As more and more software is required for new vehicle functions, the demands on control units in motor vehicles are constantly increasing. At the same time, vehicles, which typically consist of around 100 primarily monofunctional ECUs, are being driven towards centralized electrical/electronic architectures to cope with the resulting computing requirements. Accordingly, the integration of software is becoming increasingly complex as a mixture of applications from heterogeneous functional areas must be mapped to heterogeneous hardware architectures composed of different types of processors and accelerators.
The focus of Lukas Krawczyk's dissertation is on the challenges and techniques for integrating applications from heterogeneous functional domains in the context of automotive systems on centralized computing platforms in early design phases. For this purpose, one primary and three secondary research questions are formulated, which run like a red thread through the rest of this thesis. Specifically, they deal with (I) efficient techniques for optimizing the mapping of software to hardware that support the above-mentioned integration process, (II) timing analysis techniques that enable the evaluation of complex automotive systems, (III) the consideration of interference caused by memory access conflicts in the context of timing analysis, and (IV) techniques for increasing the efficiency of the mapping optimization process.
To address these research questions, the dissertation develops customized hybrid genetic algorithms tailored to specific optimization problems that optimize the software-to-hardware mapping of (I) two publicly available industrial systems, (II) a customized industrial system augmented with a network-on-chip hardware architecture, and (III) a proprietary motor management system. The methods were consistently able to determine valid mappings for all of the aforementioned case studies in the context of single and/or multi-objective optimization. In addition, they were able to improve quality attributes such as the minimum relative earliness of a system by up to 78%, its average relative earliness by up to 92% and its individual response latencies by up to 94%.
Lukas Krawczyk completed his Bachelor's and Master's degree in computer science (specializing in computer engineering) at Dortmund University of Applied Sciences and Arts. From 2011 to 2022, he worked on various ITEA research projects in the automotive environment as a research assistant at Dortmund University of Applied Sciences and Arts as part of international consortia. The results of these activities have contributed to the foundation of the open source projects Eclipse APP4MC and Eclipse Kuksa, where he is still a committer. As part of these activities, parallel and distributed real-time systems as well as timing analysis in the automotive context were the focus of his research activities. From 2016 to 2022, he completed his doctorate in engineering as part of a cooperation between Bielefeld University (CITEC) and Dortmund University of Applied Sciences and Arts (IDiAL). During this time, Lukas Krawczyk managed to publish over 24 papers in renowned and relevant conferences and journals. He successfully defended his dissertation on 16.11.2022.
Appraiser
- Prof. Dr.-Ing. Ulrich Rückert, (CITEC, Bielefeld University)
- Prof. Dr.-Ing. Carsten Wolff, (IDiAL, Dortmund University of Applied Sciences and Arts)
- Prof. Dr. Jan-Philipp Steghöfer,(University of Gothenburg)
- Dr.-Ing. Sebastian Wrede (CITEC, University of Bielefeld, PhD research assistant)