Data-Driven Development,

A Complementing Approach for Automotive Systems Engineering

Johannes Bach1 , Jacob Langner1 , Stefan Otten1 , Marc Holzäpfel2 and Eric Sax1

Abstract— Established methods and processes in the field 10%

of Automotive Systems Engineering (ASE) are challenged by

70
the rising complexity of current features. Expanding system
boundaries, tighter interconnections of functional elements, in-
creasingly complex algorithms and an ever growing operational
domain generate a multitude of different scenarios that require Environmental Perception Networking & Dependencies
consideration during specification, design, implementation and
testing. This paper reflects the current practice on the ex-
ample of the Automotive SPICE process reference for system
and software development in the automotive domain. It then
contemplates on opportunities of consistent usage of recorded
vehicle data throughout all phases of automotive development. Algorithmic Complexity Multitude of Scenarios
Our concept of data-driven development is not intended to
replace the current practice but to complement it. A summary Fig. 1: Current challenges concerning processes, methods
of our previous work demonstrates the practicability of the and tools for ASE
concept on the basis of the development of a Predictive Cruise
Control (PCC) feature. The contribution concludes with a elements for sensing and actuation. This in turn increases
scalable concept for the large scale application of data-driven the complexity and connectivity between features far beyond
development in ASE. the border of a single Electronic Control Unit (ECU). The
resulting dependencies between various features require new
I. I NTRODUCTION
ways of coordination and communication between the re-
The current trends automation, electrification and con- sponsible development teams, divisions and companies, since
nectivity [1] affect the well established development pro- descriptive and conceptual modeling are no longer sufficient
cesses in the automotive industry. Established development to carry all relevant information.
principles and methods have to be adopted to these trends
in order to meet the new and changing requirements. New To further add to the pool of current challenges, with the
features, reaching from advanced sensor systems and digital increase of algorithmic complexity, also the complexity and
maps to back-end services, impact the entire development extend of required V&V activities for assisting and automa-
process from system specification to Verification and Va- ting features is growing. Not only the functionality of each
lidation (V&V). The challenges of growing numbers, size, element on its own but the complete functional chain has to
complexity and interconnections of automotive features [2] be tested. A multitude of scenarios and situations, for ex-
are magnified by shorter time to market demands and increa- ample varying weather conditions, have to be considered for
singly complex algorithms, which are facilitated by improved sufficient test coverage [4]. Worldwide distribution demands
hardware performance. proof for compliance with various country specific rules and
Figure 1 depicts current major challenges concerning regulations. The sheer number of possible test scenarios for
processes, methods and tools for ASE. The introduction worldwide valid V&V of assisting or automating features
of digital maps and the advancement in sensor technology constitutes an enormous problem for development teams.
expands the vehicle’s environmental perception [3]. Newly In this contribution, we propose a data-driven development
emerging back-end services further enhance this sphere of approach to complement the existing methods and strategies
perception with real time traffic flow data and other informa- for ASE. The approach describes possible benefits and use
tion, magnifying the amount of data and types of information cases for application of recorded real world driving data in
available in a modern vehicle. This leads to complex depen- every phase of the development process. It represents an
dencies between data channels and fuzzy system borders, opportunity to overcome the described challenges.
which challenge current specification techniques. The paper is structured as follows. In chapter II the state of
Besides the amount of signals and sensors, also the total the art in ASE is introduced and open challenges are idetified.
number of features in a vehicle and the algorithmic com- In section III an outline of approaches is presented, that are
plexity is rising. Moreover, newly build features are often summarized under the term data-driven development. The
of a high-level nature and utilize other lower-level function application of the data-driven development is explained by
1 Johannes Bach, Jacob Langner, Stefan Otten and Eric Sax are with FZI
means of our prototype implementation for the development
Research Center for Information Technology, 76131 Karlsruhe, Germany of a PCC feature in section IV. In the subsequent section
{bach, langner, otten, sax}@fzi.de
2 Marc Holzäpfel is with Dr. Ing. hc. F. Porsche AG, 71287 Weissach, V, the necessity for scalability is discussed and an extension
Germany [email protected] concept of the prototype implementation is presented. The

978-1-5386-3403-5/17/$31.00
Authorized ©2017 IEEE
licensed use limited to: UNIVERSITÄTSBIBLIOTHEK PADERBORN. Downloaded on August 24,2023 at 11:51:41 UTC from IEEE Xplore. Restrictions apply.
 Requirements System
contribution concludes in section VI with a conclusion and Elicitation

