Smart Handover between 5G and Wi-Fi over QUIC at Network Edge for Maximizing Throughput

Multi-access edge computing (MEC) and heterogeneous access networks are critical technologies in modern 5G communication system. MEC deploys application services near base stations, while heterogeneous networks provide multiple access network, such as 5G cellular, and Wi-Fi. Additionally, edge applications can directly collect relevant information from the radio access network to optimize performance and fulfill stringent requirements in term of bandwidth and latency. However, the default configuration of mobile devices prioritizes Wi-Fi as the primary interface to avoid cellular billing, which can degrade throughput in areas with poor Wi-Fi signal strength. To address this issue, we propose a smart handover decision-making framework that facilitates cooperation between the user equipment (UE), access network, and MEC to estimate throughput for handover decisions. By employing a deep reinforcement learning model, the framework learns the difference between estimated and actual throughput, enabling dynamic selection of the best-performing access network, whether 5G or Wi-Fi. Our trained model also supports online learning, allowing it to adapt to changing environments. Experimental results demonstrate that our approach achieves, on average, over 96% of optimal throughput in various environments, with network selection accuracy ranging from 88.52% to 92.72%.