SDV Ebook

Co
m
pl
im
en
ts
of
The
Software-Defined
Vehicle
ADDigital-First
igital-F
First A
Approach
pproach to Cre
Next-Generation
N Experienc
extt Generattion E xperiienc
Dirk Slama
Slama,
Achim Nonnenmacher
& Thomas Irawan
REPORT
The Software-Defined
Vehicle
A Digital-First Approach to Creating
Next-Generation Experiences
by Dirk Slama, Achim Nonnenmacher,

and Thomas Irawan
Beijing Boston Farnham Sebastopol Tokyo

The Software-Defined Vehicle: A Digital-First Approach to Creating
Next-Generation Experiences
by Dirk Slama, Achim Nonnenmacher, and Thomas Irawan
Copyright © 2023 Dirk Slama and Achim Nonnenmacher. All rights reserved.
Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA
95472.
O’Reilly books may be purchased for educational, business, or sales promotional
use. Online editions are also available for most titles (http://oreilly.com). For more
information, contact our corporate/institutional sales department: 800-998-9938 or
[email protected].
Acquisitions Editor: Suzanne McQuade Interior Designer: David Futato

Development Editor: Gary O’Brien Cover Designer: Karen Montgomery
Production Editor: Aleeya Rahman Illustrator: Kate Dullea
Copyeditor: Paula L. Fleming
September 2023: First Edition
Revision History for the First Edition

2023-09-13: First Release
The O’Reilly logo is a registered trademark of O’Reilly Media, Inc. The Software-
Defined Vehicle, the cover image, and related trade dress are trademarks of O’Reilly
Media, Inc.
The views expressed in this work are those of the authors and do not represent the
publisher’s views. While the publisher and the authors have used good faith efforts
to ensure that the information and instructions contained in this work are accurate,
the publisher and the authors disclaim all responsibility for errors or omissions,
including without limitation responsibility for damages resulting from the use of
or reliance on this work. Use of the information and instructions contained in this
work is at your own risk. If any code samples or other technology this work contains
or describes is subject to open source licenses or the intellectual property rights of
others, it is your responsibility to ensure that your use thereof complies with such
licenses and/or rights.
This work is part of a collaboration between O’Reilly and digital.auto. See our
statement of editorial independence.
978-1-098-15778-4
[LSI]
Table of Contents
Preface and Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Goals of the SDV 2
Impediments: Why Is Automotive Software Development
Different? 6
2. What the Automotive Industry Can Learn from Other Industries. . . 15

Learning from the Smartphone Folks: Standardization,
Hardware Abstraction, and App Stores 15
Learning from the Internet Folks: Open Source and Cloud-
Native Development 18
3. Vehicle OS and Enabling Technologies. . . . . . . . . . . . . . . . . . . . . . . . . 21

Foundation: E/E Architecture 22
Vehicle APIs 22
Vehicle SOA: Encapsulating the E/E Architecture with
Vehicle APIs 24
Layers of the Vehicle SOA 25
Ensuring Functional Safety in the Vehicle SOA 26
SDV Tech Stack for the Vehicle SOA 28
OTA: Over-the-Air Updates 30
The Vehicle App Store 31
SDV and AI 32
v
4. Value Stream Management for the SDV. . . . . . . . . . . . . . . . . . . . . . . . 35
Working at Different Speeds 35
Divide and Conquer 37
Enterprise Perspective 38
5. #digitalfirst: A New Way of Working. . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Shift North 42
Shift Left 43
Virtualization 43
Summary 44
6. Next Steps. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
vi | Table of Contents
Preface and Acknowledgments
This publication is the result of the intensive work we are doing

for the digital.auto initiative, which brings together OEMs and sup‐
pliers, industry consortia members, and open source enthusiasts to
help our industry make the software-defined vehicle (SDV) a reality.
The digital.auto initiative is open to everyone, and it is vendor and
technology agnostic. We seek to advance the adoption of SDVs by
focusing on use cases and how to realize them using state-of-the-art
technologies and methodologies. Please refer to Chapter 6, “Next
Steps” if this endeavor is of interest to you.
This publication would not have been possible without support and
feedback from several people. We would like to extend a special
thank-you to the contributors of the original SDV 101 course:
• Achim Henkel, Director, Robert Bosch Group

• Alex Oyler, Director, SBD Automotive
• Alexander Djordjevic, Director of Solution Management at
RideCare, Robert Bosch GmbH
• Andre Marquis, Senior Fellow/Lecturer, UC Berkeley, Haas
School of Business, Entrepreneurship Program
• Ansgar Lindwedel, Director of SDV Ecosystem Development,
ETAS GmbH
• Daniel Riexinger, Senior Manager AI & Data-Driven Business
Models, Mercedes-Benz AG
vii
• Dominik Rose, VP Product Management & Platform Strategy,
LeanIX GmbH
• Frédéric Merceron, Transportation & Mobility Solutions Direc‐
tor, Dassault Systèmes
• Georg Hansbauer, CEO, Testbirds GmbH
• Jacek Marczyk, CEO, Ontonix
• Jann Kirchhoff, Product Success Manager, BMW
• Pei Shen, General Manager of Strategy for Tencent Intelligent
Mobility, Tencent Inc.
• Sasha Milinkovic, Manager, mm1 Consulting GmbH
• Sebastian Werner, Head of Automotive Software, Kearney &
BinaryCore
• Sven Kappel, VP Software-Defined Vehicles, ETAS GmbH
• Tom Acland, CEO, Dassault Systèmes 3DEXCITE
In addition, we would like to thank Bosch senior management for

their support of digital.auto, in particular Tanja Rückert (CDO,
Bosch group), Andreas Dempf (CSO/CMO, Bosch Mobility) and
Mathias Pillin (CTO, Bosch Mobility), as well as the leadership of
the Ferdinand-Steinbeis-Institute: Prof. Heiner Lasi and Michael
Köhnlein. Chris Cheng and the team from the Ferdinand-Steinbeis-
Institute AIoT Lab have earned our thanks for their relentless sup‐
port of digital.auto and a variety of use cases (this research was
funded by Dieter Schwarz Stiftung); Christian Seiler from Mercedes-
Benz for intense discussions about software-defined vehicles and
variant management; Damian Dyrbusch from Bosch for strategic
vision; Can Yasin Gümüş from Bosch for his input on SDV and
ADAS; Hoàng Phan Thanh, Nhân Lương Nguyen, Phong Pham
Tuan, and Tam Thai Hoang Minh from Bosch Vietnam for their
work on the digital.auto playground; Oliver Kust for his expert guid‐
ance on automotive safety; Marco Wagner for discussions around
SDV and vehicle abstraction; Imran Abdul Rahiman from ETAS (a
Bosch company) for intense discussions about AUTomotive Open
System ARchitecture (AUTOSAR); Louis Jackman for the initial
version of the digital.auto playground; Mohan B V and Babu Niran‐
jan from Bosch Global Software Technologies for their guidance
on the overall structure of this book; Grace Lee from Landing.AI
for showing us how AI can be democratized; Boris Scholl from
Microsoft for his inputs on Cloud Native; Francesco Salamida and
viii | Preface and Acknowledgments

Stefano Marzani from AWS for discussions around innovations in
the SDV space; Wieland Holfelder from Google for his input on
the digital.auto playground; and, finally, to our families for having
our backs while we were putting in the extra hours to work on this
project—thank you!
— Dr. Dirk Slama

Vice President at Bosch
Professor at Ferdinand-Steinbeis-Institute
— Dr. Achim Nonnenmacher

Senior Manager, SDV Innovations at
ETAS (a Bosch company)
— Dr. Thomas Irawan

CEO at ETAS (a Bosch company)
Berlin/Stuttgart, August 2023
Here are some key terms that are found throughout this booklet:
• API: Application Programming Interface

• AUTOSAR: AUTomotive Open System ARchitecture
• CAN: Controller Area Network
• COVESA: Connected Vehicle Systems Alliance
• CI/CD: Continuous integration and delivery
• E/E: Electrical/electronic vehicle architecture
• ECU: Electronic Control Unit
• OEM: Original Equipment Manufacturer
• SOA: Service-oriented architecture
• SOP: Start of Production
• QoS: Quality of Service
• vECU: Virtual Electronic Control Unit
• VSS: Vehicle Signal Specification (from COVESA)
Preface and Acknowledgments | ix

