Model-Driven Systems Engineering Automation with Artificial Intelligence for Robustly Writing Automotive System Requirements

<div class="section abstract"><div class="htmlview paragraph">The intensive use of software applications in modern vehicles has highlighted the critical role of Systems Engineering (SE) in the automotive industry. These “computers on wheels” are thoroughly interconnected, by their own connections and with the cloud, due to the advancement of Electronic Control Units (ECU) technologies and the widespread use of sensors transmitting real-time data. This interconnectedness and the level of software abstraction that are known today, significantly escalates the complexity of these systems. This has made it necessary to adopt an approach that is flexible to change, structured, agile, and traceable. The modern approach to SE, now model-based, offers numerous advantages over the previous paradigm, which was predominantly document-based.</div><div class="htmlview paragraph">MBSE (Model-Based Systems Engineering) emerges as a contemporary approach, providing the scalability needed for engineering teams to develop robust products. Its “model-based” essence ensures that the model acts as the primary source of truth, providing a comprehensive overview of the system. In addition, this strategy allows an atomic decomposition of the system into functions, enabled by SysML (Systems Modeling Language), which facilitates collaborative engineering and encompasses the complexity of modern systems, serving as a basis for its development.</div><div class="htmlview paragraph">During the system development process, it is essential to clearly specify system constraints, customer needs, and sensor/ECU behavior through well-written, unambiguous requirements to avoid system failures or poor implementations. Crafting well-defined requirements is challenging, particularly for complex systems with numerous connections, making the task repetitive and prone to human error. This work presents an AI-based framework for automated requirements generation from state machine diagrams. By leveraging the states and transitions contained within the diagrams, it ensures requirements completeness by systematically translating diagram elements into corresponding requirement statements. Furthermore, this framework proposes the integration of EARS (Easy Approach to Requirement Syntax) templates with Natural Language Processing (NLP) techniques to ensure requirements standardization and improve process efficiency by automating requirements writing.</div></div>