Smart Cyber-Physical Systems for Real-Time Optimization

Abstract

Mart cyber-physical systems (CPS) are consolidating as a core infrastructure of Industry 5.0, where real-time optimization must simultaneously address efficiency, resilience, cybersecurity, and sustainability across interconnected urban and industrial environments. Although AI, IoT, and digital twin technologies have accelerated CPS capabilities, prevailing approaches remain conceptually and operationally fragmented. Learning-based controllers can degrade under uncertainty and nonstationary conditions, constraint-driven optimization may become overly conservative in fast-changing regimes, and secure coordination is often implemented as an auxiliary layer rather than a design premise. This study proposes an integrated conceptual and methodological framework for real-time, cross-domain CPS orchestration that unifies (i) predictive learning through reinforcement learning, (ii) stability and constraint enforcement via model predictive control, (iii) interpretability and scenario-based evaluation through a high-fidelity digital twin environment, and (iv) trustworthy coordination through blockchain-enabled verification, traceability, and tamper resistance. Carbon-aware objectives are embedded through feedback loops that couple energy and emission signals with control decisions, enabling multi-objective optimization aligned with decarbonization requirements. Simulation-driven experimentation across mobility and multi-energy domains indicates that the hybrid RL–MPC controller yields smoother state trajectories and more consistent learning convergence than isolated baselines, while blockchain verification improves auditability and anomaly detection without compromising real-time actuation timing. The resulting architecture positions CPS as transparent, self-adaptive, and sustainability-oriented systems suitable for scalable Industry 5.0 deployments.