A Neural Network-Based Network Selection for QUIC to Enrich Gaming in NextGen Wireless Network

Online gaming on smartphones is increasingly popular, but poor Wi-Fi conditions and network handovers can lead to a subpar gaming experience. We propose ODIN (On-Device Intelligence), a novel Neural Network (NN) based cross-layer QUIC gaming proxy, to address these challenges. ODIN incorporates a network quality monitoring framework that predicts Wi-Fi contention using NN, ensuring optimal network selection for gaming applications. We evaluated ODIN through live-air experiments with top-chart Android gaming apps. By leveraging the advantages of QUIC and NN, ODIN outperforms legacy smartphones, providing a better gaming experience. ODIN surpasses Multipath TCP (MPTCP) for thin-stream applications by efficiently using mobile data and delivering seamless user experiences during network handovers, demonstrated in the BrawlStars game's zero-touch and zero-lag handover mechanism. We also introduced ODIN-LITE, a lightweight approach with 25% better power consumption efficiency than MPTCP full-mesh mode. ODIN 's efficient mobile data and power use aligns with Industry 5.0 principles, fostering harmony between human-centric gaming experiences and advanced network technologies. Its application-agnostic design and zero kernel dependency facilitate wide-scale deployment, making it a promising solution for the gaming industry's future.