/(s*s))/rad]
Ku = 0e-6 [1/N]
400
350
300
250
Understeered
200
Ku = 2.525e-6 [1/N]
150
100 Characteristic speed is

where gain curve has a
50 tangent through origin
0
0 5 10 15 20 25 30 35 40 45 50
vx [m/s]
Figure 4-33: Lateral acceleration gain ( 𝑦 ) for Steady State Cornering

𝛿
From the previous figures the responsiveness of the vehicle can be identified for different understeer
gradients. In all cases the vehicle which is understeered is the least responsive of the conditions. Both
the yaw rate and lateral acceleration cannot achieve the levels of the neutral steered or over-steered
vehicle. The over-steered vehicle is seen to exhibit instability when the critical speed is reached since
small changes in the input result in excessive output conditions. In addition, the over-steered vehicle
will have a counter-intuitive response for the driver. To maintain a constant radius curve, an increase
in speed requires that the driver turns the steering wheel opposite to the direction of desired path.
The result of these characteristics leads car manufacturers to produce understeered vehicles that are
close to neutral steering to achieve the best stability and driver feedback.
4.3.6.4 Side slip gain as function of speed

All gains above can be found from solving wz from Equation [4.15]. If instead solving the other
unknown, vy, we can draw “side slip gain” instead. Equation [4.30] shows the formula for this.
𝐶 ∙𝐶 ∙𝑙 ∙ 𝐶 ∙𝑙 ∙ ∙
= [4.30]
∙ 𝐶 ∙𝐶 ∙ (𝐶 ∙ 𝑙 𝐶 ∙𝑙 )∙ ∙
It is not solely the understeering gradient that sets the curve shape, but we can still plot for some
realistic numerical data, which are under-, neutral and over-steered, see Figure 4-34.
All cases in Figure 4-34 goes from positive side slip to negative when speed increases. This is the same
as we expected already in Figure 4-24.
186
LATERAL DYNAMICS
m = 2000 [kg]; lf = 1.3 [m]; lr = 1.5 [m]

1
understeered
lr/L neutral steered
0.5
oversteered
"side slip gain", (vy/vx)/df [1/rad]

0
-0.5
-1
Understeered
-1.5
-2
Oversteered
Neutral steered
-2.5
-3
0 5 10 15 20 25
vx [m/s]
𝑦
Figure 4-34: “Side slip gain” ( ) for Steady State Cornering.
𝑥 ∙𝛿
4.3.6.5 Limitations due to max road friction

The formulas with the gains in Eq [4.28] does not include the limitation due to maximum road friction,
i.e. the peak of the tyres’ (𝑠 ) curves. This limitation is a lateral acceleration 𝑚 where on of the
axles reach friction limit. Yaw and lateral equilibrium requires = ∙ 𝑙 ⁄ ; and = ∙
𝑙 ⁄ ;, so the limit due to front axle is ∙ 𝑚 𝑙 ⁄ = = 𝑙 ⁄ ⇒ 𝑚 =
; and corresponding for rear axle. The total limit is 𝑚 = min( ) ; Inserting this in Eq
[4.28] and eliminate , one finds:
𝑚 =√ 𝑚 𝑅 = √min( ) 𝑅; [4.31]
For example, if negotiating a curve with 𝑅 = 1 and = = 1 gives 𝑚 ≈ 31 ⁄𝑠 ≈

113 𝑘 ⁄ℎ and if either (or both) of and are just 0.5, 𝑚 ≈ ⁄𝑠 ≈ 8 𝑘 ⁄ℎ. So, we should
see the curves in Figure 4-28 and in Figure 4-31 to Figure 4-34 as invalid over certain speed 𝑚 ,
which depends on max road friction and curvature.
Reaching 𝑚 means loss of steerability if < , and loss of (yaw) stability if < . So, also a
understeered vehicle becomes (yaw) unstable, if < and driving faster than 𝑚 . And, an
oversteered vehicle might be limited from reaching its critical speed if < and driving on a certain
radius 𝑅.
4.3.7 Manoeuvrability and Stability

The overall conclusion of previous section is that all gains become higher the more over-steered (or
less understeered) the vehicle is. Higher gains are generally experienced as a sportier vehicle and it
also improves safety because it improves the manoeuvrability. A higher manoeuvrability makes it
easier for the driver to do avoidance manoeuvres. This would motivate a design for low understeering
gradient.
However, there is also the effect that a vehicle with too small understeer gradient becomes very
sensitive to the steering wheel angle input. In extreme, the driver would not be able to control the
vehicle. This limits how small the understeering gradient one can design for.
The overall design rule today is to make the understeering as small as possible, but the neutral steer is
a limit which cannot be passed.
187
LATERAL DYNAMICS
It is not impossible for a driver to keep an unstable vehicle (Ku<0 and vx>Critical Speed) on an
intended path, but it requires an active compensation with steering wheel. If adding support systems,
such as yaw damping by steering support or differentiated propulsion torques, it can be even easier. If
one could rely on a very high up time for such support systems, one could move today’s trade-off
between manoeuvrability and stability. This conceptual design step has been taken for some airplanes,
which actually are designed so that they would be unstable without active control.
4.3.8 Handling diagram

This section gives a brief introduction to handling diagram, which is a useful tool for discussing yaw
stability during different steady state cornering situations, as initiated with Equation [4.20].
There are many frequently used graphical tools or diagrams to represent vehicle characteristics. One
is the “handling diagram”, which is constructed as follows. Same simplifying assumptions are done as
in Figure 4-25, with the exception that we don’t assume linear tyre models.
Equilibrium:
𝑙 𝑙
{ ∙ = ∙ = + ; = ∙𝑙 ∙ 𝑙 ;} ⇒ = ∙ ∙ ; = ∙ ∙ ;
𝑅
Constitution:
− −
= (𝛼 ) ⇒ 𝛼 = ( ); = (𝛼 ) ⇒ 𝛼 = ( );
Solving for 𝛼 𝛼 yields:
− − − 𝑙 − 𝑙
𝛼 𝛼 = ( ) ( )= ( ∙ ∙ ) ( ∙ ∙ );
So, we can plot 𝛼 𝛼 as function of . This relation is interesting because compatibility ( +𝛼

𝛼 = ⁄𝑅 ;) yields 𝛼 𝛼 = ⁄𝑅 = 𝐴 . And 𝐴 is connected to one of the
𝜕
understandings of 𝐾𝑢 in Equation [4.20], (𝐾𝑢 = ( 𝐴 );). If we plot 𝛼 𝛼 = 𝐴 on
𝜕 𝑦
abscissa axis and on ordinate axis, we get the most common way of drawing the handling diagram,
see Figure 4-35. The axle’s constitutive relations can be used as graphical support to construct the
𝐿
diagram, but then the constitutive relations should be plotted as: (𝛼 ) = ∙ ∙ (𝛼 );. The
𝐿− 𝑖 𝑚
quantity can be seen as the lateral force on the axle, but scaled so that both axles’ values
correspond to the same vehicle lateral acceleration.
Figure 4-35 shows the construction of a handling diagram from axle slip characteristics. Figure 4-36
show examples of handling diagrams constructed via tests with simulation tools. Handling diagrams
can be designed from real vehicle tests as well. The slope in the handling diagram corresponds to
understeering gradient 𝐾𝑢 in Equation [4.18].
188
LATERAL DYNAMICS
𝐿
front axle, = = 𝛼
𝑟 𝑚
𝐿
rear axle, = =
𝑚
𝛼
𝛼 Using the axle

models above,
Handling diagram the vehicle
becomes
Linear axle models, oversteered
understeered vehicle
Normal vehicle stays

understeered
𝐴 ≈𝛼 𝛼 Try to get some

measured as well.
From VCC, Saab?
Figure 4-35: Construction of the “Handling diagram”. The axle’s slip characteristics (solid) are chosen so
that vehicle transits from understeer to over-steer with increased longitudinal speed, vx. The dashed shows
two other examples.
From: Daniel A. Fittanto, et al. “Passenger Vehicle Steady-State Directional Stability Analysis
Utilizing EDVSM and SIMON”, Copyright 2004 by Engineering Dynamics Corporation
Figure 4-36: Example of handling diagram.
(EDVSM and SIMON are two different simulation tools.)
189
LATERAL DYNAMICS
Using a one-track model of a two-axle vehicle, there is one unique such curve as drawn in Figure 4-35.
However, from a more advanced model, a vehicle with non-Ackermann steering or a real test or an
articulated vehicle, there are different curves. This is so because a certain lateral acceleration can be
reached in different ways. For instance, the sweep along the lateral acceleration axis can be made by
driving at a constant radius with increasing speed (R-handling diagram) or driving with constant
speed on a decreasing radius (V-handling diagram). An R-handling diagram can be drawn for each
combination of vehicle and radius, so that in one diagram one can often see several curves which are
for different radii. These diagrams are often seen for heavy trucks with several non-steered axles. With
data according to Section 4.2.8 and 𝐶 = 𝐶 = 𝐶 ; one can derive different handling curves for
different :
𝑏 𝐶 1 𝐶 𝑙 𝐶 𝑙
= + 𝐾𝑢 + (1 + ) ; 𝐾𝑢 = ;
𝑅 𝑅 𝐶 𝑅 𝐶 𝐶
[4.32]
𝑏 𝐶 1
= + 𝐾𝑢 + (1 + ) ;
𝑅 ⏟ 𝐶
( 𝑛 𝑢𝑒𝑛𝑐𝑒 𝑜 𝑒 𝑒𝑠)
Alternatively, one can identify an “equivalent wheelbase”:
𝑏 𝐶 1 𝑏 𝐶 1
= + 𝐾𝑢 + (1 + ) =( + (1 + )) + 𝐾𝑢 ;
𝑅 𝑅 𝐶 𝑅 ⏟ 𝐶 𝑅 𝑅 [4.33]
𝑒𝑞𝑢 𝑒𝑛𝑡 𝑤ℎ𝑒𝑒 𝑠𝑒
4.3.9 Lateral Load Transfer at Steady state cornering

at high speed
In the chapter about longitudinal dynamics we studied (vertical tyre) load transfer between front and
rear axle. The corresponding issue for lateral dynamics is load transfer between left and right side of
the vehicle. Within the steady state lateral dynamics, we will cover some of the simpler effects, but
save the more complex suspension linkage dependent effects to Sections 4.5.
The relevance to study the load transfer during steady state cornering is to limit the roll during
cornering (for comfort) and yaw balance (understeering gradient, see Section 4.3.5.2). Additionally,
the load transfer influence the transient handling; see Section 4.4 and Section 4.5.
4.3.9.1 Load transfer between vehicle sides

The total load transfer is without resolving into front and rear. It can be found without involving
suspension effects.
Moment equilibrium, around left contact with ground: ⇒

𝑤 ℎ
⇒ ∙ ∙ + ∙ ∙ℎ ∙𝑤 = ⇒ = ∙( + ∙ ); [4.34]
𝑤
ℎ
Moment equilibrium, around right contact with ground: ⇒ = ∙( ∙ 𝑤) ;
These equations confirm what we know from experience, the left side if off-loaded if turning left.
Generally, curve inner side is off-loaded.
190
LATERAL DYNAMICS
View from rear:

m*g
curve inner side m*ay curve outer side
(if ay>0, i.e. (if ay>0 , i.e.
h
turning left) Fyl turning left)
Fzl Fzr Fyr
w/2 w/2
Figure 4-37: Free Body Diagram for cornering vehicle.
The force m*ay is a fictive force. Subscript l and r here means left and right.
4.3.9.2 Body heave and roll due to lateral wheel forces

Now, we shall find out how much the vehicle rolls and heaves during steady state cornering. First, we
decide to formulate the model in “effective stiffnesses”, in the same manner as for longitudinal load
transfer in previous chapter.
displacement velocity, due to the “quasi-steady-state” assumption. As constitutive equations for the
the compliances are linear. The road is assumed to be smooth, i.e. = = .
= 0 + 𝑠 𝑑𝑒 ∙( )+ ∙ (( ) ( )) =
={ = }
= = 0 ( 𝑠 𝑑𝑒 + )∙ + ∙ ;
[4.35]
=⋯= 0 ( 𝑠 𝑑𝑒 + )∙ + ∙ ;
𝑤ℎ 𝑟 0 + 0 = ∙ ; 𝑛 0 ∙ 𝑤⁄ 0 ∙ 𝑤⁄ = ;
The stiffnesses 𝑠 𝑑𝑒 and are effective stiffnesses as measurable under the wheels. The physical
springs are mounted inside in some kind of linkage and have different stiffness values, but their effect
is captured in the effective stiffnesses. Some examples of different physical spring and linkage design
are given in Section 4.3.9.4.
-y
(Anti-roll is drawn, only in left
picture and beside the vehicle,
ay z for better clarity in drawing.)
y m*ay
zl zr
px , anti-roll bar
torsional spring
𝑠 𝑑𝑒
h 𝑠 𝑑𝑒 m*g
Fyl Fyr hRC
zlr=0 Fzr
zrr=0 Fzl
w/2 w/2 w/2 w/2
Steady-state assumed, so that
z,y,px,zl,zr, are displacements from lateral acceleration (ay) may be
a static stand-still position. non-zero, but vertical and roll
zlr=zrr=0 means that road is smooth. acceleration are zero.
Figure 4-38: Model for steady state heave and roll due to lateral acceleration. Suspension model is no
linkage (or “trivial linkage”) and without difference front and rear.
191
LATERAL DYNAMICS
We see already in free-body diagram that and always act together, so we rename + =
. This and equilibrium gives:
∙ = ;
∙ = ; [4.36]
∙ (𝑤⁄ ) ∙ (𝑤⁄ ) + ∙ℎ+ ∙ ∙ ( 𝑦) = ;
The term ∙ ∙ ( 𝑦) is taken into account, but not corresponding for ∙ , since symmetry of the
vehicle motivates that the roll takes place around a point on the vertical symmetry axis. Therefore 𝑦 is
significant but corresponding displacement in vertical direction is not. If we assume a height for the
point where the roll takes place, ℎ𝑅𝐶 ,we can express – 𝑦 = (ℎ ℎ𝑅𝐶 ) ∙ 𝑝 . We don’t know the value of
it, until below where we study the suspension design, but it can be mentioned already here that most
vehicles have an ℎ𝑅𝐶 ≪ ℎ. This causes a significant “pendulum effect”, especially for heavy trucks.
Compatibility, to introduce body displacements, z and px, gives:
= + (𝑤⁄ ) ∙ 𝑝 ; 𝑛 = (𝑤⁄ ) ∙ 𝑝 ; [4.37]
Combining constitutive relations, equilibrium and compatibility, gives, as Matlab script:

clear, syms zl zr Fzl Frz Fzl0 Frz0 Fy z px
sol=solve( ...
'Fzl=Fzl0-(cside+carb)*zl+carb*zr', ...
'Frz=Frz0-(cside+carb)*zr+carb*zl', ...
'Fzl0+Frz0=m*g', ...
'Fzl0*w/2-Frz0*w/2=0', ...
'Fy-m*ay=0', ... [4.38]
'm*g-Fzl-Frz=0', ...
'Fzl*(w/2)-Frz*(w/2)+Fy*h+m*g*(h-hRC)*px=0', ...
'zl=z+(w/2)*px', ...
'zr=z-(w/2)*px', ...
' Fzl1=-1/((Fzl/m-g/2)*w/(ay*h))', ...
zl, zr, Fzl, Frz, Fzl0, Frz0, Fy, z, px);
The results from the Matlab script in Equation [4.38]:
= ∙ ;
= ;
∙ ∙ℎ
𝑝 = ;
( 𝑠 𝑑𝑒 + ∙ )∙𝑤 ∙ ∙ (ℎ ℎ𝑅𝐶 )
∙ℎ 1
= ∙ ∙( ) ; [4.39]
𝑤 ∙ ℎ ℎ𝑅𝐶
1 ∙
𝑠 𝑑𝑒 + 𝑤
( )
∙ℎ 1
= ∙ + ∙( ) ;
𝑤 ∙ ℎ ℎ𝑅𝐶
1 ∙
𝑠 𝑑𝑒 + 𝑤
( )
In agreement with intuition and experience the body rolls with positive roll when steering to the left
(positive 𝑤 ). Further, the body centre of gravity is unchanged in heave (vertical motion, ). The
formula uses ℎ𝑅𝐶 which we cannot estimate without modelling the suspension. Since front and rear
axle normally are different, we could expect that ℎ𝑅𝐶 is expressed in some similar quantities for each
of front and rear axle, which also is the case, see Equation [4.46].
192
LATERAL DYNAMICS
4.3.9.2.1 Steady-state roll-gradient *

Function definition: Steady state roll-gradient is the body roll angle per lateral acceleration for the vehicle
during steady state cornering with a certain lateral acceleration and certain path radius on level ground.
4.3.9.3 Lateral load transfer models of suspension linkage

For longitudinal load transfer, during purely longitudinal dynamic manoeuvres, the symmetry of the
vehicle makes it reasonable to split vertical load on each axle equally between the left and right wheel
of the axle. However, for lateral dynamics it is not very realistic to assume symmetry between front
and rear axle. Hence, the suspension has to be considered separately for front and rear axle. The
properties that are important to model for each axle is not only left and right elasticity (as we
modelled the whole vehicle in Figure 4-38). It is also how the lateral tyre forces are transmitted from
road contact patches to the vehicle body. We end up with conceptually the same two possible linkage
modelling concepts as we found for longitudinal load transfer, see Figure 3-45. Either we can
introduce roll centre heights for each axle (c.f. pitch centre in Section 3.4) or we can introduce the
pivot point for each wheel (c.f. axle pivot points in Section 3.4). A difference for lateral load transfer,
compared to longitudinal load transfer, is that it is significant also at steady state (due to centrifugal
force). The two modelling ways to include the suspension in the lateral load transfer are shown in
Figure 4-39. Generally speaking, they can be combined, so that one is used on front axle and the other
on rear axle. In this compendium, we will select the roll-centre model when modelling.
One should differ between roll-centre heights and roll-centres. One can say that roll-centre heights
exists while roll-centres is only a model concept. If a vehicle was actually designed with a roll-centre as
being a real pivot point between axle and body, that vehicle would be totally rigid in vertical direction.
4.3.9.3.1 Load Transfer model with Wheel Pivot Points
This model will not be deeply presented in this compendium. However, it should be mentioned as
having quite a few advantages:
• The model has both heave and roll degree of freedom. (Roll centre model is restricted to roll
around roll centre.)
• The model does take the distribution of lateral wheel forces between left and right side into
account. (Roll centre model only uses the sum of lateral forces per axle and needs involvement
of tyre model to resolve into individual left and right side forces.)
Generally spoken, this model is more accurate and not much more computational demanding and
probably easier to intuitively understand.
Note that non-individual, rigid axles or beam axles, the pivot-point model does not have one pivot
point for each wheel, but instead two dofs for the whole axle: heave and roll. The roll-centre height
comes then in as the height where lateral force is transferred between axle and body.
Cases when this model is recommended as opposed to the model with roll centres are:
• Steady state and transient manoeuvres where the heave displacement is important.
• When large differences between lateral load on left and right wheels are present, such as:
• Large load transfer, i.e. high CoG and large lateral accelerations. One example is when studying
wheel lift and roll-over tendencies.
o Large differences between longitudinal slip, while axle skids sideways. Then one wheel
might have zero lateral force, due to that friction is used up longitudinally, while the other
can have a large lateral force.
o If individual steering within an axle would be studied. One could think of an extreme case
if actuating a sudden toe-in or toe-out, which would cause large but counter-directed
lateral forces on left and right wheel.
193
LATERAL DYNAMICS
Model with wheel pivot points Model with axle roll centres
Transversal sections from rear over front axle:
h h
ef hRCf
gf
w/2 w/2
w/2 w/2
Transversal sections from rear over rear axle:
h h
hRCr
er
gr
w/2 w/2
w/2 w/2
Figure 4-39: Two alternative models for including suspension linkage effects (kinematics) in lateral load
transfer. Anti-roll bars not drawn.
4.3.9.3.2 Load Transfer model with Axle Roll Centres
The model with axle roll centres has some drawback as listen before. To mention some advantages, it
is somewhat less computational demanding. However, the main reason why using the roll centre
based model in this compendium is that the compendium then cover two different concepts with
longitudinal and lateral load transfer.
Behold the free-body diagrams in Figure 4-40. The road is assumed to be flat, = = =
= . In free-body diagram for the front axle, Pfz and Pfy are the reaction force in the rear roll-
centre. Corresponding reaction forces are found for rear axle. Note that roll centres are free of roll
moment, which is the key assumption about roll centres! 𝑠 𝑠 𝑠 𝑛 𝑠 are the forces in the
compliances, i.e. where potential spring energy is stored. One can understand the roll-centres as also
unable to take vertical force, as opposed to constraining vertical motion (as drawn). Which of
vertically force-free or vertically motion-free depends on how one understands the concept or roll-
centre, and it does not influence the equations.
Note carefully that the “pendulum effect” is NOT included here, in section 4.3.9.3, as it was in section
4.3.9.2. The motivation is to get simpler equations for educational reasons.
194
LATERAL DYNAMICS
displacements, speeds, accelerations: forces, on body:

ay m*g
z
zfl=zrl=0 y
zfr=zrr m*ay
px
Pry
Pfy
Only one anti-roll bar

spring drawn, but there is Fsfl+Fsrl Fsfr+Fsrr
zflr=zrlr=0
+
zfrr=zrrr=0 one per axle: and .
forces, forces, on front axle: forces, on rear axle:

on vehicle: m*g Fsfl Fsfr Fsrl Fsrr
m*ay 𝑟
Pry
h Faf Pfy Faf Far
Far
Ffly+Frly Ffly Ffry Frly Frry

Ffry+Frry
Fflz Ffrz Frlz Frrz
Fflz+Frlz Ffrz+Frrz
Figure 4-40: Model for steady state heave and roll due to lateral acceleration, using roll centres, which can
be different front and rear.
displacement velocity, due to the steady-state assumption. As constitutive equations for the
the compliances are linear. Note that there are two elasticity types modelled: springs per wheel ( 𝑤
per front wheel and 𝑤 per rear wheel) and anti-roll bars per axle ( front and rear). The road is
assumed to be smooth, i.e. = = = = .
𝑠 = 𝑠 0 + 𝑤∙( );
𝑠 = 𝑠 0+ 𝑤∙( );
𝑠 = 𝑠 0+ 𝑤∙( );
𝑠 = 𝑠 0+ 𝑤∙( );
[4.40]
= + ∙ (( ) ( )) ;
= + ∙ (( ) ( ));
∙ ∙𝑙 ∙ ∙𝑙
𝑤ℎ 𝑟 𝑠 0 = 𝑠 0 = ; 𝑛 𝑠 0 = 𝑠 0 = ;
∙ ∙
The stiffnesses 𝑤 𝑤 and are the effective stiffnesses per wheel. We see already in free-
body diagram that and always act together, so we rename + = and + =
.
Equilibrium for whole vehicle (vertical, lateral, yaw, pitch, roll):
195
LATERAL DYNAMICS
+ + + = ∙ ;
∙ = + ;
= ∙𝑙 ∙𝑙 ; [4.41]
( + )∙𝑙 +( + )∙𝑙 = ;
𝑤 𝑤
( + )∙ ( + )∙ +( + )∙ℎ= ;
Equilibrium for each axle (roll, around roll centre):

𝑤 𝑤
( 𝑠 + )∙ ( 𝑠 )∙ + ∙ ℎ𝑅𝐶 = ;
𝑤 𝑤 [4.42]
( 𝑠 + )∙ ( 𝑠 )∙ + ∙ ℎ𝑅𝐶 = ;
Compatibility, to introduce body displacements, z, px and py, gives:

𝑤 𝑤
= + ∙𝑝 𝑙 ∙𝑝 ; 𝑛 = ∙𝑝 𝑙 ∙𝑝 ;
𝑤 𝑤 [4.43]
= + ∙𝑝 +𝑙 ∙𝑝 ; 𝑛 = ∙𝑝 +𝑙 ∙𝑝 ;
+ = ; 𝑛 + = ;
The measure Dh is redundant and can be connected to the other geometry measures as follows. The
geometrical interpretation is given in Figure 4-41.
𝑙 ∙ ℎ𝑅𝐶 + 𝑙 ∙ ℎ𝑅𝐶
∆ℎ = ℎ ; [4.44]
Combining Equations [4.40] to [4.44] gives, as Matlab script and solution:

clear, syms zfl zfr zrl zrr Fflz Ffrz Frlz Frrz Fsfl Fsfr Fsrl
Fsrr Faf Far Fsfl0 Fsfr0 Fsrl0 Fsrr0 Ffy Fry z px py h
sol=solve( ...
'Fsfl=Fsfl0-cfw*zfl', ...
'Fsfr=Fsfr0-cfw*zfr', ...
'Fsrl=Fsrl0-crw*zrl', ...
'Fsrr=Fsrr0-crw*zrr', ...
'Faf=0-caf*(-zfl+zfr)', ...
'Far=0-car*(-zrl+zrr)', ...
'Fsfl0=(1/2)*m*g*lr/L', ...
'Fsfr0=(1/2)*m*g*lr/L', ...
'Fsrl0=(1/2)*m*g*lf/L', ...
'Fsrr0=(1/2)*m*g*lf/L', ...
'Fflz+Ffrz+Frlz+Frrz=m*g', ...
'm*ay=Ffy+Fry', ...
'0=Ffy*lf-Fry*lr', ... [4.45]
'-(Fflz+Ffrz)*lf+(Frlz+Frrz)*lr=0', ...
'(Fflz+Frlz)*w/2-(Ffrz+Frrz)*w/2+(Ffy+Fry)*h=0', ...
'(Fflz-Fsfl+Faf)*w/2-(Ffrz-Fsfr-Faf)*w/2+Ffy*hRCf=0', ...
'(Frlz-Fsrl+Far)*w/2-(Frrz-Fsrr-Far)*w/2+Fry*hRCr=0', ...
'zfl=z+(w/2)*px-lf*py', ...
'zfr=z-(w/2)*px-lf*py', ...
'zrl=z+(w/2)*px+lr*py', ...
'zrr=z-(w/2)*px+lr*py', ...
'zfl+zfr=0', ...
'zrl+zrr=0', ...
'dh=h-(lr*hRCf+lf*hRCr)/(lf+lr)', ...
zfl, zfr, zrl, zrr, Fsfl, Fsfr, Fsrl, Fsrr, …
Faf, Far, Fsfl0, Fsfr0, Fsrl0, Fsrr0, ...
Fflz, Ffrz, Frlz, Frrz, Ffy, Fry, z, px, py, h);
196
LATERAL DYNAMICS
The result from the Matlab script in Equation [4.45], but in a prettier writing format:
𝑙 𝑙
= ∙ ∙ ; 𝑛 = ∙ ∙ ;;
∙ ∙ ∆ℎ ( + ) ∙ ∆ℎ
= ; 𝑛 𝑝 = = ; 𝑛 𝑝 = ;
𝑒ℎ 𝑐 𝑒 𝑜 𝑒ℎ 𝑐 𝑒 𝑜
∙𝑙 ℎ𝑅𝐶 ∙ 𝑙 ∆ℎ 𝑜
= ∙( ∙( + ∙ )) ;
∙ ∙𝑤 𝑤 𝑒ℎ 𝑐 𝑒 𝑜
= ∙( + ∙( + ∙ )) ;
= ∙( ∙( + ∙ )) ; [4.46]
= ∙( + ∙( + ∙ )) ;
where, roll stiffnesses in [moment/angle], are:

𝑤 𝑁
𝑜 = ∙( 𝑤 + ∙ )∙( ) ];
[
𝑟
𝑤 𝑁
𝑜 = ∙( 𝑤 + ∙ )∙( ) [ ];
𝑟
𝑁
𝑒ℎ 𝑐 𝑒 𝑜 = 𝑜 + 𝑜 [ ];
𝑟
The axle roll stiffnesses, 𝑜 and 𝑜 are identified beside vehicle roll stiffness 𝑒ℎ 𝑐 𝑒 𝑜 . We
should compare Equation [4.46] with Equation [4.39]. Equation [4.39] considers the “pendulum
effect”, but not the differentiation between front and rear suspension. Equation [4.46] does the
opposite.
Assume ℎ = ℎ𝑅𝐶 and look at the sum of vertical force on one side, in Equation [4.39]. Compare
in Equation [4.39] and + in Equation [4.46]; the equations agree if:
ℎ𝑅𝐶 ∙ 𝑙 + ℎ𝑅𝐶 ∙ 𝑙 ∆ℎ ℎ
+ = ⇒ ∙( ∙( + )) = ∙( ∙ )= ⇒
∙𝑤 𝑤 𝑤
ℎ𝑅𝐶 ∙ 𝑙 + ℎ𝑅𝐶 ∙ 𝑙 ∆ℎ ℎ
⇒ + = ⇒ ℎ𝑅𝐶 ∙ 𝑙 + ℎ𝑅𝐶 ∙ 𝑙 = (ℎ ∆ℎ) ∙ ;
∙𝑤 𝑤 𝑤
This is exactly in agreement with the definition of the redundant geometric parameter Dh, see
Equation [4.44]. This means that a consistent geometric model of the whole model is as drawn in
Figure 4-41. Here the artefact roll axis is also defined.
The terms of type ℎ𝑅𝐶 ∙ 𝑙𝑗 ⁄( 𝑤) in Equation [4.46] can be seen as the part of the lateral tyre forces
that goes via the stiff linkage. The terms of type (Δℎ⁄𝑤 ) ( 𝑤 ⁄( 𝑤 + 𝑗𝑤 )) in Equation [4.46] can be
seen as the part of the lateral tyre forces that goes via the compliance. The latter part is distributed in
proportion to roll stiffness of the studied axle, as a fraction of the vehicle roll stiffness. This should be
in agreement with intuition and experience from other preloaded mechanical systems (load
distributes as stiffness).
Body rolls with positive roll when steering to the left, as long as CoG is above roll axle. Further, the
body centre of gravity is unchanged in heave (vertical z) because the model does not allow any vertical
displacements, which is a drawback already mentioned.
197
LATERAL DYNAMICS
front axle
CoG roll centre
front axle
Dh
hRCf
rear axle
roll centre
h
rear axle
𝑙 ℎ𝑅𝐶 + 𝑙 ℎ𝑅𝐶
hRCr
∆ℎ = ℎ
Figure 4-41: Roll axis for a two axle vehicle. (Note that the picture may indicate that the roll centres and roll
axis are above wheel centre, but this is normally not the case.)
4.3.9.4 Axle suspension system

Suspension design is briefly discussed at these places in this compendium: Section 3.4.6, Section
4.3.9.4 and Section 5.2. There are axles with dependent wheel suspensions, which basically look as the
roll centre axle model in Figure 4-39 , i.e. that left right wheel are rigidly connected to each other.
Then, there are axles with dependent wheel suspensions, which look more like the model with wheel
pivot points in Figure 4-39. For these, there are no (rigid) connections between left and right wheel.
Many axles have a so called anti-roll bar, which is a elastic connection between left and right side. It is
connected such that if the wheel on one side is lifted, it lifts also the wheel on the other side. Note that,
if an anti-roll bar is added to an independent wheel suspension it is still called independent, because
the connection is not rigid.
Figure 4-42 and Figure 4-43 show design of two axles with independent wheel suspensions. Figure
4-44 shows an axle with dependent wheel suspension. These figures shows how to find wheel pivot
points and roll centre. In the McPherson suspension in Figure 4-43, one should mention that the strut
is designed to take bending moments. For the rigid axle in Figure 4-44, one should mention that the
leaf spring itself takes the lateral forces. Symmetry between left and right wheel suspension is a
reasonable assumption and it places the roll centre symmetrically between the wheels, which is
assumed in the previous models and equations regarding roll centre.
motion for points

moving with hub Pivot point
for wheel
90 deg
motion for a point

moving with hub, where From the other
the wheel which is in 90 deg wheel on same
contact with ground Roll centre
axle
for axle
90 deg
Figure 4-42: Example of how to appoint the pivot point for one wheel and roll centre for axle with double
wishbone suspension.
198
LATERAL DYNAMICS
motion for
90 deg
points moving
with hub Pivot point
for wheel
motion for a
point moving
with hub, where
the wheel which
is in contact
with ground
From the other
90 deg Roll centre wheel on same
for axle axle
90 deg
Figure 4-43: Example of how to appoint the pivot point for one
=
wheelgeneral
and motion
roll centre for axle with double
(which, in the
radial motion + McPherson
angular motionsuspension.
pivot&slide point has to be
perpendicular to the slide).
Generally, a ”rigid axle” gives roll centre height on approximately the same magnitude as wheel radius,
see Figure 4-44. With individual wheel suspension one have much larger flexibility, and typical chosen
designs are 30..90 mm front and 90..120 mm rear.
Figure 4-44: Example of how to appoint the pivot point for one wheel and roll centre for axle with rigid axle
suspended in leaf springs. From (Gillespie, 1992).
The design of roll centre height is a trade-off. On one side, high roll centre is good because it reduces
roll in steady state cornering. On the other side, low roll centre height is good because it gives small
track width variations due to vehicle heave variations. Track width variations are undesired, e.g.
because it makes the left and right tyre lateral force fight against each other, leaving less available
friction for longitudinal and lateral grip. Roll centre is normally higher rear than front. One reason for
that is that the main inertia axis leans forward, and parallelism between roll axis and main inertia axis
is desired.
4.3.10 High speed steady state vehicle functions

4.3.10.1 Steering feel *
Function definition: Steering feel is the steering wheel torque response to steering wheel angle. The function is
used in a very wide sense; on a high level, it is a measure of steering wheel torque, or its variation, for certain
driving situations. Often, it can only be subjectively assessed.
At steady state driving at high speed, there are basically three aspects of steering feel:
• Lateral steering feel feedback at cornering. The steering wheel torque is normally desired to
increase monotonously with lateral forces on the front axle. This is basically the way the
199
LATERAL DYNAMICS
mechanics work due to caster trail. Some specifications on steering assistance system is
however needed to keep the steering wheel torque low enough for comfort.
• Steering torque drop when cornering at low-friction. It is built into the mechanics of the caster
trail and the pneumatic trail that steering wheel torque drops slightly when one approaches
the friction limit on front axle. This is normally a desired behaviour because it gives driver
feedback that the vehicles is approach the limits.
• On-centre feel in straight line driving. When the vehicle is driven in straight line, the steering
wheel is normally desired to return to centre position after small perturbations. This is a
comfort function, which OEMs works a lot with and it is often rather subjectively assessed.
4.3.10.2 High Speed Steady-State Off-tracking *

Function definition: High speed steady-state off-tracking is the lateral offset between the paths of the
centre of the front axle and the centre of the most severely off-tracking axle of any unit in a steady turn at a certain
friction level and a certain constant longitudinal speed. From Reference [ (Kati, 2013)].
The function is mainly of interest for long combination vehicles, as illustrated in Figure 4-45. Off-
tracking was also mentioned in Section 4.2. It measures the lateral road space required. High speed
Off-tracking, which is an outboard off-tracking, can be either determined in a steady state turn or in a
transient manoeuvre such as lane change; the latter is termed as high speed transient off-tracking, see
Section 4.5.6.2.
Figure 4-45: Illustration of high speed off-tracking. From (Kharrazi , 2012).
4.3.10.3 Tracking-ability on straight path *

Function definition: Tracking-ability on straight path is the swept width between outer-most axle centres
when driving at a road with certain cross-fall and certain road friction at a certain speed.
The axles on any vehicle driving at a road with cross-fall will not track exactly in each other’s
trajectories. This is especially pronounced if long vehicles with many articulation points. The driving
situation is straight steady state low or high speed.
200
LATERAL DYNAMICS
Tracking ability is
road and 𝑟 𝑡𝑦
this swept width
translation
direction
cross-fall
direction
𝛽
𝛽 𝛽 𝛽
𝑤
𝛽
Figure 4-46: Tracking ability on straight path for an “A-double” (Tractor-SemiTrailer-Dolly-SemiTrailer).

