Automotive Innovation (2022) 5:180–194

https://doi.org/10.1007/s42154-022-00179-z

Impact, Challenges and Prospect of Software‑Defined Vehicles

Zongwei Liu1,2 · Wang Zhang1,2 · Fuquan Zhao1,2

Received: 7 October 2021 / Accepted: 22 February 2022 / Published online: 25 March 2022
© China Society of Automotive Engineers (China SAE) 2022

Abstract
Software-defined vehicles have been attracting increasing attentions owing to their impacts on the ecosystem of the automo-
tive industry in terms of technologies, products, services and enterprise coopetition. Starting from the technology improve-
ments of software-defined vehicles, this study systematically combs the impact of software-defined vehicles on the value
ecology of automotive products and the automotive industrial pattern. Then, based on the current situation and demand of
industrial development, the main challenges hindering the realization of software-defined vehicles are identified, including
that traditional research and development models cannot adapt to the iterative demand of new automotive products; the
transformation of enterprise capability faces multiple challenges; and many contradictions exist in the industrial division of
labor. Finally, suggestions are put forward to address these challenges and provide decision-making recommendations for
enterprises on strategy management.

Keywords Software-defined vehicle · Industrial reconstruction · Automotive industry · Strategic suggestions

Abbreviations complement each other resulting in new automotive tech-

ECU Electronic control unit nologies, infrastructure as well as research and develop-
EEA Electrical/electronic architecture ment (R&D) models, manufacturing and service and finally
ICT Information and communication technology reshaping the industrial ecosystem [1]. It can be predicted
OS Operating system that intelligent vehicles will become the core of digital life
OTA Over-the-air in the future, connecting smart transportation, smart city and
R&D Research and development smart energy and enabling the data to flow [2]. Viewed this
SDV Software-defined vehicle way, future vehicles will be defined by data in the era of the
SOA Service-oriented architecture Internet of Everything.
SoC System on chip Software, as the tool of data generation, circulation and
SOP Start of production application, will be deeply involved in the stages of R&D,
production, sale, operation and service in the whole life
cycle of vehicles. The value of automotive software will
1 Introduction become increasingly prominent [3]. According to McK-
insey's research, the software codes on each vehicle have
A new round of automotive industry reform is taking place exceeded 200 million lines, and the market scale of automo-
with the revolution of energy, connectivity and intelli- tive software and related services has exceeded 24 billion US
gence as the core driving force. The connectivity revolu- dollars, which will continue to increase at a high speed in
tion enables data to fully interact and the intelligence revo- the next 5–10 years [4]. For traditional vehicles, hardware is
lution effectively promotes the utilization of data, which both the necessary and sufficient condition, while software
is only auxiliary, while for future vehicles, hardware is only
the basic necessary condition, and software is the sufficient
* Fuquan Zhao condition to determine user experience. Compared with
[email protected]
data-defined vehicles, software-defined vehicles (SDVs) can
1
State Key Laboratory of Automotive Safety and Energy, better reflect the change of the relationship between hard-
Tsinghua University, Beijing, China ware and software and the evolution direction of automo-
2
Tsinghua Automotive Strategy Research Institute, Tsinghua tive products. It is the vehicle whose architecture design is
University, Beijing, China

13
Vol:.(1234567890)
Impact, Challenges and Prospect of Software‑Defined Vehicles 181

determined by software, and whose function, performance, operating system (OS) kernel, middleware, service layer
service and experience mainly depend on software. It should of service-oriented architecture (SOA), functional appli-
be clarified that the concept of SDV is not the same as intel- cation, service application and cloud service platform.
ligent vehicles. SDV is a measure to create intelligent vehi- Among these elements, onboard hardware includes func-
cles, and intelligent vehicles are one of the external mani- tional hardware, EEA and computing platform, which is
festations of SDV. the competency source of data generation, processing and
As an unprecedented concept, SDV requires relevant interaction [6]. Functional hardware includes sensors and
enterprises to actively explore the “no-man’s-land”. On the actuators, which are responsible for generating data and
whole, the industry is still in the state with a clear direction receiving instructions for corresponding functions. EEA
and unclear path. Different enterprises have different under- connects all functional hardware and computing platforms
standings of the connotation of SDV, so different develop- through the in-vehicle network to promote data interaction
ment strategies are formulated. Therefore, it is urgent for the and gathers data on computing platforms for processing
automotive industry to systematically and comprehensively [7]. The computing ability of platform determines the abil-
comb and analyze the impacts and challenges of SDV and ity of data processing inside the vehicle.
put forward specific and targeted suggestions. This study Onboard basic software consists of OS kernel, middle-
follows this train of thought. ware and the service layer of SOA, which can promote the
aggregation and cross-use of data among different sub-
systems and domains [8]. Typical automotive OS kernels
2 Industrial Reconstruction Caused by SDV include Linux, QNX and other real-time operating system,
each of which is responsible for the management of software
2.1 Changes in Automotive Technologies Promoted and hardware of one or more subsystems [9, 10]. Middle-
by SDV ware is a technology for communication between distrib-
uted systems, which can shield the underlying differences
One of the key technical characteristic of SDV is the decou- and provide a unified data aggregation platform upward
pling of software and hardware. Software in the component [11]. SOA is a software architecture that takes services as
subsystems of traditional vehicles is embedded in hardware. basic components and forms application software through
The result of the high binding of software and hardware is arrangement and combination of services [12]. The service
the solidification of vehicles' functions and performance. layer of SOA encapsulates the stable and repeated automo-
To improve the flexibility of automotive software for better bile functions as services, thus greatly improving the flex-
user experience, the component subsystems of traditional ibility and reusability of the upper application software [13].
vehicles will be gradually layered, and the coupling between Above the basic software is the application software for
technical elements at different levels will be reduced [5]. specific business scenarios. Application software can be
As shown in Fig. 1, in the decoupling process, the con- divided into functional applications and service applications.
notation of the original technical elements will change, Functional applications usually involve hardware control and
and new elements will emerge. The technical elements security [14, 15], and service applications are more related
of SDV are summarized as functional hardware, electri- to content service and infotainment, such as charging, vehi-
cal/electronic architecture (EEA), computing platform, cle anomaly detection and energy service [16–19].

Fig. 1  Map of technical elements of SDV

13
182 Z. Liu et al.

