Adaptive AI Driven Networks for Energy Efficient Low Latency IoT in Immersive Metaverse Platforms
Chang Wang, Xiaoling Gao, Yi Zhang · IEEE Transactions on Consumer Electronics · 2025
The increasing complexity of IoT networks requires advanced methods to efficiently manage resource allocation. Current systems struggle with scalability and resource over-provisioning, particularly in dynamic traffic conditions. This study aims to address these limitations by proposing an AI-driven adaptive resource allocation system that dynamically predicts traffic patterns and optimizes bandwidth, power, and processing resources. The system incorporates stochastic differential equations and a novel reinforcement learning algorithm to adjust resource allocation in real-time based on network fluctuations. Additionally, edge computing is integrated to reduce latency by processing data closer to devices, while federated learning enables decentralized AI model updates without data transfer to central servers. The simulation environment, built using OMNeT++, evaluated the proposed system over a period of 1000 seconds under varying traffic loads and scalability scenarios. The results show a 25% latency reduction compared to CAHRM and a 30% improvement in resource allocation efficiency over SARM, highlighting the effectiveness of the proposed approach in dynamic IoT environments.