An example is seen in Figure 4-46. If we neglect longitudinal forces and combined tyre slip effects, and
assume same cornering coefficient, 𝐶𝐶, on all axles we can derive these equations:
• 𝑤 = 𝐶𝐶 (𝛽 );
• = 𝐶𝐶 𝛽;
• = 𝐶𝐶 𝛽 ; 𝑜𝑟 = . .4
Since the levers for moment equilibria in road x-y-plane and in road x-z-plane are equal, the
distribution of the axles’ vertical forces and lateral forces becomes identical. So, relation between
𝐹
lateral force and vertical force becomes 𝐹𝑖𝑦 = tan( )≈ for all axles:
𝑖
• = 𝐶𝐶 (𝛽 );
• = 𝐶𝐶 𝛽 ; 𝑜𝑟 = 1. .4
Solving for steering angle and side slip angles:
𝜑𝑟𝑥
• 𝛽 = 𝐶𝐶
= 𝛽; 𝑜𝑟 = 1. .4;
𝜑𝑟𝑥 𝜑𝑟𝑥
• =𝛽 + = ;
𝐶𝐶 𝐶𝐶
The swept width becomes: 𝑆𝑤 𝑝𝑡𝑊 𝑡ℎ = (𝑙 +𝑙 𝑐) + (𝑙 𝑐 +𝑙 𝑐) + (𝑙 𝑐 +𝑙 𝑐) 3
𝐶𝐶 𝐶𝐶
𝐶𝐶
+ (𝑙 𝑐 +𝑙 𝑐) 4 𝐶𝐶
; This can be expressed as:
𝑆𝑤 𝑝𝑡𝑊 𝑡ℎ = ((𝑙 +𝑙 𝑐) + ∑ ((𝑙 𝑐 +𝑙 𝑐) )) ; [4.47]

𝐶𝐶
=
There is no influence of longitudinal speed. There is influence of cornering coefficient 𝐶𝐶 but not
cornering coefficient 𝐶 .
Model validity assumes that the influence of longitudinal force is small. These will affect via
articulation angles and via combined tyre slip effects. So, the model will be less valid if low road
friction and if strong up- or downhill.
4.3.11 Roll-over in steady state cornering

When going in curves, the vehicle will have roll angles of typically some degrees. At that level, the roll
is a comfort issue. However, there are manoeuvres which can cause the vehicle to roll-over, i.e. roll ≥
201
LATERAL DYNAMICS
9 so that vehicle body crashes into ground. Roll-over can be seen as a special event, but if sorting
into the chapters of this compendium it probably fits best in present chapter, about lateral dynamics.
One can categorize roll-overs in e.g. 3 different types:
• Tripped roll-over. This is when the car skids sideways and hits an edge, which causes the roll-
over. It can be an uprising edge, e.g. pavement or refuge. It can be the opposite, a ditch or loose
gravel outside road. In both these cases, it is strong lateral forces on the wheels on one side of
the vehicle that causes the roll-over.
Tripped roll-over can also be when the vehicle is exposed to large one-sided vertical wheel
forces, e.g. by running over a one-sided bump.
A third variant of tripped roll-over is when the vehicle is hit by another vehicle so hard that it
rolls over.
• Un-tripped roll-over or on-road roll-overs. These happen on the road and triggered by high
tyre lateral forces. This is why they require high road friction. For sedan passenger cars, these
event are almost impossible, since road friction seldom is higher than approximately 1. For
SUVs, un-tripped roll-overs can however occur but require dry asphalt roads, where friction is
around 1. For trucks, un-tripped roll-over, can happen already at very moderate friction, like
0.4, due to their high CoG in relation to track width. Within un-tripped roll-overs, one can differ
between:
o Steady state roll-over. If lateral acceleration is slowly increased, e.g. as running with
into a hairpin curve or a highway exit, the vehicle can slowly lift off the inner wheels
and roll-over. This is the only case of roll-over for which a model is given in this
compendium.
o Transient roll-over. This is when complex manoeuvres, like double lane changes or
sinusoidal steering, are made at high lateral accelerations. This can trigger roll eigen-
modes, which can be amplified due to unlucky timing between the turns. Models from
Section 4.5 can be used as a start, but it is required that load transfer is modelled
carefully and includes wheel lifts, suspension end-stops and bump stops.
4.3.11.1 Roll-over threshold definitions

An overall requirement on a vehicle is that the vehicle should not roll-over for certain manoeuvres.
Heavy trucks will be possible to roll-over on high-mu conditions. The requirement for those are based
on some manoeuvres which not utilize the full road friction. For passenger cars, it is often the intended
design that they should be impossible to roll-over, even at high mu. Any requirement need a definition
of what exactly roll-over is, i.e. a Roll over threshold definition. Candidates for Roll over threshold
definition are:
• One wheel lift from ground
• All wheels on one side lift from ground
• Vehicle CoG passes its highest point
Note that:
• It is the 3rd threshold which really is the limit, but other can still be useful in requirement
setting. To use the 3rd for requirement setting makes the verification much more complex, of
course in real vehicles but also in simulation.
• The 1st is not a very serious situation for a conventional vehicle with 4 wheels. However, for a
3-wheeled vehicle, such as small “Tuctucs” or a 3-wheel moped, it is still a relevant threshold.
• The 2nd threshold is probably the most useful threshold for two-tracked vehicles, because it
defines a condition from which real roll-over is an obvious risk, and still it is relatively easy to
test and simulate. For 3-wheeled vehicle, 2nd and 3rd threshold generally coincide. The 2nd
threshold will be used in this compendium.
Figure 4-47 shows how the inner wheels lift off subsequently during a slowly increasing lateral force
(or lateral acceleration) build-up. Before any wheel is lifted, the load transfer is proportional to roll-
202
LATERAL DYNAMICS
centre heights and roll stiffnesses, as shown in Equation [4.46]. But every time a wheel lifts, the
distribution changes, so that a “knee” on the curves appears, see Figure 4-47. So, the relation of type as
Equation [4.46] is no longer valid. For instance, it is not physically motivated to keep the roll-centre
model for an axle which has lifted one side. So, the prediction of critical lateral acceleration for roll-
over is not trivial, especially for heavy vehicles which has many axles, and often also a fifth wheel
which can transfer roll-moment to a certain extent. There are approximate standards for how to
calculate steady state roll-over thresholds for such vehicle, e.g. UN ECE 111
(http://www.unece.org/fileadmin/DAM/trans/main/wp29/wp29regs/r111e.pdf).
second inner
first inner
wheel lift
Vertical force on
wheel lift
inner wheels
last inner
wheel lift
Lateral force
(or lateral acceleration)
Figure 4-47: Example of 3-axle vehicle steady state roll-over wheel lift diagram.
4.3.11.2 Static Stability Factor, SSF

One very simple measure of the vehicles tendency to roll-over is the Static Stability Factor, SSF. It is
proposed by NHTSA, http://www.nhtsa.gov/cars/rules/rulings/roll_resistance/, and it is simply
defined as:
𝐻 𝑙 𝑇𝑟 𝑘𝑊 𝑡ℎ 𝑤⁄
SS = = ; [4.48]
𝐻 𝑡ℎ 𝐶𝑜𝐺 ℎ
A requirement which requires SSF>number cannot be directly interpreted in terms of certain

manoeuvre and certain roll-over threshold. It is not a performance based requirement, but a design
based requirement. However, one of many possible performance based interpretations is that the
vehicle shall not roll-over for steady-state cornering on level ground with a certain friction coefficient,
using one-sided wheel lift as threshold. Since the requirement is not truly performance based, each
interpretation will also stipulate a certain verification method; here it would be theoretical verification
using a rigid suspension model. Such model and threshold is shown in Figure 4-48.
The derivation of the SSF based requirement looks as follows:
𝑤
∙𝑤+ ∙ ∙ℎ= ∙ ∙ ;
1 ℎ∙
𝑀𝑜 𝑙: { + }⇒ = ∙ ∙( );
𝑜 = ∙ ; 𝑤 ⇒
∙ = = ∙( + 𝑜 );
𝑅 𝑞𝑢 𝑟 𝑛𝑡: ≥ ; } [4.49]
1 ℎ∙ 𝑤
⇒ 𝑅 𝑞𝑢 𝑟 𝑛𝑡: > ⇒ = 𝑆𝑆 > ;
𝑤 ∙ℎ
Maximum road friction, m, is typically 1, which is why SSF>m1 would be a reasonable. However,
typical values of SSF for passenger vehicles are between 0.95 and 1.5. For heavy trucks, it can be much
lower, maybe 0.3..0.5, much depending on how the load is placed. There are objections to use SSF as a
203
LATERAL DYNAMICS
measure, because SSF ignores suspension compliance, handling characteristics, electronic stability
control, vehicle shape and structure.
view from rear, when turning left
ay
m*g
m*ay
h
Fy
Fiz≥0
Foz
(=roll-over w/2
threshold)
w
Figure 4-48: Model for verification of requirement based on Static Stability Factor, SSF.
4.3.11.3 Steady-state cornering roll-over *

A function defined for requirement setting can be:
Function definition: Steady state cornering roll-over acceleration is the maximum lateral acceleration
the vehicle can take in steady state cornering without lifting all inner wheels. On level ground with enough road
friction and certain weight and position of payload.
For a long combination-vehicle with several articulation points, one often need to drive a long distance
after a steering angle change before steady state values on articulation angles are reached. Hence, it
can be more relevant to formulate a corresponding roll-over function in terms of curvature to follow,
total yaw angle for turn and longitudinal speed. A common way is also the, somewhat artificial, tilt-
table test, which means that one measure steady-state roll-over with a (real or virtual) tilt-table where
the maximum road pitch angle before wheel lift on one side is the measure to set requirement on. An
even simpler way to handle steady-state roll-over is to set requirement on the SSF.
Consider a roll-stiff vehicle in steady state cornering. Assume lateral acceleration is subsequently
increased. If the vehicle is a two-axle vehicle, Eq [4.46] is valid until first axle lifts its inner wheel, since
for larger lateral accelerations, the constitutive equation Eq [4.40] is invalid for the inner wheel, since
the lifted inner wheel has zero force from ground. We can identify the terms of one of the inner wheel
(if > ) equations in Eq Eq [4.46] as follows:
= ∙( ∙( + ∙ )) =
∙ ∙𝑤 𝑤 𝑜 𝑒ℎ 𝑐 𝑒
1 ∙ ∙𝑙 𝑙 ℎ𝑅𝐶 ∆ℎ 𝑜
= ∙ ∙ ∙ ⏟∙ ∙ ∙ =
⏟ ∙ ⏟ 𝑤 𝑤 ⏟𝑜 𝑒ℎ 𝑐 𝑒
𝑉 𝑟𝑡 𝑙 𝑜𝑟 𝑛𝐴𝑥𝑙 𝑡 𝑟 𝑙 𝑜𝑟 𝑛𝐴𝑥𝑙 𝑡 𝑟 𝑙 𝑜𝑟 𝑛𝑉 ℎ 𝑙
𝑅𝑜𝑙𝑙𝑆𝑡 𝑛 𝑠𝑠 𝑟 𝑡 𝑜𝑛 𝑛𝐴𝑥𝑙
ℎ𝑅𝐶 ∆ℎ 𝑜
= 𝑦( )∙ 𝑦( )∙ ∙ ;
𝑤 𝑤 𝑜 𝑒ℎ 𝑐 𝑒
If the vehicle has more axles, Eq [4.46] is generalized to Eq [4.50], which also is valid until first inner
wheel lifts from ground.
For axle of a roll-stiff vehicle:

ℎ𝑅𝐶 ∆ℎ 𝑜 [4.50]
= ( )∙ ( )∙ ∙ ;
𝑤 𝑤 𝑜 𝑒ℎ 𝑐 𝑒
204
LATERAL DYNAMICS
For a vehicle with >2 axles, the parameter Δℎ can not be calculated from Eq [4.44], but can still be
understood as the vertical distance between roll axis and the axles roll centre. It should be noted that
the pendulum effect is not included in Eq [4.50], and this is often a significant approximation if applied
on high CoG vehicles, like heavy trucks.
4.3.11.3.1 Model assuming all inner wheels lift at the same lateral acceleration
An approximation of Steady state cornering roll-over acceleration 𝑐 𝑡 (lateral acceleration when all
inner wheels lifted) can be found for vehicles where Eq [4.50] gives = for all axles at same .
Then, summing the Eq [4.50] for all axles leads to the Eq [4.51] which is the same 𝑐 𝑡 as the simple
SSF model in Figure 4-48 and Section 4.3.11.2 gives.
𝑤
𝑐 𝑡 = ; [4.51]
ℎ
In the following, we will elaborate with 4 additional effects, which marked in Figure 4-49.
• The tyre will take the vertical load on its outer edge in a roll-over situation. This suggests a
𝑤+𝒘𝒕𝒚 𝒆 𝑤+𝒘𝒕𝒚 𝒆
change of performance and requirement to: 𝑦 < ∙ℎ
and ∙ℎ
> . This effect is
accentuated when low tyre profile and/or high inflation pressure. This effect decreases the
risk for roll-over.
• Due to suspension and tyre lateral deformation, the body will translate laterally outwards,
𝑦 𝑤−𝑫𝒆𝒇𝒚 𝑤−𝑫𝒆𝒇𝒚
relative to the tyre. This could motivate < ∙ℎ
and ∙ℎ
> . This effect increases the
risk for roll-over.
• Due to suspension linkage and compliances, the body will roll. Since the CoG height above roll
𝑤−∆𝒉∙ 𝒙 𝑤−∆𝒉∙ 𝒙
axis, ∆ℎ, normally is positive, this could motivate 𝑦 < and > . This effect
∙ℎ ∙ℎ
increases the risk for roll-over. At heavy vehicle this “pendulum effect” is large.
• Due to suspension linkage and compliances, the body will also heave. This requires a
suspension model with pivot points per wheel, as opposed to roll-centre per axle, to be taken
𝑤 𝑤
into account. The heave is normally positive. This could motivate ∙(ℎ+ ) > and 𝑦 < ∙(ℎ+ ).
The effect is sometimes called “jacking” and it increases the risk for roll-over.
• Road leaning left/right (road banking), or driving with one side on a different level (e.g. out-
side road or on pavement) also influence the roll-over performance.
view from rear, when turning left
𝑥
ay m*g
∆ℎ m*ay
h
Fy
𝐷 𝑦
Fiz=0
Foz
(=roll-over w/2
threshold) wtyre/2
w
Figure 4-49: Steady-state roll-over model, with fore/aft symmetry. The measures 𝑤𝑡 𝑒 𝐷 ∆ℎ ∙ 𝑛
mark effects additional to what is covered with a simple SSF approach.
4.3.11.3.2 Model with subsequent inner wheel lift
A model which does not assume wheel lift at same lateral acceleration will be sketched. For each axle
that has lifted, the equations have to be changed. Instead of simply the constitutive equation
205
LATERAL DYNAMICS
( ) 𝑤⁄ = 𝑜 ; one need to assure = ;. The axle will then position itself so that it
keeps + = 𝑠𝑡 𝑡 𝑐 ;. That means both roll and vertical translation of the axle centre, why also
the vertical suspension compliance needs to be modelled. A new position variable has to be declared,
e.g. the lift distance of the inner wheel, 𝑙 𝑡 . This variable is contrained to 𝑙 𝑡 = ; before lift, but
after lift the constraint is = ;. So, the model is suitably implemented as a state event model with
the event “𝑤ℎ 𝑛 𝑏 𝑜 𝑠 < ”. If is swept from zero and upwards, the result will be something
like shown in Figure 4-47.
4.3.11.3.3 Using a transient model for steady-state roll-over

Another work-around to avoid complex algebra is to run a fully transient model, including suspension,
and run it until a steady state cornering conditions occur. If then, the lateral acceleration is slowly
increased, one can identify when or if the roll-over threshold is reached. Lateral acceleration increase
can be through either increase of longitudinal speed or steering angle. It should be noted that the
model should reasonably be able to manage at least lift of one wheel from the ground. This way of
verifying steady state cornering roll-over requirements has the advantage that, if using tyre models
with friction saturation, the limitation discussed in Section 4.3.11.3.2 does not have to be checked
separately.
4.3.11.4 Roll-over and understeering/propulsion

With the above formulas for roll-over there will always be a certain lateral acceleration that leads to
roll-over, because neither limitation due to road friction nor propulsion power modelled yet. Since
vehicles generally are understeered, they are limited to develop lateral acceleration, see Figure 4-33.
For propulsion-weak vehicles, there is also the limitation of lateral acceleration due to limited
longitudinal speed, which in turn is due to driving resistance from the steered wheels (=wheel lateral
force * sin(steering angle)) and loosing propulsion power due to longitudinal wheel slip. However, one
should take into account that the propulsion limitation is less in down-hill driving, which increases the
roll-over risk again. Also, if the vehicle goes relatively quickly into steady state cornering, the
longitudinal speed will not have time to decelerate to its real (longitudinal) steady state value.
For heavy trucks, the critical lateral acceleration is typically 0.3..0.4 g, which is quite possible to reach
during normal road conditions, because road friction is there around 1. For passenger cars, the critical
lateral acceleration is typically in the region of 1, so it is not obvious that it is possible to reach the roll-
over-critical lateral acceleration. This is also the case for heavy trucks on low road friction.
4.4 Stationary oscillating steering

In between steady state and transient manoeuvres, one can identify stationary oscillations as an
intermediate step. Generally, a mechanical system can be excited with a stationary oscillating
disturbance. The response of the system is, after possible transients are damped out, a stationary
oscillation. If staying within the linear region for the system and the excitation is harmonic (sinus and
cosine), the ratio between the response amplitude and the excitation amplitude is only dependent of
the frequency. The ratio is called transfer function.
For lateral vehicle dynamics, the excitation is typically steering wheel angle and the response is
amplitudes of yaw rate, curvature or lateral acceleration. The corresponding transfer functions are a
frequency version of the gains defined in Equation [4.28]. Also, there will be a delay between
excitation and response. This is another important measure, beside the amplitude ratio.
4.4.1 Stationary oscillating steering tests

When testing Stationary oscillating steering functions, one usually drives on a longer part of the test
track. It might be a high speed track, see Figure 4-8, because one generally need to find the response at
high speeds, rather than driving close to lateral limits. So, the track rather needs to be long than wide.
206
LATERAL DYNAMICS
If the available Vehicle Dynamics Area, see Figure 4-8, is long enough this can be a safer option. A
Vehicle Dynamics Area is a flat surface with typically 100..300 m diameter. It normally has entrance
roads for accelerating up to a certain speed.
Typical tests in this part of lateral vehicle dynamics are:
• Sweeping frequency and/or amplitude
• Random frequency and amplitude
There are ISO standards for both these tests. The response will be very dependent of the vehicle
longitudinal speed, why the same tests are typically done at different such speeds.
4.4.2 Transient one-track model

The model needed for stationary oscillation is only a linearization of the model needed for fully
transient handling, in Section 4.5. However, a rather complete model will be derived already in present
section, to capture the physical assumptions in a proper way. (If the reader is satisfied with the
linearized model, it can be found directly in Eq [4.55] and a less general derivation of this equation in
Figure 4-54.)
The vehicle model is sketched in Figure 4-51. The model is a development of the model for steady state
cornering in Figure 4-22, with the following changes:
• Longitudinal and lateral accelerations have both components of centripetal acceleration (𝜔
and 𝜔 ) and the derivatives ̇ and ̇ :
= 𝑟( ) 𝜔 ∙ = ̇ 𝜔 ∙ ;
[4.52]
= 𝑟( ) + 𝜔 ∙ = ̇ +𝜔 ∙ ;
• The yaw acceleration, der(wz), is no longer zero.

• The speed vx is no longer defined as a parameter, but a variable. Then, one more prescription
is needed to be a consistent model. For this purpose, an equation that sets front axle
propulsion torque to 1000 Nm is added.
The difference between acceleration ⃗ = [ ; ] and [der(vx); der(vy)]=[ ̇ ; ̇ ] can be confusing. A
way to understand those could be to think of ⃗ as the (geometric vector) acceleration in the inertial
coordinate system and [ ̇ ; ̇ ] as a (mathematical) vector, with derivatives of the scalar mathematical
variables and , which are the velocities in the vehicle fix (and moving) coordinate system, see
Section 4.4.2.2.
The model in Figure 4-51 is documented as mathematical model in Equation [4.53]. The equation is
given in Modelica format. The subscript v and w refers to vehicle coordinate system and wheel
coordinate system, respectively. The model is a development of the model for low-speed in Equation
[4.12], with the changes marked with underlined text:
207
LATERAL DYNAMICS
speeds: vy forces:
vfxv ⃗ m*ay Ffxv ⃗
vry wz
Ffyv
vfyv
Fry
vx df Frx m*ax df
vrx J*der(wz)
lr L lf lr lf
⃗ L ⃗
ax 𝒂
ay 𝒗
b vy=v*sin(b)
Figure 4-51: One-track model for transient dynamics. Dashed show fictive forces. Compare to Figure 4-22.
//(Dynamic) Equilibrium:
m*ax = Ffxv + Frx;
m*ay = Ffyv + Fry;
J*der(wz) = Ffyv*lf - Fry*lr;
ax=der(vx)-wz*vy;
ay=der(vy)+wz*vx;
//Constitutive relation (Lateral tyre force model):

Ffyw=-Cf*sfy;
Fry=-Cr*sry;
sfy=vfyw/vfxw;
sry=vry/vrx;
//Compatibility:
vfxv = vx;
vfyv = vy + lf*wz;
vrx = vx;
vry = vy - lr*wz; [4.53]

der(pz) = wz;

//Shaft torques:
Ffxw = +1000; // Front axle driven.
Frx = -100; // Rolling resistance on rear axle.
The initial longitudinal speed is a parameter, vx=100 km/h. A simulation result from the model is
shown in Figure 4-52. The manoeuvre selected is same steering wheel function of time as in Figure
4-23, for better comparison of the different characteristics of the models.
208
LATERAL DYNAMICS
df [deg]
y [m]
wz [deg/s]
vx [km/h]
beta [deg]
x[m]
time [s]
Figure 4-52: Simulation results of one-track model for transient dynamics.
Both position variables [𝑥 𝑦 =p ] and speed variables [ 𝜔 =w ] are “state variables” of this
simulation. For each continuation from one time instant, the future solution requires knowledge of the
previous states. This means that in the beginning, initial values were needed on the state variable. The
only non-zero initial value was for , which was given to 1 𝑘 ℎ.
Equation [4.53] is a complete model suitable for simulation. Eliminating some variable and rewrite in
prettier format gives:
Equilibrium:
∙( ̇ 𝜔 ∙ )= 𝑤 ∙ cos( ) 𝑤 ∙ sin( ) + ;
∙( ̇ +𝜔 ∙ )= 𝑤 ∙ sin( ) + 𝑤 ∙ cos( ) + ;
∙ 𝜔̇ = ( 𝑤 ∙ sin( ) + 𝑤 ∙ cos( )) ∙ 𝑙 ∙𝑙 ;
Constitution:
𝑤 = 𝐶 ∙𝑠 ; 𝑛 = 𝐶 ∙𝑠 ;
[4.54]
Compatibility:
𝑤 𝑙 ∙𝜔
𝑠 = ; 𝑛 𝑠 = ;
| 𝑤| | |
Transformation from vehicle to wheel coordinate system on front axle:
𝑤 =( + 𝑙 ∙ 𝜔 ) ∙ sin( ) + ∙ cos( ) ;
𝑤 =( + 𝑙 ∙ 𝜔 ) ∙ cos( ) ∙ sin( ) ;
Typically, this model is used for simulation, where , 𝑤 and are input variables. Suitable state
variables are then , and 𝜔 . It is a model suitable for arbitrary transient manoeuvres and we will
come back to this in Section 4.5. It is non-linear with respect to angles, but linear with respect to tyre
slip model. It is also non-linear with respect to that the states appear as multiplied with each other, e.g.
𝜔 ∙ .
209
LATERAL DYNAMICS
4.4.2.1 Constant speed and low friction usage

For the stationary oscillation events, this compendium will be limited to manoeuvres with constant
(⇒ ̇ = ), far from saturating tyre grip and modest cornering. We assume:
• Low usage of lateral tyre force, i.e. no saturation and small tyre slip angle (𝛼 ): 𝑠 =
𝑡 𝑛(𝛼 ) ≈ 𝛼 ;
• Low cornering, i.e.
o small steering angle: sin( ) ≈ ; cos( ) ≈ 1 ;.
𝑦𝑣 𝑦𝑣 𝑦+ ∙𝜔
o small front body slip angle (𝛽 ): 𝛼 = 𝛽 ; where 𝛽 = 𝑟 𝑡 𝑛 ( )≈ = ;
𝑥𝑣 𝑥𝑣 𝑥
We then rewrite the two last equations in Equation [4.54]:

+𝑙 ∙𝜔 𝑙 ∙𝜔
∙( ̇ +𝜔 ∙ )= 𝑤 ∙ 𝐶 ∙( ) 𝐶 ∙ ;
| | | |
⇒
+𝑙 ∙𝜔 𝑙 ∙𝜔
∙ 𝜔̇ = ( 𝑤∙ 𝐶 ∙( )) ∙ 𝑙 + 𝐶 ∙ ∙𝑙 ;
{ | | | | }
𝐶 +𝐶 𝐶 ∙𝑙 𝐶 ∙𝑙
∙ ̇ + +( + ∙ ) ∙ 𝜔 = (𝐶 + 𝑤) ∙ ;
| | | |
⇒
𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙𝑙 +𝐶 ∙𝑙
∙ 𝜔̇ + ∙ + ∙ 𝜔 = (𝐶 + 𝑤) ∙ 𝑙 ∙ ;
{ | | | | }
Write this equation on matrix form and neglecting the longitudinal equilibrium:
𝐶 +𝐶 𝐶 ∙ 𝑙 𝐶 ∙ 𝑙
+ ∙
̇ | | | | 𝐶 + 𝑤
[ ]∙[ ]+ ∙ [𝜔 ] = [ ]∙ ;
𝜔̇ 𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙ 𝑙 +𝐶 ∙ 𝑙 (𝐶 + 𝑤) ∙ 𝑙
[ | | | | ]
Assuming that propulsion forces can be neglected for lateral and yaw equilibrium:
𝐶 +𝐶 𝐶 ∙ 𝑙 𝐶 ∙ 𝑙
+ ∙
̇ | | | | 1
[ ]∙[ ]+ ∙ [𝜔 ] = [ 𝑙 ] ∙ 𝐶 ∙ ; [4.55]
𝜔̇ 𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙ 𝑙 +𝐶 ∙ 𝑙
[ | | | | ]
Eq [4.55] (or resulting equation in Figure 4-54) can be seen as a linear state-space-model:
̇
[
] = 𝑨 ∙ [𝜔 ] + 𝑩 ∙ ;
𝜔̇
𝑆𝑡 𝑡 𝑠𝑝 𝑜𝑟 : { 𝜔
[ ] = 𝑪 ∙ [𝜔 ] + 𝑫 ∙ ;
𝐶 +𝐶𝑟 𝐶 ∙ −𝐶𝑟 ∙ 𝑟
− + ∙ [4.56]
| 𝑥| | 𝑥|
where 𝑨 = [ ] ∙ [𝐶 ∙ −𝐶 ∙ 𝐶 ∙ 2 +𝐶𝑟 ∙ 𝑟 2
];
𝑟 𝑟
| 𝑥| | 𝑥|
−
1 1
and 𝑩 = [ ] ∙ [𝑙 ] ∙ 𝐶 ; and 𝑪 = [ ]∙𝑨+[ ]; and 𝑫 = 𝑩;
1
It should be noted that the assumption about small front body slip angle (𝛽 ) makes the linearization
questionable at high curvature, i.e. at small path radii.
210
LATERAL DYNAMICS
The longitudinal equilibrium (first equation in Equation [4.54]) is often not mentioned, because the
longitudinal dynamics is prescribed as a constant speed, . However, this equation can be used to
calculate required propulsion on the axles to maintain the constant longitudinal speed. Note that it is
complemented with air and grade resistance force.
𝑤 ∙ cos( ) + =
( + 𝑙 ∙ 𝜔 ) ∙ cos( ) ∙ sin( )
= + 𝑑𝑒 𝜔 ∙ 𝐶 ∙ ∙ sin( ) ;
( + 𝑙 ∙ 𝜔 ) ∙ sin( ) + ∙ cos( )
If 𝛼 𝛽 small and 𝛽 ∙ ≈ ; ∙ ≈ : 𝑤 + = + 𝑑𝑒 ∙𝜔 ∙ ; [4.57]
𝑇
If front axle driven: 𝑤 =𝑅 ; 𝑛 = ;
𝑤
𝑇𝑟
If rear axle driven: = ; 𝑛 𝑤 = ;
𝑅𝑤
4.4.2.2 Relation between accelerations in inertial system and

velocity derivatives in vehicle fix system
See Equation [4.52]. The relation between accelerations in inertial system, ⃗ = [ ; ] , and velocity
derivatives in vehicle fix system, [der(vx); der(vy)]=[ ̇ ; ̇ ], is [ ; ] = [ ̇ ; ̇ ] + [+ ; ] 𝜔 ;.
The terms proportional to 𝜔 can be understood as a centrifugal force.
Another explanation is given in left part of Figure 4-53 and the following reasoning. We can express
the velocity in direction of the x axis at time t, at the two time instants:
• Velocity at time = 𝑡:
• Velocity at time = 𝑡 + ∆𝑡: ( + ∆ ) ∙ cos(∆ ) ( + ∆ ) ∙ sin(∆ ) = ( ∙ cos(∆ ) +
∆ ∙ cos(∆ )) ( ∙ sin(∆ ) + ∆ ∙ sin(∆ )) ≈ {∆ 𝑠 𝑙𝑙} ≈ ( + ∆ )
( ∙ ∆ + ∆ ∙ ∆ ) ≈ {∆ ∙ ∆ 𝑠 𝑙𝑙} ≈ ( + ∆ ) ∙∆
Using these two expressions, we can express as the change of that speed per time unit:
{( 𝑥 +∆ 𝑥 )− 𝑦 ∙∆𝜓}− 𝑥 ∆ 𝑥 − 𝑦 ∙∆𝜓 ∆ 𝑥 ∆𝜓
• Acceleration=Velocity change per time = = = = ≈
∆𝑡 ∆𝑡 ∆𝑡 ∆𝑡
̇ ∙ ̇= ̇ ∙𝜔
Corresponding for the lateral direction gives the Equation [4.52]. In Equation [4.52], the term 𝜔 ∙ is
generally more important than the term 𝜔 ∙ . This is because is generally much larger than .
yglobal Comparison with mass

time=t+Dt: x on circular path,
with centrifugal
acceleration 𝝎𝒛 𝒗𝒙
y Dwz
wz Dvx
Dvy Δ
time=t: vy vx 𝒂𝒄𝒆𝒏𝒕 𝒊 𝒆𝒕𝒂𝒍
y
wz
wz
vy x vx
vx
xglobal 𝒂𝒄𝒆𝒏𝒕 𝒊 𝒆𝒕𝒂𝒍 =

𝑅
=𝑅 𝜔
= =𝑅 𝜔 = 𝝎𝒛 𝒗𝒙 ;
Figure 4-53: How to understand the acceleration term 𝜔 for vehicle motion in ground plane. Left: Two
consecutive time instants. Right: Comparison with circular motion, to identify the centripetal acceleration.
211
LATERAL DYNAMICS
A generalisation gives that these centripetal acceleration terms appears as 𝜔𝑗 where 𝑗, i.e. for
rotation perpendicular to translational velocity, see right part of Figure 4-53:
̇ + 𝜔 ̇ 𝜔 +𝜔
[ ]=[ ̇ ]+[ + ] [𝜔 ] = [ ̇ ] + [ +𝜔 𝜔 ] [ ]=
̇ + 𝜔 ̇ 𝜔 +𝜔
[4.58]
̇ + 𝜔 – 𝜔 𝑛 𝑜𝑠𝑡 𝑟 𝑙 𝑛𝑡 ̇
=[ ̇ 𝜔 + 𝜔 ]≈{ ℎ 𝑙 𝑜𝑝 𝑟 𝑡 𝑜𝑛𝑠: } ≈ [ ̇ ] + [ +𝜔 ] ;
̇ + 𝜔 𝜔 ≫ 𝑛 ≫ ̇ 𝜔
In most vehicle operations, the most important centripetal acceleration terms are: 𝜔 ∙ and 𝜔 ∙ .
The first comes into play in Equation [4.52], and the latter in Figure 3-46, Eq [3.34], and Section
3.4.9.1.
4.4.2.3 Simpler derivation of the same model

A simpler derivation of Equation [4.55] is given in Figure 4-54. Here, the simplifications are introduced
earlier, already in physical model, which means e.g. that the influence of 𝑤 is not found.

• Path radius >> the vehicle. Then, all forces Equilibrium:
(and centripetal acceleration) are ∙ ̇ 𝜔 ∙ ≈ + ; 𝑤ℎ 𝑟 ̇ = ;
approximately co-directed. ∙ ̇ +𝜔 ∙ ≈ + ;
• Small tyre and vehicle side slip. Then,
∙ 𝜔̇ ≈ ∙𝑙 ∙𝑙 ;
angle=sin(angle)=tan(angle).
(Angles are not drawn small, which is the reason Constitution: = 𝐶 ∙𝑠 ; = 𝐶 ∙𝑠 ;
why the forces not appear co-linear in figure.)
Compatibility:
∙ +𝑙 ∙𝜔
+𝛼 =𝛽 ; 𝛽 ≈ = ;
∙ 𝜔̇
Ffy bf 𝑙 ∙𝜔
Fry 𝛼 =𝛽 ≈ = ;
af 𝛼 ≈ 𝑠 ; 𝛼 ≈ 𝑠 ;;
b df
br ∙
Eliminate 𝑭𝒇𝒚 𝑭 𝒚 𝒇 𝜷𝒇 𝜷 yields:
wz ∙ ̇ +
𝐶 +𝐶
∙ +
𝐶 ∙𝑙 𝐶 ∙𝑙
+ ∙ ∙𝜔 ≈
lr lf ≈𝐶 ∙ ;
L ∙ 𝜔̇ +
𝐶 ∙𝑙 𝐶 𝑙
∙ +
𝐶 ∙𝑙 +𝐶 ∙𝑙
∙𝜔 ≈
= ̇ 𝜔 ∙ ; ≈𝐶 ∙𝑙 ∙ ;
= ̇ +𝜔 ∙ ; + ≈ ∙𝜔 ∙ ;
Figure 4-54: Simpler derivation final step in Equation [4.55].