In the Internet-of-Things era, most of the data generated most hardware will gradually be mastered by automotive
by software and hardware of vehicles will be transmitted enterprises instead of suppliers.
to the cloud [20]. Cloud service platform can process and
analyze the data by forming the data closed-loop, thus tap- 2.1.2 Concentration of the Computation
ping the potential of data to optimize the user experience
through over-the-air (OTA) software upgrade [21]. The With growing data generated in the car, the demand for com-
data closed-loop has been widely used in vehicle simula- puting and communication ability in vehicles is increasing.
tion, machine learning, remote repair, cloud control and Therefore, the computing unit evolves from electronic con-
other fields [22–26]. trol unit (ECU) to domain control unit to central computing
The development of SDV technology architecture can platform, and EEA evolves from distributed architecture to
be divided into three stages, starting from the “hardware- domain architecture to central centralized architecture [29].
defined stage” to the “software-controlled stage”, and Eventually, most of the computing ability in SDV will be
eventually reaching the “software-defined stage”. At dif- concentrated on a computing platform, which will be ben-
ferent stages, the software architecture, hardware archi- eficial to OTA software upgrade, improving the utilization
tecture and functional features of vehicles have different of computing resources and reducing hardware redundancy
changes. As shown in Fig. 2, three technological develop- [30].
ment trends of SDV have been illustrated. The analysis is To achieve the concentration of computation, system on
carried out below. chip (SoC) will be used as the core of computing platform to
meet the needs of different computing tasks [31]. Different
specialized chips will be integrated into the computing plat-
2.1.1 Whitening of the Functional Black Box form [32]. In addition, data interaction with large bandwidth
and low latency becomes one of the necessary conditions,
Because of the binding of hardware and software, com- for which Ethernet will be used as the backbone of the in-
ponents in traditional vehicles are black boxes for auto- vehicle network to form a new cyber-physical system [33].
motive enterprises. To support the continuous iteration
of software in the future, functional hardware needs to 2.1.3 Thickening of the Software Middle Layer
standardize the interface and abstract the function, so as
to be replaceable and upgradable [27]. Based on this, auto- To ensure the flexibility and iteration of application soft-
motive enterprises can decouple software and hardware so ware, automotive technology architecture needs a more
that data will be no longer enclosed in subsystems [28]. powerful middle layer. The basic functions required for
The original functional black boxes will be white boxes for applications need to be transferred to the service layer of
automotive enterprises. This change will continue to hap- SOA and middleware [34]. The platformization of basic soft-
pen with the development of SDV. The control models of ware supports more data interaction across subsystems and

Fig. 2  Technical development trends of SDV

13
Impact, Challenges and Prospect of Software‑Defined Vehicles 183

provides an extendible application development framework Specifically, the automotive product value ecology will
upwards [35]. undergo two major changes. On the one hand, the intelligent
Now many enterprises and researchers have invested and personalized user experience will be greatly improved,
in the R&D of service-oriented automotive middleware. such as the high-level and comfortable autonomous driving
Communication infrastructure based on IP or data distribu- [43–47]. On the other hand, costs of original automotive
tion protocol has been widely implemented [36, 37]. The products can be reduced. For example, SOA can effectively
main solutions are changing from the classic platform of reduce the complexity and cost of R&D [48], the flexible
AUTOSAR to the adaptive platform of AUTOSAR [38]. combination of software and hardware can reduce the redun-
The industry has also made a preliminary exploration on the dancy cost [49], and OTA software upgrades can also be
application of SOA in vehicles. Advanced driver assistance used to remotely repair bugs to reduce the recall cost [50].
systems and human–computer interaction systems based on As shown in Fig. 4, SDV will derive a new value chain
SOA have begun to appear in the market [39, 40]. In the along the data flow, whose core logic is to continuously uti-
future, the car-side SOA will be able to cooperate with the lize data to optimize the intelligent and personalized user
cloud-side SOA platform to provide more application ser- experience of vehicle and meet the needs of consumers. In
vices [41]. the process of data generation, with the centralization of EEA
and computation as well as the increase of sensors and digital
devices, the computing and communication capabilities of
2.2 Reshaping of the Value Ecology of Automotive the vehicle will be greatly improved so that the quantities
Products by Technological Changes and types of data generated will be increased. In the process
of data collection, layered decoupled software and service-
The key significance of the technological changes of SDV oriented application development will greatly improve the
is to promote the circulation of data in automotive industry flexibility of automotive software architecture. The improved
and fully release the potential value of data, thus reshaping software architecture enables automotive software to support
the value ecology of automotive products [42]. As shown in data integration across components, systems and domains,
Fig. 3, the value increment of traditional vehicles depends for which the potential value of data will be increased. In the
on the superposition of hardware, while the embedded soft- process of data processing, storage, analysis and utilization,
ware only assists the hardware to complete its functional cloud service platform will undertake all data-related tasks.
task. But for SDV, hardware becomes the shared resource A typical case is Tesla’s shadow model of autonomous driv-
that can be called by software, and the flexible combination ing in the cloud, which uses the real-time data of vehicles
of hardware and software can achieve more functions and to train the algorithm [51]. With the improvement of cloud
stronger performance. Based on data closed-loop, onboard platform capabilities, the utilization rate of data by related
software can even achieve self-learning and self-evolution enterprises will be increased and the data potential will be
in the stage of service. fully released. In addition, innovation will also take place in

Fig. 3  Changes in the value ecology of automotive products

13
184 Z. Liu et al.

Fig. 4  Data value chain of SDV

development of SDV in the form of smile curve. Since

software is the main measure to achieve value increment
brought by SDV, the additional value increment of the indus-
trial stages will increase with deeper software participation.
In the stage of R&D, hardware architecture is defined by
software architecture. With the increase in the quantity and
complexity of software, software will contribute the major
value. In the manufacturing stage, although hardware can
be expanded after standardization and abstraction, the value
increment is limited. In the sale and service stage, the pro-
portion of software in Bill of Materials will be constantly
increasing, and software can also make profits constantly
through OTA upgrade or subscription services. In addition,
Fig. 5  Comparison of smile curves between traditional and future relevant enterprises can provide diversified services for users
automotive industry [52] by software; thus, not only the value increment in this stage
is the most, but also the value connotation can be expanded
by broadening the automotive industrial boundary.
the services of automotive industry. The data will be applied Confronting the additional value brought by SDV, all kinds
to emerging services, such as remote repair of vehicle defects of enterprises are willing to seize the opportunities and gain
by OTA software upgrade. Through the data closed-loop, the benefits [53]. Traditional players in the industrial ecosystem,
market value of vehicles will be improved. such as automotive and component enterprises, will actively
transform, and new players will also enter the game at the same
2.3 Reconstruction of the Industrial Pattern time. The automotive industry which was fragmented and
Through New Value Ecology had high barriers will be reconstructed. As shown in Fig. 6,
the evolution of automotive industrial pattern can be divided
Figure 5 shows the change of the additional value of into three stages according to resources from an industrial the
each stage in the automotive industrial chain with the authors are working for.

