An Empirical Study of Architectural Shift from Traditional to AI-Enabled Simulink Controllers
This study compared controller structures across 62 real Simulink models, contrasting traditional and AI-enabled control systems. It used a literature-derived taxonomy to find three architectural tensions: scaffolding dominates all designs, AI controllers rely heavily on discrete dynamics, and constraint enforcement blocks are often missing in AI models. This provides an empirical baseline for understanding how design shifts between paradigms.