Space-Centric Adaptive Video Streaming with Quality of Experience Optimization in Low Earth Orbit Satellite Networks
Journal
IEEE International Conference on Communications
Journal Volume
2023-May
ISBN
9781538674628
Date Issued
2023-01-01
Author(s)
Lin, Po Hsun
Abstract
This study aims to enable space-centric adaptive video streaming through low earth orbit (LEO) satellites. Nowadays, satellite TV is sent in a broadcasting manner via geostationary orbit (GSO) satellites. However, the long distance between GSO and the Earth results in long latency, making GSO satellites less likely to provide livecast streaming services. To solve this problem, LEO satellites, which are much closer to the Earth, can relay traffic from video servers to remote users in high capacity and low latency. However, the orbiting characteristic of LEO satellites introduces new challenges such as frequent handover and fluctuating capacity that do not occur in GSO satellites. In this paper, we address the video streaming problem in LEO satellite networks, concerning adaptive bitrate, relay satellite selection and super-resolution. A novel solution called SkyTube, which is a Policy Proximal Optimization (PPO) based Deep Reinforcement Learning (DRL) method, is proposed to maximize the Quality of Experience (QoE) of the streaming user. The goal for our design is to formulate video streaming in LEO satellite networks into a resource allocation problem, thus the solution is not restricted to any special type of LEO satellite constellations. Simulation results show that SkyTube outperforms other baselines in terms of the cumulative QoE per orbit period. We also show the great performance and high convergence speed of SkyTube in all cases.
Subjects
adaptive bitrate | deep reinforcement learning | LEO satellite networks | resource allocation | SkyTube | super resolution | video streaming
Type
conference paper