CHAPTER 1
Introduction
In the past, the automotive industry stood as a testament to the

power of combustion engines and the prestige of owning a car with
the “most exhaust pipes.” Today, this old-school paradigm is under‐
going a seismic transformation. Four major innovations—electrifi‐
cation, automation, shared mobility, and connected mobility—are
happening all at once, leading to dramatic changes in the automo‐
bile landscape. Further amplifying this complexity are industry new‐
comers who are not just building top-quality cars but are specifically
focusing on the digital-savvy consumer with their “smartphones on
wheels”—vehicles equipped with large interactive screens, seamless
connectivity, and frequent updates to software features.
At the same time, customers are expressing dissatisfaction because
their vehicles lack the features and user-friendliness that are com‐
monplace in smartphones. Many are left wondering: why can’t my
$50,000 car perform the same tasks as my $300 smartphone?
From this frustration emerged the idea of a software-defined vehi‐
cle (SDV), a car that’s fully programmable. New features can be
developed and deployed within a matter of months, not years, and
there’s extra computational capacity for future updates that can be
delivered wirelessly. This is a car that keeps getting better, thanks
to the continuous delivery of valuable new software features and
updates throughout its lifetime.
However, the automotive industry has hurdles to overcome if it
wants to achieve this goal quickly. The complexity of software
is spiraling out of control, with the volume of code reaching an
1
estimated 300 million lines with level 5 autonomous driving, all
while developers cling to the software development methodologies
of yesteryear. Further, hardware and software are still tightly inte‐
grated and released simultaneously, meaning significant changes
occur only once every seven years or so.
Consider the mobile phone industry, where former leading brands
like Nokia and BlackBerry were swiftly outpaced by smartphones.
A similar fate for the automotive industry’s incumbents seems pos‐
sible. At first, many veterans of the cell phone industry insisted
that their business was fundamentally different from the comput‐
ing industry, but the smartphone revolution quickly shattered that
illusion.
This shift raises several critical questions: How can the incumbents
in the automotive industry succeed in their ongoing digital transfor‐
mation to avoid a similar fate? What makes change so challenging
within this industry, and how can we significantly accelerate auto‐
motive innovation? What will the future of revenue generation look
like? Will revenue still come predominantly from car sales, or will
digital services layered on top become the primary revenue driver?
And crucially, how do we pinpoint the “killer apps”—those standout
applications or digital features that prove so essential or attractive
that they drive the success of future vehicle generations?
While the SDV is a key technical enabler, the path to answering
these questions starts with adopting a “digital first” strategy. This
means starting with the digital customer experience and working
backward to the solution designs, engaging in early exploration,
and testing to ensure that the vehicles on the road align with the
ever-evolving needs of the modern consumer.
Goals of the SDV

Let us start our discussion by looking at why we want to build an
SDV in the first place. It isn’t about the bits and bytes or the number
of sensors and cameras—it’s about the customer experiences we
build. And these customer experiences cannot aim simply to rebuild
the smartphone experience. This is about creating a “habitat on
wheels” powered by cross-domain applications and data fusion and
delivered and constantly improved at 10 times the speed we see
today.
2 | Chapter 1: Introduction
Smartphone Habitat on Wheels
During the past decade, many car manufacturers have sought to
replicate successful smartphone applications in their cars. In many
cases, however, these in-vehicle applications could not match the
quality of the smartphone apps. In addition, consumers usually
don’t want redundant experiences, inconsistencies between their
digital ecosystems, or the irritation of cumbersome data synchroni‐
zation. Today, car makers have to walk a fine line between applica‐
tion and data ownership on the one hand and customer experience
and tighter integration with the dominant smartphone ecosystems
on the other.
This is why it is important that the SDV surpass the concept of
a “smartphone on wheels.” Instead, it has to enable a “habitat on
wheels,” utilizing the specifics of the car to provide multisensory
experiences that a smartphone could never match. With multiple
displays and a network of hundreds of sensors and actuators, the
SDV brings together domains like infotainment, autonomous driv‐
ing, intelligent body, cabin and comfort, energy, and connected car
services, crafting a unique journey.
Passengers feel recognized as they enter a vehicle personalized to
their needs, one that is a clear departure from the impersonal con‐
fines of traditional cars. Passengers experience a reassuring sense
of security while journeying under the vigilant care of cutting-edge
safety systems. Passengers step into the vehicle with a green con‐
science. Passengers reclaim their time, transforming travel into an
opportunity for meaningful engagement, whether through work,
rest, or play. Passengers with disabilities can enjoy a newfound free‐
dom and mobility thanks to autonomous driving. This technology
allows them to move with confidence, their independence undimin‐
ished. It’s not just about the journey; it’s about the freedom to
explore, to engage, and to live life unrestricted.
SDVs have revolutionized our perception of mobility. It’s no longer
merely about getting from point A to point B but about making
the journey itself enriching. Thanks to advanced driver-assistance
systems and autonomous driving, we’re embracing the transforma‐
tive, multisensory power of the SDV.
Goals of the SDV | 3

In the SDV, the digital and physical worlds merge, creating experi‐
ences that are both immersive and intuitive. This fusion allows the
car to interpret and react to physical cues with precision and sophis‐
tication, enhancing the overall experience. For instance, the vehicle’s
sensors could identify the state of health of children in the back seat
and estimate potential motion sickness based on road curvature and
speed. Responding to these subtle cues, the system could adapt the
automated driving style, adjust window positions, and regulate the
air conditioning to enhance comfort and minimize motion sickness.
Moreover, value-added digital features transform the vehicle into a
hub of productivity and entertainment. Imagine video conferences
displayed on the front car window as holograms, engaging in inter‐
active games with other drivers during traffic jams, or learning a
new language based on your surroundings during your commute—
all within the safety and comfort of your personal, mobile space.
This seamless blend of the digital and physical worlds, facilitated by
the SDV, ensures that every journey is not just a ride but an enrich‐
ing experience that adds value far beyond arrival at your destination.
Cross-Domain Applications and Data Fusion

Today, vehicle experiences often occur in isolated domains. But
the future with SDVs promises to blur the boundary between the
vehicle and the outside world. Experiences will be cross-domain,
where various vehicle functions and systems intercommunicate and
interact harmoniously to enrich the overall journey. As Figure 1-1
shows, SDVs represent a shift from disparate functions to integrated
experiences, where the vehicle works as a unified entity rather than a
collection of separate parts.
Consider the example of a digital “dog mode” that some cars already
feature. The vehicle monitors your dog in the car while you are
out shopping. Because it is hard for dogs to cool down, a hot car
interior on a summer’s day is often enough to cause serious injuries
or even death. This is a perfect illustration of customer-centric and
cross-domain functionality. It involves multiple systems: the car’s air
conditioning to maintain a comfortable temperature, the infotain‐
ment screen to display a message letting passers-by know not to
worry as the dog is safe and comfy, and the battery management
system to ensure the car has sufficient energy. All these domains are
coordinated in order to ensure the dog’s safety and comfort.
Figure 1-1. Cross-domain application services
In this connected ecosystem, your car could even become a creative

extension of your social media presence. With your permission, it
could capture a stunning sunset through its high-quality on-board
cameras during a scenic drive and propose a pre-edited post for
your approval. Your vehicle becomes an active participant in your
digital life, enabling you to share unique moments without detract‐
ing from the driving experience.
Envision your electric vehicle communicating with a smart grid,
scheduling its charging during off-peak electricity usage. This inter‐
action optimizes energy use and maximizes cost efficiency. Your
car is no longer just a consumer of energy but also an intelligent
participant in the wider energy ecosystem.
Cross-domain experiences also extend to personal wellness. Imagine
that your fitness wearable signals that you’ve had an intense work‐
out. In response, your car sets the cabin temperature to a cooler
setting, selects soothing illumination for the ambient lighting, and
plays your favorite cool-down playlist. By seamlessly integrating
with your digital devices, your car enhances your post-workout
recovery and comfort.
This fusion of digital and physical experiences makes the car an
extension of your digital life, connecting you seamlessly to the world
around you.
Goals of the SDV | 5

Accelerate Innovation by 10x
To achieve the customer-centric development and continuous
improvement required to support the usability goals outlined above,
development of SDVs must significantly speed up, shortening the
time to market for new features. Figure 1-2 shows the transforma‐
tion that must occur.
Figure 1-2. “Need for speed” in automotive software development
The goal here should be a tenfold improvement over the old ways
of delivering new features. Development costs must be significantly
reduced. Initial prototypes must be available in hours, not months.
Time to market must be reduced from years to a few weeks.
And all of this does not only apply to new features: we must be able
to constantly monitor how customers are using existing features,
learn how to improve them, and then optimize them—again, all
done 10 times faster than we do it today.
Impediments: Why Is Automotive Software