System
an outlook on future work. Requirements Qualification
Analysis Subsystem
II. S TATE - OF - THE -A RT Architectural
Integration
Design
The field of systems engineering encompasses methods,
Requirements Component
approaches and tools for ”the engineering of complex sy- Analysis Longitudinal Control Qualification

Software
stems” [5]. In the automotive domain, a widely accepted Architectural
Integration
approach to structure the necessary development activities is Design Unit
Unit calcDist()
the V-Model [6]. The process reference Automotive SPICE1 Construction
float calcDist(const float &dx, const float &dy)
{
float dist = 0.0f;
dist = sqrt(pow(dx, 2) + pow(dy, 2));
return dist;
Verification
tailors this approach according to the specific demands and
}

constraints of ASE and provides a guideline for process Fig. 2: Outline of Automotive SPICE development phases
implementations [7]. It encourages the partitioning of the and partitioning of a Cruise Control feature on system,
vehicle, seen as a system, into smaller and better mana- subsystem, component and unit levels
geable elements. Figure 2 depicts the Automotive SPICE
system and software engineering processes and an exemplary analysis. A frequent industry-wide occurrence are well honed
partitioning into system, subsystem, component and unit prototypes, which serve as an example for the detailed
levels, using the example of a PCC feature. While the V- specification of requirements.
model quickly implies a strict linear development approach, The software architectural design splits the software com-
implemented processes often realize an iterative procedure ponent into individual units, which fulfill specific parts of the
on system and subsystem levels and incremental approaches requirements. It describes the interactions of all elements and
on component and unit levels. the necessary interfaces. Modeling and simulation techniques
The development begins at system level with the elicitation support the software design with opportunities to evaluate the
of the system requirements. The aim of this process is overall product in early stages [11].
to establish a clear representation of all relevant needs The unit construction represents the implementation of
and requirements of the involved stakeholders. The applied the specified software units according to the detailed design
methods and techniques include the analysis based on sce- and the allocated requirements. During unit construction
narios, the identification of use and misuse cases and more various tools for debugging and automated assessments, such
formal techniques such as FODA (Feature Oriented Domain as static code analysis, support the developers. To provide
Analysis) [8]. them with a possibility to evaluate the interaction of the
The gathered stakeholder requirements are subsequently implemented software unit with its adjacent units, functional
analyzed to derive the more formal and technical system mock-ups and simulation based techniques are utilized.
requirements, which facilitate the system design process. The right leg of the V-model is dedicated to Verification
The requirements analysis can be supported by consistent and Validation of the implemented and integrated items [12].
usage of modeling, visualization and templating techniques. The goals of V&V differ on each level. Unit tests aim at
An exemplary approach is the Systems Modeling Language verifying the correct functionality of each software unit,
(SysML) [9] with its activity and sequence diagrams, to whereas the goal of software component tests is to verify and
cite a few. A full traceability and documentation of the validate the interaction between and integration of software
system requirements specified strongly supports all other units. On subsystem and system level the overall integration
activities. Methods for V&V of requirements range from of the implemented software in the ECU hardware and the
manual reviews and throw-away prototypes (e.g. paper mock- correct interaction with the mechanical elements need to
ups for user interfaces) to more elaborate simulation based be proven. This includes the provision of evidence for the
approaches and functional prototypes. requirement consistent realization of the chain of effects of
During the design, a system architecture is developed, the distinct features.
that fulfills the specified requirements. The system, including Varying goals require different testing methods on each
all features, is broken down into individual subsystems, abstraction layer. Unit verification is usually performed with
that implement interrelated functional elements of a specific small software testing blocks or by manual code reviews. For
feature or feature group [10]. Based on the system archi- integration tests on the component level, testers utilize mock-
tectural design, the development of mechanical, hardware ups and simulation based approaches, such as Software-in-
and software components is split up. In this contribution, the-loop (SIL) [13]. Specifically designed integration tests
we focus on the software part of the development. make use of these techniques to verify the correct interaction
The software requirements are derived from the results of between software units.
the system phases. Software specific details complement the Software components, that are integrated into an ECU
functional description of the system level requirements on the are tested with Hardware-in-the-loop (HIL) frameworks,
component level. The methods applied for this activity pretty which emulate the real system. With proceeding integration,
much correspond to the methods of the system requirements these frameworks are expanded to full subsystems. These
1 Automotive Software Process Improvement and Capability Determina- frameworks are often combined with simulation techniques,
tion that reproduce unavailable items of the system or the system