Fig. 6  Three stages of the evolution of automotive industrial pattern

13
Impact, Challenges and Prospect of Software‑Defined Vehicles 185

The industrial pattern in Stage 1.0 was in a chain-type typical industrial division of labor, which are summarized
structure. Automotive enterprises coordinated the whole from extensive literature investigation and case analysis
industrial chain and have an absolute control. In addition, the [54–63].
stages of the industrial chain were relatively independent and It can be known from Table 1 that due to the limitation of
closed. For example, chip enterprises were only responsible technological level and the game between automotive enter-
for supplying chips and would not consider collaboration with prises and suppliers, the development progress of various
upper-level algorithms, and component enterprises would not technical elements of SDV is different, and there are obvi-
consider combining with products from other suppliers. At ous weak spots in the middleware and service layer of SOA.
this stage, the automotive products were limited in functions Moreover, the industrial division of labor has not yet got rid
and performance and highly homogeneous in user experience, of the shackle of the supply relationship of the traditional
which made it difficult to meet the personalized needs of users. automotive industrial chain. These reasons result in that
The industrial pattern in Stage 2.0 is in a net-type structure. the current automotive products are far from the ideal state
There are three major changes compared with stage 1.0: of SDV, and the potential value of data has not been fully
released. Many factors are hindering the realization of SDV,
(1) Since software gradually becomes the core competi- among which the outdated R&D model, backward enterprise
tiveness of automotive products, information and com- capability and unreasonable resource combination are three
munication technology (ICT) enterprises with strong major challenges.
capabilities in software R&D will enter the automotive
industry as Tier0.5 to create collaborative innovation 3.1 The Traditional R&D Model Cannot Meet
with automotive enterprises; the Needs of Vehicle Iteration
(2) Giant component enterprises and automotive enter-
prises will gradually expand their capabilities to the As shown in Fig. 7, the R&D model of traditional vehi-
ICT field, and the business boundaries between indus- cles follows the V-process, whose requirements are decom-
trial participants will gradually blur; posed layer by layer and interrelated, and each link has strict
(3) With the increasing requirements of vehicles for the restrictions and inflexible delivery deadline, which make it
integration of computing units, the importance of difficult to achieve rapid development and change of soft-
chips has been enhanced. Chip enterprises will have ware. In addition, for traditional vehicles, software can only
the opportunity to directly supply to automotive enter- be tested after being embedded in hardware, which prolongs
prises as Tier1. the iteration cycle of software, and postpones the responds
of user needs [64].
The industrial pattern in Stage 3.0 will be the structure as To achieve agile iteration of automotive products, the
a platform and an ecosystem. The platform enterprise, as the R&D of software and hardware should be separated and
integrator of all elements, will be at the center of industrial independently verified so that software and hardware can
pattern and master the right to define the vehicle architecture evolve in different cycles [65]. Moreover, software R&D
and integrate software and hardware. Hardware suppliers and also needs to achieve self-learning and self-evolution driven
software suppliers will develop various software and hard- by data with the help of a highly automated toolchain. At
ware products based on the platform and user needs, and the the same time, it should be noted that vehicle R&D has
importance of software suppliers will be further increased the necessary safety baseline and unique logic; thus, it is
with the increment of software value. Moreover, the entry of infeasible to directly copy the R&D model in the ICT field,
third-party developers will connect the automotive industry such as agile development [66]. It is necessary to effectively
to a wider external application ecosystem, such as smart cit- integrate the traditional vehicle R&D model with the R&D
ies, and the industrial boundaries will also gradually blur. model in the ICT field.

3.2 Multiple Challenges to the Transformation

3 Challenges on the Realization of SDV of Enterprise Capability

To promote the development and implementation of SDV, Most of the capabilities of traditional automotive enterprises
it is necessary to identify the key problems faced by auto- focus on the R&D and manufacturing of hardware. However,
motive industry at present. Clarifying the industrial status the construction of internal software capability is a com-
quo is the basis of analyzing the challenges. Table 1 shows plex systematic project, which needs to be changed from top
the capability status of technical elements of SDV and the to bottom, facing challenges of talent team, organizational
structure and enterprise culture.

13
186 Z. Liu et al.

Table 1  Development status of SDV technical elements

Technical element Technical capability status Typical industrial division of labor

Functional hardware Giant component enterprises master most traditional Suppliers responsible for providing hardware products
functional hardware, such as brake Automotive enterprises responsible for the integration and
ICT enterprises and chip enterprises master most new application of functional hardware
functional hardware, such as LiDAR
EEA Automotive enterprises and some giant component enter- Automotive enterprises dominate EEA R&D
prises master the core capability in EEA R&D Giant component enterprises participate in the design of
Most advanced products still in the stage of domain EEA and provide components
architecture
Computing platform Chip enterprises master the core capability in the R&D, Chip enterprises provide the SoC
manufacturing, packaging and testing of chips Automotive enterprises or giant component enterprises
Some vehicle enterprises and giant component enterprises responsible for the system integration of software and
master the capability in the integration of computing hardware
platform
OS kernel A few ICT enterprises master the core technologies ICT enterprises provide the kernel products
Some mature OS kernels can be migrated to vehicles, but Automotive enterprises or giant component enterprises
lack of customized development for vehicle responsible for the integration and deployment
Middleware ICT enterprises master the core capability in middleware Automotive enterprises or component enterprises adopt
R&D industrial mainstream standards, such as AUTOSAR, and
Some giant automotive enterprises and component the third-party toolchain for R&D
enterprises have been jointly formulating the industrial
standards
Service layer of SOA ICT enterprises have the stronger development capability Automotive enterprises or component enterprises dominate
Automotive enterprises or component enterprises have a the SOA design
deeper understanding of the vehicle’s scene and archi- ICT enterprises undertake major development tasks
tecture design
Only a few vehicle products with cockpit SOA
Functional application Giant component enterprises master the core capability in Giant component enterprises provide software bounding
controlling hardware by software with hardware
Complex functions are difficult to mass-produce in a short Automotive enterprises are responsible for the integration
time, such as self-driving and deployment of functional applications
Service application ICT enterprises master the development capability and ICT enterprises responsible for the partial development and
operation experience operation
Automotive enterprises have not completely opened the Automotive enterprises responsible for the supervision and
application ecology partial development and operation
Cloud service platform Automotive enterprises master the ownership and man- Automotive enterprises responsible for the recycling, distri-
agement of data bution and operation management of data
Suppliers master the data analysis capability of relevant Suppliers responsible for analyzing data and developing
products software
An open cloud data ecology has not yet been built

In terms of talent team, due to the safety requirements

of automotive software, the R&D team needs cross-border
talents with abilities of both hardware and software design.
The two main sources of software talents in automotive
enterprises, i.e., the Internet field and the traditional auto-
motive electronics field, are difficult to meet the require-
ments. Although the talents from the internet field have the
skills and experience in developing large-scale software,
they are not familiar with the knowledge of automotive
hardware. The talents from the traditional automotive elec-
tronics field are familiar with the R&D of the underly-
ing hardware but lack the agile development capability of
software architecture. In addition, the salary of software
talents in automotive enterprises with low-profit margins is
Fig. 7  V-model in traditional vehicle R&D not competitive with that in ICT enterprises. As a result, it