One can add the effect of non-Ackerman between axles. E.g. a truck with double rear axle will get an
added term proportional to 𝜔 in the moment equilibrium. Alternatively, one can add the effect of non-
Ackerman within an axle. E.g. a two-axle vehicle with parallel steering, which will reduce the cornering
stiffness of that axle since the left and right tyre “fight” each other laterally.
4.4.2.4 Validity of model

When only studying the response as yaw rate and curvature response, it is easy to forget that one very
easily comes into manoeuvres where road friction is limited, i.e. where the linear tyre model is not
valid. Hence it is good to look at lateral acceleration response, because we can roughly say that for | |
in the same magnitude as ∙ ∙ , it is doubtful if the model is valid. If the wheel torques are
212
LATERAL DYNAMICS
significant, the validity limit is even lower. For high CoG vehicles, another invalidating circumstance is
∙𝑤
wheel lift, which can be approximately checked by checking that | | ≪ 𝑆𝑆 = .
∙ℎ
If one really wants to include nonlinear tyre models in stationary oscillation response, one can
simulate using time integration (same method as usually used for transient handling) over several
excitation cycles, until the response shows a clear stationary oscillation. This consumes more
computational efforts and the solutions become approximate and numerical.
4.4.2.5 Locking one axle

Locking one axle, , can be modelled by replacing 𝐶 with ;. This can be proven by linearizarion of
Eq [4.54] with the locked axle’s constitution replaced by “ = ∙ ⁄| | ;”. Since 𝐶 ≈ 𝐶𝐶
and cornering compliance, 𝐶𝐶 , is typically 5..10 (trucks) and 10..15 (passenger cars), we can
conclude that is normally ≪ 𝐶 . Realistic variation of road friction coefficient is 0.2..1. So, a
locked axle can be seen as having much (typically 5..75 times) lower cornering stiffness on the locked
axle. So, locking front gives extreme under-steering and locking rear gives extreme over-steering.
4.4.2.6 Stable and unstable conditions

Eq [4.56] can explain instability. Instability means here that vehicle gets infinitely growing solutions
although . Eq [4.81] shows the explicit solution. We set ; and a small yaw disturbance as
0
initial conditions: [𝜔 ] = [ ] ; and get the following expression for the solution:
0 𝜀
−
̂ ̂ 𝜆1 ∙𝑡 ̂ ̂ 0
[𝜔 ] = [[ ] [ ]] ∙ [ 𝜆2 ∙𝑡
] ∙ [[ ] [ ]] ∙ [𝜔 ] ;
𝜔̂ 𝜔
̂ 𝜔
̂ 𝜔̂ 0
̂ ̂ 𝜆
𝑤ℎ 𝑟 [[[ ] [ ]] [ ]] = ei (𝑨) ;
𝜔̂ 𝜔
̂ 𝜆
In Figure 4-55 we sweep for normal passenger car data and try the following cornering stiffnesses:
• Locked front axle: 𝐶 𝑙 ≈ .1 𝐶 𝑙 ; (extremely under-steered)
• Under-steered: 𝐶 𝑙 ≈ .9 𝐶 𝑙 ; (moderatly under-steered
• Neutral-steered: 𝐶 𝑙 ≈ 1 𝐶 𝑙 ;
• Over-steered: 𝐶 𝑙 ≈ 1.1 𝐶 𝑙 ; (moderatly over-steered
• Locked rear axle: 𝐶 𝑙 ≈ 1 𝐶 𝑙 ; (extremely over-steered)
In Figure 4-55 we see that only over-steered (incl. lock rear) gives an unstable vehicle. And that only
under-steered vehicles (incl. lock front) gives oscillating solutions. Also, we see that all vehicles are
stable for low enough speed.
213
LATERAL DYNAMICS
lock rear
neutral-steered
under-steered
lock front
under-steered
neutral-steered
over-steered
lock rear
lock rear
neutral-steered
Figure 4-55: Stability analysis of passenger car with varying cornering stiffnesses.
4.4.3 Steering frequency response gains *

Function definition: Steering frequency response gains are the amplifications from steering angle
amplitude to certain vehicle response measure’s amplitudes for stationary oscillating harmonic steering at a certain
longitudinal speed.
Use the linear state-space-model in [4.56] to study “stationary oscillating steering”:
̇
[ ] = 𝐴 ∙ [𝜔 ] + 𝐵 ∙ (𝑡) = 𝐴 ∙ [𝜔 ] + 𝐵 ∙ ̂ ∙ cos(𝜔 ∙ 𝑡)
𝜔̇
(Note similar notation for vehicle yaw rate, 𝜔 , and steering angular frequency, 𝜔.) Knowing ̂ and 𝜔,
it is possible to calculate the responses ̂ , 𝜔
̂ , and :
̂ 𝑦
[𝜔 ] = [ ] ∙ cos (𝜔 ∙ 𝑡 [ ]) ;
𝜔
̂ 𝜔
𝜔 𝜔̂ 𝜔
[ ] = [ ̂ ] ∙ cos (𝜔 ∙ 𝑡 [ ]) ;
𝑦
Different methods are available for calculation of the responses:

• Fourier transform
• Complex mathematics, using 𝑗∙𝜔∙𝑡
• Real trigonometry, using cos(𝜔 ∙ 𝑡 + 𝑝ℎ 𝑠 ) sin(𝜔 ∙ 𝑡 + 𝑝ℎ 𝑠 ) or cos(𝜔 ∙ 𝑡) + sin(𝜔 ∙ 𝑡).

For each of these, depending on the problem, it can be ether scalar or matrix algebra. Matrix based
Fourier transform is generally the most efficient, especially if Matlab is available and numerical
solutions can be accepted.
214
LATERAL DYNAMICS
4.4.3.1 Single frequency response

4.4.3.1.1 Solution with Fourier transform
Equation [4.56] can be transformed to the frequency domain (ℱ denotes Fourier transform, i.e.
∞
ℱ(𝜉(𝑡)) = ∞ 𝑗∙𝜔∙𝑡 ∙ 𝜉(𝑡) ∙ 𝑡):
𝑗 ∙ 𝜔 ∙ ℱ ([𝜔 ]) = 𝑨 ∙ ℱ ([𝜔 ]) + 𝑩 ∙ ℱ( );
{ 𝜔
ℱ ([ ]) = 𝑪 ∙ ℱ ([𝜔 ]) + 𝑫 ∙ ℱ( );
𝑦
Solving for states and outputs, using ℱ( ̇ ) = j ∙ ω ℱ( );, gives:
[𝜔 ] = ℱ 1
(ℱ ([𝜔 ])) = ℱ 1
((𝑗 ∙ 𝜔 ∙ 𝐼 𝑨)− ∙ 𝑩 ∙ ℱ( )) ;
𝜔 1
𝜔 1
[ ]=ℱ (ℱ ([ ])) = ℱ (𝑪 ∙ ℱ ([𝜔 ]) + 𝑫 ∙ ℱ( )) ;
𝑦 𝑦
{
Expressed as transfer functions:
ℱ ([𝜔 ]) (𝑗 ∙ 𝜔 ∙ 𝐼 𝑨)− ∙ 𝑩 ∙ ℱ( )
𝐻 𝑦 = = = (𝑗 ∙ 𝜔 ∙ 𝐼 𝑨)− ∙ 𝑩;
𝛿 [𝜔 ] ℱ( ) ℱ( )
𝜔 [4.59]
ℱ ([ ])
𝐻 𝜔 = =𝑪∙𝐻 𝑦 + 𝑫 = 𝑪 ∙ (𝑗 ∙ 𝜔 ∙ 𝐼 𝑨)− ∙ 𝑩 + 𝑫;
𝛿 [ ] ℱ( ) 𝛿 [𝜔 ]
{ 𝑦
We have derived the transfer functions. The subscript tells that the transfer function is for the vehicle
𝜔
operation with excitation=input= and response=output= [𝜔 ] and output= [ ]. The transfer
function has dimension 2x1 and it is complex. It operates as follows:
̂
] = |𝐻 [ 𝑦 |∙ ̂ ;
𝜔̂ 𝛿 [𝜔 ]
Amplitudes:
𝜔̂ ̂ ;
[ ̂ ] = |𝐻 𝜔 |∙
𝛿 [ ]
{ 𝑦
[4.60]
ar 𝑦 ( ) 1 ar ( )
[ ]=[ ] [ ] ∙ ar ( ) = { } = ar (𝐻 𝑦 );
𝜔 ar (𝜔 ) 1 = 𝛿 [𝜔 ]
Phase delays:
𝜔 ar (𝜔 ) 1 ar ( )
[ ]=[ ] [ ] ∙ ar ( ) = { } = ar (𝐻 𝜔 );
{ 𝑦 ar ( ) 1 = 𝛿 [ ]
𝑦
4.4.3.1.2 Solution with complex mathematics

This section avoids requiring skills in Fourier transform. This makes the derivation quite long to reach
the final results Equation [4.65] and Equation [4.64]. With Fourier Transform, the expression for the
Transfer Function, H, can be reached with less algebra. Knowing H, it can be used in Equation [4.64].
The fundamental situation for steering frequency response is that the excitation is: = ̂ ∙
cos( ∙ 𝜋 ∙ ∙ 𝑡) = { 𝑗∙ = cos( ) + 𝑗 ∙ sin( )} = Re( ̂ ∙ 𝑗∙ ∙𝜋∙ ∙𝑡 ), where f is the (time) frequency in Hz
and 𝑗 is the imaginary unit. We rewrite ∙ 𝜋 ∙ as 𝜔 (angular frequency), which has to be carefully
distinguished from 𝜔 (yaw rate). Insert this in Eq [4.55] and neglecting the longitudinal force 𝑤.
The full (complex) equation is used:
215
LATERAL DYNAMICS
𝐶 +𝐶 𝐶 ∙ 𝑙 𝐶 ∙ 𝑙
+ ∙
̇ 𝑐 𝑐 𝐶
Re [ ]∙[ ]+ ∙ [𝜔 ] = [
𝐶 ∙𝑙
]∙ ̂ ∙ 𝑗𝜔∙𝑡
; [4.61]
𝜔̇ 𝑐 𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙ 𝑙 +𝐶 ∙ 𝑙 𝑐
[ [ ] ]
We intend to solve the complex equation, and then find the solutions as real parts: = Re[ 𝑐 ]; and
𝜔 = Re[𝜔 𝑐 ];. (Subscript c means complex.)
If only interested in the stationary solution, which is valid after possible initial value dependent
transients are damped out, we can assume a general form for the solution.
𝑐 ̂𝑐 ̇ 𝑐 ̂𝑐
[𝜔 ] = [ ]∙ 𝑗∙𝜔∙𝑡
⇒ [ ]=𝑗∙𝜔∙[ ]∙ 𝑗∙𝜔∙𝑡
; [4.62]
𝑐 𝜔
̂𝑐 𝜔̇ 𝑐 𝜔
̂𝑐
Inserting the assumption in the differential equation gives:
̂𝑐
[ ] ∙ 𝑗 ∙ 𝜔 ∙ [ ] ∙ 𝑗∙𝜔∙𝑡 +
𝜔
̂𝑐
𝐶 +𝐶 𝐶 ∙ 𝑙 𝐶 ∙ 𝑙
+ ∙
̂𝑐 1
+ ∙[ ]∙ 𝑗∙𝜔∙𝑡
= [𝑙 ] ∙ 𝐶 ∙ ̂ ∙ 𝑗∙𝜔∙𝑡
⇒
𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙ 𝑙 +𝐶 ∙ 𝑙 𝜔
̂𝑐
[ ]
̂𝑐
⇒[ ]=
𝜔
̂𝑐 [4.63]
−
𝐶 +𝐶 𝐶 ∙ 𝑙 𝐶 ∙ 𝑙
+ ∙
1
= [ ]∙𝑗∙𝜔+ ∙ [𝑙 ] ∙ 𝐶 ∙ ̂ =
𝐶 ∙𝑙 𝐶 ∙𝑙 𝐶 ∙ 𝑙 +𝐶 ∙ 𝑙
( [ ])
=𝐻 𝑦 ∙ ̂ ;
𝛿 [𝜔 ]
Then, we can assume we know ̂ 𝑐 and 𝜔 ̂ 𝑐 from Equation [4.63], and consequently we know 𝑐 and
𝜔 𝑐 from Equation [4.62]. We have derived the transfer function, 𝐻 𝑦 . The subscript tells that the
𝛿 [𝜔 ]
transfer function is for the vehicle operation with excitation=input= and response=output= [𝜔 ]
case. This transfer function has dimension 2x1 and it is complex. It operates as follows:
̂ | |
Amplitudes: [ ]=[ ] = |𝐻 𝑦 |∙| |;
𝜔
̂ |𝜔 | 𝛿 [𝜔 ]
[4.64]
ar ( ) 1
Phase delays: [ ] [ ] ∙ ar ( ) = {ar ( ) = } = ar (𝐻 𝑦 )
ar (𝜔 ) 1 𝛿 [𝜔 ]
However, we can derive expressions for 𝑐 and 𝜔 𝑐 on a real (non-complex) form, 𝑝𝑙 𝑡𝑢 ∙

cos( 𝑛 𝑢𝑙 𝑟 𝑟 𝑞𝑢 𝑛 𝑦 ∙ 𝑡 𝑝ℎ 𝑠 𝑙 𝑦), without involving transfer function. That is done in the
following:
𝑗∙𝜔∙𝑡
= Re( 𝑐) = Re( ̂ 𝑐 ∙ ) = ⋯ = | ̂ 𝑐 | ∙ cos(𝜔 ∙ 𝑡 + ar ( ̂ 𝑐 )) ;
The same rewriting can be done with 𝜔 , so that in total:
216
LATERAL DYNAMICS
̂ ∙ cos(𝜔 ∙ 𝑡 ) | ̂ 𝑐 | ∙ cos (𝜔 ∙ 𝑡 ( ar ( ̂ 𝑐 )))

[𝜔 ] = [ ]=[ ] [4.65]
𝜔̂ ∙ cos(𝜔 ∙ 𝑡 𝜔 ) |𝜔 ̂ 𝑐 | ∙ cos(𝜔 ∙ 𝑡 ( ar (𝜔
̂ 𝑐 )))
Equation [4.63] and Equation [4.65] now gives us the possibility to find vehicle response amplitude
and phase delay. The ratios between amplitude of responses and amplitude of excitation, ̂ ⁄ ̂ and
̂ ⁄ ̂ , are called gains. The difference in argument is the phase delay.
𝜔
4.4.3.2 Lateral Velocity and Yaw Rate response

The frequency response for the two states, Lateral Velocity and Yaw Rate, can be plotted for given
vehicle data, see Equation [4.63], Equation [4.65] and Figure 4-58. The curves show response for same
vehicle, but different speed.
See Figure 4-58. The yaw rate gain curve has a knee at 0.5..1 Hz. The decrease after that is a measure of
yaw damping. The curve for high speed actually has a weak peak just before the knee. This is not
desired, because the vehicle might feel a bit nervous. Yaw damping can also be how fast yaw rate
decays after a step response, see Section 4.5.
From Equation [4.24] we can calculate that characteristic speed for the vehicle is 120 km/h. With
another understeering coefficient, we could have calculated a critical speed. Studying Figure 4-58 to
Figure 4-61, one can find these special speeds in another appearance:
• For an understeered vehicle, speeds above the characteristic speed gives a negative yaw rate
delay for low steering frequencies will be negative.
• For over-steered vehicles, speeds above the critical speed gives a yaw rate delay larger than
18 and yaw rate amplitudes which are very large for low steering frequencies.
Lateral Velocity Response Yaw Rate Response
120 10
100 8
wz gain [(rad/s)/rad]
vy gain [(m/s)/rad]
80
6
60
4
40
20 2
0 0
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
vx=150 km/h
vx=100 km/h
2 2
vx=50 km/h
1 vx=1 km/h
1.5
vy delay [rad]
wz delay [rad]
0
1
-1
0.5
-2
-3 0
-4 -0.5
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
frequency [Hz] frequency [Hz]
Figure 4-58: Vehicle response to harmonically oscillating steering angle. Vehicle data: = 𝑘 ; =
3 𝑘 ; 𝑙 = 1.3 ; 𝑙 = 1.5 ; 𝐶 = 814 𝑁 𝑟 ; 𝐶 = 78 𝑁 𝑟 ; (𝐾𝑢 = 1. 6 𝑟 𝑀𝑁).
From Equation [4.24] we can calculate that characteristic speed for the vehicle is 120 km/h.
217
LATERAL DYNAMICS
In Figure 4-59 the curves are for same speed and constant understeering gradient, but they show the
response for different sums of cornering stiffness (Cf+Cr). Increasing the stiffness increases the yaw
rate gain (agility) at high frequencies.

80 8
Cf+Cr=119 kN/rad
Cf+Cr=159 kN/rad
vy gain [(m/s)/rad]
60 Cf+Cr=199 kN/rad 6
40 4
20 2
0 0
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
2 2
1
wz delay [rad] 1.5
vy delay [rad]
-1 1
-2
0.5
-3
-4 0
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
Figure 4-59: Vehicle response to harmonically oscillating steering angle. Same vehicle data as in Figure
4-58, except varying Cf and Cr but keeping understeering gradient Ku constant. Vehicle speed = 100 km/h.
In Figure 4-60 the curves are for same speed and constant sum of cornering stiffness (Cf+Cr), but they
show the response for different values of understeering gradient (Ku). Increasing understeer gradient
decreases the yaw rate gain (agility) at low frequencies.
4.4.3.3 Lateral Acceleration response

The lateral acceleration response is another useful response to study and set requirements on.
Actually, yaw rate and lateral acceleration are the most frequently used response variables, since they
are easily measured, e.g. from ESC sensors in most vehicles.
The transfer function is found (here using Fourier transform and previous results):
̇ +𝜔 ∙ 𝑗∙𝜔∙ 𝑐 +𝜔 𝑐∙
Amplitude: ̂ = | 𝑐| = {𝑢𝑠 : = }=| |=
1
=| ∙ [𝑗 ∙ 𝜔 ]∙𝐻 𝑦 ∙ | = |[𝑗 ∙ 𝜔 ]∙𝐻 𝑦 |∙ ; [4.66]
𝛿 [𝜔 ] 𝛿 [𝜔 ]
Phase delay: ar ( ) ar ( ) = {𝑢𝑠 : ar ( ) = } = ar ([𝑗 ∙ 𝜔 ]∙𝐻 𝑦 );

𝛿 [𝜔 ]
Lateral acceleration response is plotted for different vehicle speeds in Figure 4-61.
218
LATERAL DYNAMICS

400 40
Ku= -1.26 rad/kN
Ku= -0.01 rad/kN
vy gain [(m/s)/rad]
300 30
Ku= 0.01 rad/kN
Ku= 1.26 rad/kN
200 20
100 10
0 0
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
2 2
1
1.5
wz delay [rad]
vy delay [rad]
-1 1
-2
0.5
-3
-4 0
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
Figure 4-60: Vehicle response to harmonically oscillating steering angle. Same vehicle data as in Figure
4-58, except varying understeering gradient Ku but keeping 𝐶 + 𝐶 constant. Vehicle speed = 100 km/h.
Lateral Acceleration Response Lateral Acceleration Response

250 2
vx=150 km/h
vx=100 km/h 1.5
200 vx=50 km/h
vx=1 km/h 1
ay gain [(m/(s*s))/rad]
0.5
150
ay delay [rad]
100
-0.5
-1
50
-1.5
0 -2
-2 -1 0 1 2 -2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10
Figure 4-61: Vehicle lateral acceleration response to harmonically oscillating steering angle. Vehicle data:
= 𝑘 ; =3 𝑘 ; 𝑙 = 1.3 ; 𝑙 = 1.5 ; 𝐶 = 814 𝑁 𝑟 ; 𝐶 = 78 𝑁 𝑟 ; (𝐾𝑢 =
1. 6 𝑟 𝑀𝑁).
4.4.3.4 Other responses to oscillating steering

In principle, it is possible to study a lot of other responses, such as Path Curvature response, Side slip
response and Lateral Path Width response etc. These are not very standardized in requirement setting,
but can be used as such or generally used for comparing different designs during the development
work. For combination-vehicles it is common to plot Rearward amplification (RWA) as function of
219
LATERAL DYNAMICS
̂
frequency. RWA can be either 𝑅𝑊𝐴 = max ( ̂ 𝑖 ) ; where denotes vehicle units or axles, or
= … 1
̂ 𝑖
𝜔
𝑅𝑊𝐴 = max (𝜔
̂
) ; where denotes vehicle units.
= … 1
4.4.3.5 Random frequency response

Solutions to harmonic excitation of linear dynamic systems are superimposable. This is also why the
response from a mixed frequency excitation can be spliced into separate frequencies, e.g. using Fourier
transformation. Hence, a common way to measure the frequency response diagrams is to log data from
a random steering excitation. The frequency response diagram can then be extracted from this test.
4.5 Transient handling

Transient handling in general vehicle dynamics is when the vehicle is manoeuvred in an arbitrary, but
not constant or cyclic, way. Generally, this can be turning and braking/accelerating at the same time
through a manoeuvre. In this chapter, we only cover transient handling within lateral dynamics. This
should be understood as that we assume a reasonably constant longitudinal vehicle speed and modest
longitudinal forces on the wheels. The latter assumptions make it possible to base most of the content
in this chapter on the model derived in Section 4.4.
4.5.1 Transient driving manoeuvres *

From the manoeuvres defined in this section, there are a large number of relevant function possible to
define. Here, only a few is mentioned:
Function definition: Double Lane Change Capability is the highest entry speed that a certain driver (or
driver model) can manage without hitting any cones. A certain cone pattern (e.g. as Figure 4-62 or Figure 4-63)
and a certain road friction has to be defined. A certain operation principle of pedals needs to be specified, e.g.
release both pedals at entry or full brake pedal apply.
Function definition: Over-speed into Curve Capability is the highest entry speed that a certain driver (or
driver model) can manage without hitting any cones. A certain cone pattern e.g. as Figure 4-64) and a certain road
friction has to be defined. A certain operation principle of pedals needs to be specified, e.g. release both pedals at
entry or full brake pedal apply.
Function definition: Steer-In and Release Accelerator Pedal stability is a measure (e.g. side slip rear
axle peak) after a certain simultaneous steer-in and release of accelerator pedal, starting from a steady state
cornering at a certain path radius and speed. A certain road friction has to be defined.
When testing Transient driving manoeuvres, the typical part of the test track is the Vehicle Dynamics
Area or a Handling Track, see Figure 4-8. A Handling Track is a normal width road, intentionally
curved and with safety areas beside the curves for safety in case of run-off road during tests.
Typical transient tests are:
• Step steer, where one can measure transient versions of
o Yaw rate response
o Lateral acceleration response
o Curvature response
o Yaw damping
• Lateral avoidance manoeuvres:
o Single lane change in cone track
o Double lane change in cone track, see Figure 4-26, Figure 4-62, and Figure 4-63
o Lane change while full braking
o Sine with dwell
o Steering effort in evasive manoeuvres
220
LATERAL DYNAMICS
• Tests from steady state cornering

o Brake or accelerate in curve
o Lift accelerator pedal and steer-in while cornering
o Over-speeding into curve, see Figure 4-64
• Handling type tests
o Slalom between cones
o Handling track, general driving experience
• Roll-over tests
[m]
1.1 𝑊 + . 5
𝑊+ . 5
1.3 𝑊 + . 5
1
1.
3.5
driving direction
15 30 25 25 15+15
Figure 4-62: Passenger cars - Test track for a severe lane-change manoeuvre - Part 1: Double lane-change.
Reference (ISO 3888).
1.3 𝑊 + . 5
[m]
(but not <3)

1 𝑊+ . 5
1.1 𝑊 + . 5
driving direction
12 13.5 11 12.5 12
Figure 4-63: Passenger cars - Test track for a severe lane-change manoeuvre - Part 2: Obstacle avoidance.
Reference (ISO 3888).
Figure 4-64: Cone track for one standardized test of Over-speeding into curve
221
LATERAL DYNAMICS
Standards which are relevant to these test manoeuvres are, e.g. References (ISO 3888), (ISO 7401),
(ISO 7975, 2006), (ISO 11026), (ISO 14791), (ISO 14793), (ISO 14794, 2011) and (NHTSA).
4.5.2 Transient one-track model, without lateral

load transfer
4.5.2.1 Two-axle vehicle
Transient handling within lateral vehicle dynamics should be understood as that we assume a
reasonably constant longitudinal vehicle speed and modest longitudinal forces on the wheels. The
latter assumptions makes it possible to base most of the content in this chapter on the model derived
in 4.4.2 and Figure 4-51 and Equations [4.53] and [4.54]. However, in the context of transient
dynamics it is more relevant to use the model for more violent manoeuvres, and also active control
such as ESC interventions. Hence, we extend the model in two ways:
• The constitutive relation is saturated, to reflect that each axle may reach friction limit, friction
coefficient times normal load on the axle. See max functions in Equation [4.67].
• To be able to do mentioned limitation, the longitudinal load transfer is modelled, but only in
the simplest possible way using stiff suspension models. Basically, it is the same model as given
in Equation [3.21].
• A yaw moment representing (left/right) unsymmetrical braking/propulsion. See the term
𝑀 𝑐𝑡 in yaw equilibrium in Equation [4.67].
It should be noted that the model still lacks some: Transients in load transfer and the reduced
cornering stiffness and reduced max friction due to load transfer and utilizing friction for wheel
longitudinal forces.
Equilibrium in road plane (longitudinal, lateral, yaw):

∙( ̇ 𝜔 ∙ ) = 𝑤 ∙ cos( ) 𝑤 ∙ sin( ) + ;
∙( ̇ +𝜔 ∙ )= 𝑤 ∙ sin( ) + 𝑤 ∙ cos( ) + ;
∙ 𝜔̇ = ( 𝑤 ∙ sin( ) + 𝑤 ∙ cos( )) ∙ 𝑙 ∙𝑙 +𝑀 𝑐𝑡 ;
Equilibrium out of road plane (vertical, pitch):
+ ∙ = ;
∙𝑙 + ∙𝑙 ( 𝑤 ∙ cos( ) 𝑤 ∙ sin( ) + )∙ℎ = ;
Constitution:
[4.67]
𝑤= sign(𝑠 ) ∙ min(𝐶 ∙ |𝑠 |; ∙ );
= sign(𝑠 ) ∙ min(𝐶 ∙ |𝑠 |; ∙ ) ;
Compatibility:
𝑤 𝑙 ∙𝜔
𝑠 = ; 𝑛 𝑠 = ;
𝑤
Transformation from vehicle to wheel coordinate system on front axle:
𝑤 =( + 𝑙 ∙ 𝜔 ) ∙ sin( ) + ∙ cos( ) ;
𝑤 =( + 𝑙 ∙ 𝜔 ) ∙ cos( ) ∙ sin( ) ;
The model in Modelica format is given in Equation [4.68]. Changes compared to Equations [4.53] are
marked as underlined code.
222
LATERAL DYNAMICS
//Equilibrium, in road plane:

m*(der(vx)-wz*vy) = Ffxv + Frx;
m*(der(vy)+wz*vx) = Ffyv + Fry;
J*der(wz) = Ffyv*lf - Fry*lr + Mactz;
//Equilibrium, out of road plane:
Ffz + Frz - m*g = 0;
-Ffz*lf + Frz*lr -(Ffxv + Frx)*h = 0;
//Compatibility:
vfxv = vx;
vfyv = vy + lf*wz;
vrx = vx;
vry = vy - lr*wz;
//Lateral tyre force model:

Ffyw = -sign(sfy)*min(Cf*abs(sfy), mu*Ffz); [4.68]
Fry = -sign(sry)*min(Cr*abs(sry), mu*Frz);
sfy = vfyw/vfxw;
sry = vry/vrx;

//Shaft torques
Ffxw = +1000; // Front axle driven.
Frx = -100; // Rolling resistance on rear axle.
Mactz=0;
A simulation of this model, with same steering input as used in Figure 4-52. 𝑀 𝑐𝑡 is zero. Cornering
stiffnesses are chosen so that the vehicle is understeered in steady state. Road friction coefficient is 1.
We can see that the vehicle now gets unstable and spins out with rear to the right. This is mainly
because longitudinal load transfer unloads the rear axle, since the kept steering angle decelerates the
vehicle. In this manoeuvre, it would have been reasonable to model also that the rear cornering
stiffness decreases with the decreased rear normal load, and opposite on front. Such addition to the
model would make the vehicle spin out even more. On the other hand, the longitudinal load shift is
modelled to take place immediately. With a suspension model, this load shift would require some
more time, which would calm down the spin-out. In conclusion, the manoeuvre is violent enough to
trigger a spin-out, so a further elaboration with how to control 𝑀 𝑐𝑡 could be of interest. However, it
is beyond the scope of this compendium.
The vehicle reaches zero speed already after 7 seconds, because the wide side slip decelerates the
vehicle a lot. The simulation is stopped at time=7 seconds, because the model cannot handle zero
speed. That vehicle models become singular at zero speed is very usual, since the slip definition
becomes singular due to zero speed in the denominator. The large difference compared to Figure 4-52
is due to the new constitutive equation used, which shows the importance of checking validity region
for any model one uses.
223
LATERAL DYNAMICS
df=df [deg]
y [m]
wz [deg/s]
x[m]
yz=pz [deg]
time [s]
x[m]
Figure 4-65: Simulation results of one-track model for transient dynamics. The vehicle drawn in the path
plot is not in proper scale, but the orientation is approximately correct.
A simplified version of the mathematical model in Eq [4.67] follows in Eq [4.69]; assuming constant
and small angles and small propulsion force and no 𝑀 𝑐𝑡 .
Equilibrium in road plane (lateral, yaw):

∙( ̇ +𝜔 ∙ )= + ;
∙ 𝜔̇ = ∙𝑙 ∙𝑙 ;
Equilibrium out of road plane (vertical, pitch):
+ ∙ = ;
∙𝑙 + ∙𝑙 = ;
[4.69]
Constitution:
= sign(𝑠 ) ∙ min(𝐶 ∙ |𝑠 |; ∙ );
= sign(𝑠 ) ∙ min(𝐶 ∙ |𝑠 |; ∙ );
Compatibility:
+ 𝑙 ∙𝜔 𝑙 ∙𝜔
+𝑠 = ; 𝑛 𝑠 = ;
4.5.2.2 Transient models for articulated vehicles

A disclaimer to readers which only read the compendium, without following the course: This section
might appear incomplete, while the course lectures and hand-out of model (Modelica model at course
public web page) is more complete. Still, the description in compendium is regarded as relevant to give
general feeling for the modelling challenges with articulated vehicles.
224
LATERAL DYNAMICS
A model for a combination vehicle will now be derived. It is the simplest possible model; same type as
in Equation [4.55] and Figure 4-54. It is drawn as car with trailer or rigid truck with centre-axle trailer,
but with 𝑙 𝑐 < it becomes a model of a tractor with semitrailer, which is a very common vehicle on
our roads.
= ̇ 𝜔 ;
= ̇ +𝜔 ;
𝜔̇
1 𝑦𝑤
1𝑟𝑦
Δ 1𝑟𝑥 df
Δ 𝜔
𝑙 𝐶𝑜𝐺
𝑙 𝑐
part of free body free body diagram part of free body

diagram of trailer: of coupling: diagram of trailer:
𝑐
Figure 4-66: Model of two-unit articulated vehicle.

