Search

The complexity of cyber-physical systems is increasing, driven by integration of more functionality and trends towards mass-customization. This has resulted in complex systems with many variants that require long time to develop and are difficult to adapt to changing requirements and introduction of new technology. New methodologies are hence required to reduce development time, simplify customization for a particular customer, and improve evolvability both during development and after deployment. This chapter explains how these challenges are addressed by an approach to model-based engineering (MBE) based on domainspecific languages (DSLs). However, applying the approach in industry has resulted in 5 technical research questions, namely how to: RQ1) achieve modularity and reuse in a DSL ecosystem, RQ2) achieve consistency between model and realizations, RQ3) manage an evolving DSL eco-system, RQ4) ensure model quality, RQ5) ensure quality of generated code. The five research questions are explored in the context of the published state-ofthe-art, as well as practically investigated through a case study from the defense domain.

Model-based languages such as MATLAB/Simulink play an essential role in the model-driven development of software systems. During their development, these systems can be subject to modification numerous times. For large-scale systems, to manually identify performed modifications is infeasible. However, their precise identification and subsequent validation is essential for the evolution of model-based systems. If not fully identified, modifications may cause unaccountable behavior as the system evolves and their redress can significantly delay the entire development process. In this paper, we propose a fully automated technique called Reverse Signal Propagation Analysis, which identifies and clusters variations within evolving MATLAB/Simulink models. With each cluster representing a clearly delimitable variation point between models, we allow model engineers not only to specifically focus on single variations, but by using their domain knowledge, to also relate and verify them. By identifying variation points, we assist model engineers in validating the respective parts and reduce the risk of improper system behavior as the system evolves. To assess the applicability of our technique, we present a feasibility study with real-world models from the automotive domain and show our technique to be very fast and highly precise.