Delay Optimized VNF Placement in 5G-Enabled Industry 4.0 Networks Using DRL With Wireless Reliability and Cyberattack Resilience

The increasing reliance on 5G networks in Industry 4.0 has introduced significant challenges in Virtual Network Function (VNF) placement, where low-latency communication and system resilience are critical. This study proposes a deep reinforcement learning (DRL)-based framework for optimizing VNF placement in 5G-enabled industrial environments to minimize delays while ensuring wireless connection reliability and resilience against cyberattacks. By integrating edge computing and network virtualization, the framework dynamically adapts to varying network traffic, mobility patterns, and wireless channel conditions. A custom simulation environment is designed to model industrial dynamics, including sensor traffic, Automated Guided Vehicle (AGV) mobility, and degraded gNodeB availability to simulate cyberattacks. The study benchmarks Double Deep Q Network (DQN) and Proximal Policy Optimization (PPO) against heuristic baselines, demonstrating their adaptability and scalability. Additionally, the robustness of the proposed framework is evaluated under simulated cyberattacks affecting gNode availability, highlighting its resilience in maintaining low latency communication. The simulation results show that the trained Double DQN and PPO agents achieved a mean absolute percentage error within 3.58% and 4.22% of the established baseline, demonstrating their ability to approximate optimal VNF placement decisions. Furthermore, the adversarially trained PPO agent achieved a 24.3% lower mean delay in a 10% gNodeB availability scenario compared to the normally trained agent, demonstrating enhanced robustness under adversarial conditions. This research highlights the capability of DRL for intelligent and adaptive VNF placement, contributing to the development of robust and efficient next-generation industrial networks.