Model equations for the 1st unit:
In-road-plane Equilibrium of 1st unit (x, y, rotz):
( ̇ 𝜔 ) = cos( ) 𝑤 sin( ) 𝑤+ + ;
( ̇ +𝜔 ) = sin( ) 𝑤 + cos( ) 𝑤+ + ;
𝜔̇ = (sin( ) 𝑤 + cos( ) 𝑤 ) 𝑙 𝐶𝑜𝐺 ( 𝑙 𝐶𝑜𝐺 )
(𝑙 𝑐 𝑙 𝐶𝑜𝐺 );
Compatibility, within 1st unit:
𝑤 = sin( ) + cos( ) ;
= +𝜔 𝑙 𝐶𝑜𝐺 ; [4.70]
= 𝜔 ( 𝑙 𝐶𝑜𝐺 );
𝑐 = 𝜔 (𝑙 𝑐 𝑙 𝐶𝑜𝐺 );
Constitution for axles on 1st unit:
𝑤
𝑤 = 𝐶 𝑠 ; 𝑠 = ;
= 𝐶 𝑠 ; 𝑠 = ;
225
LATERAL DYNAMICS
Model equations for the 2nd unit:

Equilibrium of 2nd unit (x, y, rotz):
( ̇ 𝜔 )= + ;
( ̇ +𝜔 )= + ;
𝜔̇ = 𝑙 𝐶𝑜𝐺 + (𝑙 𝑐 𝑙 𝐶𝑜𝐺 );
Compatibility, within 2nd unit: [4.71]
= 𝜔 𝑙 𝐶𝑜𝐺 ;
𝑐 = +𝜔 (𝑙 𝑐 𝑙 𝐶𝑜𝐺 );
Constitution for axles on 2nd unit:
= 𝐶 𝑠 ; 𝑠 = ;
Model equations for the coupling:

Equilibrium of coupling (x, y in 1st unit’s coordinates):
+ cos(Δ ) + sin(Δ ) = ;
sin(Δ ) + cos(Δ ) = ;
Compatibility of coupling: [4.72]
= + cos(Δ ) + sin(Δ ) 𝑐 ;
𝑐 = sin(Δ ) + cos(Δ ) 𝑐 ;
Δ ̇ =𝜔 𝜔 ;
Let us assume ; 𝑤; 𝑤; are known inputs. The above is then a DAE system in Equations
[4.70].. [4.72] with 19 equations and 19 unknowns: ; ;𝜔 ; ; ;𝜔 ;Δ ; 𝑤; ; ;
𝑠 ;𝑠 ;𝑠 ; 𝑐 ; 𝑐 ; ; ; ; . The equations can be written directly into a Modelica tool,
which will find the explicit form and simulations can be done. The use of a Modelica tool is much more
motivated here, compared to a non-articulated vehicle, since the articulation points makes the DAE
system of equation to be of “higher index”, meaning that some equations needs to be differentiated, cf.
Section 3.3.2.
4.5.2.2.1 Model library for articulated vehicles

The first model (Eqs [4.70].. [4.72]) for articulated vehicle is written so that each equation belongs to
either a unit or a coupling. This opens up for systematic treatment of combination vehicles with more
than one articulation point. There are basically two conceptual ways:
• Vectorised formulation, see Reference (Sundström, o.a., 2014).
• Modular library from which parts can be graphically dragged, dropped and connected. This is
briefly visualized as implemented in the Modelica-tool Dymola in Figure 4-68.
Modular library
A modular library is shown in Figure 4-68. Eq [4.70] is declared in “Unit” and Eq [4.72] in “Coupling”.
226
LATERAL DYNAMICS
drag and drop

connect
Figure 4-68: Drag, drop and connect library for heavy combination vehicles. The model example shows a so
called A-double, Tractor+Semitrailer+Dolly+Semitrailer.
An example of lane change manoeuvre, defined as lateral acceleration on 1st axle follows a single sine-
wave, is shown in Figure 4-69. The natural input is prescribed steering angle ( = sin(𝑡 );), but
since modelled in Modelica, it is as easy to prescribe something else, e.g. lateral acceleration on first
axle ( = sin(𝑡 );).
Figure 4-69: One period sinus test of a Tractor+Semitrailer.
4.5.2.3 Cambering vehicle model

The model below shows how a cambering vehicle can be modelled. The model is mainly made for
lateral dynamics, but it allows also longitudinal acceleration. The drawing shows a two-wheeler, but
227
LATERAL DYNAMICS
any or both of the axles could have two wheels, as long as the suspension linkage is such that the axle
does not take any roll moment. The model is shown in Eq [4.74] in Modelica format. It is not modelled
that driver moves within the vehicle, which is why the inertial data is constant. Also, the chassis
(frame) is modelled as stiff and steering system as massless. The model lacks two equations, which is
logical since a driver model can add prescribed steering angle and prescribed .
view from side view from rear

𝜔̇
ℎ
𝑙 𝑙
= ̇ 𝜔 ;
= ̇ +𝜔 ;
𝜔 = ̇ ;
𝜔̇
view from above
Figure 4-70: Model of cambering vehicle. (The stars marks point of moment equilibria in the “mathematical
model” derived from this “physical model”.)
//Equilibrium in road plane (x,y,rotz):
m*ax = Ffxv + Frx;
m*ay = Ffyv + Fry;
Jz*der(wz) = Ffyv*lf - Fry*lr - m*ax*h*px;
ax = der(vx) - wz*vy;
ay = der(vy) + wz*vx;
//Equilibrium out of road plane (z, rotx, roty):

m*g = Ffz + Frz;
Jx*der(wx) = m*g*h*px + m*ay*h;
0 = -Ffz*lf + Frz*lr - m*ax*h;
wx = der(px);
[4.74]
//Constitution:
Ffyw = -CC*Ffz*sfy;
Fry = -CC*Frz*sry;
//Compatibility, slip definition:

atan(sfy) = atan2(vy + lf*wz + h*wx, vx - h*px*wz) - d;
atan(sry) = atan2(vy - lr*wz + h*wx, vx - h*px*wz) - 0;
//Force coordinate transformation:

Ffxv = -Ffyw*sin(d);
Ffyv = +Ffyw*cos(d);
When entering a constant radius curve from straight driving one has to first steer out of the curve to
tilt the vehicle a suitable amount for the coming path curvature, 𝑅 . The suitable amount is hence
𝑠𝑢 𝑡 𝑒 = ⁄ = 𝜔 ⁄ = ⁄(𝑅 ) ;. Then one steers with the turn and balances
(closed loop controls) to the desired roll angle. Systems like this, which has to be operated in opposite
direction initially is called “Non-minimum phase systems”. It is generally difficult to design a controller
for such systems. The two simulations shown in Figure 4-87 are done with the above model. Initial
228
LATERAL DYNAMICS
speed is = 1 𝑠 and = . Path radius, 𝑅 (𝑡) = 𝑠𝑡 𝑝 𝑢𝑛 𝑡 𝑜𝑛 𝑡 𝑡 = .1, representing a

suddenly curving road or path.
• One simulation (veh_driver) uses a closed loop driver model which first steers outwards
( < ) and then continuously calculates the steering angle as a closed loop controller: =
𝐺 𝑛 ( 𝑒𝑞 );. It is not claimed that the driver model is representative for real drivers.
• In the other simulation (veh_inverse), the roll angle is prescribed as = ⁄(𝑅 ) ;.
To prescribe the roll angle, instead of steering angle, is a way to avoid the “Non-minimum
phase” difficulties. The system becomes a normal Minimum phase system if actuated with roll
angle instead of steering angle. The price one have to pay for this is that the model equations
has to go through a more advanced symbolic manipulation, e.g. differentiation, to solve for all
variables, including the steering angle . However, with a Modelica tool the symbolic
manipulation is done automatically. One can see this as a way to avoid a controller design and
instead use an optimal driver; optimal in the sense that it follows the path curvature with
optimal yaw agility. The road path curvature is a step function, but has to be filtered twice
(time constant 0.1 s is used in Figure 4-87) to become differentiable. veh_driver.
veh_inverse
10
The latter, optimal driver, negotiates the turn without overshot in yaw rate, so it follows a suddenly
curving path better.
8
veh_driver.d veh_inverse.d veh_driver.px veh_inverse.px
10
d= =steering angle [deg] 0
px= =roll angle [deg] 6
4
0 -10
2
-10 -20
0.0 0.4 0.8 1.2 0.0 1.6 0.4 2.0 0.8 1.2 1.6 2.0
0
veh_driver.Ffyv veh_inverse.Ffyv veh_driver.w z veh_inverse.w z
1000
1 -2
0
Ffyv= = -4
0
vehicle-lateral force
-1000 on front tyre [N] wz=𝜔 =yaw velocity [rad/s]
-6
-1
0.0 0.4 0.8 1.2 0.0 1.6 0.4 2.0 0.8 1.2 1.6 2.0 0
Figure 4-71: Simulations of entering a curve with a cambering vehicle. Blue solid curves without dot-marker
show a closed-loop driver model which actuates . Red curves with dot-marker shows an optimal/inverse
driver model which actuates .
Both driver models above only exemplify the low-level, roll-balancing, part of a driver. To run the
model in an environment with obstacles, one would also need a high-level, path selecting, part which
outputs desired, e.g., or . Additional driver model for longitudinal actuation is also needed.
It can be noted that the roll influences in two ways, compared to the non-cambering (roll-stiff) vehicles
previously modelled in the compendium:
• The roll motion itself is a dynamic motion, where the roll velocity becomes a state variable
carrying kinetic energy.
• The roll influences the tyre slip, e.g. rear: 𝑠 = ( 𝑙 𝜔 + ℎ 𝜔 )⁄( ℎ 𝜔 );. The
term ℎ 𝜔 can generally be neglected for non-cambering vehicles, but for a cambering vehicle,
such as a bike, it is essential. The term ℎ 𝜔 is only important at large roll angles, it is for
instance used as lever for longitudinal wheel forces in ESC-like control systems for motorbikes.
229
LATERAL DYNAMICS
4.5.3 Transient two-track model, with Lateral load

transfer
In model of longitudinal load transfer in Section 3.3.5, we assumed instantaneous load transfer. In
present chapter, we aim to capture the transients better.
Compared to the model presented in Figure 4-40 and Equations [4.40]..[4.46], we add the following:
• Inertial term for roll rotation, i.e. 𝑠 𝜔̇ . Note that sprung body roll inertia ( 𝑠 ), not whole
vehicle roll inertia ( ), is used in roll equilibrium because the unsprung parts does not
participate in roll motion.
• Damping forces in parallel to each spring force, i.e. adding 𝑑 𝑑 𝑑 and 𝑑 .
Due to the axle roll centre model, as opposed to wheel pivot point model, the heave acceleration is
zero. Hence, no inertial force needs to be introduced.
The free-body diagrams are given in Figure 4-72, which should be compared to Figure 4-40.
displacements, speeds, accelerations: forces, on body:

ay m*g
z
zfl=zrl y
zfr=zrr 𝑠 𝜔̇ m*ay
px
Pry
Pfy
Fsfl+Fsrl+ Fsfr+Fsrr+
zflr=zrlr=0
zfrr=zrrr=0 +Fdfl+Fdrl Pfz+Prz +Fdfr+Fdrr
forces, forces, on front axle: forces, on rear axle:

on vehicle: m*g Fsfl+Fdfl Fsfr+Fdfr Fsrl+Fdrl Fsrr+Fdrr
𝜔̇ m*ay Pfz Prz
𝑠
Pry
h Pfy
Ffly+Frly Ffly Ffry Frly Frry

Ffry+Frry
Fflz Ffrz Frlz Frrz
Fflz+Frlz Ffrz+Frrz
Figure 4-72: Model for transient load transfer due to lateral acceleration, using axle roll centres.
The constitutive equations for the compliances (or springs) are as follows, see Equation [4.39]. Note
that the anti-roll bars are not modelled. Note also that we differentiate, since we will later use the
spring forces as state variables in a simulation.
𝑠 = 𝑠 0 + 𝑤∙( ); ⇒ 𝑠̇ = 𝑤∙ ̇ = 𝑤∙ ;
𝑠 = 𝑠 0+ 𝑤∙(
̇
); ⇒ 𝑠 = 𝑤∙ ̇ = 𝑤∙ ;
= 0+ 𝑤∙(
̇
); ⇒ 𝑠 = 𝑤∙ ̇ = 𝑤∙ ;
𝑠 𝑠 [4.75]
𝑠 = 𝑠 0+ 𝑤∙( ); ⇒ 𝑠̇ = 𝑤 ∙ ̇ = 𝑤 ∙ ;
∙ ∙𝑙 ∙ ∙𝑙
𝑤ℎ 𝑟 𝑠 0 = 𝑠 0 = ; 𝑛 𝑠 0 = 𝑠 0 = ;
∙ ∙
230
LATERAL DYNAMICS
The constitutive equations for the dampers have to be added: 𝑑 = 𝑤∙ ;

= 𝑤∙ ;
𝑑
[4.76]
𝑑 = 𝑤∙ ;
𝑑 = 𝑤∙ ;
As comparable with Equation [4.40], we get the next equation to fulfil the equilibrium. The change
compared to Equation [4.40] is that we also have a roll and lateral inertia terms and 4 damper forces,
acting in parallel to each of the 4 spring forces. Actually, when setting up equations, we also
understand that a model for longitudinal load transfer is needed, which is why the simplest possible
such, which is the stiff suspension on in Equation [3.21].
In-road-plane: Equilibrium for vehicle (longitudinal, lateral, yaw):

∙ = ∙( ̇ 𝜔 ∙ )= + ;
∙ = ∙( ̇ +𝜔 ∙ )= + ;
∙ 𝜔̇ = ∙𝑙 ∙𝑙 ;
Out-of-road-plane: Equilibrium for vehicle (vertical, pitch, roll):
+ + + ∙ = ;
( + )∙𝑙 +( + )∙𝑙 ( + )∙ℎ = ; [4.77]
𝑤 𝑤
𝑠 ∙ 𝜔̇ = ( + )∙ ( + )∙ +( + ) ∙ ℎ;
Equilibrium for each axle (pitch, around roll centre):

𝑤 𝑤
( ( 𝑠 + 𝑑 )) ∙ ( ( 𝑠 + 𝑑 )) ∙ + ∙ ℎ𝑅𝐶 = ;
𝑤 𝑤
( ( 𝑠 + 𝑑 )) ∙ ( ( 𝑠 + 𝑑 )) ∙ + ∙ ℎ𝑅𝐶 = ;
Compatibility gives, keeping in mind 𝑤 𝑤

̇ = =+ ∙𝜔 ; 𝑛 ̇ = = ∙𝜔 ;
that 𝜔 is the only non-zero out-of [4.78]
𝑤 𝑤
road plane velocity (i.e. ̇ = 𝜔 = ): ̇ = =+ ∙𝜔 ; 𝑛 ̇ = = ∙𝜔 ;
Constitution for tyre forces versus slip in ground plane becomes per wheel, as opposed to per axle in
previous model:
𝐶 𝐶
𝑤 = sign(𝑠 ) ∙ (min ( ∙ |𝑠 |; ∙ ) + min ( ∙ |𝑠 |; ∙ )) ;
[4.79]
𝐶 𝐶
= sign(𝑠 ) ∙ (min ( ∙ |𝑠 |; ∙ ) + min ( ∙ |𝑠 |; ∙ )) ;
Equations [4.75]..[4.79] give a model very similar to the one in Equation [4.68]. Only the additional
equations from this new model are shown in Equation [4.80]. Note especially the new lateral tyre force
model equations, which now have one term per wheel, because one wheel on an axle can saturate
independent of the other on the same axle.
//Equilibrium, roll
Jsx*der(wx) = (Fflz + Frlz)*w/2 - (Ffrz + Frrz)*w/2 + (Ffyv + Fry)*h;
//Equilibrium for each axle (pitch, around roll centre):
(Fflz - (Fsfl + Fdfl))*w/2 - (Ffrz - (Fsfr + Fdfr))*w/2 + Ffy*hRCf = 0; [4.80]
(Frlz - (Fsrl + Fdrl))*w/2 - (Frrz - (Fsrr + Fdrr))*w/2 + Fry*hRCr = 0;
//Constitutive relation for tyres (Lateral tyre force model):
231
LATERAL DYNAMICS
Ffyw = -
sign(sfy)*(min((Cf/2)*abs(sfy), mu*Fflz) + min((Cf/2)*abs(sfy), mu*Ffrz));
Fry = -
sign(sry)*(min((Cr/2)*abs(sry), mu*Frlz) + min((Cr/2)*abs(sry), mu*Frrz));
sfy = vfyw/vfxw;
sry = vry/vrx;
//Constitution for springs:
der(Fsfl) = -cfw*vflz;
der(Fsfr) = -cfw*vfrz;
der(Fsrl) = -crw*vrlz;
der(Fsrr) = -crw*vrrz;
//Constitution for dampers:
Fdfl = -dfw*vflz;
Fdfr = -dfw*vfrz;
Fdrl = -drw*vrlz;
Fdrr = -drw*vrrz;
//Compatibility, out of road plane:

vflz = +w/2*wx;
vfrz = -w/2*wx;
vrlz = +w/2*wx;
vrrz = -w/2*wx;
A simulation of this model is shown in Figure 4-73 and should be compared with the simulation in
Figure 4-52. When comparing these, we see a slight difference, which is that the axles saturate
gradually during 3<time<3.5, instead of both at once at time=3.25.
df=df [deg]
y [m]
wz [deg/s]
x[m]
yz=pz [deg]
x[m]
time [s]
Figure 4-73: Simulation results of one-track model for transient dynamics with lateral load transfer. The
vehicle drawn in the path plot is not in proper scale, but the orientation is approximately correct.
Even if the load transfer model does not influence the vehicle path a lot in this case, it may be
important to include it to validity check the model through checking wheel lift. Wheel lift can be
identified as negative vertical wheel forces, which are why we plot some vertical wheel forces, see
Figure 4-74. In this case we see that we have no wheel lift (which would disqualify the simulation). In
the right part of the figure we can also see the separate contribution from spring ( 𝑠 ), damper ( 𝑑 )
and linkage ( 𝑛𝑘 = 𝑠 𝑑 ).
232
LATERAL DYNAMICS
4.5.3.1 Additional phenomena

It is relevant to point out the following, which are not modelled in this compendium:
• Same as pointed out as missing for longitudinal load transfer, see Section 3.4.7.5.
• Additionally, anti-roll arrangements (elastic connections between left and right wheel on one
axle, often built as torsion bar) are not modelled in this compendium. With same modelling
concept as used above, each such would be treated as a separate spring with one state variable,
e.g. (Force-antiroll-front). This force will act in parallel with 𝑠 and 𝑑 on each side. Note
that it will be added on one side and subtracted on the other.
Frrz [N]
Fsrr [N]
Ffrz [N]
Frrz [N]
Flink,rr,z [N]
Fflz [N]
Frlz [N] Fdrr [N]
time [s] time [s]

Figure 4-74: Suspension vertical force plots from simulation with one-track model for transient dynamics
with lateral load transfer dynamics (same simulation as in Figure 4-73). Left: Road contact forces for all
wheels. Right: Different forces for one wheel, rear left.
4.5.3.2 Example of explicit form model; two-axle vehicle, driver

and environment
The implementation of the model in Section 4.5.2.3 was done in Modelica. A Modelica tool
automatically transforms the model to explicit form which can be simulated, which is very efficient.
But, as mentioned in Section 1.5.7.5, explicit form models can sometimes facilitate the understanding
of the vehicle’s dynamics. This is why the following model is presented. It is implemented in the data
flow diagram tool Simulink. This section explains how the states (or state variables) together with
inputs (or input variables), influence the derivatives (or state derivatives).
The example model in this section is similar, but not identical, to the model in Section 4.5.2.3. The aim
is to model in-road-plane motion, due to transient actuation (wheel torques and wheel steering
angles). Limitations in this example model are:
• Influence from vertically uneven road is NOT modelled.
• Neither wheel lift nor suspension bump stop is modelled
• Control functions (such as ABS, TC and ESC) are NOT modelled
• The pendulum effect is NOT modelled, see 4.3.9.2
• Wheel camber and steering angle change with suspension travel is NOT modelled.
The model is a typical passenger vehicle, with driver and environment, see Figure 4-75. The driver
interface is the normal, 2 pedals, 1 steering wheel and a boolean request for direction of propulsion,
LongDir (=-1 or +1). The interface to environment is motion (variable position) in surrounding
world. To try out the vehicle model, driver and environment is also modelled. This includes also the
interface between them, which is motion of obstacles in environment, relative to subject vehicle. The
suspension is exemplified with wheel-individual suspension on both axles.
233
LATERAL DYNAMICS
Environment Driver Vehicle
Figure 4-75: Top level of model with model tree structure. The Environment is a track test with cone walls to
go left and right around. Notation “irp” and “oorp” refers to in-road-plane and out-of-road-plane,
respectively.
As an initial overview, the states are presented. There are 21 states in total, and distributed:
• Driver: 0 states
• Vehicle:
o Vehicle Control & Actuation: 0 states
o Wheels, Tyres & Suspension: 12 states (4 wheels’ rotational speed, 4 Elastic parts of
vertical wheel forces, 4 Longitudinal tyre forces, 4 Lateral tyre forces)
o Vehicle Motion: 9 states (6 velocities and 3 positions)
• Environment: 0 states
The 4+4 tyre force states arise from modelled tyre relaxation, see Section 2.5.5.
4.5.3.2.1 Submodel Environment
Generally, the environment model is where the surrounding to the driver and the vehicle should be
defined: road/road network, obstacles, other road users and the “driving task”/”driving instructions”.
In this example, it is very small and simple; it only captures stand-still point obstacles, each with
instructions whether to be passed as obstacle left or right of the vehicle. Inputs to environment model
is the (subject) vehicles position, including orientation in global coordinates. Outputs are the relative
position (x,y) to each obstacle, in (subject) vehicle coordinate system.
4.5.3.2.2 Submodel Driver
In this example, the driver model is very small and simple. Briefly described, it treats the longitudinal
dynamics very simple, as closed loop control towards a constant desired speed forward. The lateral
dynamics is divided into two parts:
• Driver planning: Based on how driver perceives the obstacles relative to the vehicle, one of the
objects is selected to mind for, which leads to where to be aim. Basically, the nearest obstacle
ahead of vehicle is selected as mind for and aim is, in principle, either half a vehicle width left
(or right) of this obstacle.
• Driver operation: Based on the driver’s motorics, the steering wheel angle is calculated. In the
example, it is simply an inverse model of an ideally tracking two-axle vehicle which calculates
which constant steering wheel angle that would be needed to make front axle run over the aim
obstacle.
It can be noted that a Driver model would also be a logical model part where to include calculation of
driver’s perception, such as steering effort and perceived safety during manoeuvre, etc. It can also be
noted that the border between Environment and Driver is sometimes not obvious, especially when it
234
LATERAL DYNAMICS
comes to modelling the “driving task” in the Environment model, which can also be seen as a “driver
high-level decision” and then be a logical part of the Driver model. If Environment model includes
surrounding vehicles (object vehicles), it also includes driver models for those. For automatic
functionality, anything from cruise control to automated driving, there should also be a “button HMI
output” from Driver model, not only pedals and steering wheel. Such interface would turn on/off such
functionality in the Vehicle model.
4.5.3.2.3 Submodel Vehicle
9 states
12 states
Figure 4-76: Vehicle submodel.

The figure shows the decomposition of the vehicle into 3 parts:
• Submodel “Vehicle Control & Actuation” models the actuators (Propulsion system, Brake
system and Steering system, including “control functions”) that respond on requests from the
driver with wheel torques (T_s and T_us) and wheel steering angles (delta_w). The
notation ending “4” refers to that the quantities are vectors with 4 components, one per wheel.
T_s is shaft torque and T_us is torque applied from unsprung parts, e.g. friction brake torque
from brake calliper. One can think of very advanced models of these actuator systems,
including e.g. propulsion system dynamics and control functions (ABS, ESC, TC, …). However, in
this example model it is only modelled very simple:
o Propulsion system outputs a fraction (determined by APed) of a certain maximum
power, distributed equally on front left and front right wheel. If brake pedal is applied,
the propulsion system outputs zero torque.
o Brake system outputs a fraction (determined by BPed) of a maximum brake torque (
𝑅𝑤 ), distributed in a certain fix fraction between front and rear axle (70/30). The
distribution within each axle is equal on left and right wheel.
o There are no states modelled in the vehicle Control & Actuator submodel.
• Submodel “Wheels, Tyre & Suspension” models the part which pushes the tyres towards the
ground and consequently transforms the wheel torques and wheel steering angles, via the tyre,
to forces on the whole vehicle. F_xyv8 is the x and y forces in each of the 4 wheels, 2x4=8.
F_z4 is the 4 vertical forces under each wheel. Submodel “Wheels, Tyre & Suspension” is
further explained in Section 4.5.3.2.5.
• Submodel “Vehicle Motion” models motion of whole vehicle in-road-plane and motion of
sprung body out-of-road-plane. The inertial effects (mass∙acceleration) of the unsprung parts
are taken into account for in-road-plane but not for out-of-road-plane. This submodel includes
integrators for the 3+3+3=9 states:
o Velocities in-road-plane, 𝜔 : v_irp3 (which is transformed to x- and y-
velocities of each wheel and then fed back as v_irpv8)
o Position in-road-plane, 𝑥 𝑦 : pos_irp3 (which is only fed forward to
“Environment”)
235
LATERAL DYNAMICS
o Velocities out-of-road-plane, 𝜔 𝜔 : v_oorp3 (which is transformed to z-velocities

of sprung body over each wheel and then fed back as v_z4)
4.5.3.2.4 Submodel Vehicle Control & Actuators
Figure 4-77: Vehicle Control & Actuators sub-model.

The example content of this sub-model is very minimalistic but can still be explained as two parts:
• Interpret pedals (including arbitrate between accelerator and brake pedal) to a sum over
wheels longitudinal force request ( 𝑒𝑞 ) and steering wheel angle to a front road wheel angle
request ( 𝑒𝑞 ).
• Coordinate and actuate propulsion and brake, i.e. allocate 𝑒𝑞 to 4 wheels’ propulsion
torques and brake torques. Also allocate 𝑒𝑞 to each of left and right front wheels.
Vehicle variables used for the control are wheel rotational speeds. Since front axle propulsion is
assumed, the front rotational speeds are also input to the propulsion actuator modelling. No state
variables are present in this minimalistic example, but in a more advanced actuation model there
could typically be states such as: engine speed, gear (discrete state), delay states for brake system and
elastic forces in steering system.
4.5.3.2.5 Submodel Wheels, Tyres & Suspension
This submodel is shown in Figure 4-78.
• The two sub-models “Coord Transf …” are straight-forward coordinate transformations, using
Eq [4.3].
• The sub-model “Wheels” is also relatively straight-forward. For each wheel, the rotational
equilibrium is used as model: ∙ 𝜔̇ = 𝑇𝑢𝑠 + 𝑇𝑠 ∙ 𝑅𝑤 si n(𝜔) ∙ 𝑅𝑅𝐶 ∙ ∙ 𝑅𝑤 ; where 𝑅𝑤 is
wheel radius and 𝑅𝑅𝐶 is rolling resistance coeeficient. The submodel will hence contain the 4
states: Rotational speeds of each wheel: w_w4.
• The sub-model “Springs, Dampers & Linkage” models the springs (incl. anti-roll-bars) and
dampers and the linkage. For each wheel:
o Four states: Elastic part of vertical tyre force under each wheel: F_s4
o The derivatives are governed by the differentiated constitution of the springs:
Conceptually 𝑠̇ = ; but involving both wheel spring and anti-roll-bar.
o The force in damper is governed by the dampers constitutive relation: 𝑑 = ;
o The contact forces are calculated in submodel “Suspension Equilibrium” in Figure
4-80. They are calculated from moment equilibrium of unsprung parts around a 3-
dimensional pivot axis. The pivot axis is defined by two points, the pivot point in
236
LATERAL DYNAMICS
longitudinal load transfer (see Figure 3-45) and the pivot point in longitudinal load
transfer (see Figure 4-39). The scalar equilibrium equation for one wheel can be
expressed, with vector (cross) and scalar (dot) products, in 𝑠 𝑑 𝑇𝑠 and
point coordinates. From this, can be solved.
marks one scalar state
Figure 4-78: Wheels, Tyres & Suspension sub-model.

• The sub-model “Tyres” models the tyre mechanics, very much like the combined slip model in
Equation [2.44] and the relaxation model in Eq [2.51]. For each wheel:
o Four states: Magnitudes of tyre forces in ground plane ( ) for each wheel: F_xyw4
o Unfortunately, the tyre forces and depend on . This could easily create algebraic
loops. However, since we also model relaxation, the tyre forces become state variables
which breaks such algebraic loops. Another way of getting rid of the algebraic loops
could have been to use “memory blocks”. “Memories” are such that value from last time
instant is used to calculate derivatives in present time instant. This is generally NOT a
recommended way of modelling.
Figure 4-79: Sub-model “Springs, Dampers & Linkage”.
237
LATERAL DYNAMICS
Figure 4-80: Sub-model “Tyres”.

4.5.3.2.6 Sub-model Vehicle Motion
The sub-model is shown in Figure 4-81. It is divided in upper part in-road-plane (irp) and lower part
out-of-road-plane (oorp). The velocities in road plane (v_irp3= [ 𝜔 ]) is needed also in Sprung
body OORP because of the centripetal term, 𝜔 identified in Section 3.3.5.1 and used in 3.4.9.1.
• state
Figure 4-81: Vehicle Motion.

4.5.3.2.7 Simulation example Double Lane Change
A double lane change between cones is used as simulation example, see Figure 4-82. The cones are run
over and even passed on the wrong side because the driver model is very simple.
238
20
LATERAL DYNAMICS
10
cone
cone
0 7
6.5
5.5
tyre force in ground plane
(vertical tyre force is radius
-10 5
of circle)
4.5
4
wheel hub translational
3.5
velocity
3
cone
-20
2.5
186 187 188 189 190 191
-30
150 160 170 180 190 200 210 220
Figure 4-82: Simulation results of a double lane change between cones.
4.5.3.3 Transient roll-over *

Function definition: Transient roll-over resistance is the most severe measure of a certain transient
manoeuvre that the vehicle can manage without lifting all wheels on one side. The manoeuvre is typically on level
ground with high road friction and certain payload, loaded high.
The manoeuvre can typically be a double lane change, since the double triggers roll oscillations. The
severity measure of the manoeuvre can be the peak lateral acceleration or a lane change width or
longitudinal speed for given lane width.
4.5.4 Step steering response *

Function definition: Step steering response is the response to a step in steering wheel angle measured in
certain vehicle measures. The step is made from a certain steady state cornering condition to a certain steering wheel
angle. The response can be the time history or certain measures on the time history, such as delay time and
overshoot.
4.5.4.1 Mild step steering response

This section is to be compared with section 4.5.4.2, which uses a more advanced model. In present section
a less advanced model will be used, which is enough for small steering steps.
239
LATERAL DYNAMICS
The model used for single frequency stationary oscillating steering can also be used for other
purposes, as long as limited lateral accelerations. Most common interpretation is to make the steering
step from an initial straight-line driving. In reality, the step will be a quick ramp. Equation [4.56]
allows an explicit solution prediction of stationary oscillating steering, but also for step response:
̇
Start from Equation [4.56]: [ ] = 𝐀 ∙ [𝜔 ] + 𝐁 ∙ ;
𝜔̇
( ) 0 0
With initial conditions: [ ] = [𝜔 ] ; or [𝜔 ] = 𝐀− ∙ 𝐁 ∙ 0 ; where 0 is before step.
𝜔 ( ) 0 0
∞ ̂ ̂ 𝜆1 ∙𝑡
Assume: [𝜔 ] = [𝜔 ] + [[ ] [ ]] ∙ [ 𝜆2 ∙𝑡
] ∙ [ ]; ⇒
∞ 𝜔̂ 𝜔
̂
̇ ̂ ̂ 𝜆 ∙ 𝜆1 ∙𝑡
⇒ [ ] = [[ ] [ ]] ∙ [ 𝜆2 ∙𝑡
]∙[ ];
𝜔̇ 𝜔
̂ 𝜔
̂ 𝜆 ∙
̂ ̂ 𝜆1 ∙𝑡
𝜆 ∙
Insert: [[ ] [ ]] ∙ [ ]∙[ ] =
𝜔
̂ 𝜔
̂ 𝜆 ∙ 𝜆2 ∙𝑡
∞ ̂ ̂ 𝜆1 ∙𝑡
= 𝐀 ∙ ([𝜔 ] + [[ ] [ ]] ∙ [ 𝜆2 ∙𝑡
]∙[ ]) + 𝐁 ∙ ;
∞ 𝜔
̂ 𝜔
̂
Solve for each time function term (constant, and 𝜆2 ∙𝑡 terms):
𝜆1 ∙𝑡
∞ ̂ ̂ 𝜆
[𝜔 ]= 𝐀− ∙ 𝐁 ∙ ; 𝑛 [[[ ] [ ]] [ ]] = ei (𝐀) ;
∞ 𝜔
̂ 𝜔
̂ 𝜆
The function ”eig” is identical to function ”eig” in Matlab. It is defined as eigenvalues and eigenvectors
for the matrix input argument.
0 ∞ ̂ ̂
Initial conditions: [𝜔 ] = [𝜔 ] + [[ ] [ ]] ∙ [ ] ; ⇒
0 ∞ 𝜔
̂ 𝜔̂
−
̂ ̂ 0 ∞
⇒[ ] = [[ ] [ ]] ∙ ([𝜔 ] [𝜔 ]) ;
𝜔̂ 𝜔
̂ 0 ∞
The solution in summary:
∞ ̂ ̂ 𝜆1 ∙𝑡
[𝜔 ] = [𝜔 ] + [[ ] [ ]] ∙ [ 𝜆2 ∙𝑡
] ∙ [ ];
{ ∞ 𝜔
̂ 𝜔
̂
= ̇ + ∙ 𝜔 = 𝜆 ∙ ̂ ∙ 𝜆1 ∙𝑡 ∙ + 𝜆 ∙ ̂ ∙ 𝜆2 ∙𝑡 ∙ + ∙𝜔 ;
∞ ̂ ̂ 𝜆 [4.81]
where: [𝜔 ]= 𝑨− ∙ 𝑩 ∙ ; and [[[ ] [ ]] [ ]] = ei (𝑨) ;
∞ 𝜔̂ 𝜔̂ 𝜆
−
̂ ̂ 0 ∞
and [ ] = [[ ] [ ]] ∙ ([𝜔 ] [𝜔 ]) ;
𝜔
̂ 𝜔
̂ 0 ∞
Results from this model for step steer to +3 deg are shown in Figure 4-83. Left diagram shows step
steer from straight line driving, while right diagram shows a step from steady state cornering with -3
deg steering angle.
4.5.4.2 Violent step steering response

This section is to be compared with section 4.5.4.1 Mild step steering response, which uses a model
with linear tyre models without saturation. In present section, a more advanced model will be used,
which might be needed when the step steering is more violent.
240
LATERAL DYNAMICS
steering step response, passenger car in vx=70 km/h steering step response, passenger car in vx=70 km/h
20 20
15 15
10 10
5 5
0 0
-5 -5
beta=arctan(vy/vx) [deg]
-10 wz [deg/s] -10 beta=arctan(vy/vx) [deg]
ay [m/(s*s)] wz [deg/s]
-15 df [deg] -15 ay [m/(s*s)]
df [deg]
-20 -20
-0.5 0 0.5 1 1.5 2 -0.5 0 0.5 1 1.5 2
vx
time [s] 28 time [s]
Figure 4-83: Steering step response. Simulation

26
with model from Equation [4.81].
Most common interpretation is to make the steering step0 from1 an initial 2 straight-line
3 4 driving.
5 In
reality, the step will be a quick ramp. In simulations, an ideal dfstep can
sfy
be used.
sry
The transients can easily be that violent that a model as

0.0 Equations [4.75]..[4.80] is needed. If ESC is to
be simulated, even more detailed models are needed (full two-track models, which are not presented
in this compendium). Anyway, if we apply a step steer-0.2to0 the model
1 in
2 Equations
3 4[4.75]..[4.80],
5 we can
simulate as in Figure 4-84.
Ffyw Fry
8000
vx
28
4000
26 vx[m/s]
0 1 2 3 4 5 0
0 1 2 3 4 5
df sfy sry
df[rad]
0.0 beta_deg wz_dps ay
15
sry[1] wz[deg/s]
-0.2 sfy[1] 10
0 1 2 3 4 5
ay[m/(s*s)]
5
Ffyw Fry
8000 Ffyw[N]
0
4000 Fry[N]
-5 beta[deg]
0
-10
0 1 2 3
time[s] 4 5 0 1 2 time[s]
3 4 5
Figure 4-84: Step steer with

beta_deg on road
wz_dps wheels
ay at = 1 km/h. Simulation using model from Eq
15
[4.75]..[4.80].
10
4.5.5 Phase portrait

5
The transient from0 one steady state to another is seen after the steps in Figure 4-83 and Figure 4-84.
Plotting several such
-5 transients as trajectories in a 𝛽 𝜔 -diagram gives a phase portrait, which is a
good graphical representation to understand how a vehicle stabilize itself, or gets unstable. For
-10
transients that stays0 within
1 unsaturated
2 3 tyre4 slip, the
5 linear one-track model can be used and the
trajectories can then be explicit time expressions using Eq[4.81]. This is exemplified in left part of
Figure 4-85. Some states can be confirmed stable already from this simple model. Simulations with
higher fidelity model is exemplified in right part of Figure 4-85. With that one can confirm some more
stable areas.
241
LATERAL DYNAMICS
model valid w.r.t. < ;

oversteering
(transient)
small negative 𝑠 𝑠
lar e ne ati e 𝑠
⁄𝑠
𝑟
𝜔
understeering
𝑠
(transient)
large negative 𝑠
small ne ati e 𝑠 model valid w.r.t. < ;
Figure 4-85: Phase portrait for constant and constant steering angle . Bottom: Using simple model in
Eq [4.56]. Only black solid trajectories credible, since they are completely within model validity. Upper: Using
a model with larger validity. From Mats Jonasson, VCC.
4.5.6 Long heavy combination vehicles high speed

functions
It is sometimes irrelevant to apply functions/measures from two axle vehicles on combinations of
units. This can be the case for passenger cars with a trailer, but it is even more obvious for long
combinations of heavy vehicles. Some typical measures for multi-unit combination vehicle are given in
this section.
4.5.6.1 Rearward Amplification *

Function definition: Rearward Amplification for long heavy combination vehicles is the ratio of
the maximum value of the motion variable of interest (e.g. yaw rate or lateral acceleration of the centre of gravity) of
the worst excited following vehicle unit to that of the first vehicle unit during a specified manoeuvre at a certain
friction level and constant speed. From Reference [ (Kati, 2013)].
Figure 4-86 illustrates Rearward Amplification, RWA. RWA is defined for a special manoeuvre, e.g. a
certain lane change or step steer. RWA is the ratio of the peak value of yaw rate or lateral acceleration
for the rearmost unit to that of the lead unit. This performance measure indicates the increased risk
for a swing out or rollover of the last unit compared to what the driver is experiencing in the lead unit.
242
LATERAL DYNAMICS
Figure 4-86: Illustration of rearward amplification, P denotes peak value of the motion variable of interest.
From (Kharrazi , 2012).
4.5.6.2 High Speed Transient Off-tracking *

