AI-Driven Consistency of SysML Diagrams

What is it about?

Designing complex systems often requires creating models, and one widely used modeling language in model-driven engineering is SysML. SysML includes various types of diagrams to capture different aspects of a system, such as its use cases (use case diagrams), architecture (block diagrams), and behavior (state-machine diagrams). Ensuring consistency between these different views is crucial, but it can be challenging in complex models. This paper introduces a framework, extending the modeling and formal verification toolkit TTool, aimed at ensuring consistency both between different SysML views and within individual views, with a focus on use case and block diagrams. The framework utilizes mathematically defined consistency rules and generative AI to detect and automatically resolve a wide range of inconsistencies.

Featured Image

Photo by Sven Mieke on Unsplash

Photo by Sven Mieke on Unsplash

Why is it important?

Over the past two years, various research efforts have introduced large language models (LLMs) based assistants to aid in modeling tasks. This work proposes an original approach to integrating LLMs into modeling by combining them with rule-based algorithmic methods, leveraging their capacity to assess whether two texts—and, by extension, two models—convey the same meaning. This capability enables the detection of inconsistencies that rule-based approaches alone cannot algorithmically resolve.