Development Different?
To achieve our goals for the SDV, we have to understand the
impediments we need to overcome. Why is automotive software
development different from development in other domains (e.g.,
smartphones)? Sure, “smartphone on wheels” is an accessible met‐
aphor that illustrates the evolving nature of SDVs. With the rise
of touchscreens, robust connectivity, and a host of app-driven func‐
tions in our cars, drawing a parallel with our handheld smart devices
seems apt. However, the simplistic charm of this metaphor can be
misleading, as it does not account for the intricacies and unique
challenges of the automotive domain.
The following discussion provides a deep dive into the key impedi‐
ments to the broad adoption of SDV in the automotive industry,
including complexity and heterogeneity, functional safety, the “clash
of two worlds,” and the need for organizational transformation.
Complexity and Heterogeneity

A key difference between the automotive industry and the smart‐
phone industry is their levels of complexity. Over the past dec‐
ade, the smartphone industry has managed to address technical
complexity through standardization and abstraction. Today’s smart‐
phones are highly integrated with multiple layers of abstractions
and interfaces. The automotive industry, on the other hand, is still
plagued by high levels of complexity and heterogeneity.
Furthermore, high levels of abstractions have enabled the smart‐
phone industry to encapsulate most of the hardware complexity,
allowing it to deal with complexity on the software level. On the
other hand, today’s vehicles are still complex systems of systems.
Each subsystem in a car, from the brakes to the powertrain, is a
complex entity, supplied by a different vendor and integrated with a
unique software architecture. The level of complexity and the need
for seamless interoperability among systems far surpasses what we
see in today’s smartphones.
Another key difference is that the automotive industry is dealing
with many more product variants. Many cars are configured accord‐
ing to individual customer preferences expressed during the sales
process. Regional differences in customer preferences and regula‐
tory requirements lead to large numbers of product variants. And
different functional requirements—think family van versus conver‐
tible versus truck—lead to many different vehicle types, models, edi‐
tions, and variants. Thus, there are usually much lower production
numbers per vehicle model than per smartphone model.
Impediments: Why Is Automotive Software Development Different? | 7

Functional Safety
The “smartphone on wheels” comparison also falls short when we
consider automotive functional safety requirements. Unlike smart‐
phones, vehicles are subject to stringent safety standards, such as
ISO 26262. This standard deals with the functional safety of electri‐
cal and electronic systems within vehicles and is fundamental to the
concept of an SDV.
The ISO 26262 standard doesn’t just offer a set of rules but embodies
a risk-based philosophy that assesses potential hazards and requires
appropriate design techniques to mitigate them. Risks are catego‐
rized according to three factors: severity (the potential harm that
can occur), exposure (the probable frequency of the risk situation),
and controllability (the ability of the driver to prevent the hazard).
By evaluating these factors, the standard effectively balances innova‐
tion and safety in automotive design.
To quantify the risk, the standard employs a framework known as
Automotive Safety Integrity Levels (ASIL), illustrated in Figure 1-3,
which classifies hazardous events that could result from a malfunc‐
tion based on their level of severity, exposure, and controllability.
Levels of risk range from ASIL A, the lowest level, to ASIL D,
the highest level. ISO 26262 defines the requirements and safety
measures to be applied at each ASIL.
Figure 1-3. Automotive safety levels
Additionally, a class called QM (Quality Managed) is assigned to

systems that, after thorough analysis, require only standard quality
management processes due to their lower potential impact on safety.
For instance, a safety-relevant function like a braking system would
be assigned a high ASIL rating, and an in-car entertainment system
would be assigned a QM rating. The software controlling the brakes
would need adhere to more rigorous safety standards than the
entertainment system. The overarching objective in vehicle design
is risk mitigation, a process that seeks to reduce potential hazards.
For example, the risk of a false speed signal due to a defective sensor
can be mitigated by introducing a redundant sensor. This additional
sensor can cross-validate signals, thus minimizing the chances of
error and enhancing the overall safety of the vehicle.
Therefore, while the “smartphone on wheels” analogy succinctly
portrays the emerging role of software in vehicles, it does not wholly
capture the stringent safety standards and rigorous risk mitigation
strategies employed in the automotive industry. The SDV is more
than a simple mobile device; it’s a sophisticated ensemble of systems
that prioritizes safety as much as functionality and convenience.
Clash of Two Worlds

Automotive engineering has traditionally focused on physical func‐
tionality, from early electronic features (such as airbags, vehicle
stabilization, and braking systems) to modern driver assistance sys‐
tems or even automated driving. The new, digital experience is
driven by software-enabled features, and the focus is on improving
the experience of using the vehicle or even providing value-added
digital features like usage-based insurance or automated payment of
parking fees. Figure 1-4 shows some of these differences.
The traditional world of engineering physical vehicle features is
characterized by a complex system of systems—conforming to func‐
tional safety, security, and real-time requirements—and homologa‐
tion, a process of obtaining government-approved certification for
product market readiness to ensure that safety and environmental
standards are met. In contrast, the new digital vehicle experience is
built on best practices from the technology sector, including Agile
methods and cloud-style development.

Figure 1-4. Digital versus physical experience
Modern vehicles are a fusion of these worlds; they must merge

the reliability of traditional engineering with the agility of modern
software development (see Figure 1-5). This presents the automotive
industry with a significant challenge: OEMs that are best able to
address this “clash of two worlds” are most likely to succeed.
Figure 1-5. Clash of two worlds
Organizational Transformation
We can observe that over the last decade, the incumbent OEMs
have undergone complex and large-scale organizational transforma‐
tion to deliver both physical and digital car features. It is vital
that OEMs enable their organizations to combine the worlds of
traditional engineering and modern software development and have
them work together in relative harmony. But this means that they
must create organizations with suborganizations that can move at
different speeds and work with different cultures, methods, and
tools. It is not trivial to create vehicle platforms and architectures
that are so modular that components with different requirements
(e.g., functional safety) can be assigned to the organizational units
that are the best fit.
This doesn’t stop with the company’s internal organization. Tradi‐
tionally, many OEMs purchased different subsystems from different
vendors, usually combining hardware and software for each subsys‐
tem. With the SDVs, however, the decoupling of hardware and

software must be considered with respect to sourcing. Established
supply chains are undergoing significant transformations. The roles
of existing vendors are changing, and new vendors are entering the
playing field.
Rethinking the Vehicle Lifecycle: Digital First

What can we learn from the discussion of impediments that can be
applied to the vehicle lifecycle? Historically, the lifecycle of a vehicle
was defined by the simultaneous production and deployment of
tightly coupled hardware and software. Once the vehicle was in the
consumer’s hands, its features remained essentially unaltered until
its end of life. However, an SDV paradigm allows for the decoupling
of hardware and software release dates—a prerequisite for a digital
first approach, which puts design and virtual validation of the digital
vehicle experience at the start of the lifecycle.
Digital first means that new ideas for the vehicle experience are ini‐
tially explored in virtual environments, getting early user feedback
long before any custom hardware must be developed or even a phys‐
ical test vehicle is provided. Digital first means that the principles
of design thinking and Lean startup, which originated in internet
culture, can now be applied in the physical world of automotive
development.
In the SDV model, new ideas are generated with approaches like
design thinking, where features are explored virtually and tested
with early customer feedback, facilitating a faster and more cost-
effective way to find a product–market fit. Product–market fit refers
to a state in which a product meets the specific needs and desires
of its target market—in this case for new vehicle applications and
services. The ability to rapidly explore, innovate, and adapt is a
cornerstone of the new automotive era.
A lot of this relates to our smartphone discussion. Just as the launch
of the iPhone opened doors to applications beyond our wildest
dreams—identifying plant species, optimizing sleep quality, study‐
ing for a driver’s license—the advent of the SDV propels us into a
future of unimagined possibilities. The applications and experiences
made possible by these intelligent habitats on wheels remain largely
untapped, inviting creative minds to explore them.
We can’t precisely predict what these revolutionary applications
might be. It’s likely that the SDV’s “killer app” won’t just be an
equivalent of Angry Birds for the car but something that leverages
the unique potential of this new mobile environment.
But we can confidently make the following statement: to discover
these groundbreaking applications and experiences, experimenta‐
tion and innovation at scale need to happen. This requires much
faster development, as we discussed earlier. However, it also requires
a new approach to validating new ideas, especially from the point
of view of desirability and usability. Building physical test vehicles is
very expensive and time-consuming. Because of this, the automotive
industry is looking at ways to virtualize usability and acceptance
testing. For example, the digital.auto playground provides an envi‐
ronment that allows us to try out new ideas for digital vehicle
features in a pure cloud environment and evaluate them against real
vehicle APIs. The emergence of spatial computing and virtual reality
will accelerate virtualized UX testing. It is important that these tests
not be limited to the physical vehicle design. The ability to test the
vehicle experience as it is enabled by SDV in virtual environments
will help significantly left-shift user testing. This in turn will help
make sure that investments are prioritized according to market
demand, and the resulting vehicle experience creates popular prod‐
ucts and high levels of customer loyalty. Figure 1-6 summarizes the
digital first vehicle lifecycle.
Figure 1-6. Digital first vehicle lifecycle