13
Impact, Challenges and Prospect of Software‑Defined Vehicles 187

is extremely difficult for automotive enterprises to recruit Besides hardware and software, the data will also cause
software talents. automotive enterprises to play games with suppliers [69].
In terms of organization, the organization structure of After the decoupling of hardware and software, to master
traditional automotive enterprises is often not flat enough the OTA software upgrade independently, automotive enter-
as it should serve the hardware R&D model. There is little prises tend to understand the software logic of components
horizontal communication between departments, and most and use the data to iterate over software. For the data of vehi-
departments follow a one-line reporting process [67]. How- cles and users belonging to automotive enterprises, when
ever, SDV requires all departments of automotive enterprises automotive enterprises do not support data backflow, the
to respond to market demands in time and carry out agile suppliers cannot iterate over their products based on big data
iteration of software, which means the organizational struc- and face the risk of losing market competitiveness. It follows
ture needs to be re-divided based on user-oriented business. that the game around data is also one of the key problems
In terms of enterprise culture, the traditional hardware hindering the realization of SDV.
R&D model leads to the rigorous engineer culture in auto-
motive enterprises. In the process of R&D, project teams
often want to make everything as controllable as possible 4 The Industrial Development Suggestions
[68]. However, there is a fundamental contradiction between for SDV
this culture and the requirement for software talents to have
innovative and agile thinking. If two kinds of talents are 4.1 Changing the Vehicle R&D Model to Adapt
blindly mixed to form a team, there will inevitably be many to SOP‑X
conflicts in thinking.
SDV requires that user experience can be continuously
3.3 Contradictions in the Industrial Division optimized through OTA software upgrade or online ser-
of Labor vice, which means that automotive enterprises must con-
tinuously operate, maintain and develop vehicles based on
Since there are too many technical elements involved in users’ needs to iterate the version of automotive products
SDV, automotive enterprises cannot do all self-R&D, and continuously, namely SOP-X. As shown in Fig. 8, this study
the division of labor and cooperation is the inevitable develops an innovative R&D model. Considering the safety
choice. At present, there are contradictions in the division requirements and integration specifications of vehicle, the
of labor and cooperation between automotive enterprises new R&D process still needs to follow the framework of
and suppliers in many fields, and if these problems are not the V-model as a whole [70]. But to achieve the agility and
properly resolved, industrial resources cannot be effectively flexibility of SOP-X, the R&D of hardware and software
combined. must be separated, and automotive enterprises should cre-
As for functional hardware, if the interface standards of ate an EEA R&D platform, a software R&D platform and
each automotive enterprise are different after standardiza- an automatic data closed-loop platform based on the char-
tion, the component enterprises need to carry out custom- acteristics, requirements and iteration cycles of hardware,
ized R&D, and the cost will be greatly increased. After the software and algorithms. The EEA R&D platform should
abstraction of the control model, the functional hardware and support the expansion and upgrade of hardware based on
functional applications can be developed independently. As standardized interface and modular design [71]. The soft-
a result, component enterprises may be deprived of the right ware R&D platform should have a complete toolchain and
of software R&D by automotive enterprises and lose market environment based on SOA to support the agile develop-
share. Therefore, it is difficult to promote the standardization ment of software. The automatic data closed-loop platform
and abstraction of functional hardware in the cooperation should support the real-time optimization of algorithms and
between automotive enterprises and component enterprises. the automation of testing through automatic recovery, clas-
As for the core software and hardware at the higher level, sification, labeling, training and evaluation of data [72].
except for the definition right of EEA and SOA, there are Before the first start of production (SOP), the main differ-
many choices for automotive enterprises in the R&D of ence between the new process and the traditional process lies
computing platform, OS kernel, middleware and applica- in the separation of R&D and testing of software and hard-
tions, such as self-R&D, outsourcing and cooperation. In ware. When the function and performance meet the basic
the division of labor of these elements, a series of problems requirements, the vehicle can be released as the initial ver-
appeared, such as how to determine the boundary of self- sion. In the use stage, the data will be fed back to the cloud
R&D, how to give full play to respective advantages in coop- service platform in real-time. To achieve SOP-X, part of data
eration and how to effectively cooperate between outsourced will flow to the software R&D platform for the R&D of new
products and products by self-R&D. functions and to the automatic data closed-loop platform

13
188 Z. Liu et al.

Fig. 8  Innovative R&D model of SDV

for the iteration of algorithms. The other part of data will The data analysis department is mainly responsible for
flow to the operation platform to help enterprises know the extracting valuable information from massive, scene-based
user’s needs, thus providing personalized mobility services and multi-sourced data [73], which can be used for the anal-
and accumulating experiences for the future R&D. ysis of user experience feedback and behavior, the monitor-
ing and diagnosis of vehicles, the training and evaluation of
4.2 Reforming the Internal Organization and Talent algorithms, and the business analysis based on big data. The
of Enterprise main talent demand of this department is data analysts who
are proficient in data processing and statistical application.
The new R&D model cannot work without the support of After a round of processing, the data that can be used
organizational structure and talents. As shown in Fig. 9, this for R&D will flow to the R&D center. The R&D center is
study designs a new organizational structure. Agile iteration divided by business applications with a flat structure. The
of software requires enterprises to use data to analyze users’ R&D of software and hardware in every team is separated
needs quickly, to develop software efficiently and to deliver but coordinated with each other, thus ensuring that software
software to users’ vehicles in time and accurately. The three and hardware can be effectively integrated to meet the func-
kinds of capabilities depend on the professional data analy- tion and performance requirements. To break the barriers
sis department, centralized R&D system and standardized between different teams and promote horizontal communi-
software management department, respectively. cation, the R&D centers should be unified at three levels:

Fig. 9  Organizational structure supporting SDV

13
Impact, Challenges and Prospect of Software‑Defined Vehicles 189

• Unified data platform: all the management, scheduling

and use of data are completed on this platform.
• Unified hardware architecture: the communication net-
work topology, hardware requirements and interfaces
have been defined in advance.
• Unified software architecture: the structure of SOA,
functional requirements and interfaces of basic software
modules have been defined in advance.

The R&D center firstly needs some architecture engineers

who have an outstanding understanding of hardware, soft-
ware and user needs to optimize the vehicle architecture.
Then, in each business team, user experience engineers are
required to give optimization suggestions to the features of
products from the user’s perspective, and software engineers
who are proficient in coding are required to upgrade the
software package. Finally, integration and test engineers are
required to integrate the software and verify the expected
function and performance.
After the development of software, the software man-
agement department is required to safely and accurately
upgrade the software on different vehicles according to dif-
ferent requirements [74]. This department is responsible for
managing software information (including software version,
update history, market stock and distribution of each vehi-
cle), software packages (including content modification,
ownership change, storage, use and deletion of software
packages), software source codes (including matching,
uploading, storage and calling of source codes) and software
defects (including safety monitoring, problem feedback,
cause finding, repair and post-optimization). Therefore,
its main talent demand is the operation and maintenance
engineer with abilities to maintain the high availability of
a series of services, and optimize the system architecture Fig. 10  Decision-making principles for industrial division of labor of
steadily to improve deployment efficiency and resource uti- automotive enterprises and suppliers
lization [75].