Function definition: High speed transient off-tracking for long heavy combination vehicles is
the overshoot in the lateral distance between the paths of the centre of the front axle and the centre of the most
severely off-tracking axle of any unit in a specified manoeuvre at a certain friction level and a certain constant
longitudinal speed. From Reference [ (Kati, 2013)].
Figure 4-45 illustrates a manoeuvre where this Off-tracking can be defined. Off-tracking can be either
determined in a steady state turn or in a transient manoeuvre such as lane change. The steady state
version is described in Section 4.3.10.2.
4.5.6.3 Yaw Damping *

Function definition: Yaw Damping for long heavy combination vehicles is the damping ratio of the least damped
articulation joint’s angle of the combination vehicle during free oscillations excited by actuating the steering wheel
with a certain pulse or a certain sine-wave steer input at a certain friction level. From Reference [ (Kati, 2013)].
Figure 4-87 illustrates Yaw Damping. It is the damping ratio of the least damped articulation joint of
the combination vehicle during free oscillations. Yaw damping ratio of an articulation joint is
determined from the amplitudes of the articulation angle of subsequent oscillations.
Figure 4-87: Illustration of yaw damping for combination vehicle, DR denotes damping ratio of the
articulation joint. From (Kharrazi , 2012).
4.5.6.4 Zero-damping speed *

Function definition: Zero-damping speed for long heavy combination vehicles is the longitudinal speed at which a
yaw disturbance leads to undamped response.
243
LATERAL DYNAMICS
The measure can be calculated from a linear model as the longitudinal speed when the damping
coefficient, for any mode involving yaw velocity, equals zero. If the measure is applied on a two-axle
vehicle it is same as critical speed.
4.6 Lateral Control Functions

Some control functions involving lateral vehicle dynamics will be presented briefly. There are more,
but the following are among the most well-established ones. But initially, some general aspects of
lateral control is given.
4.6.1 Lateral Control Design

Sensors available in production vehicles and used for lateral control are, generally those mentioned as
available for Longitudinal Control, see Section 3.5 plus some more:
Steering wheel sensors gives at least steering wheel angle, if the vehicle is equipped with ESC (which is
a legal requirement on many markets). Additionally, if the steering assistance is electrical, the steering
wheel torque can be sensed.
High specification modern vehicles also have environment sensors (camera, radar, etc) that can give
laterally interesting information, such as: Subject vehicle lateral position versus lane markers ahead
and other vehicles to the side or rear of subject vehicle.
As general considerations for actuators, one can mention that interventions with friction brake
normally have to have thresholds, because interventions are noticed by driver and also generate
energy loss. Interventions with steering are less sensitive, and can be designed without thresholds.
4.6.2 Lateral Control Functions

4.6.2.1 Lane Keeping Aid, LKA *
Function definition: Lane Keeping Aid steers the vehicle without driver having to steer, when probability for
lane departure is predicted as high. It is normally actuated as an additional steering wheel torque. Conceptually, it
can also be actuated as a steering wheel angle offset.
Lane Keeping Aid (or Lane Departure Prevention) has the purpose to guide the driver to keep in the
lane. Given the lane position from a camera, the function detects whether vehicle tends to leave the
lane. If so, the function requests a mild steering wheel torque (typically 1..2 Nm) in appropriate
direction. Driver can easily overcome the additional torque. Function does not intervene if too low
speed or turning indicator (blinker) is used. There are different concepts whether the function
continuously should aim at keeping the vehicle in centre of lane, or just intervene when close to
leaving the lane, see Figure 4-88.
steering wheel
added torque on
Two concepts:
• Keep in centre of lane
• Intervene when about
to leave lane
lateral position
in lane
Figure 4-88: Two concepts for Lane Keeping Aid.
244
LATERAL DYNAMICS
4.6.2.2 Electronic Stability Control, ESC *

Function definition: Electronic Stability Control directs the vehicle to match a desired yaw behaviour, when
the deviation from desired behaviour becomes above certain thresholds. ESC typically monitors vehicle speed,
steering angle and yaw rate to calculate a yaw rate error and uses friction brakes as actuator to reduce it.
There are 3 conceptual parts of ESC: Over-steer control, Under-steer control, Over-speed control. The
actual control error that the vehicle reacts on is typically the yaw rate error between sensed yaw rate
and desired yaw rate. Desired yaw rate is calculated from a so-called reference model. Some of today’s
advanced ESC also intervenes on sensed/estimated side slip.
Desired yaw rate (and side-slip) is calculated using a reference model. The reference model requires at
least steering angle and longitudinal velocity as input. The reference model can be either of steady
state type (approximately as Equation [4.28]) or transient (approximately as Equation [4.55]). The
vehicle modelled by the reference model should rather be the desired vehicle than the actually
controlled vehicle. But, in order to avoid too much friction brake interventions; the reference model
cannot be too different. Also, in order to avoid that vehicle yaws more than its path curvature; the
reference model has to be almost as understeered as the controlled vehicle, which typically can be
fixed by saturating lateral tyre forces on the front axle of the reference model. This requires some kind
of friction estimation, especially for low-mu driving.
When controlling yaw via wheel torques, one can identify some different concepts such as direct and
in-direct yaw moment, see Section 4.3.5.6. For ESC there is also a “pre-cautious yaw control” which
aims at reducing speed, see Section 4.6.2.2.3.
4.6.2.2.1 Over-steer Control
Over-steer control was the first and most efficient concept in ESC. When a vehicle over-steers, ESC will
brake the outer front wheel. It can brake to deep slip levels (typically -50%) since losing side grip on
front axle is positive in an over-steer situation. More advanced ESC variants also brakes outer rear, but
less and not to same deep slip level.
For combination vehicles with trailers that have controllable brakes, also the trailer is braked to avoid
jack-knife effect, see upper part of Figure 4-89, or swing-out of the towed units.
4.6.2.2.2 Under-steer Control
Under-steer control means that inner rear is braked when vehicle under-steers. This helps the vehicle
turn-in. This intervention is most efficient on low mu, because on high mu the inner rear wheel
normally has very little normal load. Also, the slip levels are not usually as deep as corresponding
over-steer intervention, but rather -10%. This is because there is always a danger in braking too much
on rear axle, since it can cause over-steering. More advanced ESC variants also brakes inner front,
which makes the function very similar to the function in Section 4.6.2.2.3.
245
LATERAL DYNAMICS
Figure 4-89: ESC when over-steer and understeer, on a truck with trailer
4.6.2.2.3 Over-speed Control
Over-speed control is not always recognised as a separate concept, but a part of under-steer control. It
means that more than just inner rear wheels are braked. In this text, we identify this as done to
decrease speed, which has a positive effect later in the curve.
4.6.2.2.4 Wheel-level control
A pre-requisite for all controls mentioned above in Section 4.6.2.2 is that the wheel torque actuator
responds to a torque request, but it is a normal design that it is also responding to a longitudinal slip
request. The slip request is generally used as a sort of “safety net” to not lock-up the wheel more than
10-20%, but at RSC interventions (see Section 4.6.2.3) the lateral grip should be braked away, so a
deeper slip request is then used, typically 50-70%.
TorqueRequest,
𝑇 𝑒𝑞 Torque Fx
𝑒𝑞
controller hi-mu
ESC & RSC Fax Fax
𝑒𝑞
controllers Min Actuator
lo-mu (or large sy)
LongSlipRequest, Closed loop Wheel rotational
slip controller 𝑒𝑞
𝑠 𝑒𝑞 speed sensor sx
𝑠
vehicle state Calculate 𝜔
normal request RSC request
estimator slip sx-max = 10..20% sx-max = 50..60%
Figure 4-90: Individual wheel control by friction brakes for ESC-type functions
4.6.2.2.5 Other intervention than individual wheel brakes
Balancing with Propulsion per axle
For vehicles with controllable distribution of propulsion torque between the axles, ESC can intervene
also with a request for redistribution of the propulsion torque. If over-steering, the propulsion should
be redistributed towards front and opposite for understeering.
Torque Vectoring
246
LATERAL DYNAMICS
For vehicles with controllable distribution of propulsion torque between the left and right, ESC can
intervene also with a request for redistribution of the propulsion torque. If over-steering, the
propulsion should be redistributed towards inner side and opposite for understeering.
Steering guidance
For vehicles with controllable steering wheel torque, ESC can intervene also with a request for
additional steering wheel torque. The most obvious function is to guide driver to open up steering
(counter-steer) when the vehicle over-steers. Such functions are on market in passenger cars today.
Less obvious is how to guide the driver when vehicle is under-steering.
4.6.2.2.6 Environment-information-based-ESC and/or ESC-for-the-virtual-driver
A prognosis of the future development of ESC like functions is that environment sensors can be used to
better predict what driver tries to do; presently ESC can only look at steering wheel angle.
Related to this, but still somewhat different, would be to utilize the automated driving development by
utilizing that a “virtual driver” can be much better predicted than a “manual driver”. So, a predictive
ESC control is more possible.
4.6.2.3 Roll Stability Control, RSC *

Function definition: Roll Stability Control, RSC, prohibits vehicle to roll-over due to lateral wheel forces from
road friction. RSC uses friction brake as actuator.
The purpose of RSC is to avoid un-tripped roll-overs. The actuator used is the friction brake system.
When roll-over risk is detected, via lateral acceleration sensor (or in some advanced RSC
implementations, also roll rate sensor), the outer front wheel is braked. RSC can brake to deep slip
levels (typically -70%..-50%) since losing side grip on front axle is positive in this situation. To lock the
wheels (slip= ∞) is undesired since wheel rotational inertia makes it difficult to quickly regain
lateral grip when needed after the intervention.
On heavy vehicles, RSC intervenes earlier and similar to function described in Section 4.6.2.2.3 Over-
speed Control.
In future, RSC could be developed towards also using steering.
4.6.2.4 Lateral Collision Avoidance, LCA *

Function definition: Lateral Collision Avoidance support the driver when he has to do late lateral obstacle
avoidance, when probability for forward collision is predicted as high.
There are systems on the market for Automatic Emergency Brake, see Section 3.5. If these are seen as
Longitudinal Collision Avoidance, one could also think of Lateral Collision Avoidance functions, which
would automatically steer away laterally from an obstacle ahead of subject vehicle. There are not yet
any such functions on market. One possible first market introduction could be that triggering requires
driver to initiate steering. Another would be to trigger on a first collision impact, see Reference (Yang,
2013).
4.6.2.5 Automated Driving (AD)

Combining longitudinal control (such as ACC, in Section 3.5.2.2) with a lateral control (such as LKA, see
Section 4.6.2.1) results in functionality which clearly approaches automated driving (AD). AD is a very
general expression but are sometimes interpreted as more specific, but specific in different ways
depending on context. In a way, AD is already reality since there are vehicles on the road which can be
driven (or drive themselves) with ACC and LKA active at the same time. On the other hand, AD can be
seen as very futuristic, since completely driverless vehicle which can operate in all situations is far
from produced and sold.
247
LATERAL DYNAMICS
248
VERTICAL DYNAMICS
5 VERTICAL DYNAMICS
5.1 Introduction
The vertical dynamics are needed since vehicles are operated on real roads, and real roads are not
perfectly smooth. Also, vehicle can be operated off-road, where the ground unevenness is even larger.
The irregularities of the road can be categorized. A transient disturbance, such as a pothole or bump,
can be represented as a step input or ramp. Undulating surfaces like grooves across the road may be a
type of sinusoidal or other stationary oscillating (or periodic) input. More natural input like the
random surface texture of the road may be a random noise distribution. In all cases, the same
mechanical system must react when the vehicle travels over the road at varying speeds including
doing manoeuvres in longitudinal and lateral directions.
The chapter is organised around the 3 complete vehicle functions: 5.6.1 Ride comfort *, 5.6.2 Fatigue
life *, and 5.6.3 Road grip *. It is, to a larger extent than Chapters 3 and 4, organised with mathematical
theory first followed by the vehicle functions. In Figure 5-1 shows the 3 main functions. It explains the
importance of the vehicle’s dynamic structure. The vehicle’s dynamic structure calls for a pretty
extensive theory base, described mainly in Section 5.3.
Models in this chapter focus the disturbance from vertical irregularities from the road, i.e. only the
vertical forces on the tyre from the road and not the forces in road plane. This enables the use of
simple models which are independent of exact wheel and axle suspension, such as pivot axes and roll
centres. Only the wheel stiffness rate (effective stiffness) and wheel damping rate (effective damping),
see Figure 3-43, influence. This has the benefit that the chapter becomes relatively independent of
previous chapters, but it has the drawback that the presented models are not really suitable for
studies of steep road irregularities (which have longitudinal components) and sudden changes in
wheel torque or tyre side forces. Also, noise is not covered in this compendium.

• “Chapter 21 Suspension Systems” in Reference (Ploechl, 2013).
• “Chapter 29 Ride Comfort and Road Holding” in Reference (Ploechl, 2013).
𝒛̇ 𝒔 ⋯
𝒛𝒔
Fatigue Life
Stresses in
𝒛
Material
𝒛̇
vehicle
fatigue
Road surface structure
irregularities Ride Comfort
𝒛𝒔 Vehicle’s
speed Vehicle’s 𝒛̈ 𝒔 Human
Vibrations of
𝒛 dynamic
human
perception of
structure vibrations
𝑭 𝒛 Vehicle
created
disturbances Contact
Compression 𝑭 𝒛
between tyres
of tyre
and road
Road Grip
Figure 5-1: Different types of knowledge and functions in the area of vertical vehicle dynamics, organised
around the vehicle’s dynamic structure.
249
VERTICAL DYNAMICS
5.2 Suspension System

Suspension in a vehicle may refer to suspension of wheels (or axles), suspension of sub-frame and
drivetrain and suspension of cabin (for heavy trucks). In this compendium, only wheel (or axle)
suspension is considered.
Suspension design is briefly discussed at the following section: 3.4.63.4.7.4, 4.3.9.4 and 5.2. It
influences ride comfort, load on suspension and road grip through how vertical forces and camber
and steering angles on each wheel changes with body motion (heave, roll, pitch), road unevenness
(bumps, potholes, waviness) and wheel forces in ground plane (from Propulsion, Braking and
Steering subsystems).
A wheel suspension has the purpose to constrain the wheel from 6 degrees of freedom, dofs, relative
the body, to 2 or 3 dofs. A steered wheel needs 3 dofs (pitch rotation, heave and yaw rotation). For an
un-steered wheel, also the yaw rotation is constrained. For most steering, left and right wheels on the
same axle have dependent steering angles, which could be seen as 2.5 dofs/wheel or 5 dofs/axle.
driver,
environment (sprung) body
prop, brk, ste

systems (HW&SW)
wheel suspen-
& tyre sion
multiple
Fy
Fz Fx
road
Figure 5-2: Individual wheel suspension described as one modular sub-model per wheel. It may be noted that
a both wheel model (main geometry such as wheel radius) and tyre model (how and vary with tyre slip
and ) is a part of each such sub-model.
An (axle) suspension system mainly consists of:
• Linkage, which has the purpose to constrain the relative motion between wheel and body via
kinematics. An alternative way to express this is that the linkage defines how longitudinal and
lateral wheel forces are brought into the body (body = sprung mass). Effective pivot points and
roll/pitch centres, mentioned earlier in this compendium, are defined through the linkage. In the
real pivot points in the linkage, there are bushings with stiffness and damping. The bushings
stiffness is much larger than the stiffness of the compliances mentioned below. For steered
wheels, the linkage also has the purpose to allow steering.
There are two main concepts: Individual wheel suspension and rigid axle suspension.
• Compliances (or springs), which has the purpose to allow temporary vertical displacement of
the wheels relative to the body. There are often one spring per wheel but also a spring per axle.
The second is called anti-roll bar and connects left and right wheel to each other to reduce body
roll. Compliances often have a rather linear relation between the vertical displacement and force
of each wheel, but there are exceptions:
o Anti-roll bars make two wheels dependent of each other (still linear). Anti-roll bars can be
used on both individual wheel suspensions and rigid axle suspensions.
250
VERTICAL DYNAMICS
o The compliances can be intentionally designed to be non-linear in the outer end of the
stroke, e.g. bump stops.
o The compliances can be non-linear during the whole stroke, e.g. air-springs and leaf-
springs.
o The compliances can be intentionally designed to be controllable during operation of the
vehicle. This can be to change the pre-load level to adjust for varying roads or varying
weight of vehicle cargo or to be controllable in a shorter time scale for compensating in each
oscillation cycle. The latter is very energy consuming and no such “active suspension” is
available on market.
• Dampers, which has the purpose to dissipate energy from any oscillations of the vertical
displacement of the wheel relative to the body. Dampers often has a rather linear relation
between the vertical deformation speed and force of each wheel, but there are exceptions:
o The dampers can be intentionally designed to be different in different deformation
direction. This is actually the normal design for dampers of hydraulic piston type, and it
means that damping coefficient is different in compression and rebound. Typical values are
about 3 times more damping in rebound (extension) than compression (bump). This can be
motivated from that driving over a steep bump requires low damping to reduce upward
jerk in vehicle, especially since there is a hard bump stop in the end of the spring stroke. In
the other direction, driving over equally steep hole the downward jerk is limited by that the
wheel cannot develop pulling forces on the ground; instead it lifts from ground if hole is too
steep. So larger damping can be allowed in rebound. A reflection here is that high damping
damps oscillations, but high damping also increases the shock transmittance (with this
reasoning, the name “shock-absorber” is misleading).
o Damping in leaf springs is non-linear since they work with dry friction.
o Damping in air-springs is non-linear due to the nature of compressing gas.
o The dampers can be designed to be controllable during operation of the vehicle. This can
be used to change the damping characteristics to adjust for varying roads or varying weight
of vehicle cargo or to be controllable in a shorter time scale for compensating in each
oscillation cycle. The latter is called “semi-active suspension” and is available on some high-
end vehicles on market.
The simplest view we can have of a suspension system is that it is an individual suspension between
the vehicle body and each wheel, consisting of one linear spring and one linear damper in parallel.
Chapter 5 uses this simple view for analysis models, because it facilitates understanding and it is
enough for a first order evaluation of the functions studied (comfort, road grip and fatigue load)
during normal driving on normal roads.
5.3 Stationary oscillations theory

Many vehicle functions in this chapter will be studied using stationary oscillations (cyclic repeating),
as opposed to transiently varying. An example of transiently varying quantity is a single step function
or single square pulse. A stationary oscillation can be as a sum of several harmonic terms, a multiple
frequency harmonic stationary oscillation. The special case with only one frequency is called a single
frequency harmonic stationary oscillation. See Figure 5-3 and Equation [5.1].
𝑯𝒂 𝒎𝒐𝒏𝒊𝒄 𝒔𝒕𝒂𝒕𝒊𝒐𝒏𝒂 𝒚 𝒐𝒔𝒄𝒊𝒍𝒍𝒂𝒕𝒊𝒐𝒏𝒔:

𝑆 𝑛 𝑙 𝑟 𝑞𝑢 𝑛 𝑦 ∶ (𝜉) = ̂ ∙ cos(𝜔 ∙ 𝜉 + ) ;
[5.1]
𝑀𝑢𝑙𝑡 𝑝𝑙 𝑟 𝑞𝑢 𝑛 𝑠 ∶ (𝜉) = ∑ ̂ ∙ cos(𝜔 ∙ 𝜉 + );
=
𝑤ℎ 𝑟 𝜉 𝑠 𝑡ℎ 𝑛 𝑝 𝑛 𝑛𝑡 𝑟 𝑏𝑙 .
251
VERTICAL DYNAMICS
m
zz1 zz1 zz1 zz1
k c
t t t t
F1
Step, as one example of Harmonic None-harmonic Random noise
transient (none- stationary stationary oscillation,
stationary oscillating) oscillation, single single frequency (or multiple
frequency random-frequency
other examples can be (or multiple-frequency harmonic
step down, square pulse, harmonic stationary stationary
ramp etc. oscillation) oscillation)
Figure 5-3: Different types of variables, both transient and stationary oscillating. The independent variable 𝜉
can, typically, be either time or distance.
The most intuitive is probably to think of time as the independent variable, i.e. that the variation takes
place as function of time and that 𝜉 = 𝑡 in Equation [5.1]. However, for one specific road, the vertical
displacement varies with longitudinal position, rather than with time. This is why we can either do
analysis in time domain (𝜉 = 𝑡) and space domain (𝜉 = 𝑥).
Since the same oscillation can be described either as a function of 𝜉 ( = (𝜉)) or as a function of
frequency 𝜔 ( ̂ = ̂ (𝜔)), we can do analysis either in the independent variable domain (𝜉) or in
frequency domain (𝜔).
The four combinations of domains are shown in Figure 5-4.
time Fourier Transform (time) frequency

domain, t domain, 𝒐 𝒇
𝒙 = 𝒗𝒙 𝒕; 𝒇𝒔 = 𝒇⁄𝒗𝒙 ;
𝝎𝒔 = 𝝎⁄𝒗𝒙 ;
space Fourier Transform spatial frequency

domain, x domain, 𝝎𝒔 𝒐 𝒇𝒔
Figure 5-4: Four combinations of domains

Time and space domains are treated in section 5.3.1 and 0. In addition to the domains, we also need to
differ between discrete and continuous representations in the frequency domains, see 5.3.1 and 5.3.2.
5.3.1 Time as independent variable

With time as independent variable, the frequency has the meaning of “how often per time”. Even so,
there are two relevant ways to measure frequency: angular (time) frequency, and (time) frequency.
𝜔 = ∙𝜋∙ ;
𝑤ℎ 𝑟 𝜔 [𝑟 ⁄𝑠] = 𝑛 𝑢𝑙 𝑟 (𝑡 ) 𝑟 𝑞𝑢 𝑛 𝑦; [5.2]
𝑛 [1⁄𝑠 = 𝑜𝑠 𝑙𝑙 𝑡 𝑜𝑛𝑠⁄𝑠] = (𝑡 ) 𝑟 𝑞𝑢 𝑛 𝑦;
The time for one oscillation is called the period time. It is denoted 𝑇:
𝑇 = 1⁄ = ∙ 𝜋 ⁄𝜔 ; [5.3]
252
VERTICAL DYNAMICS
5.3.1.1 Mean Square (MS) and Root Mean Square (RMS) of

variable
For a variable, z, we can define MS and RMS values as follows:
𝑉 𝑟 𝑏𝑙 : = (𝑡);
𝑡 𝑛𝑑
0
∙ 𝑡
𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑀𝑆( ) = ;
𝑡𝑒𝑛𝑑 [5.4]
𝑡 𝑛𝑑
∙ 𝑡
𝑅𝑜𝑜𝑡𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑅𝑀𝑆( ) = √ 0
;
𝑡𝑒𝑛𝑑
If the variable is written as a single frequency harmonic stationary oscillation:

𝑉 𝑟 𝑏𝑙 : = ̂ ∙ cos(𝜔 ∙ 𝑡 + ) ;
𝑡 𝑛𝑑 𝑡 𝑛𝑑
0
∙ 𝑡 0
( ̂ ∙ cos(𝜔 ∙ 𝑡 + )) ∙ 𝑡
𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑀𝑆( ) = = =
𝑡𝑒𝑛𝑑 𝑡𝑒𝑛𝑑
𝑡 sin( ∙ 𝜔 ∙ 𝑡) 𝑡=𝑡 𝑛𝑑 𝑡 sin( ∙ 𝜔 ∙ 𝑡𝑒𝑛𝑑 )
̂ ∙[ + 4∙𝜔 ] 𝑡=0 ̂ ∙ ( 𝑒𝑛𝑑 + 4∙𝜔 ) ̂ [5.5]
= = → ;
𝑡𝑒𝑛𝑑 𝑡𝑒𝑛𝑑 𝑡 𝑛𝑑 ∞
𝑡 𝑛𝑑
∙ 𝑡 | ̂|
𝑅𝑜𝑜𝑡𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑅𝑀𝑆( ) = √ 0
= √𝑀𝑆( ) = ;
𝑡𝑒𝑛𝑑 √
If the variable is written as a multiple frequency harmonic stationary oscillation:
𝑉 𝑟 𝑏𝑙 : =∑ = ∑ ̂ ∙ cos(𝜔 ∙ 𝑡 + );
= =
𝑡 𝑛𝑑 𝑡 𝑛𝑑
0
∙ 𝑡 0
(∑ = ) ∙ 𝑡
𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑀𝑆( ) = = =
𝑡𝑒𝑛𝑑 𝑡𝑒𝑛𝑑
𝑡 𝑛𝑑
0
(∑ = ̂ ∙ cos(𝜔 ∙ 𝑡 + )) ∙ 𝑡
= ⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝑡𝑒𝑛𝑑 ∞
𝑡𝑒𝑛𝑑 [5.6]
𝑡 𝑛𝑑
0
∑= ̂ ∙ (cos(𝜔 ∙ 𝑡 + )) ∙ 𝑡 ̂
⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗⃗
𝑡𝑒𝑛𝑑 ∞ = ∑ 𝑀𝑆( ) = ∑ ;
𝑡𝑒𝑛𝑑
= =
̂
𝑅𝑜𝑜𝑡𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑅𝑀𝑆( ) = √𝑀𝑆( ) = √∑ = √∑ 𝑀𝑆( ) = √∑(𝑅𝑀𝑆( )) ;
= = =
5.3.1.2 Power Spectral Density and Frequency bands

So far, the frequency has been a discrete number of frequencies, 𝜔 𝜔 ⋯ 𝜔 . There are reasons to
treat the frequency as a continuous variable instead. The discrete amplitudes, ̂ ̂ ⋯ ̂ , should then
be thought of as integrals of a “continuous amplitude curve”, 𝑐̂ , where the integration is done over a
small frequency interval, centred around a each mid frequency, 𝜔 :
𝜔𝑖 +𝜔𝑖+1
𝜔+ 𝜔− ̂
̂ = ∫ ̂𝑐 ∙ 𝜔 = ̂𝑐 (𝜔 ) ∙ = ̂𝑐 (𝜔 ) ∙ ∆𝜔 ; ⇒ ̂𝑐 (𝜔 ) = ; [5.7]
∆𝜔
𝜔 +𝜔𝑖
𝜔= 𝑖−1
253
VERTICAL DYNAMICS
We realize that the unit of ̂𝑐 has to be same as for , but per [rad/s]. So, if z is a displacement in [m], ̂𝑐
has the unit [m/(rad/s)]. Now, ̂𝑐 is a way to understand the concept of a spectral density. A similar
value, but more used, is the Power Spectral Density, PSD (also called Mean Square Spectral Density).
𝑆𝐷(𝜔) is a continuous function, while ̂ is a discrete function. That means that 𝑆𝐷(𝜔) is fully
determined by a certain measured or calculated variable (𝑡), while ̂ depends on which
discretization (which 𝜔 or which ∆𝜔) that is chosen.
𝑀𝑆( 𝑙𝑡 𝑟( (𝑡) 𝜔 ∆𝜔))

𝑆𝐷( (𝑡) 𝜔 ∆𝜔) = = 𝐺(𝜔); [5.8]
∆𝜔
𝑤ℎ 𝑟 filter 𝑠 𝑏 𝑛 𝑝 𝑠𝑠 𝑙𝑡 𝑟 𝑛𝑡 𝑟 𝑟𝑜𝑢𝑛 𝜔 𝑛 𝑤 𝑡ℎ 𝑏 𝑛 𝑤 𝑡ℎ ∆𝜔;
PSD can also be defined with band width in time frequency instead of angular frequency. Eq [5.8] is the
same but replacing ∆𝜔 with ∆ .
When the variable to study (z) is known and the band width is known, one often write simply 𝑆𝐷(𝜔)
or 𝐺(𝜔). G has the same unit as , but per [rad/s] or per [oscillations/s]. So, if z is a displacement in
[m], G has the unit [ ⁄(𝑟 ⁄𝑠)] or [ ⁄(1⁄𝑠) = ∙ 𝑠].
RMS is square root of the area under the PSD curve:
𝑅𝑀𝑆( ) = √∑ 𝑀𝑆( ) = √∑ 𝐺(𝜔 ) ∙ ∆𝜔 = √ ∫ 𝐺(𝜔) ∙ 𝜔 ; [5.9]

= = 𝜔=0
5.3.1.2.1 Differentiation of PSD

Knowing the PSD of a variable, we can easily obtain the PSD for the derivative of the same variable:
𝐺 ̇ (𝜔) = 𝜔 ∙ 𝐺 (𝜔); [5.10]
5.3.1.3 Transfer function

In a minimum model for vertical dynamics there is at least one excitation, often road vertical
displacement, , and one response, e.g. vertical displacement of sprung mass (=vehicle body), 𝑠 . A
Transfer function, 𝐻 = 𝐻(𝑗 ∙ 𝜔), is the function which we can use to find the response, given the
excitation:
𝑍𝑠 (𝜔) = 𝐻(𝜔) ∙ 𝑍 (𝜔); ⟺ ℱ( 𝑠 (𝑡)) = 𝐻(𝜔) ∙ ℱ( (𝑡));
∞
−𝑗∙𝜔∙𝑡 [5.11]
𝑤ℎ 𝑟 ℱ 𝑠 𝑡ℎ 𝑜𝑢𝑟 𝑟 𝑜𝑝 𝑟 𝑡𝑜𝑟: 𝑍(𝜔) = ℱ( (𝑡)) = ∫ ∙ (𝑡) ∙ 𝑡 ;
0
𝐻 is complex, with magnitude, |𝐻| = √(Re(𝐻)) + (Im(𝐻)) , and phase, ar (𝐻(𝜔)) =

arctan(Im(𝐻)⁄Re(𝐻)).
𝐴 𝑝𝑙 𝑡𝑢 : ̂𝑠 (𝜔) = |𝐻(𝜔)| ∙ ̂ (𝜔);

ℎ 𝑠 : 𝑠 (𝜔) (𝜔) = ar (𝐻(𝜔)) ;
[5.12]
𝑤ℎ 𝑟 = ∑ ̂ (𝜔 ) ∙ cos(𝜔 ∙ 𝑡 + );
=
Since there can be different excitations and responses in a system, there are several transfer functions.
To distinguish between those, a subscripting of 𝐻 is often used: 𝐻𝑒 𝑐 𝑡 𝑡 𝑜𝑛 𝑒𝑠 𝑜𝑛𝑠𝑒 , which would be
254
VERTICAL DYNAMICS
𝐻 𝑟 𝑠 = 𝐻 𝑜 𝑑 𝑑 𝑠 𝑐𝑒𝑚𝑒𝑛𝑡 𝑠 𝑢𝑛 𝑚 𝑠𝑠 𝑑 𝑠 𝑐𝑒𝑚𝑒𝑛𝑡 in the example above. Other examples of relevant

transfer functions in vertical vehicle dynamics are:
• 𝐻 𝑜 𝑑𝑑 𝑠 𝑐𝑒𝑚𝑒𝑛𝑡 𝑠 𝑢𝑛 𝑚 𝑠𝑠 𝑐𝑐𝑒 𝑒 [( ⁄𝑠 )⁄ ], see Section 5.6
𝑡 𝑜𝑛
• 𝐻 𝑜 𝑑𝑑 𝑠 𝑐𝑒𝑚𝑒𝑛𝑡 𝑠𝑢𝑠 𝑒𝑛𝑠 𝑜𝑛 𝑑𝑒 𝑜 𝑚 𝑡 𝑜𝑛 [ ⁄ ] , see Section 5.6.2
• 𝐻 𝑜 𝑑𝑑 𝑠 𝑐𝑒𝑚𝑒𝑛𝑡 𝑡 𝑒 𝑜 𝑐𝑒 [𝑁⁄ ] , see Section 5.6.3
When transfer function for one derivative is found, it is often easy to convert it to another:
𝐻 1 ̇2 = 𝑗 ∙ 𝜔 ∙ 𝐻 1 2;
𝐻 1 ̈2 = 𝑗 ∙ 𝜔 ∙ 𝑗 ∙ 𝜔 ∙ 𝐻 1 2
= 𝜔 ∙𝐻 1 2
; [5.13]
𝐻 1 2− 3
=𝐻1 2 𝐻1 3
;
Note that these relations are valid for the complex transfer function. They corresponding relations can
be used with |𝐻| replacing 𝐻, in cases where the phase is same for all involve variables, which often
can be the case when damping is neglected.
The usage of the transfer function is, primarily, to easily obtain the response from the excitation, as
shown in Equation [5.12]. Also, the transfer function can operate on the Power Spectral Density,
PSD=G, as shown in the following:
𝑀𝑆( 𝑠 (𝑡) 𝜔) ( ̂𝑠 (𝜔)) ⁄ (|𝐻(𝜔)| ∙ ̂ (𝜔)) ⁄

𝐺 𝑠 (𝜔) = = = =
∆𝜔 ∆𝜔 ∆𝜔 [5.14]
( ̂ (𝜔)) ⁄
= |𝐻 (𝜔)| ∙ = |𝐻 𝑟 𝑠 (𝜔)| ∙ 𝐺 𝑟 (𝜔);
𝑟 𝑠
∆𝜔
Using Equation [5.9], we can then express 𝑅𝑀𝑆( 𝑠 ) (sprung mass), from knowing 𝐺 𝑟 (𝜔) (road):
∞
𝑅𝑀𝑆( 𝑠 ) = √ |𝐻 𝑟 (𝜔)| ∙ 𝐺 𝑟 (𝜔) ∙ 𝜔 ; [5.15]
𝜔=0 𝑠
5.3.2 Space as independent variable

All transformations, in this compendium, between time domain and space domain requires a constant
longitudinal speed, , so that:
𝑥= ∙ 𝑡 + 𝑥0 ; [5.16]
The offset (𝑥0 ) is the phase (spatial) offset (𝑥0 ) is the correspondence to the phase angle ( ).
The corresponding formulas as given in Equations [5.2]..[5.13] can be formulated when changing to
space domain, or spatial domain. It is generally a good idea to use a separate set of notations for the
spatial domain. Hence the formulas are repeated with new notations, which is basically what will be
done in present section.
In space domain, the frequency has the common understanding of “how often per distance”. Even so,
there are two relevant ways to measure frequency: spatial angular frequency and spatial frequency.
= ∙ 𝜋 ∙ 𝑠;
𝑤ℎ 𝑟 [𝑟 ⁄ ] = 𝑛 𝑢𝑙 𝑟 𝑠𝑝 𝑡 𝑙 𝑟 𝑞𝑢 𝑛 𝑦; [5.17]
𝑛 𝑠 [1 ⁄ = 𝑜𝑠 𝑙𝑙 𝑡 𝑜𝑛𝑠⁄ ] = 𝑠𝑝 𝑡 𝑙 𝑟 𝑞𝑢 𝑛 𝑦;
The correspondence to period time is wave length, denoted 𝜆:
𝜆[ ] = 1⁄ 𝑠 = ∙ 𝜋⁄ ; [5.18]
255
VERTICAL DYNAMICS
Now, the basic assumption in Equation [5.16] and definitions of frequencies gives:
𝜔= ∙ ; 𝑛 = ∙ 𝑠; [5.19]
The relation between the phase (spatial) offset (𝑥0 ) and the phase angle ( ) is:
𝜆∙
𝑥0 = ; [5.20]
∙𝜋
5.3.2.1 Spatial Mean Square (MS) and spatial Root Mean Square
(RMS) of variable
In space domain, a variable, z, varies 𝑉 𝑟 𝑏𝑙 : = (𝑥);
with distance, x. We can define Mean 𝑛𝑑
∙ 𝑥
0
Square and Root Mean Square values 𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑀𝑆𝑠 ( ) = ;
𝑥𝑒𝑛𝑑 [5.21]
also in space domain. We subscript
𝑛𝑑
these with s for space. 0
∙ 𝑥
𝑅𝑜𝑜𝑡𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑅𝑀𝑆𝑠 ( ) = √ ;
𝑥𝑒𝑛𝑑
Because 𝑥 is constant, the Mean Square and Root Mean Square will be the same in time and space
domain. If the variable is written as a single frequency harmonic stationary oscillation, these values
becomes as follows:
𝑉 𝑟 𝑏𝑙 : = ̂ ∙ cos( ∙ 𝑥 + 𝑥0 ) ;
̂
𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑀𝑆𝑠 ( ) = ⋯ = = 𝑀𝑆( );
[5.22]
| ̂|
𝑅𝑜𝑜𝑡𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑅𝑀𝑆𝑠 ( ) = ⋯ = = 𝑅𝑀𝑆( );
√
If the variable is written as a multiple frequency harmonic stationary oscillation:
𝑉 𝑟 𝑏𝑙 : =∑ = ∑ ̂ ∙ cos( ∙ 𝑥 + 𝑥0 ) ;
= =
̂
𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑀𝑆𝑠 ( ) = ∑ = 𝑀𝑆( );
[5.23]
=
𝑅𝑜𝑜𝑡𝑀 𝑛𝑆𝑞𝑢 𝑟 : 𝑅𝑀𝑆𝑠 ( ) = √𝑀𝑆𝑠 ( ) = √∑(𝑅𝑀𝑆( )) = 𝑅𝑀𝑆( );

