Self-Evolving Adaptive Resource Management Framework for Resilient 6G-IoT Architectures Using Federated Edge Intelligence and Semantic Communication

Abstract

The massive rise of hyperconnected devices, ultra-reliable low-latency communications (URLLC), and autonomous systems on the 6G-IoT front line, have called for a drastic change to resource management methodologies. In this paper we present a pioneering work the S-EARM (Self-Evolving Adaptive Resource Management) framework that utilises the combination of federated edge intelligence and semantic-driven communications protocols that leverages an optimal distribution of resources for ultra-dense heterogeneous IoT systems in a real- time manner. In contrast to conventional centralised strategies, S-EARM utilises distributed deep reinforcement learning (DRL) as an autonomous prediction, allocation and reconfiguration control mechanism at edge to adaptively and automatically determine communication and computing resources according to real-time contexts and intent-level semantics. The proposed system adopts a dual-level optimized procedure—edge-local prediction and global policy synchronization through privacy-preserving federated learning—to achieve low response time, adaptive load balancing and energy-efficient energy-aware operation. Finally, communication cost is further reduced using semantic compression, while ensuring that the context integrity between the communicated data is maintained. The extensive simulations on a synthetic 6G-IoT testbed reveal that the proposed S-EARM reduces the end-to-end latency by 42.8%, while enhancing the energy efficiency by 38.4%, and the resource utilization by 53.2% over the existing heuristic-based state-of-the-arts. The results place S-EARM as a key enabler for resilient, context-aware, self-optimizing IoT systems, in the 6G timeframe, with application in critical areas, including autonomous mobility, smart healthcare, and industrial automation. This work set up a scalable and future-proof prototype enabling AI in next generation of IoT networks.