Unlike in the past when car manufacturers had limited insight into
the actual use of certain features, SDVs can, provided the driver
gives consent, measure and assess actual usage of vehicle apps and
their features. This data-driven approach fosters a fast learning
cycle, facilitating innovations that are tailored to customer use and
preferences. Chapter 5, “#digitalfirst: A New Way of Working” pro‐
vides more details on how all of this can be applied in practice.
CHAPTER 2
What the Automotive Industry Can
Learn from Other Industries
In Chapter 1, we explored the vision for the modern automotive

software realm, diving deep into the concept of a “smartphone habi‐
tat on wheels” and addressing the unique challenges and opportuni‐
ties it introduces. We’ve demonstrated the imperative to innovate at
a rapid pace and the complexities of ensuring functional safety in
this hybrid landscape.
Now we shift our focus to the foundational concepts and technolo‐
gies that underpin this vision. We will take a look at what made
the smartphone and internet revolution successful and how we can
apply these lessons to the SDV.
Learning from the Smartphone Folks:

Standardization, Hardware Abstraction,
and App Stores
Nokia’s journey into the mobile world offers a significant lesson for
the SDV. By 2009, Nokia had created a maze with 57 different and
incompatible versions of its operating system. The consequences
were devastating. For developers, it became a daunting task to create
apps because of the vast fragmentation. For users, the resultant
limited application ecosystem made Nokia’s platform less attractive.
This tale mirrors a current challenge in the automotive world.
Today, almost every car model, even those from a single manufac‐
15
turer, employs custom hardware and software components sourced
from various suppliers. The result: extreme fragmentation combined
with monolithic programming frameworks, where creating a “vehi‐
cle app” that can run across multiple models of the same manufac‐
turer seems nearly impossible.
However, the smartphone industry offers a blueprint for overcom‐
ing this fragmentation. Its solution was multipronged:
• Standardized Application Programming Interfaces (APIs) provide

a consistent interface for developers, regardless of underlying
system differences.
• An easy-to-use programming language and app development
environment ensures that developers experience a flat learning
curve and encapsulated environment for their applications.
• An app store acts as a centralized distribution hub, allowing
third-party developers to get their applications seamlessly into
the hands of end users.
So, what should the SDV industry learn from this? There are at least
three lessons here:
1. Unified APIs. A set of standardized vehicle APIs would greatly

simplify the process of creating software for vehicles. By ensur‐
ing that these APIs have minimal fragmentation, developers
could write software once and have it work across multiple
vehicle models. Based on industry trends, we believe that, in
the long run, there will be a natural gravitation toward a select
few dominant API sets, minimizing fragmentation across the
industry. Within each OEM, we anticipate a more unified API
approach, leading to consistency across their individual product
lines.
2. Hardware abstraction layer (HAL). This acts as a bridge between
the software applications and the multitude of vehicle hardware
variations. It ensures that software can run irrespective of the
underlying hardware differences, adding a layer of consistency
and predictability.
16 | Chapter 2: What the Automotive Industry Can Learn from Other Industries
3. Supportive software stack (vehicle OS). A robust software stack
that’s in harmony with the standardized APIs and HAL ensures
that software can interact seamlessly with a vehicle’s compo‐
nents, making software-driven innovations easier to introduce
and adopt.
Figure 2-1 provides an overview of how best practices in the

smartphone industry can be transferred to the automotive industry.
The overview takes into consideration the fact that the automotive
domain includes areas with requirements for real-time processing
and functional safety.
Figure 2-1. Applying smartphone best practices to automotive

development
One key obstacle is the need for OEMs to differentiate themselves.

Large OEMs in particular will want to keep large parts of their
development proprietary while smaller OEMs might have a stronger
interest in standardization and industry-wide reuse. The solution
here can be to standardize frameworks and allow differentiation on
the content side. For example, an OEM might want to standardize
and open up the 70% to 80% of APIs that cover commodity func‐
tions. Using the same technical framework, this OEM could still
manage the 20% to 30% of APIs that it wants to differentiate in a
proprietary way.
Learning from the Smartphone Folks: Standardization, Hardware Abstraction, and App Stores |
17
Mohan B.V., Technology Head of Strategy, Mobility Next at Bosch
BGSW, says:
We are convinced that in the future, a lot of obligatory features will
be cross-OEM. Does it make sense to build OEM-specific APIs?
Seamless interoperability for data between vehicle, service provider,
and driver is key, and harmonized API will enable this—this is what
we have learned from the smartphone industry.
Can Yasin Gümüş, Senior Project Manager of Innovation Manage‐

ment at Bosch XC, adds:
Defining standardized API opens the market for a community
strategy. As an example, solving automated driving and driver
assistance are huge challenges. One reason is the regional, complex
problems the systems face. Opening up access to the solution space,
developers around the globe can contribute to solving these kinds
of problems. SDV opens a marketplace, allowing democratization
of the development of automated driving and driver assistance.
Learning from the Internet Folks: Open Source

and Cloud-Native Development
Even before the emergence of smartphones, the internet trans‐
formed the world by revolutionizing communication, access to
information, commerce, and social interaction. Key enablers of this
change include cloud-native technologies, the democratization of
knowledge, and open source development. Cloud native refers to an
approach to software architecture and development that leverages
cloud-computing principles and services to enable fast adaptation
to market demands, internet-level scalability, and resilience. Today,
many of the software elements required for cloud native are sup‐
ported by a large, global open source community.
As shown in Figure 2-2, key elements of cloud-native include micro‐
services and APIs, containerization, continuous integration and
delivery (CI/CD), and development and operations of IT systems
(DevOps).
Microservices and APIs
Microservices are software components that encapsulate their
data and business logic and make these available through well-
defined APIs. Because microservice architectures are loosely
coupled, they are ideally suited to support cross-organizational
teams working on multiple microservices, evolving at different
speeds.
Containerization
Containers provide the cloud-native runtime for microservices.
Containers provide virtualization on the application level,
which is more lightweight than the virtualization of an entire
operating system. Containers are usually deployed on multiple
network nodes in the cloud, providing scalability and resilience
for the microservices running on them. They also provide addi‐
tional levels of isolation, security, and systems management.
Continuous delivery and DevOps
Continuous delivery and DevOps are modern software devel‐
opment practices that ensure that the complexities, dynamics,
and uncertainties of today’s markets can be supported by fre‐
quent and reliable incremental code changes by cross-functional
DevOps teams that collaborate throughout the product lifecycle
and jointly take ownership of the deliverables.
Figure 2-2. Key elements of cloud native and open source development
Learning from the Internet Folks: Open Source and Cloud-Native Development | 19
The success of the internet would not have been possible without
open source, which has evolved from grassroots community projects
to a mainstream movement. Today’s thriving open source ecosystem
has delivered a secure and scalable infrastructure, which is the back‐
bone of the internet and most modern enterprise application land‐
scapes. The open source community has delivered operating systems
(e.g., Linux), container infrastructure (e.g., Kubernetes and Docker),
middleware for microservices (e.g., Swagger), and the toolchain for
CI/CD. Leading open source organizations in this space include the
Linux Foundation, the Eclipse Foundation, and the Cloud Native
Computing Foundation. The IT industry learned that by collabo‐
rating on non-differentiating parts like a Linux Kernel (shared by
everything from smart TVs to cloud servers), more resources can be
spent on the differentiating, customer-facing parts of their products.
Today, open source has evolved into a multi-billion-dollar market,
including commercial enterprise-grade support, consulting, custom‐
izing, training, hosting services, dual licensing, and building com‐
mercial products on top of open source foundations.
CHAPTER 3
Vehicle OS and Enabling
Technologies
Almost all OEMs—and many of their suppliers—have been working