Authorized licensed use limited to: UNIVERSITÄTSBIBLIOTHEK PADERBORN. Downloaded on August 24,2023 at 11:51:41 UTC from IEEE Xplore. Restrictions apply.
 Prototypes Field Trials Test Campaigns
environment. Models replace these items with abstract repre-

Sources
sentations during the tests. In addition to the described test
techniques, different stages of prototype evaluation models

Elict. & Analys.

Requirements
¾ improved statistical

Integration &
Qualification
are constructed and integrated. They are used in field trials ¾ Contextual information
significance
of system & environment
and test campaigns for real world validation of the developed ¾ shortened iterations
¾ statistical analysis
¾ increased effectiveness
¾ identification of patterns
items. and efficiency of testing

The described V&V methods mainly employ specification Recorded Scenarios

& Verification
¾ detailed understanding ¾ Support of debugging

Construction
SW Design
System &
and in-depth analysis of signal on component and subsystem
based techniques. With the introduced increase of product interfaces level during implementation
complexity, the detailed manual specification of a sufficient ¾ support of critical design choices ¾ analysis of detailed design and
¾ improved coordination implementation decisions
number of variants becomes more and more difficult to
muster. Sophisticated methods for automated test case ge- Fig. 3: Applications of data-driven development to support
neration often fail due to the inherent complexity of these different development phases.
approaches. In computer vision, usage of video data recorded
engineers and developers with a deeper understanding of the
in real world is a very common approach [14]. Besides the
targeted feature. During the implementation phase, specific
simplicity of gathering this data during anyway inevitable
scenarios in the recorded data are selected for simulation
real world tests, an important reason is the immense variety
based verification of the functional behavior on system level.
of optical effects within a single frame and the associated
In the later Automotive SPICE stages integration and quali-
modeling effort for sufficiently accurate generation.
fication, the exploitation of large amounts of data in system
Up to now, the data gathered during real world validation
simulation widens the space of covered variants significantly
is mainly used for specific analyses of the individual feature’s
and complements the currently applied specification based
functional behavior. The information included within these
approaches for V&V.
data can be further exploited to complement the established
specification based techniques and to cover a broader range
A. Requirements Elicitation and Analysis
of variants. In the following, we provide an overview of
possible usage scenarios for the individual development During requirements elicitation and analysis recorded data
activities. can be used to better understand the system as well as
its environment and contextual purpose. With this deeper
III. DATA -D RIVEN D EVELOPMENT understanding of the feature and its surrounding elements
The term data-driven development appears in context of and inputs, the specification of requirements and constraints
product and software development2 as well as economics3 . on system and subsystem level is facilitated. In this phase a
It pools approaches, which utilize data and data analytics broad variety of test scenarios with limited depth is suitable
to substantially influence business or design decisions. This to be analyzed by a generic set of statistical methods to
section presents usage scenarios of data-driven development answer simple questions like why, when, and where certain
for ASE. things happen to whom or what. Recorded driving data
During real world tests with prototype vehicles and series contains data from multiple sensors and input signals relevant
cars for quality assessment, usually the entire communication to the developed system, which allows parameter value
of the vehicle is consistently recorded with dedicated har- observation in a multidimensional space. Data visualization
dware for data logging. The captured communication mes- methods facilitate the analysis of the complex dynamical
sages include measured states of the mechanical components, behavior of input signals and allow exploration of the dyna-
such as velocity and momentum, as well as objects detected mics and ranges to be expected. An analysis of dependencies
in the vehicle environment by vehicle sensors. The goal of between signals might uncover other useful information like
our approach is to skim the benefits that a consistent usage correlations between allegedly independent parameters or
of recorded real world driving data can offer throughout the refutation of expected dependencies.
development process. As shown in Figure 3, we grouped For example an analysis of steering wheel angle, lateral
the Automotive SPICE development process in four groups acceleration and the detected lane marking curvature from a
for which we identified well suited methods that help to
deal with the typical challenges in each specific Automotive
SPICE development stage. In the requirements elicitation
and analysis phases, data analysis supports the definition
of requirements and helps to understand the system and
its environment. Descriptive preparation of significant data
facilitates the design process and provides the involved
2 Data Driven Development/Lean Programming,
The Practical Dev, https://dev.to/topriddy/
data-driven-developmentlean-programming
3 Data-Driven Development, World Economic Forum Fig. 4: Angle of steeringwheel and lateral acceleration com-
Switzerland, https://www.weforum.org/projects/
data-driven-development pared with curvature of detected lanes.

