🤖 AI Summary
This study addresses the insufficient understanding of cross-paradigm architectural discrepancies and the absence of design knowledge embedding when AI components replace traditional control laws. Employing a literature-based taxonomy, we conducted static analysis on 62 Simulink models alongside an empirical survey of 13 practitioners. This work quantitatively reveals, for the first time, three major architectural contradictions: subsystem-organization-dominated structures, discrete dynamic dependencies, and the disappearance of constraint-enforcement modules, highlighting traceability ruptures caused by the implicitization of safety mechanisms. Furthermore, it identifies that core logic constitutes a minimal proportion within AI controllers, while user-defined abstractions are absent and explicit safety structures degrade. Collectively, these findings systematically delineate the architectural migration risks inherent in transitioning from conventional control to AI-driven paradigms.
📝 Abstract
Effective AI adoption in cyber-physical systems (CPS) depends on embedding design knowledge into engineering practice. Yet as AI-enabled components increasingly replace analytically derived control laws, this occurs without a systematic understanding of how controller architectures differ or remain similar across paradigms. We address this gap with an empirical study of traditional and AI-enabled Simulink controllers, guided by a literature-derived taxonomy of ten structural categories and nine functional roles. The study analyzes 62 real-world models spanning 8 controller types and 10 application domains, and surveys 13 practitioners, identifying three architectural tensions. First, subsystem organization dominates all controller structures regardless of paradigm, occupying 68-72% of controller footprint, while core control logic occupies minimal space. Second, AI-enabled controllers rely heavily on discrete dynamics and user-defined abstraction, categories largely absent from AI literature, exposing a gap between described and implemented architectures. Third, constraint enforcement blocks largely disappear from AI-enabled models despite practitioner expectations. This reveals a misalignment where safety mechanisms shift from explicit structure to implicit training-time artifacts, breaking traceability.