on creating a modern vehicle operating system (OS) for some time
now. There is still no comprehensive and widely accepted definition
of what comprises a vehicle OS. However, most definitions contain
the following key components: specialized software runtimes for
different functional domains, hardware and software decoupling via
standardized APIs, and over-the-air (OTA) software updates.
When following this definition, an SDV is a key part of the vehicle
OS—or maybe even synonymous with it. SDVs are enabled by a
set of technologies that are efficient and continuous, even after the
start of production (SOP) delivery of new, digital vehicle features,
combining on-board and off-board components into an integrated
end-to-end software architecture. We will start by looking at emerg‐
ing electrical and electronic (E/E) architecture and how it can work
with a service-oriented architecture (SOA). Key elements of E/E
and SOA are hardware abstraction, vehicle APIs, and the SDV
tech stack. Modern vehicles use OTA updates to support post-SOP
updates, which will eventually be the foundation for vehicle app
stores. Finally, we will look at how artificial intelligence (AI) can
augment the software-centric vehicle.
21
Foundation: E/E Architecture
Today, so-called E/E architecture describes the overall design and
layout of electrical and electronic systems in a vehicle. This archi‐
tecture encompasses the distribution of power, data, and control
signals throughout the vehicle as well as the integration and inter‐
connection of various E/E components and systems. Many vehicles
still implement a domain-centric E/E architecture where different
vehicle domains, such as the powertrain, chassis, passenger com‐
partment, and body, are logically grouped together and connected
by dedicated bus systems, such as the Controller Area Network
(CAN) bus. The CAN bus is a signal-based protocol designed to
allow electronic control units (ECUs) and other compute nodes in a
vehicle to communicate with each other in a reliable, priority-driven
way.
Since the domain E/E architecture results in very complex and
heavy vehicle-wiring harnesses, OEMs are using so-called zonal
architectures, which aim to group different vehicle sensors and
actuators according to their physical location in the vehicle. In a
zonal E/E architecture, wiring harnesses become less redundant,
allowing for simplified connections within individual zones, reduc‐
ing complexity and weight, and enabling easier integration of new
features and technologies. Zonal architectures usually combine
dedicated zone controllers with high-performance vehicle comput‐
ers. The zone controllers are locally connected to various sensors
and actuators, often using different legacy/heritage bus systems,
such as CAN, Local Interconnect Network (LIN), and FlexRay. The
zone controllers are then connected with each other and with the
high-performance vehicle computers via new on-board, high-speed
networks based on Ethernet (the foundation of today’s internet).
Vehicle APIs
The first step toward a service-oriented architecture for digital on-
and off-board services is to provide a hardware abstraction via vehi‐
cle APIs. Today, developing new on-board features usually involves
a complex and lengthy alignment process among many departments
of the OEM. This is because all signals within a given on-board
domain are communicated via a shared bus system (e.g., the CAN
bus). For each vehicle type, a CAN matrix defines which ECUs send
which message under which conditions and with which cycle time,
22 | Chapter 3: Vehicle OS and Enabling Technologies

which ECUs receive which messages, and how the messages are
structured and prioritized. This results in a tightly coupled architec‐
ture that requires very close alignment on technical and organiza‐
tional levels. To get from here to a loosely coupled, service-oriented
architecture, a new level of hardware abstraction is required.
From the perspective of the service consumer (i.e., user of digital
applications), the vehicle should provide a set of well-defined APIs
that provide an abstraction of the vehicle’s functions. On the lowest
level, these are the sensors and actuators of the vehicle. A good
example of this open industry standard is the Vehicle Signal Speci‐
fication (VSS) defined by the Connected Vehicle Systems Alliance
(COVESA VSS). COVESA VSS defines a tree-like API structure for
accessing vehicle sensors and actuators as signals (see Figure 3-1).
Figure 3-1. Vehicle signal API tree (based on COVESA VS)
Vehicle APIs | 23
For example, Vehicle.Cabin.Seat.Row1.Pos1.Headrest.Angle
would give an application developer access to the actuator that
controls the angle of the headrest of the front left seat. This is a level
of abstraction suitable for developers used to cloud-native or smart-
phone development. Of course, these relatively low-level signal-to-
service APIs have to be augmented with higher-level orchestration
services over time.
However, there are some issues with this approach. First, as dis‐
cussed before, it can be challenging to map such a high-level soft‐
ware API to the underlying, complex E/E architecture. Second, there
is the question of how to ensure the functional safety of APIs that
may impact vehicle physics. And third, it is necessary to solve for
Quality of Service (QoS) aspects of such an API (e.g., real-time
requirements). We will look at all of these aspects in the following
sections.
Vehicle SOA: Encapsulating the E/E

Architecture with Vehicle APIs
The issue with the E/E architecture perspective is that it is modeled
after the physical design of the vehicle, including actuators, sensors,
on-board networks, and compute nodes. A key prerequisite for
SDVs will be to encapsulate the hardware-focused E/E architectures
with a vehicle service-oriented architecture (SOA), as depicted in
Figure 3-2.
Figure 3-2. SOA encapsulating the E/E architecture
On the E/E hardware side, high-performance computers run the

SDV functions. These computers are physically connected with
the other on-board components, including smaller compute nodes,
sensors, and actuators. The translation from the hardware side
to the software side can happen on these high-performance com‐

puters; hardware functions are exposed via APIs. These APIs are
the foundation for the communication between application-level
microservices and the underlying hardware. A microservice is an
encapsulated piece of software that implements a specific function.
Interaction between microservices always happens via APIs. These
can either be APIs abstracting a specific hardware function or
application-level APIs. An application is a collection of microser‐
vices that are orchestrated to provide a specific feature or digital
service.
Layers of the Vehicle SOA

SOAs introduce a level of architectural layering that helps deal with
the individual characteristics of the different microservices involved.
In the world of the internet, for example, the frontend layer of an
application would include microservices that change frequently due
to continuous optimization of the user interface while the basic
services layer would include more stable, data-centric services that
change much less frequently. This type of layering is essential for the
efficient evolution of complex systems.
In the automotive world and SDV, this layering is different and
has multiple dimensions. The first dimension we need to look at
is on-board versus off-board. The second dimension is defined by
the functional safety and real-time requirements of the contained
microservices and the vehicle event chains encapsulated by those
microservices.
Putting these dimensions together with what we discussed earlier
about E/E architectures and vehicle APIs, we get a service-oriented
architecture for the SDV as described in Figure 3-3. At the top, the
green layer indicates the environment for QM-only microservices
hosted in the cloud. A set of vehicle-to-cloud APIs connects on-
board services with off-board services. In the middle part, the SDV
layer includes both QM-only and ASIL A/B microservices, which
are hosted in different environments. Further south in this architec‐
ture, the physical vehicle functions can be found, made accessible
through a signal-to-service API (e.g., based on COVESA VSS). The
signal-to-service API must encapsulate the mapping of the APIs to
the E/E architecture of the vehicle. For example, the API must know
which zone controller would actually host the functions required
to support Vehicle.Cabin.Seat.Row1.Pos1.Headrest.Angle (see
Layers of the Vehicle SOA | 25

Figure 3-1). The zone controller (in this case the “Zone FL” of the
front left) would, in turn, need to provide a software proxy that can
translate the API onto the matching CAN signal so that the angle of
the front left headrest is changed.
Figure 3-3. Service-oriented architecture for the software-defined

vehicle
Ensuring Functional Safety in the Vehicle SOA

A key benefit of the multilayered vehicle service architecture is that
we now have several options for functional safety. Let’s look at a
concrete example in Figure 3-4, a function in a smartphone app that
remotely opens the trunk of a vehicle.
The problem with the “open trunk” function is that it can only be
safely executed if the vehicle is in a safe state (i.e., not moving).
So, the question is: who will perform this check, and how? The
caller itself cannot perform the check—this would violate the loose
coupling principles of the service-oriented architecture, as the ser‐
vice implementation cannot assume clients always properly follow
a certain protocol. This is certainly true for clients calling from the
cloud, but also for clients calling from a QM environment on board
the vehicle. Indeed, this is the whole point of the QM environment;
it provides none of the ASIL QoS properties. This means that the
on-board “open trunk” service has to perform the required safety

checks itself. Assuming that the service interface is exposed to QM
clients, there are now at least two options.
Figure 3-4. Vehicle SOA and Functional Safety–two examples
In the first option (Figure 3-4, left), the on-board “open trunk”
service (Vehicle.Body.Trunk.Rear.IsOpen) runs in a QM environ‐
ment and is therefore not considered safe. It might provide some
additional services, but eventually the safety check must be done
from within the ASIL safety environment. In our case, this would
be the zone controller for the trunk. Here, the system must check
the current vehicle speed before opening the trunk. However, in this
example, a different zone controller manages the speed signal (i.e.,
the zone controller for the powertrain). So the zone controller that
manages the trunk must access the zone controller that manages the
speed signal before communicating with the low-level ECU control‐
ling the trunk. And all of this must happen in real time (which is
why this check is performed on the zone controller, not in the QM
layer above). This does not necessarily mean it has to happen in a
fraction of a millisecond, but it has to happen within a guaranteed
time interval. For this to happen, the two zone controllers must be
in more direct communication, such as a direct link supporting so-
called Time-Sensitive Networking (TSN) via high-speed Ethernet.
Ensuring Functional Safety in the Vehicle SOA | 27

The second option (Figure 3-4, right) assumes that the on-board
“open trunk” service is executed in a safety environment matching
the ASIL levels required for this function. This means the exposed
interface can still be called from a QM client (e.g., an event sequence
originating in the cloud), but the implementation behind the inter‐
face is safe. If there is another safety-compliant Vehicle.Speed ser‐
vice (and a safe way for the trunk service to call the speed service),
then this can now all be done in the same environment. The benefit
is that the speed service is provided as a real microservice that can
be reused by QM as well as ASIL services. However, this requires
a service architecture capable of orchestrating microservices with
ASIL QoS levels.
Both options presented here are valid. The first option makes
fewer demands on the integration levels within the environment
and might be easier to implement with current technologies. The
second option has more potential to enable advanced cross-domain
use cases, but requires a more advanced ASIL-compliant runtime
environment. The following discussion will look at the SDV tech
stack required to support both options.
SDV Tech Stack for the Vehicle SOA

To understand how a vehicle SOA can support architectural lay‐
ers with different QoS levels, we need to add another dimension—
adding the hardware, operating system, and middleware/application
environment. This is shown in Figure 3-5.
The cloud layer shown here is as to be expected. The hardware
includes generic central processing units (CPUs) and graphics pro‐
cessing units (GPUs), which are used today for processing machine
learning workloads in the cloud. On the operating systems (OS)
level, we usually find a general-purpose OS like Linux. Hypervisors
or virtual machines are used for running multiple OS instances on
shared hardware, which is important for scalability and enabling
cloud elasticity. Modern clouds also provide very rich middleware
and application runtimes.