Authorized licensed use limited to: UNIVERSITÄTSBIBLIOTHEK PADERBORN. Downloaded on August 24,2023 at 11:51:41 UTC from IEEE Xplore. Restrictions apply.
 camera was conducted. The expected dependency between context of wanted and unwanted changes in the function’s
steering wheel angle, lateral acceleration and road curvature behavior.
is visible in the bivariate histograms shown in figure 4, Unit verification with one or just a few inputs can easily
but also strong outliers are visible. These cases depict lane be done with generated data. However, with an increasing
swaps, banked road surfaces and detected road curvature number of input dimensions test cases skyrocket. While
from branching lanes, which otherwise might have been corner cases and boundary values are already covered by
overlooked in the requirements specification. Knowledge regular unit tests, the usage of recorded data provides a broad
about cases like these is beneficial for a comprehensive range of additional test coverage of the intervening spaces.
requirements definition and system specification. This effect amplifies with correlated parameters and time
dependent signals. The tedious specification of adequate test
B. System and Software Design vectors, that cover a reasonable part of combinations and
signal curves is reduced by usage of the recorded data. Data-
In the system and software design phase the definition
driven simulations on component and subsystem level can
of interfaces and the allocation of the functional elements
be used to extract the initial states of units and signal values
of the feature under development to specific components is
of interfaces required for unit test specification. These tests
an essential task. Thus, recorded driving data, which can be
however should not fully replace the existing testing methods
used to determine the expectable characteristics and quality
but rather add an exhaustive coverage of realistic test cases.
of input data is highly valuable. The in-depth analysis of
recorded signals on a broad data base provides information D. Software and System Integration and Qualification
about the overall availability of certain signals and the types
The aim of the integration and qualification phases for
and frequency of invalid value occurrences. Such informa-
software and system items is to provide evidence for specifi-
tion is beneficial for critical design decisions concerning
cation compliant implementation and interaction. The simu-
redundancy, required filter stages and according computing
lation of the components, subsystems and the full system
performance.
is not only viable for testing and debugging during the
Especially early on knowledge about invalid values and
implementation phase, but also an appropriate tool for wide-
outliers can prove beneficial to the system design and the abi-
spread V&V tests. With a vast amount of automotive full
lity to predict and avoid errors in the implementation phase.
logs and the above mentioned methods to facilitate the usage
Suitable visualization of specific dependencies, uncertainties
of the recorded driving data in simulation environments,
and erroneous signals with time lines, maps or statistical
the system can be executed for thousands of kilometers
charts improves the overall system knowledge on which
with realistic inputs. This again does not aim to replace
design choices are made. Also timing aspects in the design
existing test methods like HIL and prototype vehicles, but
phase can be sharpened with the help of the recorded data,
enhance the existing tool set of the developers. Being able
allowing for a better coordination with other development
to simulate with recorded data and suitably analyze specific
teams.
scenarios and simulations of the real world before, during
and after prototype and field tests benefits the effectiveness
C. Unit Construction and Verification
and efficiency of V&V activities. The extend of the possible
During software implementation, programmers can apply analysis is increased and real world activity can be better
different methods for usage of the data to test and verify the focused based on the analysis.
functionality of the developed software components on diffe- For one, the virtual nature of these tests shortens the
rent levels. Section IV presents an example for usage of data iteration time of software development and testing drastically
records in open-loop simulation as well as an example for and for another the extended test sets can simulate thousands
usage in closed-loop simulation. Based on the results of these of kilometers with different scenario settings to increase the
ad-hoc tests, better and faster decisions for software design statistical significance of the executed tests. Different scena-
and implementation can be made. A hand-picked test drive, rio settings include different road types, weather conditions,
one well known to the developer, can for example help to traffic scenarios or countries. These extended tests can be
identify weaknesses and failures in the behavior of individual implemented on a nightly base and thoroughly evaluate the
software units and support identification and understanding performance of the current software release with a broad
of the failure’s root cause. A fast and neat possibility to amount of test data. Automated reports and continuous
simulate the implemented software unit in interaction with integration of the data-driven simulation gives experts a neat
the other units and components provides the developer with overview and, if necessary, detailed in-depth information to
a possibility to analyze the impact on the overall feature. validate the expected behavior or inspect undesired effects.
The use of recorded data ensures realistic stimulation and
complements the usually used generic scenarios. Comparing IV. A PPLICATON IN C ONTEXT OF P REDICTIVE C RUISE
the recorded driving data based simulation output of different C ONTROL (PCC)
software versions gives the developer an overview of the We implemented the data-driven development approach
effects of the changes made. Geographic visualizations of for the development of a Predictive Cruise Control (PCC)
the changed behavior allow for a quick evaluation of the feature. The PCC constitutes an advanced Adaptive Cruise