=
5.3.2.2 Spatial Power Spectral Density and Frequency bands

A correspondence to Power Spectral Density in space domain is denoted 𝑆𝐷𝑠 in the following:
𝑀𝑆( 𝑡𝑒 ( ( ) Ω ∆𝜆))
𝑆𝐷𝑠 ( (𝑥) ∆𝜆) = = Φ( );
∆𝜆 [5.24]
where "filter" is a band pass filter centred around ω and with band width ∆ ;
When the variable to study (z) is known and the band width is known, one often write simply 𝑆𝐷𝑠 ( )
or Φ( ). Φ has the same unit as , but per [rad/m] or per [oscillations/m]. So, if z is a displacement in
𝑚2 𝑚3 𝑚2
[m], Φ has the unit [ 𝑑 ⁄𝑚
= 𝑑
] or [ ⁄𝑚
= ].
256
VERTICAL DYNAMICS
5.4 Road models

In general, a road model can include ground properties such as coefficient of friction,
damping/elasticity of ground and vertical position. The independent variable is either one, along an
assumed path, or generally two, x and y in ground plane. In vertical dynamics in this compendium, we
only assume vertical displacement as function of a path. We use x as independent variable along the
path, meaning that the road model is: = (𝑥). The function (𝑥) can be either of the types in
Figure 5-3. We will concentrate on stationary oscillations, which by Fourier series, always can be
expressed as multiple (spatial) frequency harmonic stationary oscillation. This can be specialized to
either single (spatial) frequency or random (spatial) frequency. Hence, the general form of the road
model is multiple (spatial) frequencies:
= (𝑥) = ∑ ̂ ∙ cos( ∙ 𝑥 + 𝑥0 ) ; [5.25]

=
5.4.1 One frequency road model

For certain roads, such as roads built with concrete blocks, a single (spatial) frequency can be a
relevant approximation to study a certain single wave length. Also, the single (spatial) frequency road
model is good for learning the different concepts. A single (spatial) frequency model is the same as a
single wave length model (𝜆 = ∙ 𝜋⁄ , from Equation [5.17]) and it can be described as:
= (𝑥) = ̂ ∙ cos( ∙ 𝑥 + 𝑥0 ) ; [5.26]
5.4.2 Multiple frequency road models

Based on the general format in Equation [5.25], we will now specialise to models for different road
qualities. In Figure 5-5, there are 4 types of road types defined. Approximately, the 3 upper of those
are roads and they are also defined as PSD-plots in Figure 5-6. The mathematical formula is given in
Equation [5.27] and numerical parameter values are given in Equation [5.28].
𝛺 −𝑤 𝑀𝑆𝑠 ( 𝛺)
𝛷 = 𝛷(𝛺) = 𝛷0 ∙ ( ) = ;
𝛺0 ∆𝛺
m
𝑤ℎ 𝑟 𝛷0 = road se erity [ ]; [5.27]
rad⁄m
w = road wa iness [1];
𝛺 = spatial an ular frequency [rad⁄m];
𝛺0 = 1 [rad m];
Typical values are:
−6
m
Very ood road: 𝛷0 = 1 ∙ 1 [ ];
rad⁄m
m
Bad road ∶ 𝛷0 = 1 ∙ 1 −6 [ ];
rad⁄m
[5.28]
m
Very bad road ∶ 𝛷0 = 1 ∙ 1 −6 [ ];
rad⁄m
The waviness is normally in the range of 𝑤 = . .3 [1]
where smooth roads have larger waviness than bad roads.
257
VERTICAL DYNAMICS
The decreasing amplitude for higher (spatial) frequencies (i.e. for smaller wave length) can be
explained by that height variation over a short distance requires large gradients. On micro-level, in the
granular level in the asphalt, there can of course be steep slopes on the each small stone in the asphalt.
These are of less interest for vehicle vertical dynamics, since the wheel dimensions filter out wave
length << tyre contact length, see Figure 2-43. A certain road can be described with:
• 𝛺 ⋯ 𝛺
• ̂ ⋯ ̂
• 𝑥0 ⋯ 𝑥0
Figure 5-5: Four typical road types. From (AB Volvo, 2011).
-2
Φ = Φ0 ⁄ 0 −
PHI=PHI0.*((OMEGA/OMEGA0).^(-waviness))
10
very rough very bad road
-3
rough bad road
10 smooth very good road
]
𝑑 ⁄𝑚
-4
𝑚2
10
𝑆𝐷𝑠PHI /[m*m/(rad/m)]
in [
-5
10
=Φ
-6
10
-7
10 2
very rough 𝟎 = 𝟏𝟎𝟎 𝟏𝟎−𝟔 ⁄m
; 𝒘=𝟐 𝟏;
2
-8 rough 𝟎 = 𝟏𝟎 𝟏𝟎−𝟔 ⁄m
; 𝒘 = 𝟐. 𝟏;
10 2
smooth 𝟎 = 𝟏 𝟏𝟎−𝟔 ⁄m
; 𝒘= 𝟏;
𝟎 = 𝟏 rad⁄m for all;
-9
10
-1 0 1
10 10 in [𝑟 ⁄ ] 10
OMEGA /[rad/m]
𝜆 5 1 5 1
Figure 5-6: PSD spectra for the three typical roads in Figure 5-5.
258
VERTICAL DYNAMICS
Number of frequency components, N, to select is a matter of accuracy or experience. The offsets,

𝑥0 ⋯ 𝑥0 , can often be assumed to be zero. If phase is to be studied, as in Figure 5-6, a random
generation of offsets is suitable. See also Reference (ISO 8608).
We can generate (𝑥) curves for the 3 road types in Figure 5-6 as shown in Figure 5-7. To generate
those plots, we have assumed different number of harmonic components (N in Equation [5.25]) and
also randomly generate the phase for each component (each 𝑥0 ).
very rough
RoadQuality=1
0.05
N=10
zr /[m]
N=20
0 N=100
-0.05
0 1 2 3 4 5 6 7 8 9 10
x/[m]
rough
RoadQuality=2
0.05
N=10
zr /[m]
N=20
0 N=100
-0.05
0 1 2 3 4 5 6 7 8 9 10
x/[m]
smooth
RoadQuality=3
0.05
N=10
zr /[m]
N=20
0 N=100
-0.05
0 1 2 3 4 5 6 7 8 9 10
x/[m]
Figure 5-7: Road profiles, (𝑥), for the three typical roads in Figure 5-5.
5.4.2.1 Transfer function from road spectrum in spatial

domain to system response in time domain
Since we assume constant longitudinal velocity, , the road spectrum can be transformed to the time-
frequency domain:
𝑀𝑆( 𝜔) 𝑀𝑆𝑠 ( 𝛺)
𝐺 𝑟 (𝜔) = = {𝑢𝑠 : 𝜔 = ∙ 𝛺} = =
∆𝜔 ∙ ∆𝛺
−𝑤
𝛺
𝛺 −𝑤 𝑀𝑆𝑠 ( 𝛺) 𝛷0 ∙ (𝛺 )
0
= {𝑢𝑠 : 𝛷0 ∙ ( ) = }= = [5.29]
𝛺0 ∆𝛺
𝜔 −𝑤
𝛷0 ∙ 𝛺−𝑤 𝛷0 ( ) 𝛷0 𝑤−
= −𝑤 = −𝑤 ∙ = −𝑤 ∙ ∙ 𝜔−𝑤 ;
𝛺0 ∙ 𝛺0 𝛺0
Then, we can use Equation [5.14] to obtain the response 𝑠 :

𝛷
𝐺 𝑠 (𝜔) = |𝐻 𝑟 𝑠
(𝜔)| ∙ 𝐺 𝑟 (𝜔) = |𝐻 𝑟 𝑠
(𝜔)| ∙ −𝑤 ∙ 𝑤−
∙ 𝜔−𝑤 ; [5.30]
𝛺0
Then we can use Equation [5.9] to obtain the RMS of the response 𝑠:
259
VERTICAL DYNAMICS
𝛷0 𝑤− −𝑤
𝑅𝑀𝑆( 𝑠 ) = √∑ 𝐺 𝑠 (𝜔 ) ∙ ∆𝜔 = √ ∙ ∙ ∑|𝐻 (𝜔 )| ∙ 𝜔 ∙ ∆𝜔 ;
𝛺0 −𝑤 𝑟 𝑠
= =
or [5.31]
∞ ∞
𝛷0 𝑤−
𝑅𝑀𝑆( 𝑠 ) = √ ∫ 𝐺 𝑠 (𝜔) ∙ 𝜔 = √ −𝑤 ∙ ∙ ∫ |𝐻 𝑟 𝑠
(𝜔)| ∙ 𝜔 −𝑤 ∙ 𝜔 ;
𝛺0
𝜔=0 𝜔=0
5.5 One-dimensional vehicle models

“One-dimensional” refers to pure vertical motion, i.e. that the vehicle heaves without pitch and without
roll. The tyre is stiff and massless.
This can be seen as that the whole vehicle mass, m, is modelled as suspended by the sum of all wheels’
vertical forces, = + + + . However, the model can sometimes be referred to as a
“quarter-car-model”. That is because one can see the model as a quarter of the vehicle mass, 4,
which is suspended by one of the wheel’s vertical force, 𝑗 . The exact physical interpolation of a
quarter car is less obvious, since one can argue whether the fraction ¼ of the vehicle mass is the
proper fraction or from which point of view it is proper. Using the fraction ¼ is as least debatable if the
vehicle is completely symmetrical, both left/right and front/rear.
5.5.1 One-dimensional model without dynamic

degree of freedom
“Without dynamic degree of freedom” refers to that the (axle) suspension is modelled as ideally stiff.
The model can be visualised as in Figure 5-8.
real vehicle model

zs
z m
m x ̈
py
zr
Frz
Figure 5-8: One-dimensional model without dynamic degree of freedom
We could find the equations without very much formalism ( ∙ ̈𝑠 = ; 𝑛 𝑠 = (𝑡);), but the
following equations exemplifies a formalism which will be useful when we expand the model later.
𝑬𝒒𝒖𝒊𝒍𝒊𝒃 𝒊𝒖𝒎: ∙ ̈+ ∙ = ;
𝑪𝒐𝒎 𝒂𝒕𝒊𝒃𝒊𝒍𝒊𝒕𝒚: = ; [5.32]
𝑬𝒙𝒄𝒊𝒕𝒂𝒕𝒊𝒐𝒏: = (𝑡);
5.5.1.1 Response to a single frequency excitation

Assume that the road has only one (spatial) frequency, i.e. one wave length. Then the excitation is as
follows:
260
VERTICAL DYNAMICS
∙π
= (𝑥) = ̂ ∙ cos( ∙ 𝑥 + 𝑥0 ) = ̂ ∙ cos ( ∙ 𝑥 + 𝑥0 ) ;
{ 𝜆 }⇒
𝑥 = ∙ 𝑡;
𝐴𝑠𝑠𝑢 𝑥0 = ;
∙π∙
⇒ (𝑡) = ̂ ∙ cos ( ∙ 𝑡) = ̂ ∙ cos(ω ∙ 𝑡) ; ⇒ [5.33]
𝜆
∙π∙ ∙π∙
⇒ ̇ (𝑡) = ∙ ̂ ∙ sin ( ∙ 𝑡) = ω ∙ ̂ ∙ sin(ω ∙ 𝑡) ; ⇒
𝜆 𝜆
∙π∙ ∙π∙
⇒ ̈ (𝑡) = ( ) ∙ ̂ ∙ cos ( ∙ 𝑡) = ω ∙ ̂ ∙ cos(ω ∙ 𝑡) ;
𝜆 𝜆
Insertion in the model in Equation [5.32] gives directly the solution:
(𝑡) = ∙ + ∆ (𝑡) = ∙ + ̂ ∙ cos(ω ∙ 𝑡) ;

{ (𝑡) = 𝑠 (𝑡) = ̂ ∙ cos (ω ∙ 𝑡);
̈ (𝑡) = ̈𝑠 (𝑡) = ̂ ∙ cos(ω ∙ 𝑡) ; [5.34]
𝑤ℎ 𝑟 ̂ = ∙ω ∙ ̂ ; 𝑛 ̂= ω ∙ ̂ ;
5.5.1.1.1 Analysis of solution

We can identify the magnitude of the transfer functions. The negative sign in Equation [5.34] means
180 degrees phase shift:
ℱ( 𝑠 )
𝐻 = {𝐻 = }=1+𝑗∙ ;
𝑟 𝑠 𝑟 𝑠
ℱ( )
𝐻 𝑟 𝑟− 𝑠
= {𝐻 𝑟 𝑟 − 𝑠 = 𝐻 𝑟 𝑠 𝐻 𝑠 𝑠 = 𝐻 𝑟 𝑠 1} = + 𝑗 ∙ ; [5.35]
𝐻 𝑟 ̈𝑠 = {𝐻 𝑟 ̈𝑠 = (𝑗 ∙ 𝜔) ∙ 𝐻 𝑟 𝑠 = ω ∙ 𝐻 𝑟 𝑠 } = ω + 𝑗 ∙ ;
𝐻 𝑟 ∆𝐹𝑟 = {𝐻 𝑟 ∆𝐹𝑟 = ∙ 𝐻 𝑟 ̈𝑠 } = ∙ω +𝑗∙ ;
The motivation to choose exactly those transfer functions is revealed later, in Section5.6, 5.6.2 and
5.6.3. For now, we simply conclude that various transfer functions can be identified and plotted. The
plots are found in Figure 5-9. Numerical values for m and 𝜆 has been chosen.
So, for example, we can use the transfer function diagram as follows: If we have a road displacement
amplitude of 1 cm ( ̂ = . 1 ) and a speed of 50 km/h ( ≈ 14 𝑠=̂ ≈ .8𝐻 ), we can read out,
e.g.:
• |𝐻 𝑟 ̈𝑠 ( )| ≈ 3 5; ⇒ | ̂| = 31 ∙ ̂𝑟 = 3 5 ∙ . 1 = 3. 5 ⁄𝑠 ;. From this we can
calculate 𝑅𝑀𝑆( ̈ 𝑠 ) = |3. 5|⁄√ ≈ .16 ⁄𝑠 . The RMS value of acceleration will later be
related to ride comfort, see Section 5.6.
• |𝐻 𝑟 𝑟 − 𝑠 ( )| = ;, i.e. no deformation, which is not strange, since model is stiff. The
deformation of suspension will later be related to fatigue life, see Section 5.6.2.
• |𝐻 𝑟 ∆ 𝑟 ( )| ≈ 487 ; ⇒ | ̂ | = 487 ∙ ̂𝑟 = 487 ∙ . 1 = 487 𝑁;. If | ̂ | had been >
∙ ≈ 16 𝑁, the model would have been outside its validity region, because it would
require pulling forces between tyre and road, which is not possible. The variation in tyre road
contact force will be related to road grip, see Section 5.6.3.
The phases for the studied variables can be found in Equation [5.35]. With this model, the phases
becomes constant and ±9 .
261
VERTICAL DYNAMICS
7
Driving at road with wave-length, lambda = 5 [m].
10
6
10
5
10
4
10
H_zr_zs (==1)
abs(H)
3
10 H_zr_zr-zs (==0)
H_zr_derderzs
2 H_zr_Frz
10
m = ms+mu = 1600[kg];
1
10 f=1 Hz at vx=5[m/s]; f=10 Hz at vx=50[m/s]
0
10
-1
10
0 1 10 Hz 2 3
10 1 Hz 10 10 10
vx [m/s]
Figure 5-9: Transfer functions from model in Figure 5-8, excited with single frequencies.
5.5.1.2 Response to a multiple frequency excitation

Using Equation [5.31], Equation [5.45] and values for road type “rough” in Figure 5-6, we can
conclude:
∞
𝛷 𝑤 1
𝑅𝑀𝑆( 𝑠 ) = √ ∙ 𝑥 ∙ ∫|𝐻 𝑟 (𝜔)| ∙ 𝜔 𝑤 ∙ 𝜔 =
𝛺0 −𝑤 𝑠
𝜔=
[5.36]
6 ∞
1 ∙1 .5 1 .5 ∙ 𝜔 ;
=√ ∙ 𝑥 ∙ ∫|𝐻 𝑟 (𝜔)| ∙ 𝜔
1 𝑠
𝜔=
For now, we simply note that it is possible to calculate this (scalar) RMS value for each vehicle speed
over the assumed road. In corresponding way, an RMS value can be calculated for any of the oscillating
variables, such as 𝑠̈ , 𝑠 and . We will come back to Equation [5.36] in Section 5.6.1.2.
5.5.2 One dimensional model with one dynamic

degree of freedom
“With one dynamic degree of freedom” refers to that the axle suspension is modelled as a linear spring
and linear (viscous) damper in parallel. The tyre is still stiff and massless as in model in Section 5.5.1.
The model can be visualised as in Figure 5-10.
Note that the mass is now divided into sprung and unsprung mass, 𝑠 and 𝑢. The mathematical
model becomes as follows:
262
VERTICAL DYNAMICS
Equilibrium sprung mass: 𝑠 ∙ ̈𝑠 𝑠 ∙ = ;

Equilibrium unsprung mass: 𝑢 ∙ ̈ 𝑢 ∙ = ;
[5.37]
Constitution: = ∙( 𝑠) + ∙( ̇ ̇𝑠 ) + 𝑠 ∙ ;
Excitation: = (𝑡);
real vehicle model
̈𝑠
𝑠 𝑠
m
𝑠
z
̈𝑠
𝑠
𝑠
m x
py
c d
̈
𝑢 𝑢
̈𝑢
𝑢
𝑢
𝑢
𝑢
̈𝑢 = ̇ 𝑢 ; ̈𝑠 = ̇ 𝑠 ;
Figure 5-10: One-dimensional model with one dynamic degree of freedom

Eliminating and gives:
𝑠 ∙ ̈𝑠 = ∙ ( (𝑡) 𝑠) + ∙ ( ̇ (𝑡) ̇𝑠 );
Δ = ∙ ( (𝑡) 𝑠) + ∙ ( ̇ (𝑡) ̇𝑠 ) + 𝑢 ∙ ̈ (𝑡); [5.38]
𝑤ℎ 𝑟 =Δ +( 𝑠 + 𝑢) ∙ ;
Note that since we measure 𝑢 and 𝑠 from the static equilibrium, the static load, 𝑠 ∙ , disappears
when constitution is inserted in equilibrium.
Assume that the road has only one (spatial) frequency, i.e. one wave length. Then the excitation is as in
Equation [5.26], in which we assume 𝑥0 = . So, we can insert (𝑡) = ̂ ∙ cos(ω ∙ 𝑡) ⇒ ̇ (𝑡) = 𝜔
̂ ∙ sin(ω ∙ 𝑡) ⇒ ̈ (𝑡) = 𝜔 ̂ ∙ cos(ω ∙ 𝑡) ; in Equation [5.38] and solve it for 𝑠 (𝑡) and Δ (𝑡) with
trigonometry or Fourier transform.
Fourier transform of Equation [5.38] yields:
𝑠 ∙ ( ω ∙ ℱ( 𝑠 )) = ∙ (ℱ( ) ℱ( 𝑠 )) + ∙ (𝑗 ∙ 𝜔 ∙ ℱ( ) 𝑗 ∙ 𝜔 ∙ ℱ( 𝑠 ));
[5.39]
ℱ(∆ 𝑟 ) = ∙ (ℱ( 𝑟 ) ℱ( 𝑠 )) + ∙ (𝑗 ∙ 𝜔 ∙ ℱ( 𝑟 ) 𝑗 ∙ 𝜔 ∙ ℱ( 𝑠 )) 𝑢 ω ∙ ℱ( 𝑟 );
From this, we can then solve the transfer functions:
263
VERTICAL DYNAMICS
ℱ( 𝑠 ) +𝑗∙ ∙𝜔
𝐻 = = ;
𝑟 𝑠
ℱ( ) +𝑗∙ ∙𝜔 𝑠∙ω
ℱ(∆ )
𝐻 ∆𝐹𝑟 = =( 𝑗∙𝜔∙ )∙𝐻 + +𝑗∙𝜔∙ 𝑢 ω = [5.40]
𝑟
ℱ( ) 𝑟 𝑠
𝜔 ∙
= + +𝑗∙𝜔∙ 𝑢 ω ;
+𝑗∙𝜔∙ 𝑠 ∙ω
5.5.2.1.2 Solution with trigonometry

One way to solve the mathematical model in Equation [5.37] is to assume a real solution, insert the
assumption and find expressions for the coefficients in the assumption. Assume such solution (and
that 𝑥0 = in Equation [5.26]):
∙ ̈𝑠 + ∙ ̇𝑠 + ∙ 𝑠 = ̂ ∙ ( ∙ cos(𝜔 ∙ 𝑡) ∙ 𝜔 ∙ sin(𝜔 ∙ 𝑡)); [5.41]
Assumed solution:
𝑠 = ̂𝑠 ∙ cos(𝜔 ∙ 𝑡 ); ⇒
⇒ {𝑢𝑠 : cos( 𝑏) = cos ∙ cos 𝑏 + sin ∙ sin 𝑏} ⇒
⇒ 𝑠 = ̂𝑠 ∙ [cos(𝜔 ∙ 𝑡) ∙ cos + sin(𝜔 ∙ 𝑡) ∙ sin ]; ⇒ [5.42]
⇒ ̇𝑠 = ̂𝑠 ∙ 𝜔 ∙ [ sin(𝜔 ∙ 𝑡) ∙ cos + cos(𝜔 ∙ 𝑡) ∙ sin ]; ⇒
⇒ ̈𝑠 = ̂𝑠 ∙ 𝜔 ∙ [cos(𝜔 ∙ 𝑡) ∙ cos + sin(𝜔 ∙ 𝑡) ∙ sin ];
Insertion:
∙ ̂𝑠 ∙ 𝜔 ∙ [cos(𝜔 ∙ 𝑡) ∙ cos + sin(𝜔 ∙ 𝑡) ∙ sin ] +
+ ∙ ̂𝑠 ∙ 𝜔 ∙ [ sin(𝜔 ∙ 𝑡) ∙ cos + cos(𝜔 ∙ 𝑡) ∙ sin ] +
+ ∙ ̂𝑠 ∙ [cos(𝜔 ∙ 𝑡) ∙ cos + sin(𝜔 ∙ 𝑡) ∙ sin ] =
= ̂ ∙ ( ∙ cos(𝜔 ∙ 𝑡) ∙ 𝜔 ∙ sin(𝜔 ∙ 𝑡)); ⇒
𝒄𝒐𝒔 𝒕𝒆 𝒎𝒔: ∙ ̂𝑠 ∙ 𝜔 ∙ cos + ∙ ̂𝑠 ∙ 𝜔 ∙ sin + ∙ ̂𝑠 ∙ cos = ̂ ∙ ;
⇒ { ⇒
𝒔𝒊𝒏 𝒕𝒆 𝒎𝒔: ∙ 𝑠̂ ∙ 𝜔 ∙ sin ∙ ̂𝑠 ∙ 𝜔 ∙ cos + ∙ ̂𝑠 ∙ sin = ̂ ∙ ∙ 𝜔;
∙ ∙𝜔 [5.43]
= arctan ( );
∙ ∙𝜔 + ∙𝜔
⇒
𝑠̂ ∙𝜔
= = |𝐻 𝑟 𝑠 |;
{ ̂ ( ∙𝜔 ) ∙ sin + ∙ 𝜔 ∙ cos
∙π∙ 𝑥
𝑤ℎ 𝑟 𝜔 = ;
𝜆
We have identified |𝐻 𝑟 𝑠 ( )|, which can be compared to |𝐻 𝑟 𝑠 ( )| in Equation [5.35]. The other
transfer functions in Equation [5.35] are more difficult to express using the method with real algebra.
We leave them to next section.
We can elaborate further with Equation [5.40]:
264
VERTICAL DYNAMICS
Amplitude:
̂𝑠 +𝑗∙ ∙ω
|𝐻 |= =| |=
𝑟 𝑠
̂ ( ∙ω )+𝑗∙ ∙ω
+𝑗∙ ∙ω
𝐴𝑠𝑠𝑢 = 𝑅 𝐻 + 𝑗 ∙ 𝐼 𝐻;
( ∙ω )+𝑗∙ ∙ω + ∙ω
= 𝑆𝑜𝑙 𝑜𝑟 𝑅 𝐻 𝑛 𝐼 𝐻; =⋯=√ ;
( ∙ω ) + ∙ω
{ |𝐻 𝑟 𝑠
| = √𝑅 𝐻 + 𝐼 𝐻 ; }
Phase: [5.44]
+𝑗∙ ∙ω
𝑠 (𝜔) (𝜔) = ar ( )=
∙ω + +𝑗∙ ∙ω
+𝑗∙ ∙ω
𝐴𝑠𝑠𝑢 = 𝑅 𝐻 + 𝑗 ∙ 𝐼 𝐻;
( ∙ω )+𝑗∙ ∙ω
= =⋯=
𝑆𝑜𝑙 𝑜𝑟 𝑅 𝐻 𝑛 𝐼 𝐻;
{ tan(ar (𝐻 𝑟 𝑠 )) = 𝐼 𝐻⁄𝑅 𝐻 ; }
∙ ∙𝜔 ∙𝜋∙
= arctan ( ); where 𝜔 = ;
∙ ∙𝜔 + ∙𝜔 𝜆
Equation [5.13] now allows us to get the magnitudes of the other transfer functions as well:
𝐻 𝑟 𝑠
= from Equation [554 ];
𝐻 𝑟 𝑟− 𝑠
= 𝐻 𝑟 𝑟 𝐻 𝑟 𝑠 = 1 𝐻 𝑟 𝑠;
𝐻 𝑟 ̈𝑠 = 𝜔 ∙ 𝐻 𝑟 𝑠 ;
𝐻 𝑟 ∆ 𝑟 = {∆ 𝑟 = ∙ ( 𝑟 𝑠) + ∙ ( ̇ 𝑟 ̇ 𝑠 )} = [5.45]
= ∙ (𝐻 𝑟 𝑟 𝐻 𝑟 𝑠 ) + ∙ (𝐻 𝑟 ̇ 𝑟 𝐻 𝑟 ̇𝑠) =
= ( + 𝑗 ∙ ∙ 𝜔) ∙ (𝐻 𝑟 𝑟 𝐻 𝑟 𝑠 ) =
= ( + 𝑗 ∙ ∙ 𝜔) ∙ (1 𝐻 𝑟 𝑠 );
The motivation to choose exactly those transfer functions is revealed later, in 5.6. Some of those
magnitudes are easily expressed in reel (non-complex) mathematics using Equation [5.44]:
+ ∙ω
|𝐻 𝑟 𝑠
|=√ ;
( ∙ω ) + ∙ω
[5.46]
+ ∙ω
|𝐻 𝑟 ̈𝑠 | =𝜔 ∙√ ;
( ∙ω ) + ∙ω
The transfer functions in Equation [5.44] are plotted in Figure 5-11 and Figure 5-12. Numerical values
for m and 𝜆 has been chosen.
For example, we can use Figure 5-11 as follows, cf. end of Section 5.5.1.1.1: If we have a certain road,
with displacement amplitude of 1 cm ( ̂ = . 1 ) and the vehicle should drive on it at a speed of 50
km/h ( ≈ 14 ̂ ≈ .8𝐻 ), we can read out, e.g.:
𝑠=
• Ride comfort related: |𝐻 𝑟 ̈𝑠 ( )| ≈ 1 ; ⇒ | ̂| = 1 ∙ ̂𝑟 = 1 ∙ . 1 = 1. ⁄𝑠 ;.
From this we can calculate 𝑅𝑀𝑆( ̈ 𝑠 ) = |1. |⁄√ ≈ .8485 ⁄𝑠 .
• Fatigue life related: |𝐻 𝑟 𝑟 − 𝑠 ( )| ≈ 1.11; ⇒ | ̂ ̂𝑠 | = 1.11 ∙ ̂𝑟 = 1.11 ∙ . 1 =
. 111 = 1.11 ;.
• Road grip related: |𝐻 𝑟 ∆ 𝑟 ( )| ≈ 59795; ⇒ | ̂ 𝑟 | = 59795 ∙ ̂𝑟 = 59795 ∙ . 1 = 598 𝑁;.
265
VERTICAL DYNAMICS
7
10
6
10
5
10
4 m = ms = 1415[kg]; c = cs = 68 [kN/m]; d = ds = 8[kN/(m/s)];

10
mu = 0 [kg]; ct = inf [kN/m];
abs(H)
3 f=1 Hz at vx=5[m/s]; f=10 Hz at vx=50[m/s]

10
2
10
1
10 H_zr_zs
H_zr_zr-zs
0 H_zr_derderzs
10
H_zr_Frz
-1
10
0 1 2 3
10 10 10 Hz 10 10
1 Hz vx [m/s]
Figure 5-11: Transfer functions for amplitudes from model in Figure 5-10, excited with single frequencies.
Thin lines are without damping. Notation: 𝐻 is denoted H_a_b.
200
H_zr_zs
H_zr_zr-zs
150
H_zr_derderzs
H_zr_Frz
phase delay=arg(H) [deg]
100
50
-50
-100
0 1 2 3
10 10 10 10
vx [m/s]
Figure 5-12: Transfer functions for phase delays from model in Figure 5-10, excited with single frequencies.
We compare these numbers with the corresponding numbers for the simpler model, see Sections
5.5.1.1.1. The comfort is better. The fatigue life and road grip have become more realistic.
Figure 5-11 also shows the curves for the undamped system (d=0). The highest peaks appear at
approximately = 5. .6 m/s. This corresponds to the speed where the natural (=undamped) eigen
frequency appears ( 𝑐 𝑡 = 𝜆 ∙ 𝑐 𝑡 = 𝜆 ∙ 𝜔𝑐 𝑡 ⁄( ∙ 𝜋) = 𝜆 ∙ √ ⁄ ⁄( ∙ 𝜋) ≈ 5.5 𝑠).
Figure 5-12 shows the phase angles for the different responses.
266
VERTICAL DYNAMICS
5.5.3 One dimensional model with two degrees of

freedom
The expression “two dynamic degree of freedom” in section the heading refers to that both elasticity
between road and wheel as well as between wheel and sprung mass is modelled. Alternatively, one can
describe the two degrees of freedom as the suspension spring deformation and the tyre spring
deformation.
The model can be visualised as in Figure 5-13. No damping is modelled in tyre (in parallel with
elasticity 𝑡 ) because it is generally relatively low.
real vehicle model

𝑠 ∙ ̈𝑠 𝑠 ∙
zs
ms Fsz
Fsz
z zs
cs ds
x Fsz
zu
py zu mu
∙ ̈𝑢 ∙ Fsz
𝑢 𝑢
ct dt
Frz Frz
zr zu
zr
Frz
Figure 5-13: One-dimensional model with two dynamic degrees of freedom
The corresponding mathematical model becomes as follows:
Equilibrium:
𝑠 ∙ ̈𝑠 = 𝑠 𝑠 ∙ ;
𝑢 ∙ ̈𝑢 = 𝑠 𝑢 ∙ ;
Constitution (displacements counted from static equilibrium):
[5.47]
𝑠 = 𝑠 ∙( 𝑢 𝑠) + 𝑠 ∙ ( ̇𝑢 ̇𝑠 ) + 𝑠 ∙ ;
= 𝑡 ∙( 𝑢) + 𝑡 ∙( ̇ ̇𝑢 ) + ( 𝑠 + 𝑢) ∙ ;
Excitation:
= (𝑡);
The same can be formulated with matrices and Fourier transforms:
267
VERTICAL DYNAMICS
𝑠 ̈ ̇ 𝑠 𝑠 𝑠
[ ] ∙ [ 𝑠] + [ 𝑠 𝑠
] ∙ [ 𝑠] + [ + ]∙[ ]= [ ]∙ ̇ +[ ]∙ ;
𝑢 ̈ 𝑢 𝑠 𝑠+ 𝑡 ̇ 𝑢 𝑠 𝑠 𝑡 𝑢 𝑡 𝑡
̈ ̇ 𝑠
⇒ 𝑴 ∙ [ 𝑠] + 𝑫 ∙ [ 𝑠] + 𝑪 ∙ [ ] = 𝑫 ∙ ̇ + 𝑪 ∙ ;⇒
̈𝑢 ̇𝑢 𝑢
⇒ 𝑴 ∙ ( 𝜔 ∙ 𝑍𝑠𝑢 ) + 𝑫 ∙ (𝑗 ∙ 𝜔 ∙ 𝑍𝑠𝑢 ) + 𝑪 ∙ 𝑍𝑠𝑢 = 𝑫 ∙ (𝑗 ∙ 𝜔 ∙ 𝑍 ) + 𝑪 ∙ 𝑍 ; ⇒ [5.48]
⇒ ( 𝜔 ∙ 𝑴 + 𝑗 ∙ 𝜔 ∙ 𝑫 + 𝑪) ∙ 𝑍𝑠𝑢 = (𝑗 ∙ 𝜔 ∙ 𝑫 + 𝑪 ) ∙ 𝑍 ;
𝑍𝑠 ℱ( 𝑠 )
𝑤ℎ 𝑟 𝑍𝑠𝑢 = [ ]= [ ] ; 𝑛 𝑍 = ℱ( );
𝑍𝑢 ℱ( 𝑢 )
The matrix 𝑫 is zero, but marks a more general form, typical for modelling damping also in the tyre.

