Towards a safe future in an age of increasing complexity.

The Autonomous is a collaborative platform that brings together top executives and experts in the mobility industry to tackle safety challenges in autonomous driving, with its Safety & Architecture Working Group developing a system-level conceptual architecture for automated vehicles. Through extensive research and expert review spanning from 2021 to 2023, the initiative provides valuable architectural insights for SAE L4 Highway Pilot implementation, offering system owners comprehensive guidance on design decisions and safety considerations across multiple abstraction levels.

Covering the intention of the doctoral research project, we address the challenges of reusing existing software in safety-critical automotive systems, where traditional re-engineering often proves more time-consuming than starting from scratch. The study aims to aggregate information about key obstacles in software reuse and develop novel approaches to enable the use of pre-existing software while maintaining high safety standards. By aligning with ISO/AWI PAS 8926 standards, the research seeks to facilitate cross-company collaboration and leverage modern development tools to provide robust safety evidence comparable to conventional development methods.

This research presents an innovative approach to transforming existing C code into an ANSYS SCADE model, enabling original equipment manufacturers (OEMs) to maintain and reuse legacy code in new development environments. The model transformation process is manually performed, but the testing is fully automated, allowing for the transfer of existing test cases to the SCADE Test Environment. By extending the original code to generate test scenarios during runtime, the approach supports white-box testing and empirical validation to ensure functional equivalence between the original code and the new model.

This Master's thesis explores tool-assisted approaches for developing safety-critical automotive control software in the era of autonomous driving, focusing on the evaluation of a toolchain using CATIA STIMULUS and ANSYS SCADE Suite for requirements specification and model-based development. The research systematically examines the transfer of existing artifacts to new development environments, comparing usability, verification capabilities, and testing methodologies while highlighting the potential for reducing manual work. By presenting recommended practices and proposing potential extensions, the study aims to advance towards a more integrated and automated model-based development process for safety-critical automotive systems.

As autonomous systems become increasingly complex, ensuring the safety and reliability of their software is critical, particularly when dealing with systems that incorporate data from varying sources and with uncertain reliability. This thesis explores multiple formal methods for verifying properties of such reactive systems, focusing on a detailed case study based on an industrial automotive function with features like floating point calculations, error models, and stochastic input data. By comparing different programming languages, model checkers, and verification approaches, the research provides insights into the current challenges and methodologies for validating safety-critical software in autonomous systems.

The automotive industry is facing increasing complexity in control software, with safety-critical functions running on interconnected networks of control units that require precise real-time performance. Decomposing and verifying requirements for distributed functions across multiple control units can introduce challenging inconsistencies that may impact system reliability. This research presents an automated verification method to analyze requirements early in the development process, helping identify potential problems before they escalate.

This thesis addresses the critical challenge of verifying the consistency of complex real-time systems in automotive software development, where hundreds of interconnected systems with safety-critical functions are deployed. The proposed approach enables automated verification of system requirements at the end of the planning stage, helping developers identify potential inconsistencies early in the process. By providing detailed insights into requirement feasibility, the method aims to reduce development delays, resource expenditure, and potential risks associated with software system design.