VR-Based Dynamic Traffic Offloading in XL-MIMO Systems

Abstract

Traditional heterogeneous networks (HetNets) rely on auxiliary localization techniques, such as GPS and radio-frequency (RF) fingerprinting, to estimate active users per region during the small cell selection and activation phase. However, these methods introduce high costs, excessive energy consumption, and signaling overhead. This letter proposes a novel visibility region (VR)-based framework for dynamic traffic offloading in extra-large MIMO (XL-MIMO) HetNets. By leveraging spatial VR characteristics, the system accurately estimates the number of active user per region and adaptively activates small cells without external localization. The XL-MIMO node maintains an updated VR map, dynamically associating small cells with users. During congestion, it evaluates VR similarity between users and small cells to optimize the activation, efficiently redistributing traffic. Simulation results of key performance indicators, such as association error probability, false wake-up probability, and the system sum rate, demonstrate that the proposed approach offers a scalable and cost-effective solution for crowded regions identification and small-cell activation for next-generation HetNet deployments.