Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LL-GABR: Energy Efficient Live Video Streaming Using Reinforcement Learning

Adithya Raman, Bekir Turkkan, Tevfik Kosar

TL;DR

This paper tackles the energy inefficiency of live video streaming on mobile by redefining QoE through perceptual quality, measured with VMAF, rather than bitrate. It introduces LL-GABR, a deep reinforcement learning agent that jointly optimizes playback bitrate and speed under a lightweight bandwidth predictor and an energy-aware QoE reward, generalizable across per-title encoding ladders without retraining. Using a distributed Soft-Actor-Critic framework with Priority Experience Replay, LL-GABR achieves up to 44% higher perceptual QoE and up to 73% better energy efficiency, while reducing total energy consumption by 11% across trace-driven evaluations. The approach demonstrates the potential for scalable, energy-conscious live ABR in mobile environments, balancing perceptual quality, latency, and energy use in real time.

Abstract

Over the recent years, research and development in adaptive bitrate (ABR) algorithms for live video streaming have been successful in improving users' quality of experience (QoE) by reducing latency to near real-time levels while delivering higher bitrate videos with minimal rebuffering time. However, the QoE models used by these ABR algorithms do not take into account that a large portion of live video streaming clients use mobile devices where a higher bitrate does not necessarily translate into higher perceived quality. Ignoring perceived quality results in playing videos at higher bitrates without a significant increase in perceptual video quality and becomes a burden for battery-constrained mobile devices due to higher energy consumption. In this paper, we propose LL-GABR, a deep reinforcement learning approach that models the QoE using perceived video quality instead of bitrate and uses energy consumption along with other metrics like latency, rebuffering events, and smoothness. LL-GABR makes no assumptions about the underlying video, environment, or network settings and can operate flexibly on different video titles, each having a different bitrate encoding ladder without additional re-training, unlike existing learning-based ABRs. Trace-driven experimental results show that LL-GABR outperforms the state-of-the-art approaches by up to 44% in terms of perceptual QoE and a 73% increase in energy efficiency as a result of reducing net energy consumption by 11%.

LL-GABR: Energy Efficient Live Video Streaming Using Reinforcement Learning

TL;DR

This paper tackles the energy inefficiency of live video streaming on mobile by redefining QoE through perceptual quality, measured with VMAF, rather than bitrate. It introduces LL-GABR, a deep reinforcement learning agent that jointly optimizes playback bitrate and speed under a lightweight bandwidth predictor and an energy-aware QoE reward, generalizable across per-title encoding ladders without retraining. Using a distributed Soft-Actor-Critic framework with Priority Experience Replay, LL-GABR achieves up to 44% higher perceptual QoE and up to 73% better energy efficiency, while reducing total energy consumption by 11% across trace-driven evaluations. The approach demonstrates the potential for scalable, energy-conscious live ABR in mobile environments, balancing perceptual quality, latency, and energy use in real time.

Abstract

Over the recent years, research and development in adaptive bitrate (ABR) algorithms for live video streaming have been successful in improving users' quality of experience (QoE) by reducing latency to near real-time levels while delivering higher bitrate videos with minimal rebuffering time. However, the QoE models used by these ABR algorithms do not take into account that a large portion of live video streaming clients use mobile devices where a higher bitrate does not necessarily translate into higher perceived quality. Ignoring perceived quality results in playing videos at higher bitrates without a significant increase in perceptual video quality and becomes a burden for battery-constrained mobile devices due to higher energy consumption. In this paper, we propose LL-GABR, a deep reinforcement learning approach that models the QoE using perceived video quality instead of bitrate and uses energy consumption along with other metrics like latency, rebuffering events, and smoothness. LL-GABR makes no assumptions about the underlying video, environment, or network settings and can operate flexibly on different video titles, each having a different bitrate encoding ladder without additional re-training, unlike existing learning-based ABRs. Trace-driven experimental results show that LL-GABR outperforms the state-of-the-art approaches by up to 44% in terms of perceptual QoE and a 73% increase in energy efficiency as a result of reducing net energy consumption by 11%.
Paper Structure (30 sections, 6 equations, 3 figures, 4 tables, 1 algorithm)

This paper contains 30 sections, 6 equations, 3 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background and Related Work
  3. Adaptive Bitrate (ABR) Streaming over HTTP
  4. Live ABR Streaming
  5. Energy Aware ABR Streaming
  6. Limitations of Existing Solutions
  7. LL-GABR Design
  8. Design Goals
  9. Bandwidth Prediction
  10. Energy Model
  11. Local Playback Energy:
  12. Data Acquisition Energy:
  13. Reinforcement Learning for Live Streaming
  14. Why RL?:
  15. State Space:
  16. ...and 15 more sections

Figures (3)

  • Figure 1: Overview of LL-GABR's distributed SAC training.
  • Figure 2: Testing results averages on all traces across all videos.
  • Figure 3: Trace simulation for a fluctuating network.