AI-Driven Self-Learning Network Management for Industrial IoT Wireless Sensor Networks Using Autoencoder and Reinforcement Learning

Abstract

Industrial Wireless Sensor Networks (IWSNs) are essential to Industrial Internet of Things (IIoT) systems because they allow automation, data-driven decision-making, and monitoring in smart industrial settings. However, these networks encounter several challenges such as energy limitations, abnormal node behavior, packet loss, and unpredictable communication links. To discover these problems, this paper introduces an Artificial Intelligence (AI)-driven self-learning network management system that combines Autoencoder-based anomaly detection and Reinforcement Learning (RL)-based routing optimization. The Autoencoder algorithm learns normal network behavior patterns and identifies anomalous sensor nodes based on reconstruction error, allowing early detection of anomalous activities or faults. Using network characteristics like residual energy, link quality, and congestion level, reinforcement learning is used to dynamically choose the best routing path. This hybrid AI approach assists adaptive decision-making and nonstop learning in dynamic industrial environments. Simulation results illustrate that the proposed system enhances anomaly detection accuracy and increases network throughput compared with traditional approaches. This intelligent approach offers a scalable solution for dynamic network management in IIoT environments.