4.3 Combination of Industrial Ecological Resources pursuing differentiation as much as possible with sufficient
internal resources and capabilities. The differentiation poten-
SDV cannot be realized by automotive enterprises only, tial of a technical element specifically refers to whether
and it is the only way to rationally combine industrial eco- automotive enterprises can make differentiated products
logical resources, whose premise is that automotive enter- and create high added value for consumers by investing in
prises and suppliers can reasonably divide work. This study this element. The internal resources and capabilities of an
extracted the decision-making principles for industrial divi- enterprise include internal knowledge, talents, R&D tools,
sion of labor of automotive enterprises and suppliers from organizational structure and management capabilities, which
resources belonging to an industrial the authors are working can be used to distinguish between "strong automotive enter-
for (according to Fig. 10). Some feasible suggestions on the prises" and "weak automotive enterprises". There are usually
industrial division of labor are put forward based on these four choices for automotive enterprises: outsourcing, col-
principles. laboration, cooperation and self-R&D [76].
Automotive enterprises consider the industrial division Suppliers mainly consider the industrial division of
of labor from two major dimensions, i.e., the differentiation labor from two dimensions, which are the market potential
potential and the internal resources and capabilities. Their and the sustainable competitiveness. Their decision-mak-
decision-making principle can be summarized as follows: ing principle can be summarized as follows: improving

13
190 Z. Liu et al.

product competitiveness as much as possible under the 4.3.3 Industrial Division of Labor Around OS Kernel
premise of ensuring the market share. The market potential and Middleware
of a technical element specifically refers to the demand
of automotive enterprises for supplier’s products. While Since OS kernel has limited influence on the differentia-
sustainable competitiveness reflects whether suppliers tion of upper-level applications, and ICT enterprises have
can keep products differentiated, leading in the market obvious technological and cost advantages in R&D, strong
and irreplaceable for a long time to ensure survival in the automotive enterprises should consider outsourcing to sup-
industrial transformation, which can be used to distinguish pliers for customized development. While weak automotive
between "strong suppliers" and "weak suppliers". There enterprises can consider appropriately reducing technologi-
are usually four choices for suppliers: deep binding, cre- cal requirements and continue to use existing products.
ating barriers, large-scale development and customized Middleware plays a key role in the interaction between
development. functions, which has great differentiation potential, but the
There will be different choices of the industrial division complexity and cost of R&D are in a high level. Therefore,
of labor when different enterprises face the same technical strong automotive enterprises should give priority to self-
element and the same enterprise faces different technical ele- R&D of parts without mature solutions, such as self-driving
ments. Here are some specific suggestions for the industrial middleware, and other parts can adopt third-party solutions,
division of labor around each element. while weak automotive enterprises should actively seek
cooperation with strong suppliers.

4.3.1 Industrial Division of Labor Around Functional 4.3.4 Industrial Division of Labor Around EEA
Hardware and the Service Layer of SOA

From the perspective of industrial development, the stand- EEA and the service layer of SOA determine the integration
ardization and abstraction of functional hardware is an of hardware and software of the whole vehicle, respectively,
inevitable trend and the common demand of all automotive which has a great influence on the differentiation of vehi-
enterprises. Therefore, suppliers should not refuse to change, cles. Thus, most automotive enterprises cannot give up the
instead, they should actively strive for reasonable benefits definition right of these two elements, and suppliers should
in the cooperation with automotive enterprises. Strong sup- actively cooperate, such as providing the automotive enter-
pliers can choose to work with automotive enterprises to prises with R&D tools and architecture design ideas.
formulate interface standards and cooperate in the R&D
of corresponding software solutions. Weak suppliers can 4.3.5 Industrial Division of Labor Around Applications
maintain a long-term supply relationship with automotive
enterprises to make up for the extra costs. In the long run, For functional applications related to safety and hardware
the whole automotive industry can formulate a unified func- control, strong automotive enterprises should strive for self-
tional interface standard and create an industrial hardware R&D as much as possible. For those parts that cannot be
platform, thus promoting the division of labor and coopera- self-developed at present, enterprises can temporarily seek
tion around functional hardware. help from suppliers and gradually achieve self-substitution
in the future. Weak automotive enterprises should maintain
a long-term strategic cooperative relationship with the sup-
4.3.2 Industrial Division of Labor Around Computing pliers to ensure the continuous optimization of functions
Platform and performance. For service applications, all automotive
enterprises should create an open ecosystem and get third-
The industrial division of labor around computing platforms party developers from the field of consumer electronics or
can be divided into R&D and manufacturing. In the R&D Internet involved so that ecological resources can effectively
phase, automotive enterprises and chip enterprises should meet the individual needs of users.
cooperate deeply to achieve the complementarity of scene
understanding and design ability. Strong enterprises should 4.3.6 Industrial Division of Labor Around Cloud Service
focus on improving abilities of chip design to master more Platform
definition rights in the cooperation. In the manufacturing
phase, regardless of the manufacturing process or cost con- Automotive enterprises should be responsible for the opera-
trol, chip enterprises have absolute advantages over automo- tion, management and supervision of the cloud service plat-
tive enterprises [77]; thus, automotive enterprises should form. Thus, the key problem of the industrial division of
outsource to suppliers. labor around cloud service platforms is to avoid the game

13
Impact, Challenges and Prospect of Software‑Defined Vehicles 191

around the data between automotive enterprises and suppli- Continuous sales of software and services with high prof-
ers. This problem needs to be solved in different scenarios: itability have become a new business model of SDV, and
For strong automotive enterprises and strong suppliers, brand-new supply relationships are needed. As shown in
both have strong capabilities, and there may be competition Fig. 11, for the consumer-oriented business, automotive
between them in the R&D of software. Both sides should try enterprises can let users choose software at the first sale,
their best to avoid the complete closure of data because the and continuously sell new software or subscription services
necessary data flow is of positive significance to the indus- through OTA upgrade after delivery. Even in the future, not
trial ecosystem [78]. The two sides can exchange reasonable only automotive enterprises will face consumers directly,
interests, for example, the suppliers can pay some authoriza- third-party developers can also directly connect with con-
tion fees to the automotive enterprises to get data. sumers through APP store [79]. For the enterprise-oriented
For strong automotive enterprises and weak suppliers, the business, suppliers can sell codes, software packages and
analysis and utilization of data will inevitably be dominated supporting technical services to automotive enterprises.
by automotive enterprises, while automotive enterprises Besides one-time buyout, suppliers can also charge licens-
should selectively share some data with cooperative suppli- ing fees or sales shares.
ers, thus full taking suppliers’ technological advantages in The new supply relationships and business models will
some elements to achieve the cost reduction or experience effectively promote the implementation of the industrial
improvement. division of labor and turn it into real income, thus supporting
For weak automotive enterprises and strong suppliers, the rational combination of industrial ecological resources
the two sides should maintain a strategic cooperation rela- and further pushing forward the change of the automotive
tionship. Automotive enterprises should share the suppliers industrial ecosystem.
related data after desensitization, and suppliers should make
customized development plans according to the needs of
automotive enterprises so that the products can be continu- 5 Conclusions
ously improved and automotive enterprises have a relatively
strong position in the product definition. This study identifies three major technological trends of
For weak automotive enterprises and weak suppliers, SDV, which are the concentration of the computation, whit-
the two sides should not be antagonistic, rather they should ening of the functional black box and thickening of the soft-
achieve data sharing, cooperative analysis and utilization of ware middle layer. As a result, software and hardware are
data, and even team integration. decoupled, and software can freely call and control hard-
ware, thus realizing the continuous evolution of automo-
4.3.7 The Supply Relationships and Business Models tive products driven by data and continuously creating new
Supporting the Industrial Division of Labor experiences for users. Under this background, the value ecol-
ogy of automotive products will gradually be dominated by
The above-mentioned industrial division of labor cannot be software and attract old players to actively transform and
sustained for a long time without a money guarantee, and new players to enter the market.
the traditional business model of one-off sales of vehicles However, because it is difficult to accomplish the
is unable to provide such guarantee for relevant enterprises. improvement of technological level, the change of R&D