Figure 3-5. The SDV tech stack
Next we have the on-board environment for QM-only microservices

and applications. This type of environment aims to replicate as
much as possible the rich tech stack found today in the cloud
while adding some vehicle-specific aspects. These will include, for
example, faster start-up times, energy efficiency, and support for
managing highly distributed instances that are not always online
(e.g., large vehicle fleets, as opposed to central cloud data centers).
The OS in this layer will usually be a general-purpose OS combined
with hypervisors for virtualization. This will run on a generic high-
performance vehicle computer, potentially with additional GPUs for
AI/ML workloads.
The next layer has to support high QoS levels and consequently
requires a more traditional high-availability environment, usually
with real-time support. A widely used OS in this area is QNX from
BlackBerry. The zone controllers will probably share a similar setup.
A real-time-capable middleware will be required to link microser‐
vices from the environment north of the S2S API with those resid‐
ing south of it (e.g., on zone controllers).
Finally, the zone controllers connect to lower-level ECUs. Especially
for the control of specific vehicle sensors and actuators, microcon‐
trollers are used. These are integrated circuits (ICs), usually repre‐
SDV Tech Stack for the Vehicle SOA | 29

senting a complete system on chip (SoC), including the processor
core, memory, and IO, all inside one discrete package. Specialized
environments for low-level but highly efficient and very targeted
embedded functions are used here. Between the zone controller
and the different low-level ECUs and microcontrollers, different bus
systems have to be supported, including CAN, LIN, and FlexRay.
OTA: Over-the-Air Updates

In the past, the automotive industry aimed to freeze the hardware
and supporting software of a new vehicle generation at the SOP.
Post-SOP changes to the supporting software were usually made
only if serious problems needed to be fixed.
Of the 70 to 100 ECUs and controllers usually found on a traditional
vehicle, the majority would never be updated after the initial soft‐
ware flashing as part of the manufacturing process. If updates were
needed, a dedicated update campaign would be run for each affected
variant (“push”). Updates are usually manually packaged based on
extensive variant research and validation.
In the future, many OEMs envision a rich application ecosystem
developing for their vehicles. Software updates will no longer be
done only for quality reasons but can happen on demand. Custom‐
ers will select which new applications and features they want to
download and activate (“pull”). These new applications can span
different domains, ranging from infotainment and well-being to
cabin comfort and driving performance. However, this increasing
individualization will also lead to a dramatic rise in variants of
hardware/software combinations. The next generation of OTA will
have to address this (Figure 3-6).
Figure 3-6. Modern over-the-air update architecture

The Vehicle App Store
The holy grail of the SDV is the vehicle app store. The smartphone
industry has proven the huge potential of applications and content
generated by partners and external developers, made accessible to
customers on demand and post-SOP. Now OEMs are jumping on
the bandwagon. Replicating the success story of smartphone app
stores in the automotive industry requires several things. Figure 3-7
gives an overview of the technical prerequisites.
Figure 3-7. Vehicle app store
First, vehicles need a secure on-board application runtime environ‐

ment for the execution of applications downloaded via the vehicle
app store. This runtime is often referred to as a sandbox. It is
important that this sandbox be secure, not allowing code to break
out and access other parts of the vehicle. A virus that takes control
of a vehicle can have deadly consequences for the vehicle’s occupants
or bystanders.
Second, the system architecture must ensure that applications run‐
ning inside the sandbox can only interact with other parts of the
vehicle via a set of controlled APIs. Most likely, OEMs will have to
define different trust levels for their APIs, differentiating between
API access through OEM-developed applications, trusted partner
apps, and potentially apps developed by completely unknown third
parties. Already, some vehicles available today provide app stores
The Vehicle App Store | 31

for in-vehicle infotainment. In the future, vehicle app stores should
also provide access to sensor APIs to unleash the creativity of the
global developer community. They may even offer safe access, so
selected actuators might be an option (e.g., in the body and comfort
domain).
While the smartphone juggernauts were already able to position
apps via Apple Car Play and Android Automotive, many OEMs are
now starting to offer app stores for applications running natively in
the vehicle. This journey starts in the infotainment area, introducing
apps designed to run natively on the built-in infotainment system
screen. Examples include apps for music and podcasting, video
conferencing, weather, gaming, news, parking, and electric vehicle
(EV) charging. It remains to be seen how well these vehicle infotain‐
ment applications will be able to compete with their smartphone
counterparts.
To be better differentiated, the next generation of in-vehicle apps is
likely to make use of vehicle-specific features. Once apps are able to
interact in a meaningful and safe way with the vehicle, a new experi‐
ence can be created. Being able to access data from vehicle sensors—
and potentially even control some of the vehicle actuators—will
create a new type of application. For example, being able to access
in-vehicle sensors, such as cameras, radar, and microphones, could
create a new generation of health and well-being apps. This would
help OEMs come closer to the vision of a “habitat on wheels.”
SDV and AI
While AI has already become a hot topic, the release of ChatGTP
has moved the discussion from “software will eat the world” to “AI
will eat the world.” AI is disruptive on many levels, and the automo‐
tive world is no exception. So, do we need an “AI-defined vehicle”
instead of (or in addition) to a software-defined one? Where in the
SDV architecture does AI play a role? The answer is: potentially in
all the layers of the SDV architecture (see Figure 3-8).

Figure 3-8. SDV and AI
On the smartphone, AI can be used to control access to the vehicle

or specific vehicle functions (e.g., via face recognition). In the cloud,
AI can operate on vehicle fleet data (e.g., to support vehicle rout‐
ing, optimize the EV charging experience, perform traffic analysis,
improve battery performance, etc.).
On-board, we are seeing different types of AI. For highly automated
driving, AI is deeply integrated with the vehicle’s E/E architecture,
providing object identification (e.g., recognizing a child on a bicycle
in front of the car) and controlling the vehicle’s trajectory (e.g.,
braking for the aforementioned child). In addition, there seems
to be a huge potential for AI-enabled long-tail applications. These
applications require much less investment for their development
and will probably have a lesser overall impact individually but still
have a huge impact overall. The smartphone app stores have proven
that if the boundaries of entry are sufficiently low, a plethora of new,
creative applications will be created. This should also be true for
AI-enabled vehicle apps.
Combining AI with data from in-vehicle sensors alone could be a
game changer (e.g., for infotainment and well-being). If this was
then combined with SDV-enabled functions, a whole new genera‐
tion of in-vehicle applications could be created (e.g., cabin comfort
applications that use sensor data to change the vehicle’s ambience
by accessing in-cabin light, HVAC, seat massage functions, and so
SDV and AI | 33
forth). The use of AI will reach far beyond vehicle apps. From
voice assistance to predictive maintenance, from fleet operations to
automated driving to AI-assisted development toolchains, AI will be
a game changer for the way we design, develop, operate, and interact
with vehicles.