Authorized licensed use limited to: UNIVERSITÄTSBIBLIOTHEK PADERBORN. Downloaded on August 24,2023 at 11:51:41 UTC from IEEE Xplore. Restrictions apply.
 Data Records Desktop System Results &
Control (ACC) feature, which uses a model predictive control
MATLAB Reports
algorithm to operate the speed of the vehicle not only to

lelization
Preprocessing & Execution

Paral-
preceding traffic objects, but also to the road topology and SW-Items
speed limits [15]. The feature includes open- and closed-loop MATLAB/Simulink Assessment & Visualization
functional elements on component and subsystem level. We ANSI-C

present two use cases to demonstrate the applicability of the Cloud Computing System of N Clusters
data-driven development approach for both open- and closed- Cloud Management Framework
loop functions on component and subsystem level. This Cluster 1 Cluster 2 Cluster N-1 Cluster N
includes the evaluation of constraints and requirements for Parallel Workers
newly introduced enhancements of the functional capability Data Storage
and simulation based evaluation of implemented software
items. All dynamic testing described in this contribution Fig. 5: Scalability concept for parallel tests of vehicle soft-
is performed in a SIL environment on regular desktop PC ware with a large amount of recorded data
systems without utilization of virtual hardware and targeted situation. The potentially large amount of roundabout drive-
compiler tool chains. through scenarios in the database facilitates the simulation
The first use case is the evaluation of an open-loop and assessment of a statistically significant sample by control
function to estimate the driver state [16]. The function’s experts.
purpose is to anticipate the control inputs of the driver in Furthermore, the data-driven approach in combination
the foreseeable future depending on the perceived driver with a large test data pool can be used for the preparation
actions, vehicle state and environment. Since the actually of country specific field campaigns. Being able to prepone a
realized control variables, like the vehicle’s velocity and field campaign with data from a specific country in a virtual
brake applications, are included in the recorded data, the test environment improves the effectiveness of the upcoming
ground truth to assess the prediction output is already avai- real world campaign, as less faults can be expected. Changes
lable. Therefore, the data used for testing is already labeled. in the parametrization of the developed features can be
Time line based visualizations help the developers to tweak evaluated on a broad data base, regarding the achievement of
the parametrization of the function and find weak spots in the desired improvement and the absence of undesired side
the prediction model using their desktop computers. This effects.
reduces required real world tests with prototypes and enables
the developers to draw a statistical significant sample. Geo- V. U SABILITY AND S CALABILITY
graphical visualizations of the faulty predictions allow visual To cover a sufficient amount of variants during V&V
pattern detection, e.g. repetitive faulty behavior in motorway activities of ASE, large scale simulations are necessary. The
entry situations. data-driven simulation approaches in combination with a
The second use case supports the test and validation of the continuously growing data pool enable large scale simula-
full PCC feature on subsystem level. To enable the simulation tions to a certain extend. However, conventional hardware
of the closed-loop control feature with recorded driving data, and single threaded simulations don’t scale to required data
we developed the Reactive-Replay approach [17]. It supports volumes and a flexible and scalable approach is required.
the developers with fast and easy configurable testing capa- We propose a parallelization approach utilizing MATLAB’s
bilities for V&V of the implemented software. The approach built-in features, that scales from local PCs and workstations
opens up a large data pool of recorded data from field trials up to cluster and cloud computing networks. Figure 5 out-
and test campaigns with the included variants of different lines our approach of a scalable framework that supports
situations and scenarios for reuse during development. Spe- data analysis and evaluation of software items via open-
cific scenarios can be extracted for conceptual assessments and closed-loop simulations. Integration of proprietary C-
as well as for the continuous tracking of software maturity code functions via dynamic linking in MATLAB gives access
[18]. Test sets, carefully selected by control system experts, to the visualization and data handling tools well known to
facilitate holistic feature tests in the simulated environment most development teams in the automotive industry. Thereby,
and regression based test strategies. Due to the system’s granting easy customization of existing analyses, access
closed-loop characteristic, the ground truth for regression to various toolboxes and the possibility for defining new,
testing has to be generated for assessment of the functional custom-made assessment algorithms, which are tailored for
behavior. This is achieved by comparison of the new software specific problems.
release with the simulation result of the latest release of the However, with the growing data pool, new challenges
software, that was successfully validated in the real world arise. Over time, the database of test data will become more
[18]. and more heterogeneous, seeing as test files from previous
If a faulty situation is discovered during real world tests, prototypes and older vehicles may not include all of the cur-
e.g. undesired behavior during roundabout drive-throughs, rent functionality and therefore not support all of the signals
the Reactive-Replay toolkit enables extraction of similar in the current build process. Signal names may deviate and
scenarios from the data pool. These can be used for repeated signal formats might be changed. Without a flexible method
in-depth testing of scenarios resembling the identified faulty to address these changes, old data records will become