Fig. 11  New supply relationships and business models of SDV

13
192 Z. Liu et al.

model, the transformation of enterprise capabilities and the chains to functional networks. In: Proceedings of the 3rd Inter-
industrial division of labor, the effective combination of national Conference on Vehicle Technology and Intelligent
Transport Systems, INSTICC, Porto (2017)
industrial ecological resources has not yet been achieved. 6. Li, X., Yu, K.: Moving towards super vehicle central com-
The development process of various technical elements of puter—the innovation of intelligent vehicle electronic and
SDV is different, and the ideal products cannot be effec- electrical architecture to meet the digital transformation. Micro
tively created. The following suggestions are put forward for Nano Electron. Intell. Manuf. 1(02), 62–71 (2019)
7. Shao, N., Zhang, Q., Wang, Z., et al.: The evolution of automo-
automotive enterprises, suppliers, government and industrial tive electronic and electrical architectures. Sci. Technol. Innov.
organizations: 2020(35), 98–100 (2020)
8. Bjelica, M., Lukac, Z.: Central vehicle computer design: soft-
(1) As the owner of automotive products and brands, auto- ware taking over. IEEE Consum. Electron. Mag. 8(6), 84–90
(2019)
motive enterprises should take the lead in the develop- 9. Wang, Q., Su, D.: Research on the development of smart and
ment of SDV. It is essential to focus on the creation connected vehicle operating system. Inform. Commun. Technol.
of different user experiences based on the brand-new Policy. 2019(09), 57–60 (2019)
organization, team and R&D mechanism. Choosing 10. Lingga, W., Budiman, B., Sambegoro, P.: Automotive real-
time operating system in vehicular technology progress review.
suitable partners and business models to jointly create Paper presented at the 6th International Conference on Electric
an open ecosystem is also important. Vehicular Technology. IEEE, Bali (2019)
(2) On the basis of increasing investment in software, sup- 11. Iorio, M., Buttiglieri, A., Reineri, M., et al.: Protecting in-vehi-
pliers should make full use of their technological or cle services: security-enabled SOME/IP middleware. IEEE Veh.
Technol. Mag. 15(3), 77–85 (2020)
cost advantages in some elements to meet the needs of 12. Kugele, S., Obergfell, P., Broy, M., et al.: On service-orientation
automotive enterprises from customization, standardi- for automotive software. Paper Presented at 2017 IEEE Interna-
zation to other dimensions. Besides, suppliers should tional Conference on Software Architecture. IEEE, Gothenburg
also actively participate in the cooperative industrial (2017)
13. Cebotari, V., Kugele, S.: On the nature of automotive service
ecosystem and try to play an indispensable role. architectures. Paper Presented at the 2019 IEEE International
(3) The government or industrial organizations should pro- Conference on Software Architecture Companion. IEEE, Ham-
mote the technical development and industrial coop- burg (2019)
eration of SDV from the macro or medium level, such 14. Li, K., Dai, Y., Li, S., et al.: State-of-the-art and technical trends
of intelligent and connected vehicles. J. Automot. Saf. Energy.
as promoting the standard setting of common technical 8(1), 1–14 (2017)
elements like basic software and building the ecologi- 15. China Society of Automotive Engineering: Strategic Advisory
cal cooperation platform to help enterprises connect. Committee of Energy-saving and New Energy Vehicle Technol-
ogy Roadmap. China Machine Press, Beijing (2020)
16. Fleming, B.: Smarter cars: incredible infotainment, wireless
device charging, satellite-based road taxes, and better EV bat-
Acknowledgements This work was supported by National Natural Sci- teries. IEEE Veh. Technol. Mag. 8(2), 5–13 (2013)
ence Foundation of China (U1764265). 17. Cao, Y., Song, H., Kaiwartya, O., et al.: Mobile edge computing
for big-data-enabled electric vehicle charging. IEEE Commun.
Declarations Mag. 56(3), 150–156 (2018)
18. Zhang, M., Chen, C., Wo, T., et al.: Safedrive: online driving
Conflict of interest On behalf of all the authors, the corresponding au- anomaly detection from large-scale vehicle data. IEEE Trans.
thor states that there is no conflict of interest. Ind. Inf. 13(4), 2087–2096 (2017)
19. Grée Laznikova, V., Kim, B., Garcia, G., Gao, B.: Cloud-based
big data platform for vehicle-to-grid (v2g). World Electric. Veh.
J. 11(2), 30 (2020)
20. Kim, Y., Oh, H., Kang, S.: Proof of concept of home IoT con-
References nected vehicles. Sensors 17(6), 1289–1301 (2017)
21. Wang, Z., Han, J., Miao, T.: An efficient and dependable
1. Zhao, F., Liu, Z., Hao, H., Shi, T.: Characteristics, trends and FOTA-based upgrade mechanism for in-vehicle systems. Paper
opportunities in changing automotive industry. J. Automot. Saf. Presented at the 2019 International Conference on Internet of
Energy 9(3), 233–249 (2018) Things, IEEE, Atlanta (2019)
2. Liu, Z., Song, H., Hao, H., Zhao, F.: Innovation and development 22. Zhang, Y., Lu, S., Yang, Y., Guo, Q.: Internet-distributed vehi-
strategies of China’s new-generation smart vehicles based on 4S cle-in-the-loop simulation for HEVS. IEEE Trans. Veh. Tech-
integration. Strateg. Study CAE 23(03), 153–162 (2021) nol. 67(5), 3729–3739 (2018)
3. Vdovic, H., Babic, J., Podobnik, V.: Automotive software in con- 23. Yang, Z., He, Z.: Application of improved genetic algorithm in
nected and autonomous electric vehicles: a review. IEEE Access vehicle networked cloud data platform. Paper Presented at the
7, 166365–166379 (2019) International Conference on Intelligent Transportation. IEEE,
4. McKinsey.: Automotive software and electronics 2030. https://​ Xiamen (2018)
max.​book1​18.​com/​html/​2019/​0924/​60121​12021​002110.​shtm 24. Eichel, J.A., Mishra, A., Miller, N., et al.: Large-scale machine
(2020). Accessed 11 Nov 2021 learning and evaluation platform for real-time traffic surveil-
5. Bach, J., Otten, S., Sax, E.: A taxonomy and systematic lance. J. Electron. Imaging 25(5), 1–14 (2016)
approach for automotive system architectures—from functional