CHAPTER 4
Value Stream Management
for the SDV
The internet is built on sophisticated, constantly evolving technol‐

ogies. Consequently, many successful internet businesses are built
on a technology culture. However, this reliance on technology
can sometimes make it difficult to focus on building customer-
centric products, increasing competitiveness and generating rev‐
enue. Therefore, many companies have adopted Value Stream
Management (VSM) as a best practice for their digital businesses.
VSM is a set of practices designed to ensure that new digital features
are delivered fast and efficiently and that they deliver clear customer
value.
So how can VSM be applied to the software-defined vehicle? Let’s
have a look.
Working at Different Speeds

What is maybe the most important realization in the context of the
SDV is that there cannot be a single value stream. Our discussion
of the “clash of two worlds” and the following technical discussions
have shown that different requirements need different, specialized
approaches. The digital vehicle experience must be delivered by a
digital value stream, which adheres to Agile best practices, while
the physical vehicle experience must be delivered by a value stream
that adheres to the rigorous and “first time right” thinking required
35
for areas with high levels of functional safety requirements (see
Figure 4-1).
Figure 4-1. High-level value stream perspective for the SDV
The digital value stream must be able to address uncharted territory,

deal with vague requirements and constantly changing ideas, and be
able to quickly react to customer feedback. Building an explorative
and feedback-based approach to the digital value stream is key. The
technical delivery pipelines of the digital value stream will be out‐
putting artifacts that are deployed either in the cloud or on-board.
The on-board artifacts will use SDV, containerization, OTA, and
vehicle APIs. Mechanisms for measuring customer success can be
natively built into this technology stack, similar to the customer
journey analysis tools familiar from internet applications.
The physical value stream must be aligned much more closely with
long-term planning and the enterprise architecture. The methods
applied here will adhere to established best practices for functional
safety-related features (e.g., the well-established V-model of soft‐
ware development, which has formal verification and validation
mechanisms built in). The technical outputs of the physical value
stream will include a combination of hardware and software, such
as embedded software and ECUs, as well as mechatronic system
components. Implementing mechanisms for measuring customer
success in the physical value stream will be more difficult.
36 | Chapter 4: Value Stream Management for the SDV

Establishing an effective VSM strategy that enables OEMs to work at
different speeds in different areas will be a key success factor in the
future.
Divide and Conquer

The beauty of SOA and the API-centric way of working that we
introduced earlier is that these approaches give us the capability to
create integration not only on the technical level but also on the
organizational level. To support value streams that are moving at
different speeds, a loose coupling is required both on the technical
as well as on the organizational level. And this is exactly what can be
achieved with APIs and hardware abstraction.
Figure 4-2 shows an example of loose coupling. At the top, we have
technical artifacts created as output of the digital value stream. First,
prototypes are created and implemented against the vehicle API—
e.g., using the digital.auto playground for rapid online prototyping
of SDV features. These prototypes can be used to get very early
feedback from key stakeholders, including customers and manage‐
ment. Next, the early SDV prototypes are refined and brought closer
to real applications. In the early development phases, these SDV
applications can utilize vehicle simulations behind the APIs to create
realistic test environments. Once real vehicle hardware is available,
this hardware can replace the vehicle simulation. Since the APIs are
not changing (at least in an ideal world) and SDV application is
implemented against the API, this shift from a vehicle simulation
to real vehicle hardware should be seamless from the point of view
of the SDV application. This means that the API becomes the mech‐
anism that ensures loose coupling not only between the technical
artifacts but also between the different organizations involved. The
teams developing the digital value stream are doing so against the
APIs representing the real hardware underneath (hence “hardware
abstraction”), and by using simulators, they can move at their own
speed, independent of the availability of the real hardware. This
approach takes the well-established concepts of hardware-in-the-
loop (HIL) and software-in-the-loop (SIL) to another level.
Divide and Conquer | 37

Figure 4-2. Value streams with technical artifacts moving at different
speeds
Being able to split up a complex body of work using a divide and

conquer strategy is a huge benefit, reducing complexity to a man‐
ageable level. Notice that the APIs are likely to take on the role of
a “master clock,” helping to synchronize work between the different
teams in the different value streams.
Enterprise Perspective
OEMs have traditionally addressed the complexity they face with
enterprise architecture management (EAM) and Model-Based Sys‐
tems Engineering (MBSE), shown in Figure 4-3. EAM helps manage
the dependencies between the systems-of-systems perspective (e.g.,
the vehicle in the context of its environment), the system perspec‐
tive (the vehicle itself), as well as the subsystems, including key
components and features. Of course, all of this must be seen in the
context of many different vehicle variants and vehicle types. Finally,
managing the reuse of vehicle platforms is key, including hardware
platforms, E/E platforms, and software platforms. MBSE plays an
increasingly important role in the detailed design of many system
components and their interdependencies.

Figure 4-3. The enterprise perspective
However, coming back to the “clash of two worlds” discussion, it

is important to notice that these kinds of tools and methods are
often seen as very controversial in the software world, which has
spent the last 20 years adopting an Agile culture. This becomes clear
in Table 4-1, which compares how the Agile values defined in the
famous manifesto for Agile software development map to model-
centric and code-centric approaches to automotive development.
Table 4-1. Model-centric versus code-centric development

Values from the Agile Model-centric (EAM/MBSE) Code-centric/SDV
Manifesto
Individuals and interactionsTools and processes required, Well suited to Agile
over processes and tools especially for hardware and software processes for QM
with ASIL requirements features
Working software over Can be achieved via code generated Well supported
comprehensive documentation from models (but not a trivial task)
Customer collaboration over Model as contract; early customer APIs as contracts, but
contract negotiation validation requires virtual exploration features developed in
close alignment with
customers
Responding to change over Long-term planning required for Supported by Agile
following a plan hardware and ASIL software approach
Enterprise Perspective | 39
As we have discussed before, the ability to work with different value
streams that embody different approaches and methods is the key
to success. On the enterprise-level, methods and mechanisms must
be established to keep the different value streams in sync, supported
by a loose-coupling approach on the organizational level. Again,
APIs can play a key role here, for example, by providing a loose cou‐
pling between the Agile/code-centric and the model-centric/MBSE
perspective.

CHAPTER 5
#digitalfirst: A New
Way of Working
Taking everything into account that we have discussed so far, we

propose a new way of working for OEMs, which we are simply
calling #digitalfirst.
The fluid nature of today’s consumer preferences means we can’t
definitively predict today which features will be in demand tomor‐
row. However, a key certainty emerges: OEMs that can’t swiftly
explore, test, and deliver new features at scale may find themselves
incapable of crafting the engaging customer experiences today’s con‐
sumers demand. In our rapidly evolving digital age, the agility to
innovate quickly and effectively is not just desirable—it’s essential to
staying relevant in the automotive industry.
Consequently, #digitalfirst starts with customer experiences and
works backward to the technology. We have co-founded the digi‐
tal.auto initiative to support this. “Digital first, auto second” is not
just a tagline, but the core ethos guiding this transformation. As
shown in Figure 5-1, #digitalfirst assumes that an OEM has to
undergo three tectonic shifts: the shift north (of the vehicle API),
the shift left (toward early-stage testing), and the shift toward vir‐
tualized development. We will look at each of these shifts in more
detail.
41
Figure 5-1. The three tectonic shifts underlying #digitalfirst
Shift North
The shift north involves moving functionalities from the safety
domain to the QM domain, creating a separation between the
domains via well-defined APIs. This is driven by the extra effort that
will remain for every development in the safety domain—validation,
homologation, and extra documentation—that is usually required to
a lesser degree in the QM world.
Shifting code north into the QM world, as shown in Figure 5-2,
means that modern software engineering techniques and tools can
be used, speeding up development and making updates post-SOP
much easier. In addition, in this domain, even software develop‐
ers who don’t have years of experience in the automotive sector
(modern software engineers who also know ISO 26262 are rare) can
work productively, thus accelerating development significantly.
Figure 5-2. Shift north
42 | Chapter 5: #digitalfirst: A New Way of Working

Shift Left
The shift left refers to exploring customer features and conducting
user testing as early as possible in the development process, as
shown in Figure 5-3. Many of today’s cars offer an abundance of
buttons and features, many of which remain unknown or unused
by users. Hence, the objective is to discern truly desirable features
and maximize customer satisfaction through early exploration and
testing.
Figure 5-3. Shift-left testing
According to a report from the National Institute of Standards and

Technology (NIST), an agency of the US Department of Commerce,
the cost of fixing problems downstream in the development process
can be up to 640 times the original development cost. And this does
not factor in the cost for losing business due to unhappy customers,
or customers who simply don’t care about certain features that the
developers were sure the customers would love but never cared to
ask.
Virtualization
Virtualization emphasizes the development and testing of systems
in virtual cloud environments. An interesting organization in this
space is SOAFEE, which is developing virtualization concepts in the
context for ARM-centric hardware architectures. The main motiva‐
tion here is to overcome the traditional tight coupling between
hardware and software in automotive development, where software
engineers must wait for expensive, early versions of hardware for
Virtualization | 43
development and testing. Integration and testing of components are
costly and complex due to limited prototypes and numerous vehicle
variations, such as different engines, trim levels, or country-specific
requirements.
The capacity that cloud environments offer for infinite scalability
and cost reduction, along with the use of virtual electronic control
units (vECUs) or virtualized cars, presents a solution to these chal‐
lenges, as shown in Figure 5-4.
Figure 5-4. Virtualization
Summary
To deliver on the vision of the “habitat on wheels,” with rich cross-
domain applications and data fusion delivered 10 times as fast,
OEMs must overcome significant impediments presented by func‐
tional safety requirements, technical constraints, and organizational
constraints (“clash of two worlds”).
The vehicle OS can be a powerful platform, enabling rapid appli‐
cation development in the QM world via service-oriented architec‐
tures, OTA, and vehicle app stores. Combining the SDV and AI is
important for data-driven applications.
To manage “development at different speeds,” OEMs must embrace
VSM and use hardware abstraction and vehicle APIs to create a
loose coupling, not only on the technical level, but also on the
organizational level between the digital and the physical value
streams, as shown in Figure 5-5.
44 | Chapter 5: #digitalfirst: A New Way of Working

