Dagstuhl Seminar 19502
Future Automotive HW/SW Platform Design
( Dec 08 – Dec 11, 2019 )
(Click in the middle of the image to enlarge)
Permalink
Please use the following short url to reference this page:
https://www.dagstuhl.de/19502
Organizers
- Xiaobo Sharon Hu (University of Notre Dame, US)
- Selma Saidi (TU Dortmund, DE)
- Sebastian Steinhorst (TU München, DE)
- Dirk Ziegenbein (Robert Bosch GmbH - Stuttgart, DE)
Contact
- Michael Gerke (for scientific matters)
Schedule
Driven by new functionality and applications (such as automated driving and vehicle-to-X-connectivity) and fueled by the entry of new players from the IT industry, automotive systems are currently undergoing a radical shift in the way they are designed, implemented, and deployed. The trend towards automation and connectivity imposes an increased complexity and requires unprecedented computing resources, while, at the same time, the demanding requirements regarding cost-efficiency and dependability still need to be fulfilled. One of the most visible changes is the integration of formerly separated function domains onto centralized computing platforms. This leads to a heterogeneous mix of applications with different models of computation (e.g., control, stream processing, and cognition) on heterogeneous, specialized hardware platforms (comprising, e.g., application cores, safety cores, GPUs, deep learning accelerators) to accommodate advanced functionalities such as automated driving and on-line optimization of operating strategies for electrified powertrains.
The adoption of these novel heterogeneous platforms raises several challenges. In particular, many of their components stem from embedded consumer devices and have never been designed for application in safety-critical real-time systems. Therefore, while their computational capabilities are well understood, there is an increased need to comprehend these platforms from the perspective of extra-functional requirements such as predictability, determinism, and freedom-from-interference. This process deeply impacts the core design aspects of automotive E/E architectures and heavily challenges established methods and methodologies in HW/SW automotive design.
The goal of this Dagstuhl Seminar is to gather researchers and practitioners from academia and industry to discuss key industrial challenges, existing solutions and research directions in the HW/SW design of future automotive platforms. The seminar will, in particular, focus on
- predictability of systems regarding extra-functional properties,
- safe integration of hardware and software components and
- programmability and optimization of emerging heterogeneous platforms.
These inter-dependent challenges require the interaction between multiple disciplines, combining resource-constrained embedded, cyber-physical, and real-time aspects.
Another important aspect of the seminar is to provide insight into novel automotive functionalities (such as automated driving, online optimization, or over-the-air-update) and their software architectures and requirements as well as into the HW/SW platforms they are executed on. Application kernels and models will be made available to the participants.
The seminar provides a unique opportunity for participants from the automotive industry to present their challenges and constraints and receive feedback and ideas from academia. At the same time, it allows researchers to confront their own ideas and/or solutions with industrial reality and together identify new research directions in order to make an impact in the automotive industry.
Creative Commons BY 3.0 DE
Xiaobo Sharon Hu, Selma Saidi, Sebastian Steinhorst, and Dirk Ziegenbein
Driven by new functionality and applications (such as automated driving and vehicle-to-X-connectivity) and fueled by the entry of new players from the IT industry, automotive systems are currently undergoing a radical shift in the way they are designed, implemented, and deployed. The trend towards automation and connectivity imposes an increased complexity and requires unprecedented computing resources, while, at the same time, the demanding requirements regarding cost-efficiency and dependability still need to be fulfilled. One of the most visible changes is the integration of formerly separated function domains onto centralized computing platforms. This leads to a heterogeneous mix of applications with different models of computation (e.g., control, stream processing, and cognition) on heterogeneous, specialized hardware platforms (comprising, e.g., application cores, safety cores, GPUs, deep learning accelerators) to accommodate advanced functionalities such as automated driving and on-line optimization of operating strategies for electrified powertrains.
The adoption of these novel heterogeneous platforms raises several challenges. In particular, many of their components stem from embedded consumer devices and have never been designed for application in safety-critical real-time systems. Therefore, while their computational capabilities are well understood, there is an increased need to comprehend these platforms from the perspective of extra-functional requirements such as predictability, determinism, and freedom-from-interference. This process deeply impacts the core design aspects of automotive E/E architectures and heavily challenges established methods and methodologies in HW/SW automotive design.
The goal of this Dagstuhl Seminar was to gather researchers and practitioners from academia and industry to discuss key industrial challenges, existing solutions and research directions in the HW/SW design of future automotive platforms. The seminar focussed, in particular, on
- predictability of systems regarding extra-functional properties,
- safe integration of hardware and software components and
- programmability and optimization of emerging heterogeneous platforms.
These inter-dependent challenges require the interaction between multiple disciplines, combining resource-constrained embedded, cyber-physical, and real-time aspects. Another important aspect of the seminar was to provide insight into novel automotive functionalities (such as automated driving, online optimization, or over-the-air-update) and their software architectures and requirements as well as into the HW/SW platforms they are executed on.
The seminar provided a unique opportunity for participants from the automotive industry to present their challenges and constraints and receive feedback and ideas from academia. At the same time, it allowed researchers to confront their own ideas and/or solutions with industrial reality and together identify new research directions in order to make an impact in the automotive industry.