Authorized licensed use limited to: UNIVERSITÄTSBIBLIOTHEK PADERBORN. Downloaded on August 24,2023 at 11:51:41 UTC from IEEE Xplore. Restrictions apply.
 obsolete. Our framework handles these challenges with the the available data for use with many types of mainly software
help of generalization in the preprocessing step. The mapping based features. We hope, that our consistent development
from concrete to abstract signal representations, the sampling framework based on a tooling well known in the application
of signals to the function’s cycle as well as the substitution domain increases the usability and acceptance of data-driven
of missing signals are systematically supported, detaching development approaches.
the simulation from implementation details, variants and Future work needs to evaluate the feasibility of the descri-
versions. bed framework concept and prove its effectiveness, efficiency
To achieve the described scalability, our concept integrates and the transferability to further features. A balanced set of
cloud computing capabilities to distribute the required com- methods and tools to analyze and assess the time series data
putations. The underlying cloud management framework pro- needs to be evolved, that considers the special constraints in
vides each machine with the individually required data and the automotive domain.
gathers the results for combined analysis. It further deploys
independent, portable and encapsulated simulation kernels, R EFERENCES
that enable the processing of even large scale data sets in [1] K. Bengler, K. Dietmayer, B. Färber, M. Maurer, C. Stiller, and
a feasible time span. The flexibility of cluster and cloud H. Winner, “Three decades of driver assistance systems,” IEEE In-
telligent Transportation Systems Magazine, vol. 6, no. 4, pp. 6–22,
resources leaves almost no limits to resource expansion. 2014.
To allow adoption to further features, the detailed design of [2] J. Bach, S. Otten, and E. Sax, “A taxonomy and systematic approach
the framework should be as modular as possible and support for automotive system architectures - from functional chains to functi-
onal networks,” in 3rd International Conference on Vehicle Technology
the flexible exchange of the function under test, evaluation and Intelligent Transport Systems (VEHITS), 2017.
methods and input data. Our concept aims at establishing [3] M. Aeberhard, S. Rauch, M. Bahram, G. Tanzmeister, J. Thomas,
a fundamental analysis toolkit with common methods for Y. Pilat, F. Homm, W. Huber, and N. Kaempchen, “Experience, results
and lessons learned from automated driving on germany’s highways,”
ASE. Though, in-depth function specific evaluation metrics IEEE Intelligent Transportation Systems Magazine, vol. 7, no. 1, pp.
will have to be implemented on a case by case basis. 42–57, 2015.
[4] T. Helmer, L. Wang, K. Kompass, and R. Kates, “Safety performance
VI. C ONCLUSION assessment of assisted and automated driving by virtual experiments:
Stochastic microscopic traffic simulation as knowledge synthesis,” in
We introduced this contribution with an analysis of current Intelligent Transportation Systems (ITSC), 2015 IEEE 18th Internati-
challenges in Automotive Systems Engineering. Rising inner onal Conference on. IEEE, 2015, pp. 2019–2023.
product complexity due to denser cross-linkage between [5] A. Korsiakoff, W. N. Sweet, S. J. Seymour, and S. M. Biemer, Systems
Engineering Principles and Practice. John Wiley & Sons, Inc., 2011.
former individual system elements, the expanding percep- [6] J. Weber, Automotive Development Process. Springer-Verlag, 2009.
tion capabilities of upcoming vehicles and the increase of [7] Automotive SPICE Process Assessment / Reference Model, 3rd ed.,
