LiveStream Meta-DAMS: Multipath Scheduler Using Hybrid Meta Reinforcement Learning for Live Video Streaming

Amir Sepahi, Lin Cai, Wenjun Yang, Jianping Pan

Abstract: Overcoming challenges in mobile environments, such as bandwidth constraints, user mobility, and network hand-offs, is crucial for video streaming applications. To address these challenges, we can use multiple network paths to mitigate bandwidth limitations and guarantee end-to-end delay, enhancing the overall quality of experience for the users. This paper presents LiveStream Meta Learning-based Delay Aware Multipath Scheduler (LSMeta-DAMS), a novel learning-based multipath scheduler explicitly designed for live streaming applications. LSMeta-DAMS employs a hybrid meta-reinforcement learning architecture, incorporating both online and offline phases to enhance speed and accuracy for training and decision making. Prioritizing packet scheduling based on frame types and considering the video coding features like group of pictures (GOP), scalable video coding (SVC), and Dynamic Adaptive Streaming over HTTP (MPEG-DASH), LSMeta-DAMS offers a tailored solution for multipath video streaming. Trace-driven emulations highlight its superior performance, demonstrating up to 32% improvement in learning, up to 25% reduction in download time, up to 15% enhancement in video quality assessment, and up to 35% reduction in stalling time compared to the state-of-the-art multipath schedulers. These findings underscore LSMeta-DAMS’s potential to substantially enhance video streaming experiences in highly dynamic network conditions.

External IDs:dblp:journals/tccn/SepahiCYP25