Organization of the seminar
The seminar took place from 8th to 11th December 2019. The seminar started with an overview of current trends and challenges in the design of future automotive HW/SW platforms by the organizers. After that the agenda was structured along the previously mentioned challenges. Monday's talk sessions were focused on dependability and predictability of HW/SW systems. The sessions on Tuesday dealt with the safe integration of heterogeneous software applications covering aspects of software architectures, networks and cyber-physical systems in the automotive domain and touched societal issues as well. On Wednesday, the talks focused on the programmability and optimization of heterogeneous platforms. All talks were restricted to 15 minutes, leaving ample time for discussions as well as breakout sessions on the following topics:
- Modeling hardware and software dependencies
- Weakly hard real-time models
- Machine learning in cyber-physical systems
- HW/SW architecture exchange
- Benchmarking efforts for future HW/SW platforms
- Modularizing control systems
- Automotive software lifecycle
- Programming vs. execution models
More details on breakout sessions are available in a dedicated section of this document, after the overview of the talks given during the seminar.
Outcome
The seminar succeeded in bringing together participants from different communities who were engaged in very intensive, interdisciplinary group discussions. Not surprisingly, many participants stated that they were able to learn a lot from adjacent fields. As many of the industrial challenges at hand require interdisplinary approaches, the organizers consider this a significant success of the seminar. One example that became evident during the course of the seminar was that terms like execution model are quite differently used in e.g. the high performance computing domain and in the embedded systems community. A group formed in one of the breakout sessions intends to write a whitepaper on unifying terminology and formulating a common understanding of the different layers of models used in designing automotive HW/SW systems. A first follow-up meeting already took place in February 2020.
Several industrial presentations gave valuable insights in the industrial state-of-the-practice and outlined challenges for future research. A very good example for this was the breakout session ``HW/SW Architecture Exchange'' which discussed current architectural patterns and open challenges in the context of designing dependable systems and achieving deterministic behavior on heterogeneous high-performance HW platforms.
Another breakout session provided an overview of current automotive benchmarks and performance models that can be used as a basis for research activities. This session also raised the awareness that industry needs to be more active in providing relevant benchmarks in order to enable researchers to validate the industrial viability of their solutions.
Overall, the feedback of the participants showed that they made a lot of new contacts in academia and industry and a follow-up seminar in about two years was requested by many participants. The seminar inspired several new collaborations including contributions to the Autonomous Systems Design workshop at DATE 2020, ideas for special sessions at DAC 2020 and ESWEEK 2020 and also a student project on automotive HW/SW platform simulation between a students' project group and an industrial partner.
Dirk Ziegenbein, Selma Saidi, Xiaobo Sharon Hu, and Sebastian Steinhorst
- Bart Besselink (University of Groningen, NL) [dblp]
- Alessandro Biondi (Sant'Anna School of Advanced Studies - Pisa, IT) [dblp]
- Jerónimo Castrillón-Mazo (TU Dresden, DE) [dblp]
- Lulu Chan (NXP Semiconductors - Eindhoven, NL)
- Wanli Chang (University of York, GB) [dblp]
- Thidapat (Tam) Chantem (Virginia Polytechnic Institute - Arlington, US) [dblp]
- Jyotirmoy Deshmukh (USC - Los Angeles, US) [dblp]
- Rolf Ernst (TU Braunschweig, DE) [dblp]
- Sabine Glesner (TU Berlin, DE) [dblp]
- Masaki Gondo (eSOL - Tokyo, JP)
- Baik Hoh (Toyota Motors North America - Mountain View, US) [dblp]
- Oliver Kopp (Mercedes Benz AG - Stuttgart, DE) [dblp]
- Peter Gorm Larsen (Aarhus University, DK) [dblp]
- Mark Lawford (McMaster University - Hamilton, CA) [dblp]
- Roland Leißa (Universität des Saarlandes - Saarbrücken, DE) [dblp]
- Chung-Wei Lin (National Taiwan University - Taipei, TW) [dblp]
- Martina Maggio (Lund University, SE) [dblp]
- Albrecht Mayer (Infineon Technologies - München, DE) [dblp]
- Frank Mueller (North Carolina State University - Raleigh, US) [dblp]
- Philipp Mundhenk (Autonomous Intelligent Driving GmbH - München, DE) [dblp]
- Alessandra Nardi (Cadence - San Jose, US) [dblp]
- Moritz Neukirchner (Elektrobit Automotive - Erlangen, DE) [dblp]
- Philipp Obergfell (BMW AG - München, DE) [dblp]
- Maximilian Odendahl (Silexica - Köln, DE) [dblp]
- Zhu Qi (Northwestern University - Evanston, US) [dblp]
- Eduardo Quinones (Barcelona Supercomputing Center, ES) [dblp]
- Sophie Quinton (INRIA - Grenoble, FR) [dblp]
- Selma Saidi (TU Dortmund, DE) [dblp]
- Ignacio Sanudo Olmedo (University of Modena, IT) [dblp]
- Lea Schönberger (TU Dortmund, DE)
- Lukas Sommer (TU Darmstadt, DE)
- Wilfried Steiner (TTTech Computertechnik - Wien, AT) [dblp]
- Seyhan Uçar (Toyota Motors North America - Mountain View, US) [dblp]
- Dirk Ziegenbein (Robert Bosch GmbH - Stuttgart, DE) [dblp]
Classification
- modelling / simulation
- optimization / scheduling
- semantics / formal methods
Keywords
- Automotive
- HW/SW Platforms
- Real-Time Systems
- Systems Design Automation
