Controller architectures change as traditional control switches to ai models
An Empirical Study of Architectural Shift from Traditional to AI-Enabled Simulink Controllers
Software EngineeringArtificial Intelligence
Summary
Controllers in machines and systems are shifting from old-style rules to ones powered by AI. The authors studied many real controller designs and found that while the overall structure stays mostly the same, AI controllers use different building blocks and hide safety checks inside training instead of making them visible in the design. This makes tracking how safety works harder. Their work helps us see what really differs between traditional and AI-based controller designs.
What this means in practice
- •For embedded systems engineers: Design controller architectures that combine traditional structure with AI components while maintaining traceable safety mechanisms.
- •For industrial automation teams: Update control systems using AI-enabled models that support complex discrete dynamics and user-defined abstractions to improve system adaptability.
Authors
Hadiza Umar Yusuf, Khouloud Gaaloul
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.