Smart Control Systems Using Edge-Deployed Neural Networks for Cyber-Physical Infrastructures

Abstract

Cyber-physical systems (CPS) require intelligent control mechanisms that guarantee reliability, low latency, and adaptability to rapidly changing environmental conditions. Cloud-centric control architectures often introduce delays, pose security risks, and depend heavily on network availability. This paper proposes a smart control system leveraging edge-deployed neural networks to enable decentralized, real-time decision-making in CPS infrastructures. The architecture supports online learning, local anomaly prediction, and resource-aware scheduling while minimizing communication overhead. A specialized neural compression technique ensures compact models suitable for low-power edge devices. Case studies in smart grids, automated transportation, and environmental monitoring show that the system achieves superior response time, reduced network congestion, and enhanced security resilience. By shifting intelligence toward the edge, the proposed system significantly enhances the autonomy and robustness of modern CPS installations.

KEYWORDS: Cyber-Physical Systems, Edge Intelligence, Neural Networks, Smart Control, Real-Time Computing