Open AccessAdaptive streaming algorithm based on reinforcement learning
Author(s) -
Suliu Feng,
Caihong Wang,
Xiuhua Jiang
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/768/7/072069
Subject(s) - computer science , arithmetic underflow , reinforcement learning , dash , quality of experience , dynamic adaptive streaming over http , trace (psycholinguistics) , internet video , quality (philosophy) , real time computing , the internet , artificial intelligence , machine learning , multimedia , computer network , quality of service , world wide web , linguistics , philosophy , epistemology , programming language , operating system
As video streaming occupies a large part of the Internet traffic, and DASH gradually becomes the dominant standard for video transmission recent years, a lot of researches about client-side bitrate adaptive algorithms are constantly emerging.This paper mainly adopts the DRL(Deep Reinforcement learning) algorithm that combines Deep Learning and Reinforcement Learning techniques to optimize the Quality of Experience (QoE) of DASH.This method learns to make decisions by observing the network environment and adjust the strategy to get more reward based on feedback,which doesn’t rely on pre-programmed models or assumptions about the environment. After experiments on the real network trace datasets, we can find that the algorithm this paper adopted can obtain higher quality of user experience than the existing researches.Moreover, the experimental results show that less buffer underflow occurs and smoother video playback is guaranteed.