Randomized trial shows improved quality of experience with a new adaptive bitrate algorithm, suggesting better performance across various network conditions.
Key Points
This research aims to develop an adaptive bitrate selection algorithm using meta-learning to enhance users’ quality of experience in video streaming.
Proposed MetaABR framework using meta-reinforcement learning for adaptive bitrate selection.
Jointly trained multiple tasks with a shared meta-critic to enhance bitrate optimization.
Implemented and tested MetaABR on an emulation platform connected to the Linux network protocol stack.
MetaABR demonstrated superior quality of experience compared to existing state-of-the-art ABR algorithms.
Achieved effective bitrate optimization in a variety of network environments.
Successfully adapted to new tasks with only a few trials, showcasing efficiency in unseen environments.