Figure 5-5. #digitalfirst and VSM for the digital.auto
#digitalfirst is a new way of working that combines three tectonic

shifts:
Shift north
Breaking up event chains and focusing on QM application
development north of the vehicle API
Shift left
Left-shifting user tests and getting early customer validation of
the integrated digital/physical experience
Shift toward virtualization
Further decoupling of hardware and software development,
thus increasing agility
Together, the SDV and #digitalfirst can deliver a number of benefits
for OEMs and the entire supply chain. Early and continuous user
feedback helps ensure that digital investments are paying off as
intended. Avoiding investments in unwanted digital features helps
ensure that development capacities are used where they create
the highest customer value. Significantly increasing development
speed (10x) provides the agility that is required to react quickly to
changing customer needs and preferences. The divide and conquer
approach enabled via hardware abstraction and specialized value
streams helps manage the complexity of today’s automotive world.
Virtualization helps reduce the cost and complexity of managing too
many hardware prototypes.
Summary | 45
CHAPTER 6
Next Steps
We hope you find the concepts outlined here helpful for your daily
work. Should you be interested in finding out more about how we
work and how you can get involved, please visit digital.auto.
We helped co-initiate digital.auto as an open and vendor-neutral
community to enable our industry to use the SDV to deliver all the
exciting use cases we have been talking about here. We do this via
different collaboration and thought leadership activities, including
this publication. In addition, the digital.auto community has worked
together to create a number of open source activities.
Figure 6-1 maps the digital.auto focus areas against the #digitalfirst
SDV lifecycle introduced in Chapter 1.
First, we focus on advancing methods and tools to support early-
stage virtual exploration of SDV experiences. In particular, we
worked together to create a cloud-based, rapid-prototyping environ‐
ment for the SDV. This digital.auto playground can be used to rap‐
idly try out new ideas for the SDV against real vehicle APIs, which
are simulated in the backend to get realistic test data. The resulting
prototypes can be used to get early customer feedback and learn
more about requirements, including the APIs needed for the new
application. Prototypes developed in the playground can be directly
deployed onto real SDV platforms (e.g., the Eclipse Velocitas open
source SDV runtime). The digital.auto playground is open source
and free to use. You can try it out at playground.digital.auto. We
look forward to your feedback!
47
Figure 6-1. The digital.auto focus areas
Second, we are fostering interoperability along the SDV value chain.

To support a #digitalfirst approach, a number of tools and technical
platforms must be seamlessly integrated, from early-stage testing
over virtualized development in the cloud to deployment on the
target vehicle hardware. The site interop.digital.auto summarizes
our current activities in this space and provides a number of interop
proof points we have implemented in the digital.auto community.
Third, we believe that to unlock the full potential of the SDV, we
must create a vibrant ecosystem of OEMs, start-ups, developers,
and innovators, all working together to try out new, exciting use
cases. This will require a test infrastructure for SDVs that is not yet
available. Our vision is to work together on an open test fleet for
SDVs, including an open app store for experimental new features,
executed in a safe environment. If this sounds intriguing to you, visit
www.digital.auto and contact us to help make this vision become a
reality.
48 | Chapter 6: Next Steps

About the Authors
Dirk Slama is a vice president at Robert Bosch GmbH and chair
of the digital.auto initiative. He is also the director of the AIoT
Lab at the Ferdinand-Steinbeis-Institute, where he holds a full pro‐
fessorship. Dirk has 25 years experience in large-scale IT projects in
automotive, manufacturing, finance, and telecoms. He is coauthor
of four books, with more than 50,000 copies sold. His academic
credentials include a PhD in information systems, an MBA, and a
diploma in computer science.
Achim Nonnenmacher is driving software-defined vehicle innova‐
tions as a senior manager at ETAS GmbH (a Robert Bosch Subsid‐
iary). As cochair of the digital.auto initiative, he helps speeding up
the use case-driven adoption of new technologies in the automotive
industry. Before this, he led product innovations at scale in the
mobility sector by validating business, technology hypotheses and
user needs for strategic projects. Achim holds a PhD from Swiss
Institute of Technology (EPFL), an M.Sc. in physics, and an execu‐
tive degree on innovation acceleration from UC Berkeley.
Thomas Irawan is president of ETAS and chairman of the Board
of Management. Before joining ETAS, Thomas spent 16 years at
Robert Bosch GmbH in a number of manufacturing, quality, devel‐
opment, and engineering leadership positions. He was technical
plant manager of Bosch Thailand and served as SVP of quality man‐
agement in the Chassis Systems Control division. Most recently, he
headed the driver experience business unit for assisted, automated,
and connected driving in the Cross-Domain Computing Solutions
division. Thomas earned his doctorate in physics from the Univer‐
sity of Dortmund, Germany.

SDV Ebook

Uploaded by

Copyright:

Available Formats

SDV Ebook

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

SDV Ebook

Uploaded by

Copyright:

Available Formats

Co

by Dirk Slama, Achim Nonnenmacher,

Beijing Boston Farnham Sebastopol Tokyo

Acquisitions Editor: Suzanne McQuade Interior Designer: David Futato

September 2023: First Edition

Revision History for the First Edition

Preface and Acknowledgments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

2. What the Automotive Industry Can Learn from Other Industries. . . 15

3. Vehicle OS and Enabling Technologies. . . . . . . . . . . . . . . . . . . . . . . . . 21

5. #digitalfirst: A New Way of Working. . . . . . . . . . . . . . . . . . . . . . . . . . . 41

This publication is the result of the intensive work we are doing

• Achim Henkel, Director, Robert Bosch Group

In addition, we would like to thank Bosch senior management for

viii | Preface and Acknowledgments

— Dr. Dirk Slama

— Dr. Achim Nonnenmacher

— Dr. Thomas Irawan

Berlin/Stuttgart, August 2023

• API: Application Programming Interface

Preface and Acknowledgments | ix

In the past, the automotive industry stood as a testament to the

Goals of the SDV

Goals of the SDV | 3

Cross-Domain Applications and Data Fusion

In this connected ecosystem, your car could even become a creative

Goals of the SDV | 5

Figure 1-2. “Need for speed” in automotive software development

Impediments: Why Is Automotive Software

Complexity and Heterogeneity

Impediments: Why Is Automotive Software Development Different? | 7

Figure 1-3. Automotive safety levels

Additionally, a class called QM (Quality Managed) is assigned to

Clash of Two Worlds

Impediments: Why Is Automotive Software Development Different? | 9

Modern vehicles are a fusion of these worlds; they must merge

Impediments: Why Is Automotive Software Development Different? | 11

Rethinking the Vehicle Lifecycle: Digital First

Figure 1-6. Digital first vehicle lifecycle

Impediments: Why Is Automotive Software Development Different? | 13

In Chapter 1, we explored the vision for the modern automotive

Learning from the Smartphone Folks:

• Standardized Application Programming Interfaces (APIs) provide

1. Unified APIs. A set of standardized vehicle APIs would greatly

Figure 2-1 provides an overview of how best practices in the

Figure 2-1. Applying smartphone best practices to automotive

One key obstacle is the need for OEMs to differentiate themselves.

Can Yasin Gümüş, Senior Project Manager of Innovation Manage‐

Learning from the Internet Folks: Open Source

Almost all OEMs—and many of their suppliers—have been working

22 | Chapter 3: Vehicle OS and Enabling Technologies

Figure 3-1. Vehicle signal API tree (based on COVESA VS)

Vehicle SOA: Encapsulating the E/E

Figure 3-2. SOA encapsulating the E/E architecture

On the E/E hardware side, high-performance computers run the

24 | Chapter 3: Vehicle OS and Enabling Technologies

Layers of the Vehicle SOA

Layers of the Vehicle SOA | 25

Figure 3-3. Service-oriented architecture for the software-defined

Ensuring Functional Safety in the Vehicle SOA