A Machine Learning Solution to Mitigate Co-Tier Interference and Improve QoE for Video Delivery in 5G HetNets

Abstract

The exponential growth of rich media services across the globe has led to a massive increase in data traffic. The recent COVID-19 pandemic has also contributed to this surge as user traffic patterns have witnessed a sharp growth in demand for rich media services, particularly video conferencing (e.g., Zoom, Skype, Teams) and entertainment (e.g., Netflix, Hulu, Amazon). This has put a significant pressure on the current Heterogeneous Network (HetNet) environments, impacting end users' Quality of Experience (QoE). One of the promising solutions to deal with this issue is the introduction of 5th Generation (5G) networks within HetNets and the deployment of small cells (i.e., femtocells) to shift the load from the traditional macrocells. Yet, the big challenge with this approach is the co-tier interference that can occur between different femtocell users. To mitigate this problem, we propose a Machine Learning Interference Classification and Offloading Scheme (MLICOS) that classifies users' traffic based on the level of experienced co-tier interference and offloads the most affected traffic to nearby femtocells, with the ultimate goal of improving the users' QoE. MLICOS performance was evaluated using various QoE metrics, including Peak signal-to-noise ratio (PSNR), Structural Similarity Index Measure (SSIM), and Video Multi method Assessment Fusion (VMAF), and was compared to Proportional Fair (PF) scheduling algorithm, Variable Radius and Proportional Fair scheduling (VR+PF) algorithm, and a Cognitive Approach (CA). Simulation results show that MLICOS generates the highest PSNR, SSIM, and VMAF compared to the other schemes, therefore providing high user QoE.