Balanced5G - Fair Traffic Steering Technique using Data-Driven Learning in Beyond 5G Systems

The rapid adoption of the fifth generation (5G) communication systems and increasing data consumption per device is exponentially growing mobile data traffic. Traffic Steering (TS), Load Balancing (LB), and Handover (HO) are the critical techniques for maintaining network traffic loads between over-crowded and less-crowded cells, which must be appropriately managed to avoid service degradation and interruption. In diverse 5G frequency spectrum deployment, steering traffic across frequency bands and base stations is critically important, if not done efficiently, resulting in decreased network throughput and user experience. Network operators can ensure better quality-of-service (QoS) by employing optimal mobility and balanced traffic steering policies. This work proposes Balanced5G, a data-driven traffic steering framework that takes proactive actions during the HO stage to steer traffic fairly between different frequencies (low, mid, and high). Balanced5G ensures that UEs (fixed or mobile) do not select a frequency solely based on the most robust signal strength but consider other network key metrics such as network load, active network slices, and the type of service for which the UE is requesting resources. Our simulation results show that the proposed traffic steering scheme achieves 19.5% better predictions against traditional supervised learning models.