We can find the transfer functions via Fourier transform, starting from Equation [4.61]:
𝐻𝑟 𝑍𝑠 1 1
[ 𝑠
]= [ ]∙ = 𝑍𝑠𝑢 ∙ = ( 𝜔 ∙ 𝑴 + 𝑗 ∙ 𝜔 ∙ 𝑫 + 𝑪)− ∙ (𝑗 ∙ 𝜔 ∙ 𝑫 + 𝑪 ); [5.49]
𝐻𝑟 𝑢
𝑍𝑢 𝑍 𝑍
This format is very compact, since it includes both transfer functions for amplitude and phase. For
numerical analyses, the expression in Equation [5.52] is explicit enough, since there are tools, e.g.
Matlab, which supports matrix inversion and complex mathematics.
For analytic solution, one would need symbolic tools, e.g. Mathematica or Matlab with symbolic
toolbox, or careful manual algebraic operations. With 𝑡 = , the explicit forms become as follows:
( 𝑢 ∙ω +𝑗∙ + 𝑠) ∙ 𝑡
𝑠 ∙ω+ 𝑡
( 𝑠 + 𝑗 ∙ 𝑠 ∙ ω) 𝑡
𝑍𝑠 𝑢∙ω +𝑗∙ 𝑠 ∙ω+ 𝑡 + 𝑠
𝑠∙ω +𝑗∙ 𝑠∙ω+ 𝑠
𝐻 = = ;
𝑟 𝑠
𝑍 𝑠 + 𝑗 ∙ 𝑠 ∙ ω
𝑍𝑢 𝑡
[5.50]
𝐻 = = ;
𝑟 𝑢
𝑍 ( 𝑠 + 𝑗 ∙ 𝑠 ∙ ω)
𝑢 ∙ω +𝑗∙ 𝑠 ∙ω+ 𝑡 + 𝑠
𝑠 ∙ω +𝑗∙ 𝑠∙ω+ 𝑠
∙π∙ 𝑥
where 𝜔 = 𝜆
;

Equation [5.13] now allows us to get the magnitudes of the other transfer functions as well:
𝐻 𝑟
= see Equation [555 ];
𝑠
𝐻 𝑟 𝑢
= see Equation [555 ];
𝐻 𝑟 𝑟− 𝑢
= 𝐻 𝑟 𝑟 𝐻 𝑟 𝑢 = 1 𝐻 𝑟 𝑢;
𝐻 𝑟 𝑢− 𝑠
= 𝐻 𝑟 𝑢 𝐻 𝑟 𝑠;
𝐻 𝑟 ̈𝑠 = 𝜔 ∙ 𝐻 𝑟 𝑠 ; [5.51]
𝐻 𝑟 Δ𝐹𝑠 = {Δ 𝑠 = 𝑠 ∙ ( 𝑢 𝑠 ) + 𝑠 ∙ ( ̇𝑢 ̇𝑠 )} =
= 𝑠 ∙ (𝐻 𝑟 𝑢 𝐻 𝑟 𝑠 ) + 𝑠 ∙ 𝑗 ∙ 𝜔 ∙ (𝐻 𝑟 𝑢
𝐻 𝑟 𝑠
)=
= ( 𝑠 + 𝑗 ∙ 𝑠 ∙ 𝜔) ∙ (𝐻 𝑟 𝑢 𝐻 𝑟 𝑠 );
𝐻 𝑟 Δ𝐹𝑟 = {Δ = 𝑡 ∙( 𝑢 )} = 𝑡 ∙ (𝐻 𝑟 𝑟 𝐻𝑟 𝑢
)= 𝑡 ∙ (1 𝐻 𝑟 𝑢
);
268
VERTICAL DYNAMICS
Expression in real can be derived, see Equation [5.52]. Note that it is more general than Equation
[5.51], having a tyre damper modelled, 𝑡 . The derivation is not documented in present
compendium.
√( 𝑠 ∙ 𝑡 𝑠 ∙ 𝑡 ∙ 𝜔 ) + (𝜔 ∙ ( 𝑠 ∙ 𝑡 + 𝑡 ∙ 𝑠 ))
|𝐻 𝑟 ̈𝑠 | =𝜔 ∙ ;
√𝐴 + 𝐵
𝑠 ∙ √( 𝑡 ∙ 𝜔 ) + ( 𝑡 ∙ 𝜔 )
|𝐻 𝑟 𝑢 − 𝑠 | = ;
√𝐴 + 𝐵 [5.52]
√( 𝑠 ∙ 𝑢 ∙ 𝜔 + 𝜔 ∙ ( 𝑠 + 𝑢 ) ∙ 𝑠 ) + (𝜔 ∙ ( 𝑠 + 𝑢) ∙ 𝑠)
|𝐻 𝑟 𝑟 − 𝑢 | = ;
√𝐴 + 𝐵
𝐴 = 𝜔 ∙ 𝑠 ∙ 𝑢 𝜔 ∙ ( 𝑠 ∙ 𝑡 + 𝑠 ∙ 𝑠 + 𝑠 ∙ 𝑡 + 𝑠 ∙ 𝑢) + 𝑠 ∙ 𝑡;
𝐵 = 𝜔 ∙ ( 𝑠 ∙ 𝑡 + 𝑠 ∙ 𝑠 + 𝑢 ∙ 𝑠 ) 𝜔 ∙ ( 𝑠 ∙ 𝑡 + 𝑡 ∙ 𝑠 );
The transfer functions in Equation [5.51] are plotted in Figure 5-14. If we have a certain road, with
displacement amplitude of 1 cm ( ̂ = . 1 ) and the vehicle should drive on it at a speed of 50 km/h
𝑚
( ≈ 14 𝑠 = ̂ ≈ .8𝐻 ), we can read out, e.g.:
• Ride comfort related: |𝐻 𝑟 ̈𝑠 ( )| ≈ 1 3; ⇒ | ̂| = 1 3 ∙ ̂𝑟 = 1 3 ∙ . 1 = 1. 3 ⁄𝑠 ;.

From this we can calculate 𝑅𝑀𝑆( ̈ 𝑠 ) = |1. 3|⁄√ ≈ .8697 ⁄𝑠 .
• Fatigue life related: |𝐻 𝑟 𝑢 − 𝑠 ( )| ≈ 1.14; ⇒ | ̂𝑢 ̂𝑠 | = 1.14 ∙ ̂𝑟 = 1.14 ∙ . 1 =
. 114 = 1.14 ;.
• Road grip related: |𝐻 𝑟 Δ 𝑟 ( )| ≈ 17747 ; ⇒ | ̂ 𝑟 | = 17747 ∙ ̂𝑟 = 17747 ∙ . 1 =
1775 𝑁;.
This analysis can be compared with the analysis in Section 5.5.2.1.3. Ride comfort and fatigue does not
change a lot, but road grip does. This indicates that the more advanced model is only needed for road
grip evaluation.
Figure 5-15 shows the phase angles for the different responses.
Figure 5-16, shows the amplitude gains for the corresponding un-damped system. We identify natural
frequencies at around 5 m/s and 50 m/s. These two speeds correspond to frequencies 𝑐 𝑡 𝜆, i.e.
approximately 1 Hz and 10 Hz. The lower frequency is an oscillation mode where the both masses
move in phase with each other, the so called “heave mode” or “bounce mode”. The higher frequency
comes from the mode where the masses are in counter-phase to each other, the so called “wheel hop
mode”. In the wheel hop mode, the sprung mass is almost not moving at all. We will come back to these
modes in Section 5.5.3.2.
269
VERTICAL DYNAMICS
7
10
6
10
5
10
ms = 1415[kg]; cs = 68 [kN/m]; ds = 8[kN/(m/s)];
4
10 mu = 185[kg]; ct = 676 [kN/m];
f=1 Hz at vx=5[m/s]; f=10 Hz at vx=50[m/s]
abs(H)
3
10
2
10
1
10 H_zr_zs
H_zr_zu-zs
0
10 H_zr_derderzs
H_zr_Frz
-1
10
0 1 10 Hz 2 3
10 10 10 10
1 Hz vx [m/s]
Figure 5-14: Transfer functions for amplitudes from model in Figure 5-13, excited with single frequencies.
Notation: 𝐻 is denoted H_a_b.
200
150
100
50
arg(H)
-50
-100
H_zr_zs
H_zr_zu-zs
-150
H_zr_derderzs
H_zr_Frz
-200
0 1 2 3
10 10 10 10
vx [m/s]
Figure 5-15: Transfer functions for phase delays. Same model and data as in Figure 5-14.
270
VERTICAL DYNAMICS
7
10
6
10
5
10
4
10 H_zr_zs
H_zr_zu-zs
abs(H)
3
10 H_zr_derderzs
H_zr_Frz
2
10
1
10
0
10
-1
10
0 1 2 3
10 10 10 10
vx [m/s]
Figure 5-16: Un-damped transfer functions for amplitudes. Same model and data as in Figure 5-14,
except 𝑠 = .
5.5.3.2 Simplified model

A more practical way to approach the system is to consider the properties of the system. The sprung
mass is typically an order of magnitude greater than the unsprung mass and the suspension spring is
usually an order of magnitude lower than the tyre stiffness. Then, one can split the model in to models,
which explains one mode each, see Figure 5-17. The modes were identified already in Section 5.5.3.1.1.
Modes Models
(Arrows marks displacement amplitudes) heave (or bounce) mode wheel hop mode
zs
ms
cs ds
cs ds zu mu
ct ct
heave (or bounce) wheel hop
(sprung & unsprung (sprung & unsprung
Figure 5-17: Modes and approximate models.

We will now derive the natural frequencies for the two models, and compared with the natural
frequencies ( 𝑠 = ) found for the combined model, in Figure 5-16. Both models are one degree of
freedom models with mass and spring, why the Eigen frequency is √stiffness mass.
For the heave model, the mass is 𝑠 . Stiffnesses 𝑠 and 𝑡 are series connected, which means that the
total stiffness= 1⁄((1⁄ 𝑠 ) + (1⁄ 𝑡 )).
For the wheel hop model, the mass is 𝑢. Stiffnesses 𝑠 and 𝑡 are parallel connected, which means
that the total stiffness= 𝑠 + 𝑡 .
271
VERTICAL DYNAMICS
1 1
1⁄( + ) 𝑟
𝜔𝐵𝑜𝑢𝑛𝑐𝑒 =√ 𝑠 𝑡
= 15.3 ;
𝑠
𝑠 [5.53]
𝑠 + 𝑡 𝑟
𝜔𝑊ℎ𝑒𝑒 𝐻𝑜 = √ = 86. ;
𝑢 𝑠
We see that the numbers match well with the more advanced model, which gave 15. 𝑟 𝑠 and
86.7 𝑟 𝑠, respectively.
Heave (or Bounce) refers to the mode where the sprung mass has the greatest amplitude and wheel
hop is related to the case when the unsprung mass exhibits the greatest amplitude. For a passenger
car, the spring mass has the lowest frequency, typically around 1 Hz while tyre hop is more prevalent
at frequencies around 10 Hz.
5.6 Functions for stationary oscillations

5.6.1 Ride comfort *
Function definition: (Stationary) Ride comfort is the comfort that vehicle occupants experience from
stationary oscillations when the vehicle travels over a road with certain vertical irregularity in a certain speed. The
measure is defined at least including driver (or driver seat) vertical acceleration amplitudes.
Ride comfort is sometimes divided into:
• Primary Ride – the vehicle body on its suspension. Heave (Bounce), Pitch and Roll 0..4 Hz
• Secondary Ride – same but above body natural frequencies, i.e. 4..25 Hz
5.6.1.1 Single frequency

It is generally accepted for stationary vibrations, that humans are sensitive to the RMS value of the
acceleration. However, the sensitivity is frequency dependent, so that highest discomfort appears for a
certain range of frequencies. Some human tolerance curves are shown in Figure 5-18 and Figure 5-19.
The curves can be considered a threshold for acceptance where everything above the line is
unacceptable and points below the curve are acceptable. Discomfort is a subjective measure, and this
is why the different diagrams cannot be directly compared to each other. The SAE has suggested that
frequencies from 4 to 8 Hz are the most sensitive and the accepted accelerations for these are no
higher than 0.025 g (RMS).
The curves in Figure 5-18 mostly represent an extended exposure to the vibration. As one can expect, a
human can endure exposure to more severe conditions for short periods of time. The SAE limits
presented are indicative of 8 hours of continuous exposure. Curves for different exposure times can
also be obtained from ISO, (ISO 2631). The ISO curves are from the first version of ISO 2631 and were
later modified, see Figure 5-19.
272
VERTICAL DYNAMICS
How to use the diagram
RMS not
accele- OK
ration
OK
4 Hz 8 Hz Frequency
Figure 5-18: Various Human Sensitivity Curves to Vertical Vibration, (Gillespie, 1992)
Figure 5-19: ISO 2631 Tolerance Curves
5.6.1.2 Multiple frequencies

The curves in Figure 5-18 and Figure 5-19 can be interpreted as a filter, where the response of the
human is influenced by the frequencies they are exposed to. This leads to the concept of a Human
Filter Function 𝑊𝑘 ( ). (𝑊𝑘 refers to vertical whole human body vibration sensitivity, while there are
other for sensitivities for other directions and human parts.) This approach uses the concept of a
273
VERTICAL DYNAMICS
transfer function from driver seat to somewhere inside the drivers brain, where discomfort is
perceived.
1
ISO2631 Asymptotic approximations of ISO 2631 weighting curves
10
)
0
0
𝑛𝑢𝑚 𝑒
0 -5
10
⇒ 𝑊𝑘 = 1
-10
0 𝑊𝑘 (dB)
Wk in dB
Wk [1]
-1
10 -15
∙ lo Gain
-20
-2
10 -25 Wd (horizontal)
(number=
Wk (vertical)
-30
-3
10 0 1
-1
10
0
10
1
10 10
2
10
3 10 10
f [Hz] Frequency (Hz)
Figure 5-20: Human Filter Function. From (ISO 2631). Right: Asymptotic approximation
Figure 5-21: Human Filter Function for vertical vibrations. Table from (ISO 2631).
With formulas from earlier in this chapter we can calculate an RMS value of a signal with multiple
frequencies, see Equation [5.6]. Consequently, we can calculate RMS of multiple frequency
acceleration. Since humans are sensitive to acceleration, it would give one measure of human
discomfort. However, to get a measure which is useful for comparing accelerations with different
frequency content, the measure has to take the human filter function into account. The Weighted RMS
Acceleration, aw, in the following formula is such measure:
274
VERTICAL DYNAMICS
̂̈ (𝑊𝑘 (𝜔 ) ∙ ̂̈ )
𝑤 = 𝑤( ̈ (𝑡)) = 𝑢𝑠 : 𝑅𝑀𝑆( ̈ (𝑡)) = √∑ = √∑ ;
= =
{ }
or
∞
[5.54]
𝑤 = 𝑤( ̈ (𝑡)) = {𝑢𝑠 : 𝑅𝑀𝑆( ̈ (𝑡)) = √ ∫ 𝐺 ̈ (𝜔) ∙ 𝜔 } =
𝜔=0
= √ ∫ (𝑊𝑘 (𝜔)) ∙ 𝐺 ̈ (𝜔) ∙ 𝜔 ;

𝜔=0
Equation [5.54] is written for a case with only vertical vibrations, hence 𝑊𝑘 and 𝐺 ̈ . If vibrations in
several directions, a total 𝑤 can still be calculated, see (ISO 2631).
In (ISO 2631) one can also find the following equation, which
weights together several time periods, with different vibrations ∑ 𝑤 ∙𝑇
=√ ; [5.55]
spectra. Time averaged whole-body vibration exposure value is 𝑤
∑ 𝑇
denoted 𝑤 .
The 𝑤 in Eq [5.55] is used both for vehicle customer requirement setting at OEMs and governmental
legislation. One example of legislation is (DIRECTIVE 2002/44/EC, 2002). This directive stipulates that
𝑤 in Eq [5.55] in any direction, normalized to 8 hours, may not exceed 1.15 m/s , and if the value
2
exceeds 0.5 m/s action must be taken.

2
5.6.1.2.1 Certain combination of road, vehicle and speed

Now we can use Equation [5.36] without assuming road type. However, we have to identify ̂𝑠̈ and
multiply it with 𝑊𝑘 (𝜔), according to Equation [5.54]. Then we get [5.56].
Using Equation [5.56], we can calculate the weighted RMS value for the different models in Sections
5.5.1, 5.5.2 and 5.5.3. For each model, it will vary with speed, vx. A plot, assuming a certain road type
(“Rough” from Figure 5-6) is shown in Figure 5-28. We can see that the simplest model “stiff tyre, no
unsprung mass” gives much different comfort value than the two others, so the simplest is not goof to
estimate comfort. However, the two other models give approximately same result, which indicates that
the medium model, “stiff tyre, no unsprung mass”, is enough for comfort evaluation. This is no general
truth but an indication that the most advanced model, “two masses, elastic tyre”, is not needed for
comfort on normal roads. The advanced model is more needed for road grip.
We can also see that the comfort decreases, the faster the vehicle drives. If we read out at which speed
we reach 𝑤 = 1 ⁄𝑠 (which is a reasonable value for long time exposure) we get around ≈
7 ⁄𝑠 ≈ 5 𝑘 ℎ on this road type (“Rough”) with the medium (and advanced) model. With the
simplest model, we get ≈ 1 ⁄𝑠 ≈ 3. .4 𝑘 ℎ.
5.6.1.3 Other excitation sources

Present chapter focuses on the influence of excitation from vertical displacement of the road.
Examples of other, but often co-operating, excitation sources are:
• Powertrain vibrations, non-uniform rotation in engine. Frequencies will be proportional to
engine speed
• Wheel vibrations, e.g. due to non-round wheels or otherwise unbalanced wheels. Frequencies
will be proportional to vehicle speed.
• Special machineries mounted on vehicles (e.g. climate systems or concrete mixers)
275
VERTICAL DYNAMICS
∞
𝛷 𝑤 1
𝑢𝑠 : 𝐻 ̈𝑠 =
∙𝜔 𝑤∙ 𝜔 ={
𝑟
𝑅𝑀𝑆( ̈𝑠 ) = √ −𝑤 ∙ 𝑥 ∙ ∫|𝐻 𝑟 ̈𝑠 (𝜔)| }=
𝛺0 = 𝜔 ∙𝐻 𝑟 𝑠
𝜔=
∞
𝛷 𝑤 1
=√ ∙ 𝑥 ∙ ∫ |𝜔 ∙ 𝐻 𝑟 (𝜔)| ∙ 𝜔 𝑤 ∙ 𝜔 =
𝛺0 −𝑤 𝑠
𝜔=
∞ 𝑢𝑠 :
𝛷 𝑤 1
=√ ∙ 𝑥 ∙ ∫|𝐻 𝑟 (𝜔)| ∙ 𝜔4 𝑤 ∙ 𝜔 ; ⇒ {Equation } ⇒ [5.56]
𝛺0 −𝑤 𝑠
[5554]
𝜔=
∞
𝛷 𝑤 1
⇒ =√ ∙ 𝑥 ∙ ∫(𝑊𝑘 (𝜔)) ∙ |𝐻 𝑟 (𝜔)| ∙ 𝜔4 𝑤 ∙ 𝜔 =
𝑤
𝛺0 −𝑤 𝑠
𝜔=
∞
𝛷 𝑤 1
=√ ∙ 𝑥 ∙ ∫(𝑊𝑘 (𝜔)) ∙ |𝐻 𝑟 (𝜔)| ∙ 𝜔4 𝑤 ∙ 𝜔 ;
𝛺0 −𝑤 𝑠
𝜔=
2
Ride Comfort. For road type "Rough"
10
1
10
Weigthed RMS value, aw [m/(s*s)]
0
10
-1
10
-2
10
-3
10
-2 -1 0 1 2
10 10 10 10 10
vx [m/s]
Figure 5-22: Weighted RMS values for road type “Smooth” from Figure 5-6. The 3 curves show 3 different
models: Simplest (from Sections 1.5.1), Medium (from Section 1.5.2) and Most advanced (from Section 1.5.3).
5.6.2 Fatigue life *

Function definition: (Vehicle) Fatigue life is the life that the vehicle, mainly suspension, can reach due to
stationary oscillations when vehicle travels over a road with certain vertical irregularity in a certain speed. The
measure is defined at least including suspension vertical deformation amplitudes.
Beside human comfort, the fatigue of the vehicle structure itself is one issue to consider in vertical
vehicle dynamics.
276
VERTICAL DYNAMICS
5.6.2.1 Single frequency

5.6.2.1.1 Loads on suspension spring
In particular, the suspension spring may be subject to fatigue. The variation in spring material stress is
dimensioning, which is why the force variation or amplitude in the springs should be under
observation. Since spring force is proportional to deformation, the suspension deformation amplitude
is proposed as a good measure. This is the explanation to why the amplitude of 𝑢 𝑠 is plotted in
Figure 5-11.
Beside fatigue loads, 𝑢 𝑠 is also relevant for judging whether suspension bump-stops become
engaged or not. At normal driving, that limit should be far from reached, except possibly at high loads
(many persons/much payload).
It can be noted that 𝑢 𝑠 represents the variation in material stress only if spring is not changed. So,
if different spring designs are compared, it is not sufficient to study only 𝑢 𝑠.
5.6.2.1.2 Fatigue of other components

Fatigue of other parts may require other amplitudes.
One other relevant example can be the damper fatigue. Damper fatigue would be more relevant to
judge from amplitude of ̇𝑢 ̇𝑠 , which determines the force level and hence the stress level.
Another example is the load of the road itself. For heavy trucks, it is important to consider how much
they wear the road. At some roads with legislated maximum (static) axle load, one can be allowed to
exceed that limit if the vehicle has especially road friendly suspensions. For these judgements, it is the
contact force between tyre and road, ̂ , which is the most relevant variable to study.

If the excitation is of one single frequency, the stress amplitude can be used when comparing two
designs. However, for spectra of multiple frequencies, one cannot look at amplitudes solely,
[ ̂ ̂ ⋯ ̂ ], because the amplitudes will depend on how the discretization is done, i.e. the number N.
Some kind of integral of a spectral density is more reasonable. In this compendium, it is proposed that
a very approximate measure of fatigue load is calculated as follows, exemplified for the case of fatigue
of the spring:
𝑀 𝑠𝑢𝑟 𝑜𝑟 𝑠𝑝𝑟 𝑛 𝑡 𝑢 𝑙 = 𝑅𝑀𝑆( 𝑢 (𝑡) 𝑠 (𝑡)) =

∞
𝛷 𝑤 1
𝑢𝑠 : 𝐻 𝑟 =
=√ −𝑤 ∙ 𝑥 ∙ ∫|𝐻 𝑟 (𝜔)| ∙ 𝜔 𝑤 ∙ 𝜔 = { 𝑢 𝑠
}=
𝛺0 𝑢 𝑠 =𝐻𝑟 𝑢 𝐻 𝑟 𝑠
𝜔= [5.57]
∞
𝛷 𝑤 1
=√ −𝑤 ∙ 𝑥 ∙ ∫|𝐻 𝑟 (𝜔) 𝐻 𝑟
(𝜔)| ∙ 𝜔 𝑤 ∙ 𝜔 ;
𝛺0 𝑢 𝑠
𝜔=
Equation is written for application to a known road spectra (𝛷0 𝑤 ) and vehicle dynamic structure
(𝐻 𝑟 𝑢 𝐻 𝑟 𝑠 ), but the first expression (𝑅𝑀𝑆( 𝑢 (𝑡) 𝑠 (𝑡))) is applicable on a measured or
simulated time domain solution.
277
VERTICAL DYNAMICS
Road grip *
Function definition: Road grip (on undulated roads) is how well the longitudinal and lateral grip between
tyres and road is retained due to stationary oscillations when the vehicle travels over a road with certain vertical
irregularity in a certain speed.
Sections 3.3 and 3.4 show the brush model explain how the tyre forces in the ground plane appears. It
is a physical model where the contact length influences how stiff the tyre is for longitudinal and lateral
slip. There is also a brief description of relaxation models for tyres. This together motivates that a tyre
has more difficult to build up forces in ground plane if the vertical force varies. We can understand it
as when contact length varies, the shear stress builds up has to start all over again. As an average
effect, the tyre will lose more and more grip, the more the vertical force varies.

If the excitation is of one single frequency, the stress amplitude can be used when comparing two
designs. However, for spectra of multiple frequencies, one cannot look at amplitudes solely,
[ ̂ ̂ ⋯ ̂ ], because the amplitudes will depend on how the discretization is done, i.e. the number N.
Some kind of integral of a spectral density is more reasonable. In this compendium, it is proposed that
a very approximate measure of road grip is calculated as follows:
𝑀 𝑠𝑢𝑟 𝑜𝑟 (𝑏 )𝑟𝑜 𝑟 𝑝 = 𝑅𝑀𝑆(Δ 𝑟 (𝑡)) =

∞
𝛷 𝑤 1
=√ ∙ 𝑥 ∙ ∫|𝐻 𝑟 (𝜔)| ∙ 𝜔 𝑤 ∙ 𝜔 =
𝛺0 −𝑤 Δ 𝑟
𝜔=
[5.58]
= {𝑢𝑠 : 𝐻 𝑟 𝑢 𝑠
= 𝑡 ∙ (1 𝐻 𝑟 𝑢
)} =
∞
𝛷 𝑤 1
=√ ∙ 𝑥 ∙ ∫| ∙ (1 𝐻 )| ∙ 𝜔 𝑤 ∙ 𝜔 ;
𝛺0 −𝑤 𝑡 𝑟 𝑢
𝜔=
Equation is written for application to a known road spectrum (𝛷0 𝑤 ) and vehicle dynamic structure
(𝐻 𝑟 𝑢 ), but the first expression (𝑅𝑀𝑆(Δ 𝑟 (𝑡))) is applicable on a measured or simulated time
domain solution.
5.6.4 Other functions

Present chapter focuses on the functions, (vertical) ride comfort, fatigue and road grip. Examples of
other functions are:
• An area of functions that encompasses the vertical dynamics is Noise, Vibration, and
Harshness – NVH. It is similar to ride comfort, but the frequencies are higher, stretching up to
sound which is heard by humans.
• Ground clearance (static and dynamic) between vehicle body and ground. Typically,
important for off-road situations.
• Longitudinal comfort, due to drive line oscillations and/or vertical road displacements.
Especially critical when driver cabin is separately suspended to the body. This is the case for
heavy trucks.
• Disturbances in steering wheel feel, due to one-sided bumps. Especially critical for rigid
steered axles. This is often the design of the front axle in heavy trucks.
• There are of course an infinite number of combined manoeuvres, in which functions with
requirements can be found. Examples can be bump during strong cornering (possibly
278
VERTICAL DYNAMICS
destabilizing vehicle) or one-sided bump (exciting both heave=bounce, pitch and roll modes).
When studying such transients, the vertical dynamics is not enough to capture the comfort, but
one often need to involve also longitudinal dynamics; the linkage with ant-dive/anti-squat
geometry from Chapter 3 becomes important as well as tyre vertical (radial) deflection
characteristics.
• Energy is dissipated in suspension dampers, which influence energy consumption for the
vehicle. This energy loss is much related, but not same as, to (tyre) rolling resistance.
Suspension characteristics do influence this energy loss, but it is normally negligible, unless
driving very fast on very uneven road.
5.7 Variation of suspension design

The influence of design parameters on vehicle functions Ride comfort, Suspension fatigue and Road
grip can now be made. E.g., it is important to not only use the transfer function, but also take the road
and human sensitivity into account, which calls for different weighting for different frequencies.
Transfer function for the model in Section 5.5.3 is shown as dashed lines in
Figure 5-23. Same figure also shows the Road- and Human-weighted versions. Studying how these
curves change with design parameters gives a quantitative understanding of how different suspension
design parameters influence. Such variations will be done in Sections 5.7.1 to 5.7.4.
There are two particular frequency intervals of the graphs to observe. These are the 2 peaks around
the two the natural frequencies of the sprung and unsprung masses, the peak at lower frequency is
mainly a resonance in heave mode, while the higher one is in wheel hop mode.
3
Comfort as der(der(zs)) 1
Suspension Fatigue as zu-zs 1
Road Grip as zr-zu
10 10 10
2 0
10 10
0
10
1 -1
10 10
amplification /[(m/(s*s))/(m)]
-1
amplification /[m/m]
amplification /[m/m]
10
0 -2
10 10
-2
10
-1 -3
10 10
-3
amplification /[m/m] 10
-2 -4
10 10 Road-weighted
amplification /[(m/(s*s))/(m)]
Road-weighted amplification /[m/m]
Road- and Human-weighted Road-weighted
-3 -5 -4
10 10 10
-1 0 1 2 -1 0 1 2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
f /[Hz] f /[Hz] f /[Hz]
Figure 5-23: For a passenger car with 𝑠 = 16 [𝑘 ] 𝑢 = [𝑘 ] 𝑠 = 76 [𝑘𝑁⁄ ] 𝑠 =

9 [𝑘𝑁⁄( ⁄𝑠 )] 𝑡 = 764 [𝑘𝑁⁄ ] 𝑡 = . Left is vertical acceleration (amplitude) of sprung mass for Ride
Comfort. Middle is relative displacement (amplitude) between sprung and unsprung mass for Suspension
Fatigue. Right is deformation (amplitude) of tyre spring for Road Grip. Weightings for typical road and for
human sensitivity is shown.
279
VERTICAL DYNAMICS
5.7.1 Varying suspension stiffness

In Figure 5-24 the benefits of the low suspension stiffness (1 Hz) is seen for suspension travel and
comfort without much change in the road grip performance.
Suspension Fatigue as zu-zs Road Grip as zr-zu
2 0
10 10 10
0
10
Road- and Human-weighted amplification /[(m/(s*s))/(m)]
1
10

-1 -1
10 10
0
10
-2
10
-1
10
-3 -2
10 10
-2
10
-4
10
smaller cs = 38 [kN/m]
reference cs = 76 [kN/m]
larger cs = 153 [kN/m]
-3 -5 -3
10 10 10
-1 0 1 2 -1 0 1 2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
f /[Hz] f /[Hz] f /[Hz]
Figure 5-24: Result from varying suspension stiffness, 𝑠
Regarding Figure 5-24 and Figure 5-25 we see that there is a large influence of the acceleration gain at
low frequencies with little change at the wheel hop and higher frequencies. The suspension stiffness
and damping was seen to have little influence on the ride comfort / road grip response around 10 Hz.
5.7.2 Varying suspension damping

In Figure 5-25, we see that the changes in suspension damping have opposite effects for the heave and
wheel hop frequency responses. High damping is good for reducing heave, but not so effective for
wheel hop.
5.7.3 Varying unsprung mass

In Figure 5-26, we see that if the response around the wheel hop frequency is to be changed, the
unsprung mass is one of the most influential parameters. The unsprung mass is usually in the range of
10% of the sprung mass. Opposite to the suspension parameters, the unsprung mass influences
frequencies around the wheel hop frequency with little influence around the heave frequency.
In Figure 5-26, the case with 𝑢 = is added. This is to demonstrate what a model with neglected
mass gives and can be nearly compared with the model in Section 5.5.2.
5.7.4 Varying tyre stiffness

In Figure 5-27 a general observation is that low sprung mass natural frequencies are preferred for
comfort considerations. Another parameter that has a strong affect near the wheel hop frequency is
the tyre stiffness. The strongest response is noticed for the road grip function. (Note that, since 𝑡 is
now varying, we have to express road grip as 𝑡 ∙ ( 𝑢 ); only 𝑢 does not give a fair
comparison.)
280
VERTICAL DYNAMICS
2 0
10 10 10
0
10
1
10

-1 -1
10 10
0
10
-2
10
-1
10
-3 -2
10 10
-2
10
-4
10
smaller ds = 4 [kN/m]
reference ds = 9 [kN/m]
larger ds = 18 [kN/m]
-3 -5 -3
10 10 10
-1 0 1 2 -1 0 1 2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
f /[Hz] f /[Hz] f /[Hz]
Figure 5-25: Result from varying suspension damping, 𝑠
2 0
10 10 10
0
10
1
10
-1 -1
10 10
0
10
-2
10
-1
10
-3 -2
10 10
-2
10
-4
10 smaller mu = 100 [kg]
reference mu = 200 [kg]
larger mu = 400 [kg]
zero mu = 0 [kg]
-3 -5 -3
10 10 10
-1 0 1 2 -1 0 1 2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
f /[Hz] f /[Hz] f /[Hz]
Figure 5-26: Result from varying unsprung mass, 𝑢
281
VERTICAL DYNAMICS
2
Suspension Fatigue as zu-zs 6
Road Grip as ct*(zr-zu)
10 10 10
0
10
1
10

-1 5
10 10
0
10
-2
10
-1
10
-3 4
10 10
-2
10
-4
10 smaller ct = 382 [kN/m]
reference ct = 764 [kN/m]
larger ct = 1529 [kN/m]
-3 -5 3
10 10 10
-1 0 1 2 -1 0 1 2 -1 0 1 2
10 10 10 10 10 10 10 10 10 10 10 10
f /[Hz] f /[Hz] f /[Hz]
Figure 5-27: Result from varying Tyre Stiffness, 𝑡
5.8 Two dimensional oscillations

The one-dimensional model is useful for analysing the response of one wheel/suspension assembly.
Some phenomena do connect other vehicle body motions than the vertical translation, especially pitch
and roll. Here, other models are needed, such as Figure 5-28 and Figure 5-31.
5.8.1 Heave and pitch

A model like in Figure 5-28 is proposed. We have been studying heave (or bounce) and pitch before, in
Section 3.4.7. Hence compare with corresponding model in Figure 3-46. In Chapter 3, the excitation
was longitudinal tyre forces, while the vertical displacement of the road was assumed to be zero. In
vertical vehicle dynamics, it is the opposite. The importance of model with linkage geometry (pitch
centre or axle pivot points above ground level) is that tyre forces are transferred correctly to the body.
That means that the linkage geometry is not so relevant for vertical vehicle dynamics in Chapter 5. So
the model can be somewhat simpler.
=
= ̇ 𝜔
𝜔 𝜔̇
𝑙 𝑙
Figure 5-28: Heave and pitch physical model.
282
VERTICAL DYNAMICS
A mathematical model would typically show two different modes, see Figure 5-29. The heave eigen-
frequency is typically 1-1.5 Hz for a passenger car. The pitch frequency is somewhat higher.
We should reflect on that the models in Section 5.5 and 5.8.1 refer to the same bounce mode. But the
models will most likely give different numbers of, e.g., Eigen frequency. A total model, with all degrees
of freedom, would align those values, but the larger a model is the more data it produces which often
leads to less easy design decisions.
Pitch Heave (or Bounce)

mode mode
Motion node Motion node for
for Pitch mode Heave mode
Figure 5-29: Oscillation modes of a Heave and Pitch model.
5.8.1.1 Wheel base filtering

See Figure 5-30. If the wheel base is a integer multiple of the wave length, only heave (bounce) will be
excited. If wave length is in the middle between those, only pitch will be excited. This phenomenon is
called “wheel base filtering”.
HEAVE PITCH
Figure 5-30: Response of the Vehicle to Different Road Wavelengths. (Gillespie, 1992)
5.8.1.2 Mathematical model

From Figure 5-28 we can derive the following mathematical model:
Equilibrium: + ( ̇ 𝜔 )= ; 𝑛 𝑙 + 𝑙 𝜔̇ = ;
Compatibility: = 𝑙 𝜔 ; 𝑛 = +𝑙 𝜔 ;
283
VERTICAL DYNAMICS
Constitution: = 𝑠+ ( ); ̇ 𝑠 = ( );
= 𝑠+ ( ); 𝑛 ̇𝑠= ( );
Note that we chose to formulate the model on first order form ( ̇ ⋯ ; and ̇ = (⋯ );) to show an
alternative to second order differential equation form ( ̈ ⋯ ;).
On matrix form:
̇ 𝑙 𝑙 + 1 1
𝜔̇ 𝑙 𝑙 𝑙 𝑙 𝑙 𝑙 𝜔
1⁄ ̇𝑠 = 1 𝑙
[
𝑠
]+
[
⏟ 1⁄ ] [ ̇ 𝑠 ] [⏟ 1 𝑙 ] 𝑠
𝑴𝑪 𝑫
𝑙 𝑙
+[ ]+[ ] [ ];
1
⏟ ⏟ 1
𝒈 𝑫
With [ 𝜔 𝑠 𝑠] =𝒛 𝑫−𝟏 𝒈; and 𝒛 = [ ] ; one gets: 𝑴𝑪 𝒛̇ = 𝑫 𝒛 + 𝑫 𝒛 ;.
Fourier transform gives: 𝑗 𝜔 𝑴𝑪 𝓕(𝒛) = 𝑫 𝓕(𝒛) + 𝑫 𝓕(𝒛 );