13
Impact, Challenges and Prospect of Software‑Defined Vehicles 193

25. Giannetti, V.: Srinivasan: the cloud and its silver lining: nega- 42. Soley, A., Siegel, J., Suo, D., et al.: Value in vehicles: economic
tive and positive spillovers from automotive recalls. Mark. Lett. assessment of automotive data. Digital Policy, Regulation and
32(4), 397–409 (2021) Governance 20(6), 513–527 (2018)
26. Esen, H., Adachi, M., Bernardini, D., et al.: Control as a service 43. Li, J., Cheng, H., Guo, H., et al.: Survey on artificial intelligence
(CaaS): cloud-based software architecture for automotive con- for vehicles. Automot. Innov. 1, 2–14 (2018)
trol applications. In: Proceedings of the Second International 44. Yang, D., Jiao, X., Jiang, K., et al.: Driving space for autono-
Workshop on the Swarm at the Edge of the Cloud. ACM, Seattle mous vehicles. Automot. Innov. 2, 241–253 (2019)
(2015) 45. Clark, J., Stanton, N., Revell, K.: Automated vehicle handover
27. Sutopo, W., Kadir, E.: Designing framework for standardization interface design: focus groups with learner, intermediate and
case study: lithium-ion battery module in electric vehicle applica- advanced drivers. Automot. Innov. 3, 14–29 (2020)
tion. Int. J. Electric. Comput. Eng. 8(1), 220–226 (2018) 46. Hu, J., Cai, S., Huang, T., et al.: Vehicle travel destination pre-
28. Mckinsey&Company.: The case for an end to end automotive diction method based on multi-source data. Automot. Innov. 4,
software platform. https://​www.​mckin​sey.​com/​~/​media/​McKin​ 315–327 (2021)
sey/I​ ndust​ ries/A
​ utomo​ tive%2​ 0and%2​ 0Asse​ mbly/O
​ ur%2​ 0Insi​ ghts/​ 47. Stevic, S., Lazic, Bjelica, M.Z., Lukic, N.: IoT-based software
The%2​ 0case%2​ 0for%2​ 0an%2​ 0end%2​ 0to%2​ 0end%2​ 0auto​ motiv​ e%​ update proposal for next generation automotive middleware
20sof​tware%​20pla​tform/​The-​case-​for-​an-​end-​to-​end-​autom​otive-​ stacks. Paper Presented at the 8th International Conference on
softw​are-​platf​orm.​ashx (2020) Accessed 11 Nov 2021 Consumer Electronics. IEEE, Berlin (2018)
29. Navale, V., Williams, K., Lagospiris, A., et al.: Revolution of E/E 48. Zheng, M., Zada, I., Shahzad, S., et al.: Key performance indi-
architectures. SAE Int. J. Passenger Cars Electron. Electric. Syst. cators for the integration of the service-oriented architecture
8(2), 282–288 (2015) and scrum process model for IOT. Sci. Program. 2021(1), 1–11
30. Ayres, N., Deka, L., Passow, B.: Virtualisation as a means for (2021). https://​doi.​org/​10.​1155/​2021/​66135​79
dynamic software update within the automotive E/E architecture. 49. Larin, S.: Exploiting program redundancy to improve perfor-
Paper Presented at the 2019 IEEE SmartWorld. IEEE, Leicester mance, cost and power consumtion in embedded systems. Dis-
(2019) sertation, North Carolina State University (2000)
31. Tabani, H., Mazzocchetti, F., Benedicte, P., et al.: Performance 50. Bauwens, J., Ruckebusch, P., Giannoulis, S., et al.: Over-the-air
analysis and optimization opportunities for NVIDA automotive software updates in the internet of things: an overview of key
GPUs. J. Parallel Distrib. Comput. 152, 21–32 (2021) principles. IEEE Commun. Mag. 58(2), 35–41 (2020)
32. Iwabuchi, K., Uchida, D., Ishida, Y., et al.: The collaboration with 51. Hardman, S., Chakraborty, K.E.: A quantitative investigation
FPGA and RT-Middleware by AP SoC. Paper Presented at the into the impact of partially automated vehicles on vehicle miles
JSME Annual Conference on Robotics and Mechatronics. Japa- travelled in California. Institute of Transportation Studies,
nese Society of Mechanical Engineers, Tokyo (2018) Davis (2021)
33. Poudel, B., Munir, A.: Design and evaluation of a reconfigurable 52. Kuang, X., Zhao, F., Hao, H., et al.: Intelligent connected vehi-
ECU architecture for secure and dependable automotive CPS. cles: the industrial practices and impacts on automotive value-
IEEE Trans. Depend. Secure Comput. 18(1), 235–252 (2018) chains in China. Asia Pac. Bus. Rev. 24(1), 1–21 (2018)
34. Gopu, G., Kavitha, K., et al.: Service oriented architecture based 53. Maldonado, G., Garza, R.: Eco-innovation practices’ adoption in
connectivity of automotive ECUs. Paper Presented at the 2016 the automotive industry. Int. J. Innov. Sci. 12(1), 80–98 (2020)
International Conference on Circuit, Power and Computing Tech- 54. Liu, Z., Shi, T., Hao, H., et al.: Current situation, development
nologies. IEEE, Nagercoil (2016) demand and future trend of automotive technologies in China.
35. Becker, M., Lu, Z., Chen, D.: Towards QoS-aware service-ori- Automob. Technol. 1, 1–6 (2017)
ented communication in E/E automotive architectures. Paper 55. Bello, L., Mariani, R., Mubeen, S., et al.: Recent advances and
Presented at the 44th Annual Conference of the IEEE Industrial trends in on-board embedded and networked automotive sys-
Electronics Society. IEEE, Washington (2018) tems. IEEE Trans. Ind. Inf. 15(2), 1038–1051 (2019)
36. Nichitelea, T.C., Unguritu, M.G.: Automotive ethernet applica- 56. Zhou, Z., Lee, J., Berger, M.S., et al.: Simulating TSN traffic
tions using scalable service-oriented middleware over IP: service scheduling and shaping for future automotive Ethernet. J. Com-
discovery. Paper Presented at the 24th International Conference mun. Netw. 23(1), 53–62 (2021)
on Methods and Models in Automation and Robotics. IEEE, 57. Kaiser, C., Festl, A., Pucher, G., et al.: The vehicle data value
Miedzyzdroje (2019) chain as a lightweight model to describe digital vehicle ser-
37. Takrouni, M., Hasnaoui, A., Mejri, I., Hasnaoui, S.: A new meth- vices. Paper Presented at the 15th International Conference on
odology for implementing the data distribution service on top Web Information Systems and Technologies. Delft University
of gigabit ethernet for automotive applications. J. Circuits Syst. of Technology, Vienna (2019)
Comput. 29(13), 205–210 (2020) 58. Detlef, Z., Darren, B.: Paradigm shift in the market for automo-
38. Gaglio, S., Re, G., Martorella, G., Peri, D.: A middleware to tive software. ATZ Worldwide 121, 28–33 (2019)
develop and test vehicular sensor network applications. Paper 59. Gal, M., Kifor, C.: Human resources assignment in R&D depart-
Presented at the 2019 AEIT International Conference of Electrical ments from automotive industry. Paper Presented at the Man-
and Electronic Technologies for Automotive. IEEE, Turin (2019) agement, Knowledge and Learning International Conference
39. Lotz, J., Vogelsang, A., Benderius, O., et al.: Microservice Archi- 2020. Expanding Horizons Business, Management and Tech-
tectures for Advanced Driver Assistance Systems: A Case-Study. nology for Better Society (2020)
Paper Presented at the 2019 IEEE International Conference on 60. Deloitte.: Software is transforming the automotive world—four
Software Architecture Companion. IEEE, Hamburg (2019) strategic options for pure-play software companies merging
40. Sharma, H., Kuvedulibla, R., Ramani, A.K.: Component oriented into the automotive lane. https://​mp.​weixin.​qq.​com/s/​QO5L3-
human machine interface for in-vehicle infotainment applications. I-​IPY0M​LOk8S​gh7Q (2020). Accessed 11 Nov 2021
Lect. Notes Eng. Comput. Sci. 2170(1), 1–4 (2008) 61. Tiwari, J.D.: Strategic implications of standardized software
41. Iwai, A., Aoyama, M.: Automotive cloud service systems based on platforms in automotive industry: impact of adaptive AUTO-
service-oriented architecture and its evaluation. Paper Presented SAR in industry 4.0. Dissertation, Coventry University (2019)
at the 2011 IEEE International Conference on Cloud Computing.
IEEE, Washington (2011)

