Adaptive Offloading Control in 6G Cell-Free O-RAN Using Reinforcement Learning

Abstract

Open Radio Access Networks (O-RAN) enable programmable radio access architectures in which learning-based intelligence at the RAN Intelligent Controller (RIC) supports adaptive resource management in future 6G systems. To address the dynamic and heterogeneous traffic conditions observed at the regional edge, we propose a Reinforcement Learning (RL)–driven offloading framework to adaptively offload wireless services between SDN-controlled network segments. Simulation results demonstrate that the proposed approach consistently outperforms a threshold-based heuristic baseline, achieving an average reduction of approximately 34% in overall blocking across different traffic loads. These results convey the effectiveness of learning-based offloading control for improving service performance in dynamic 6G network environments.