5.8.1.2.1 Uncorrelated
We now assume that front and rear are excited “uncorrelated”. This is wrong if driving on a road
where rear axle follows front axle, but it could be achieved in a shake rig with independent excitation
under each axle.
ℱ( ) 1 ℱ( )
𝓕(𝒛 ) = [ ]= [ ];
ℱ( ) 𝑗 𝜔 ℱ( )
𝐻𝑟 𝑢𝑛𝑐𝑜 𝐻 𝑟𝑟 𝑢𝑛𝑐𝑜
1 ℱ( ) 𝐻𝑟 𝜔𝑦 𝑢𝑛𝑐𝑜 𝐻 𝑟𝑟 𝜔𝑦 𝑢𝑛𝑐𝑜 ℱ( )
𝓕(𝒛) = (𝑗 𝜔 𝑴𝑪 𝑫)− 𝑫 [ ]= [ ];
𝑗 𝜔 ℱ( ) 𝐻𝑟 𝐹 𝑠 𝑢𝑛𝑐𝑜 𝐻 𝑟𝑟 𝐹 𝑠 𝑢𝑛𝑐𝑜 ℱ( )
[𝐻 𝑟 𝐹𝑟𝑠 𝑢𝑛𝑐𝑜 𝐻 𝑟𝑟 𝐹𝑟𝑠 𝑢𝑛𝑐𝑜 ]
The elements in 𝐻 depend on 𝜔 (or 𝜆) and .
5.8.1.2.2 Correlated or Wheel base filtered

If we instead assume that front and rear are excited with same frequencies, but delayed at the rear:
ℱ( ) ̂ cos(𝜔 𝑡)
𝓕(𝒛 ) = [ ] = 𝓕 ([ ]) = 𝓕 ( [ ]) = 𝓕 ( [ ])
ℱ( ) 𝑡 𝑡 ̂ cos(𝜔 (𝑡 ⁄ ))
1
=[ ] ℱ( )=
cos( 𝜋 ⁄ ) + 𝑗 sin( 𝜋 ⁄ )
1 1 1
= [ ] ℱ( ) = 𝒅 ℱ( );
𝑗 𝜔 cos( 𝜋 ⁄ ) + 𝑗 sin( 𝜋 ⁄ ) 𝑗 𝜔 𝑳
1
𝑗 𝜔 𝑴𝑪 𝓕(𝒛) = 𝑫 𝓕(𝒛) + 𝑫 𝒅𝑳 ℱ( );
𝑗 𝜔
𝐻𝑟
1 𝐻𝑟 𝜔𝑦
𝓕(𝒛) = (𝑗 𝜔 𝑴𝑪 𝑫)− 𝑫 𝒅𝑳 ℱ( ) = 𝑯 ℱ( )= ℱ( );
𝑗 𝜔 𝐻𝑟 𝐹 𝑠
[𝐻 𝑟 𝐹𝑟𝑠 ]
284
VERTICAL DYNAMICS
All 𝐻 depend on 𝜔 (or 𝜆) and .
5.8.2 Heave and Roll

A model like in Figure 5-31 is proposed. We have been studying Heave (bounce) and pitch before, in
Sections 4.3.9 and 4.5.2.3. Hence compare with corresponding model in Figure 4-72. In Chapter 4, the
excitation was lateral tyre/axle forces, while the vertical displacement of the road was assumed to be
zero. In vertical vehicle dynamics, it is the opposite. That means that the linkage geometry (roll centre
or wheel pivot points) is not so relevant here. So, the model can be somewhat simpler.
m*az m*g
z
zl zr J*der(wx)
px
zrl zrr
zrl
zrr Flz Frrz
x x
Figure 5-31: Heave and roll model.
No equations is formulated for this model in this compendium, but a model will typically show two
different modes, the heave and roll. Heave Eigen frequency is typically 1-1.5 Hz for a passenger car, as
mentioned before. The roll frequency is similar or somewhat higher.
If modelling unsprung masses without inertia, we still get 2 state variables, heave and roll . Using
same mathematical form of equations as in Eq [5.48] we get this model (subscripts rl for “road left”
and rr for “road right”):
𝑴 ∙ 𝒛̈ + 𝑫 ∙ 𝒛̇ + 𝑪 ∙ 𝒛 = 𝑫 ∙ 𝒛̇ + 𝑪 ∙ 𝒛 ;
𝑤ℎ 𝑟 𝒛 = [ ] 𝑛 𝒛 =[ ];
The disturbances from the road are two independent ones, so [5.59]
the transfer functions will be a × matrix:
ℱ( ) ℱ( ) 𝐻 𝐻 𝑟𝑟 ℱ( )
[ ]=𝑯 [ ] = [ 𝑟𝑙 ] [ ];
ℱ( ) ℱ( ) 𝐻 𝑟𝑙 𝜑𝑥 𝐻 𝑟𝑟 𝜑𝑥 ℱ( )
Note that the restoring matrix 𝑪 might need to include both elastic restoring and pendulum effects, see
4.3.9.2 and Reference (Mägi, 1988). For high-loaded trucks, the pendulum effect is really relevant,
while it often can be omitted for a low sportscar.
5.8.3 Three dimensional oscillations

A real road generates motion in all out-of-road-plane dimensions: heave , roll and pitch . One
can, for instance, use [5.27] to generate a vertical profile for each side of the vehicle, but then
randomly generate different phases for each side. Another way is to record a certain piece of a road
where one does testing, typically at the vehicle manufacturer’s test track. One then gets wheel base
filtering one each side and a roll influence due to that left and right side are not in phase with each
other.
285
VERTICAL DYNAMICS
If modelling unsprung masses without inertia, we still get 3 state variables: heave , pitch and roll
. Using same form of equations as in Eq [5.48] we get this model (subscripts 𝑟 𝑙 for “road front left”
and so on):
𝑴 ∙ 𝒛̈ + 𝑫 ∙ 𝒛̇ + 𝑪 ∙ 𝒛 = 𝑫 ∙ 𝒛̇ + 𝑪 ∙ 𝒛 ;
𝑤ℎ 𝑟 𝒛 = [ ];
1
1
𝑛 𝒛 =[ ] = [cos( ) + 𝑗 sin( ) ] [ ];
cos( ) + 𝑗 sin( ) [5.60]
𝑤ℎ 𝑟 = 𝜋 ⁄ ;
The disturbances from the road are two independent ones, so the transfer
functions will be a 3× matrix:
ℱ( ) 𝐻𝑟𝑙 𝐻𝑟𝑟
ℱ( ) ℱ( )
[ ℱ( )] = 𝑯 [ ] = [𝐻 𝑟 𝑙 𝜑𝑥 𝐻𝑟𝑟 𝜑𝑥 ] [ ];
ℱ( ) ℱ( )
ℱ( ) 𝐻𝑟𝑙 𝜑𝑦 𝐻𝑟𝑟 𝜑𝑦
5.9 Transient vertical dynamics

The majority of the chapter you read now, considers normal driving which is how the vehicle is driven
for during long time periods. Functions are then suitably analysed using theory for stationary
oscillations.
Vertical vehicle dynamics also have transient disturbances to consider. Test cases can be one-sided or
two sides road bumps or pot-holes. Two-sided bump is envisioned in
Figure 5-32. It can represent driving over a speed bump or an object/low animal on a road.
Models from earlier in this chapter are all relevant for two-sided bumps/pot-holes, but one might need
to consider non-linearities such as bump stops or wheel lift as well as different damping in
compression and rebound. For one-sided bumps/potholes, the models from earlier in this chapter are
generally not enough. The evaluation criteria should be shifted somewhat:
• Human comfort for transients are often better described as time derivative of acceleration
(called “jerk”).
• The material loads are more of maximum load type than fatigue life dimensioning, i.e. higher
material stress but fewer load cycles during vehicle life time.
• Road grip studies over road bumps and pot-holes are challenging. Qualitatively, the tyre
models have to include relaxation, because that is the mechanism which reduces road grip
when vertical load shifts. To get quantitatively correct tyre models is beyond the goal of the
compendium you presently read.
• Roll-over can be tripped by large one-sided bumps. This kind of roll-overs is unusual and
requires complex models.
286
VERTICAL DYNAMICS
Figure 5-32: Response of Vehicle for Front and Rear Axle Impulses, (Gillespie, 1992)
Models for studying transient vertical dynamics can, in general be categorized as the stationary
oscillation models, 1D, 2D and 3D. But they cannot generally be linear, so they require simulation, not
frequency analysis. One typically need to add inertia of unsprung parts and vertical elasticities in each
tyre. And “trivial linkage” suspension is generally not enough if sharp road unevenness, but instead
one might identify the pivot axis in space for each wheel linkage.
A 3D model according to these concepts gets the states 𝒛 containing 𝑛 if
modelled with a second order differential equation (𝒇(𝒛̈ 𝒛̇ 𝒛 𝒖 𝑡) = ;). If modelled with first order
differential equations (𝒇(𝒛̇ 𝒛 𝒖 𝑡) = ;) and the concept of using forces in elasticities as states, see
1.5.2.5.2.1, the states 𝒛 will instead contain 𝑠 𝑠 𝑠 𝑛 𝑠 ,
where 𝑗 is vertical velocity of unsprung mass in wheel 𝑗 and 𝑠 𝑗 is elastic part of vertical force
under wheel 𝑗. The inputs (disturbance) 𝒖 = 𝒛 will contain 𝑛 .
287
VERTICAL DYNAMICS
288
Bibliography
AB Volvo. 2011. Global Transport Application. 2011.
Andreasson, Johan. 2007. On Generic Road Vehicle Motion Modelling and Control. Stockholm : KTH,
2007. Dissertation.
Bakker, E., Nyborg, L., Pacejka, H.B. 1987. Tyre Modelling for use in Vehicle Dynamics Studies. 1987.
SAE Paper No.870495.
Barnard, R. H. Road Vehicle Aerodynamic Design. u.o. : Mechaero Publishing. ISBN 0954073479.
—. 2010. Road Vehicle Aerodynamic Design. u.o. : Mechaero Publishing, 2010.
Boerboom, Max. 2012. Electric Vehicle Blended Braking maximizing energy recovery while
maintaining vehicle stability and maneuverability. Göteborg, Sweden : Chalmers University of
Technology, 2012. Diploma work - Department of Applied Mechanics. ISSN 1652-8557; no 2012:01.
Clark, S, (ed). 1971. Mechanics of Pneumatic Tires, Monograph 122. u.o. : National Bureau of
Standards, USA, 1971.
Cooper Tire & Rubber Co. 2007. Passenger radial tire cutaway. Online Art. [Online] Cooper Tire &
Rubber Co., den 16 September 2007. http://deantires.com/us/en/information/info-construction.asp.
DIRECTIVE 2002/44/EC. 2002. DIRECTIVE 2002/44/EC OF THE EUROPEAN PARLIAMENT AND OF
THE COUNCIL. 2002.
http://europa.eu/legislation_summaries/employment_and_social_policy/health_hygiene_safety_at_wo
rk/c11145_en.htm.
Drenth, Edo F. 1993. Brake Stability of Front Wheel Driven Cars at High Speed. Delft, Netherlands :
Delft University of Technology., 1993. Master thesis.
Encyclopædia Britannica Online. 2007. belted tire: tire designs." Online Art. Encyclopædia
Britannica Online. [Online] 16 September 2007. http://www.britannica.com/eb/art-7786.
Gillespie, T. 1992. Fundamentals of Vehicle Dynamics. s.l. : Society of Automotive Engineers, 1992.
Grosch, K.A. and Schallamach, A. 1961. Tyre wear at controlled slip. Wear. September–October
1961, Vol. Volume 4, Issue 5, pp. Pages 356–371.
Happian-Smith, Julian. 2002. An Introduction to Modern Vehicle Design. u.o. : Butterwoth-Heinemann,
ISBN 0-7506-5044-3, 2002.
Hirschberg, W., Rill, G. and Weinfurter, H. 2007. Tire model TMeasy. Vehicle System Dynamics. 2007,
Vol. 45:1, pp. 101 — 119.
Hucho, Wolf-Heinric. 1998. Aerodynamics of Road Vehicles. u.o. : SAE International, 1998. R-177.
ISO 11026. ISO 11026 Heavy commercial vehicles and buses - Test method for roll stability - Closing-
curve test. u.o. : International Organization for Standardization, Genève, Switzerland.
ISO 14791. ISO 14791 Road vehicles - Heavy commercial vehicle combinations and articulated buses -
Lateral stability test methods. u.o. : International Organization for Standardization, Genève,
Switzerland.
ISO 14792. ISO 14792 Road vehicles - Heavy commercial vehicles and buses - Steady-state circular tests.
u.o. : International Organization for Standardization, Genève, Switzerland. ISO 14792.
ISO 14793. ISO 14793 Road vehicles – Heavy commercial vehicles and buses – Lateral transient response
test methods. u.o. : International Organization for Standardization, Genève, Switzerland.
289
ISO 14794. 2011. ISO 14794 Heavy commercial vehicles and buses - Braking in a turn - Open-loop test
methods. u.o. : International Organization for Standardization, Genève, Switzerland, 2011. ISO 14794.
ISO 26262. 2011-2012. Road vehicles – Functional safety, 1-10. u.o. : International Organization for
Standardization, Genève, Switzerland, 2011-2012.
ISO 2631. ISO 2631 – Evaluation of Human Exposure to Whole-Body Vibration. s.l. : International
Organization for Standardization, Genève, Switzerland.
ISO 3888. ISO 3888 Passenenger Cars -- Test track for severe lane change manouvre -- Part 2: Obstacle
Avoidance. u.o. : International Organization for Standardization, Genève, Switzerland. ISO 3888.
ISO 4138. ISO 4138 Passenger cars – Steady-state circular driving behaviour - Open-loop test methods.
u.o. : International Organization for Standardization, Genève, Switzerland. ISO 4138.
ISO 7401. ISO 7401 Lateral transient response test methods - Open-loop test methods. u.o. :
International Organization for Standardization, Genève, Switzerland.
ISO 7975. 2006. ISO 7975 Passenger cars – Braking in a turn – Open-loop test method. u.o. :
International Organization for Standardization, Genève, Switzerland, 2006. ISO 7975.
ISO 8608. ISO 8608 Mechanical vibration - Road surface profiles - Reporting of measured data. u.o. :
ISO 8855. ISO 8855 – Road vehicle – Vehicle dynamics and road holding ability – Vocabulary. s.l. :
ISO. 2011. ISO 14794 Heavy commercial vehicles and buses - Braking in a turn - Open-loop test methods.
u.o. : ISO, 2011. ISO 14794.
—. 2006. Passenger cars – Braking in a turn – Open-loop test method. u.o. : ISO, 2006. ISO 7975.
Kati, Maliheh Sadeghi. 2013. Definitions of Performance Based Characteristics for Long Heavy Vehicle
Combinations. Signals and Systems. u.o. : Chalmers University of Technology, 2013. ISSN 1403-266x.
Kharrazi , Sogol. 2012. Steering Based Lateral Performance Control of Long Heavy Vehicle
Combinations. Göteborg, Sweden : Chalmers University of Technology, 2012. ISBN/ISSN: 978-91-7385-
724-6.
Kiencke, U. and Nielsen, L. 2005 . Automotive Control Systems. 2005 .
Lex, Cornelia. 2015. Estimation of the Maximum Coefficient of Friction between Tire and Road based on
Vehicle State measurements. Graz : Graz University of Technology, 2015. Doctoral thesis.
Michelin. 2003. The tyre Rolling resistance and fuel savings. u.o. : Michelin, 2003.
Mägi, Mart. 1988. The Significance of System Pre-Load at Modal Analysis of Low-Resonant Mechanical
Systems. Århus, Denmark : Modal Testing & FEM Seminar in Århus, 1988.
NHTSA. Electronic Stability Control Systems. u.o. : NHTSA. FMVSS 126.
Pacejka, H. 2005. Tyre and Vehicle Dynamics, 2nd ed. u.o. : Elsevier, 2005.
Ploechl, Manfred, [red.]. 2013. Road and Off-Road Vehicle System Dynamics Handbook. u.o. : CRC
Press, 2013. Print ISBN: 978-0-8493-3322-4; eBook ISBN: 978-1-4200-0490-8;
http://www.crcnetbase.com/isbn/9781420004908.
Rajamani, Rajesh. 2012. Vehicle Dynamics and Control. u.o. : Springer, 2012. ISSN 0941-5122.
Rhyne, T. B. 2005. Development of a Vertical Stiffness Relationship for Belted Radial Tires. u.o. : Tire
Science and Technology, TSTCA, Vol. 33, no. 3, July-September 2005, 2005. ss. pp. 136-155.
Rill, Georg. 2006. FIRST ORDER TIRE DYNAMICS. Lisbon, Portugal : III European Conference on
Computational Mechanics Solids, Structures and Coupled Problems in Engineering, 5–8 June 2006,
2006.
290
Robert Bosch GmbH. 2004. Bosch Automotive Handbook 6th Edition. s.l. : Bentley Publishes, 2004.
Ross, I. Michael. 2015. A Primer on Pontryagin's Principle in Optimal Control. 2015.
Rösth, Markus. 2007. Hydraulic Power Steering System Design in Road Vehicles: Analysis, Testing and
Enhanced Functionality. Linköping : Linköping Univeristy, Department of Management and
Engineering, Fluid and Mechatronic Systems, 2007. Dissertation. Dissertations, ISSN 0345-7524; 1068.
SAE_J3016. 2016. SAE J3016 Taxonomy and Definitions for Terms Related toDriving Automation
Systems for On-Road Motor Vehicles. u.o. : SAE International, 2016. Tech. rep. 2016.
SAEJ670. SAE J670e - Vehicle Dynamics Terminology. u.o. : Society of Automotive Engineers,
Warrendale, PA, USA.
Sundström, Peter, Jacobson, Bengt och Laine, Leo. 2014. Vectorized single-track model in Modelica
for articulated vehicles with arbitrary number of units and axles. Lund, Sweden : Modelica conference
2014, March 10-12, 2014, 2014.
Svendenius, Jacob. 2007. Tire Modeling and Friction Estimation. Lund, Sweden : Department of
Automatic Control, Lund University, 2007. Dissertation.
Tagesson, Kristoffer. 2017. Driver-centred Motion Control of Heavy Trucks. Gothenburg, Sweden :
Chalmers University of Technology, 2017. Dissertation.
Tagesson, Krsitoffer och Cole, David. 2017. Advanced emergency braking under split friction
conditions and the influence of a destabilising steering wheel torque. Vehicle System Dynamics. 2017,
Vol. 55.
Tiller, Michael. 2016 (continuously updated). Modelica by example. xogeny. [Online] 2016
(continuously updated). [Citat: den 6 January 2017.] http://book.xogeny.com/.
UN ECE 111. 2001. UNIFORM PROVISIONS CONCERNING THE APPROVAL OF TANK VEHICLES OF
CATEGORIES N AND O WITH REGARD TO ROLLOVER STABILITY. u.o. : UNITED NATIONS, 2001.
Regulation No. 111.
Wedlin, Johan, Tillback, Lars-Runo och Bane, Olof. 1992. Combining Properties for Driving Pleasure
and Driving Safety: A Challenge for the Chassis Engineer. u.o. : SAE, 1992. SAE Technical paper 921595.
Wipke, Keith B, Cuddy, Matthew R. och Burch, Steven D. 1999. ADVISOR 2.1: A User-Friendly
Advanced Powertrain Simulation Using a Combined Backward/Forward Approach. IEEE
TRANSACTIONS ON VEHICULAR TECHNOLOGY. 1999, Vol. 48, 6, ss. 1751-1761.
Wong, J.Y. 2001. Theory of Ground Vehicles (3rd ed.). s.l. : John Wiley and Sons, Inc., New York, 2001.
Yang, Derong. 2013. Vehicle Dynamics Control after Impacts in Multiple-Event Accidents. Göteborg :
Chalmers University of Technology, 2013. PhD thesis. ISBN 978-91-7385-887-8.
291
Index
(discrete) state diagram ..........................................21 counter-steer .............................................................177
Ackermann steering angle ................................... 176 Critical Speed .............................................................179
Ackermann steering geometry .......................... 155 data flow diagram ...................................................... 21
Active safety .................................................................42 Differential-Algebraic system of Equations,
DAE .................................................................................. 17
actual toe angle ...........................................................52
direct yaw moment .................................................183
actuation ........................................................................89
discrete dynamics ...................................................... 18
Advanced Driver Assistance Systems (ADAS) 42
effective damping ......................................... 138, 141
aerodynamic drag ......................................................96
effective stiffness .....................................................138
agility ............................................................................ 218
Electronic Brake Distribution, EBD ....... 149, 295
Algebraic loops ............................................................19
Electronic Control Units, ECUs ............................. 43
angular (time) frequency ..................................... 252
Electronic Stability Control, ESC............. 245, 295
Anti-lock Braking System, ABS ................ 148, 295
Engine Drag Torque Control, EDC ....................150
architecture ..................................................................43
Equivalent wheelbase ............................................168
Attributes .......................................................................11
flow charts .................................................................... 21
Automatic Emergency Brake, AEB ................... 150
Friction Circle .............................................................. 88
bicycle model ............................................................ 159
Functions ....................................................................... 12
bounce mode ............................................................. 269
Gain controller ............................................................ 32
brush model..................................................................64
handling diagram .....................................................188
Cambering Vehicle .....................................................47
heave mode ................................................................269
castor offset at ground .......................................... 157
High index ..................................................................... 19
Castor offset at wheel center .............................. 157
higher index problems............................................. 38
Characteristic Speed .............................................. 179
Human Filter Function ..........................................273
Closed loop controller...........................................31
hybrid dynamics .......................................................... 19
Closed-loop Manoeuvres .........................................40
ideal tracking .............................................................160
coefficient of rolling resistance ............................60
indirect yaw moment .............................................184
combined (tyre) slip..................................................88
inertial coordinate system ..................................... 38
Continuous Variable ratio Transmission, CVT
......................................................................................... 117 internal combustion engine (ICE).....................104
Control allocation, CA ...............................................33 Inverse model controller ........................................ 32
Controller Area Network, CAN..............................43 kinematic models .....................................................160
Cornering .................................................................... 153 Kingpin axis ................................................................157
cornering coefficient .................................................84 Lane Departure Prevention ..................................244
Cornering Coefficient ................................................68 Lane Keeping Aid, LKA ..........................................244
Cornering Reistance Coefficient ........................ 115 Lateral Collision Avoidance, LCA.......................247
292
lateral slip ......................................................................78 Rigid ring tyre model ............................................... 75
Leaning vehicle............................................................47 roll axis .........................................................................197
longitudinal (tyre) slip .............................................58 Roll Stability Control, RSC ......................... 247, 295
longitudinal cornering coefficient .......................68 rolling resistance ....................................................... 60
low speed steering angle ...................................... 176 Roll-stiff Vehicle ......................................................... 47
Magic Formula .............................................................73 Scrub radius ................................................................157
Mean Square Spectral Density ........................... 254 Secondary Ride ........................................................272
Model Predictive Control, MPC .............................33 side slip .......................................................................... 78
momentaneous control .............................................31 Simultaneous equations .......................................... 19
neutral steered vehicle.......................................... 176 single-track model ...................................................159
Normal steering axis offset at wheel centre ... 157 slip angle........................................................................ 78
object vehicle ..............................................................42 spectral density ........................................................254
Off-tracking ...................................................... 200, 243 sprung mass ...............................................................137
one-track model ....................................................... 159 static toe angle ............................................................ 52
Open loop controller..............................................31 Steering axis ...............................................................157
Open-loop Manoeuvres............................................40 steering axis offset at ground .............................157
Ordinary Differential Equations, ODE................36 subject vehicle ........................................................... 42
oversteered vehicle ................................................ 176 Traction Control, TC ...............................................149
overturning moment.................................................86 Traction diagram .....................................................113
parallell steering geometry ................................. 155 tyre slip .......................................................................... 58
pendulum effect ....................................................... 205 understeer gradient ..................................... 173, 175
phase portrait ........................................................... 241 understeered vehicle..............................................176
physical stiffness ..................................................... 138 unsprung mass ..........................................................137
platform..........................................................................43 Weighted RMS Acceleration ................................274
pneumatic trail ............................................................86 wheel base filtering.................................................283
Power Spectral Density, PSD .............................. 254 wheel hop mode .......................................................269
predictive control ........................................................31 wheel spring rate ......................................................138
Primary Ride ............................................................ 272 yaw damping .............................................................243
Rearward Amplification, RWA ........................... 242
293
6 Appendix: References for
Lectures
6.1 Course MMF062 Vehicle Dynamics
Table 6.1: Lectures and recommended reading in course “MMF062 Vehicle Dynamics”.
Lecture (≈45 min) Recommended pre-studies in compendium
Subject Vehicle Dynamics

• Browse: 1.1 Definition of Vehicle Dynamics - 1.5.1 Engineering
1:1 in its Context and Course
• Course memo
administration
• 1.5.8 Coordinate Systems
• 1.5.9 Terms with special meaning
Vehicle types,
1:2 • Browse: 1.5.10 Architectures
Terminology
• 1.6 Heavy trucks
• Browse: 1.7 Smaller vehicles
• 1.5.2 Modelling, drawing and mathematics
1:3 Modelling
• 1.5.3 Mechanical engineering
• 1.5.5 Verification methods with real vehicle
• 1.5.6 Verification methods with virtual vehicle
1:4 Virtual verification
• 1.5.7 (Computer) Tools & Methods
• 1.5.10.3 Virtual vehicle architecture
• Browse: 2.2 Wheel & Tyre Terminology & Design
2:1 Tyre Design and Rolling • 2.3.1 Tyre Rolling and Radii
• 2.3.2 Rolling Resistance of Tyres
• 2.3.3.1.1 Uniform pressure distribution and known contact length
• 2.3.3.1.2 Longitudinal tyre (slip) stiffness
• 2.3.3.1.5 Brush model with parabolic pressure distribution
2:2 Tyre Longitudinal
• 2.3.3.2 Empirical tyre models
• 2.3.3.3 Tyre with both rolling resistance and slip
• Browse: 2.3.3.4.2 Transients due to Relaxation in Contact patch modelled as filter
• 2.4.1 Tyre brush model for lateral slip
• 2.4.2 Lateral tyre (slip) stiffness
Tyre Lateral and
2:3 • Browse: 2.4.6 Other lateral effects
Combined
• 2.5.1 Isotropic combined slip, any 𝑭=𝒇(𝒔);
• 2.5.2 Isotropic brush model for combined slip
• 2.6 Vertical Properties of Tyres
Tyre vertical and other
2:4 • Browse: 2.8 Vehicle Aerodynamics
vehicle interactions
• Browse: 2.9 Driving and Vehicle Operation
• 3.2.1 Propulsion System
• 3.2.1.1 Prime movers
• 3.2.1.2 Transmissions
• 3.2.2 Traction diagram
• 3.2.3 Driving Resistance
3:1 Longitudinal Steady State
• 3.2.4 Functions after start
• 3.2.5 Starting with slipping clutch
• 3.2.6 Steady state vertical force distribution over axles
• 3.2.7 Friction limit
• 3.2.8 Start functions
• Browse: 3.3.1 Description formats of Vehicle Operation
• 3.3.2 Rotating inertia effects
3:2 Longitudinal Long Cycles • 3.3.4 Functions over cycles
• 3.3.5 Load Transfer with rigid suspension
• 3.3.6 Acceleration
294
• Browse: 3.4.1 Typical test manoeuvres

• 3.4.2 Deceleration performance
3:3 Longitudinal Short Events • 3.4.3 Brake System
• Browse: 3.4.4 Pedal functions
• 3.4.5 Brake proportioning
• 3.4.6 Steady state heave and pitch
• 3.4.7 Load Transfer with suspension compliance and linkage
Longitudinal Pitch
3:4 • 3.4.8 Pitch functions at quick force changes
Dynamics
• Browse: 3.4.8.3 Anti-dive and Anti-squat designs
• 3.4.9 Acceleration and Deceleration
Longitudinal Control
3:5 • Browse: 3.5 Control functions
Functions
Longitudinal Heavy
3:6 • 3.4.3 Brake System
Vehicles Brake System
• Browse: 4.2.1 Path with orientation
Transient one-track
• 4.2.2 Vehicle and wheel orientations
4:1 model, two-axle, low CoG,
• 4.2.4 One-track models
good grip
• 4.4.2 Transient one-track model
• Browse: 4.2.3 Steering System
• 4.2.3.1 Chassis steering geometry
Lateral Low Speed
4:2 • 4.2.5 Ideally tracking wheels and axles
Ackerman
• 4.2.6 One-track model for low speeds, with Ackermann geometry
• 4.2.7 Low speed functions
Lateral Low Speed Non- • 4.2.8 One-track model for low speeds, with non-Ackerman geometry
4:3
Ackerman • 4.2.9 One-track model for low speeds, with trailer
• 4.3 Steady state cornering at high speed
Steady State Cornering
4:4 • 4.3.1 Steady state driving manoeuvres
without load transfer
• 4.3.2 Steady state one-track model
• 4.3.3 Under-, Neutral- and Over-steering *
• 4.3.4 Normalized required steering angle
Steady State Cornering • 4.3.5 How to design for understeer gradient
4:5
functions • 4.3.6 Steady state cornering gains *
• Browse: 4.3.7 Manoeuvrability and Stability
• Browse: 4.3.8 Handling diagram
Vehicle Simulator as
4:6 • 1.5.6.1.1 Simulator = Driver in the loop simulation
verification tool
Steady State Cornering
• 4.3.9.1 Load transfer between vehicle sides
4:7 with Load Transfer
• 4.3.9.2 Body heave and roll due to lateral wheel forces
without Roll Centre
• 4.3.9.3 Lateral load transfer
Steady State Cornering • Browse: 4.3.9.4 Axle suspension system
4:8
with Roll Centre • 4.3.10 High speed steady state vehicle functions
• Browse: 4.3.11 Roll-over in steady state cornering
• 3.5.2.3 Anti-lock Braking System, ABS *
Heavy vehicle, ABS, ESC, • 3.5.2.4 Electronic Brake Distribution, EBD *
4:9
RSC • 4.6.2.2 Electronic Stability Control, ESC *
• 4.6.2.3 Roll Stability Control, RSC *
• 4.4.1 Stationary oscillating steering tests
Stationary oscillating
4:10 • Repetition: 4.4.2 Transient one-track model
steering
• 4.4.3 Steering frequency response gains *
• Browse: 4.5.1 Transient driving manoeuvres *
• 4.5.2 Transient one-track model, without lateral load transfer
Transient lateral • Browse: 4.5.2.2 Transient models for articulated vehicles
4:11
manoeuvres, arbitrary • Browse: 4.5.2.3 Cambering vehicle model
• 4.5.3 Transient two-track model, with Lateral load transfer
• 4.5.3.2 Example of explicit form model; two-axle vehicle, driver and environment
• Browse: 4.5.3.3 Transient roll-over *
4:12 Step steering • 4.5.4 Step steering response *
• Browse: 4.5.6 Long heavy combination vehicles
4:13 Lateral Control Functions • 4.6 Lateral Control Functions
295
• Browse: 5.1 Introduction

Vertical Stationary
5:1 • Browse: 5.2 Suspension System
oscillation theory
• 5.3 Stationary oscillations theory
• Repetition: 3.4.7 Load Transfer with suspension compliance and linkage
• Repetition: 3.4.8 Pitch functions at quick force changes
5:2 Heavy Vehicle Suspension • Repetition: 4.3.9.3 Lateral load transfer
• Repetition: 4.3.9.4 Axle suspension system
• 5.2 Suspension System
Road Models, 1D Vehicle • 5.4 Road models
5:3
Models • 5.5 One-dimensional vehicle models
Comfort, Grip, Fatigue, • 5.6 Functions for stationary oscillations
5:4
2D, 3D • Browse: 5.7 Variation of suspension design
Vertical multi-
• 5.8 Two dimensional oscillations
5:5 dimensional and
• Browse: 5.9 Transient vertical dynamics
transient
296

Vehicle Dynamics - 128709302 PDF

Uploaded by

Copyright:

Available Formats

Vehicle Dynamics - 128709302 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vehicle Dynamics - 128709302 PDF

Uploaded by

Copyright:

Available Formats

What is the purpose and scope of the document?

What is the purpose and scope of the document?

What topics are covered in the document?

What topics are covered in the document?

3 LONGITUDINAL DYNAMICS 103

4 LATERAL DYNAMICS 153

5 VERTICAL DYNAMICS 249

6 APPENDIX: REFERENCES FOR LECTURES 294

1.2 About this compendium

1.3 Automotive engineering

1.4 Attributes and Functions

Decompose in subsystem. Design loops per

1.4.2.1 Organise functionality in functions

1.4.4 Models, methods and tools

1.5 Technical References

Engineering design (synthesis)

Requirements Test the design, with Analysis with a model:

1.5.1.1 Vehicle Dynamics role in industry

Figure 1-4: An example of vehicle programme.

1.5.2 Modelling, drawing and mathematics

Figure 1-5: Modelling and Real versus Theoretical world.

1.5.2.2 Discrete dynamics

1.5.2.3 Simultaneous equations or Algebraic loops

3 equations, 3 unknowns ( ): Solved by symbolic manipulation of equations:

Figure 1-6: An example of Simultaneous equations, also called Algebraic loop.

1.5.2.4 High index

High index No working causality:

3 equations, 3 unknowns ( ): Solved by symbolic manipulations:

1.5.2.5 Constants, Parameters, Variables and Signals

Chose of state variables

1.5.2.8 Notation conventions

1.5.2.9 Notation list

1.5.3 Mechanical engineering

1.5.3.1 Free-body diagrams

Table 1.2: Notation

1. General Basic Gravity g g m/s 2 9.80665 average on Earth

1. General Basic Height h h H

2 Subscript can be given to

1. General Basic Moment (of forces) M M T Nm Not for shaft torque

1. General Dynamics Output variables in a dynamic system y y <various>

1. General Dynamics Power spectral density G F PSD

1. General Dynamics States variables in a dynamic system x x <various>

The independent variable

2. Vehicle 1. General Wheel radius R R r, Rw m

Sign convention might be

The angle required for a

1.5.3.2 Operating conditions

1.5.3.3 Equation types

m*acceleration air resistance

acceleration selecting the free bodies

air resistance m*g 𝑭𝒇𝒛𝟎 𝑭𝒇𝒛𝟎

Equilibrium for “unsprung axles”:

1.5.4 Control engineering

Vehicle variables (positions, velocities, forces, etc.) and

1.5.4.1 Decision making or Execution

1.5.4.2 Momentaneous or Predictive control

1.5.4.3 Open or closed loop control

1.5.4.4 Degree of over-actuation

Unladen radius R0 = (Dr25.4+2(H/S)*S)/2