13
194 Z. Liu et al.

62. Zhang, D., Lv, C., Yang, T., et al.: Cyber-attack detection for 78. Turgut, D., Boloni, L.: Value of information and cost of privacy in
autonomous driving using vehicle dynamic state estimation. Auto- the internet of things. IEEE Commun. Mag. 55(9), 62–66 (2017)
mot. Innov. 4, 262–273 (2021) 79. Macario, G., Torchiano, M., Violante, M.: An in-vehicle info-
63. AUTOSAR Partnership.: Achievements and exploitation of the tainment software architecture based on Google Android. Paper
AUTOSAR development partnership. Paper presented at Euro Presented at the IEEE International Symposium on Industrial
forum conference. SAE, Detroit (2006). https://w ​ ww.s​ ae.o​ rg/p​ ubli​ Embedded Systems. IEEE, Lausanne (2009)
catio​ns/​techn​ical-​papers/​conte​nt/​2006-​21-​0019/
64. Wei, X., Dai, H., Sun, Z.: Methology, architicture and develop-
ment flow of automotive embedded systems. J. Tongji Univ. (Nat. Zongwei Liu is the associate professor of the
Sci.) 40(07), 1064–1070 (2012) School of Vehicle and Mobility at Tsinghua
65. Staron, M.: Automotive software development. ATZelectronics University and assistant director of Tsinghua
Worldwide (2020) Automotive Strategy Research Institute. His
66. Appello Bernardi, P., Bugeja, C., Pollaccia, G., et al.: An opti- research is on automotive industry develop-
mized test during burn-in for automotive SoC. IEEE Des. Test ment strategies, technology innovation sys-
35(3), 46–53 (2018) tem and technology path decisions. He has
67. Marx, T.: The impacts of business strategy on organizational published more than 60 papers and won 15
structure. J. Manag. Hist. 22(3), 249–268 (2016) granted patents. He serves as the associate
68. Briody, E.K., Trotter, R.T., Meerwarth, T.L.: Significant Cultural editor-in-chief of Automotive Innovation.
Transformations in the Automotive Industry. Palgrave Macmillan
US (2010)
69. Beier, G., Kiefer, J., Knopf, J.: Potentials of big data for corporate
environmental management: a case study from the German auto-
motive industry. J. Ind. Ecol. 24(4), 1–14 (2020). https://​doi.​org/​ Wang Zhang is a Ph.D. candidate at Tsinghua
10.​1111/​jiec.​13062 University. He obtained the B.E. degree at
70. Nazareth, D., Siwy, R.: Development of an AUTOSAR software Tsinghua University of Automotive Engi-
component based on the V-model. Paper Presented at the FISITA neering in 2021. His research interests
2012 World Automotive Congress. Springer, Berlin (2013) include technology strategy of intelligent
71. Zhao, F., Liu, Z., Li, Z.: Development mode and implementation connected vehicles, automotive industry eco-
strategy of automotive product platform and modularity. Auto- system and software-defined vehicles.
mob. Technol. 2017(6), 1–6 (2017)
72. Huang, Y., Mcmurran, R., Amor-Segan, M., et al.: Development
of an automated testing system for vehicle infotainment system.
Int. J. Adv. Manuf. Technol. 51(4), 233–246 (2010)
73. Placho, T., Schmittner, C., Bonitz, A., et al.: Management of
automotive software updates. Microprocess. Microsyst. 78(1), Fuquan Zhao is the professor of the School of
287–295 (2020) Vehicle and Mobility at Tsinghua University,
74. Baouya, A., Mohamed, O., Ouchani, S., et al.: Reliability-driven the director of Tsinghua Automotive Strategy
automotive software deployment based on a parametrizable Research Institute and the honorable lifetime
probabilistic model checking. Expert Syst. Appl. 174(1), 114–132 president of the International Federation of
(2021). https://​doi.​org/​10.​1016/j.​eswa.​2021.​114572 Automotive Engineers Societies. He serves as
75. Protzmann, R., Hübner, A., Bauknecht, U., Witt, A.: Large-scale the Editor-in-chief of Automotive Innovation
modeling of future automotive data traffic towards the edge cloud. and editorial board member of 6 internation-
Paper Presented at the 20th ITG-Symposium. IEEE, Leipzig ally renowned academic journals. His research
(2019) interests include automotive industry develop-
76. RolandBerger.: The changes of automotive supply chain under ment, corporate operation and management,
the trend of SDV. https://​www.​sohu.​com/a/​42614​0581_​372592 technology development roadmaps.
(2020). Accessed 11 Nov 2021
77. Lee, C., Kim, S.W., Yoo, C.: VADI: GPU virtualization for an
automotive platform. IEEE Trans. Ind. Inf. 12(1), 277–290 (2017)

13