algorithmic complexity render new approaches for ASE VDA QMC Working Group 13 and Automotive SIG, Berlin, Germany,
7 2015. [Online]. Available: http://www.automotivespice.com/
necessary. Especially the large amount of relevant situation [8] K. C. Kang, S. G. Cohen, J. A. Hess, W. E. Novak, and A. S.
and scenario variants that need to be considered, included Peterson, “Feature-oriented domain analysis (foda) feasibility study,”
and validated were identified to exacerbate the development DTIC Document, Tech. Rep., 1990.
[9] OMG Systems Modeling Language (OMG SysML), Object Manage-
of high level features. ment Group, 2012.
We analyzed and described the current procedure of sy- [10] J. Schäuffele and T. Zurawka, Automotive Software Engineering -
stem and software development in the automotive domain Grundlagen, Prozesse, Methoden und Werzeuge effizient einsetzen,
5th ed. Springer Fachmedien Wiesbaden GmbH, 2012.
based on the Automotive SPICE process reference and [11] P. Bourque, R. E. Fairley, et al., Guide to the software engineering
identified the demand for new approaches to supplement the body of knowledge (SWEBOK (R)): Version 3.0. IEEE Computer
specification based engineering process. Society Press, 2014.
[12] E. Sax, Automatisiertes Testen Eingebetteter Systeme in der Automo-
In Section III we introduced what we call data-driven bilindustrie. München: Hanser, Carl, 2008.
development. We presented beneficial approaches of exten- [13] H. Shokry and M. Hinchey, “Model-based verification of embedded
sive data usage for all stages and activities of the product software,” Computer, vol. 42, no. 4, pp. 53–59, 4 2009.
[14] M. Cordts, M. Omran, S. Ramos, T. Rehfeld, M. Enzweiler,
development. These include the analysis of the contextual R. Benenson, U. Franke, S. Roth, and B. Schiele, “The cityscapes
background of a feature, fact based design decisions and dataset for semantic urban scene understanding,” in IEEE Conference
simplified debugging as well as an increase in overall Ve- on Computer Vision and Pattern Recognition (CVPR), 2016. [Online].
Available: https://www.cityscapes-dataset.com/
rification and Validation effectiveness and efficiency. [15] K.-L. Bauer and F. Gauterin, “A two-layer approach for predictive
The approaches were explained from a more applied optimal cruise control,” in SAE 2016 World Congress and Exhibition,
point of view using the example of PCC development for 2016.
[16] S. Fünfgeld, M. Holzäpfel, M. Frey, and F. Gauterin, “Driver state
which we developed and implemented them. The benefit estimation for prediction of vehicle states within control systems,” in
of complementing existing and established approaches with Intelligent Vehicles Symposium (IV), 2016 IEEE. IEEE, 2016, pp.
data-driven methods were clearly outlined. 1386–1391.
[17] J. Bach, K.-L. Bauer, M. Holzäpfel, M. Hillenbrand, and E. Sax,
In order to fully exploit the data-driven development “Control based driving assistant functions test using recorded in field
approach we identified a demand for scalability and pa- data,” in 7. Tagung Fahrerassistenzsysteme, 2015. [Online]. Available:
rallelization. We presented a concept, that is a first step https://mediatum.ub.tum.de/node?id=1285215
[18] J. Bach, S. Otten, M. Holzäpfel, and E. Sax, “Reactive-replay approach
towards meeting this demand. Abstraction of variant and for verification and validation of closed-loop control systems in early
version specific information facilitates the generalization of development,” in SAE Technical Paper 2017-01-1671, 2017.

Authorized licensed use limited to: UNIVERSITÄTSBIBLIOTHEK PADERBORN. Downloaded on August 24,2023 at 11:51:41 UTC from IEEE Xplore. Restrictions apply.