F-IRAN: Performance Analysis of 6G Fog Intelligent Radio Access Network

The next generation of wireless systems–Sixth Generation (6G) and beyond–is expected to support ultra-massive connectivity through a dense composition of small- and micro-scale networks with high traffic loads. To address the resulting demands for network flexibility, low-latency access, and energy-aware operation, conventional Radio Access Network (RAN) architectures require fundamental redesign. This paper proposes a novel Fog-based Intelligent Radio Access Network (F-IRAN) architecture for 6G, integrating Artificial Intelligence (AI), Tiny Machine Learning (TinyML), and Fog Radio Access Network (F-RAN) concepts across diverse deployment scenarios with varying numbers of Intelligent Remote Radio Heads (IRRHs).The proposed F-IRAN framework is realized through two deployment strategies: a virtualization-based design that leverages distributed computing resources to reduce power consumption and enhance energy efficiency, and a hardware separation-based design that embeds lightweight TinyML models at the fog layer to accelerate processing and reduce execution latency. Simulation results confirm that both strategies outperform conventional F-RAN in key performance indicators. The separation-based approach achieves Energy Efficiency (EE) gains of up to 70% and latency reductions exceeding 36%, while the virtualization-based approach delivers over 50% improvement in EE along with consistent Spectral Efficiency (SE) performance, even under varying IRRH densities and a latency reduction of roughly 11%. Compared to centralized architectures such as Cloud RAN (C-RAN) and Heterogeneous Cloud RAN (H-CRAN), F-IRAN significantly reduces execution delay by offloading computation to the network edge and alleviating fronthaul bottlenecks. Furthermore, when evaluated against recent architectures such as Open RAN (O-RAN) and Virtualized RAN (V-RAN), the proposed architecture demonstrates lower energy consumption and latency by avoiding coordination and virtualization overhead. These findings validate F-IRAN as a scalable, low-latency, and energy-efficient solution for deployment in Beyond 5G (B5G) and 6G network environments.