3B-ARA: Bandwidth, Buffer, and Battery Aware Rate Adaptation for Dynamic HTTP Streaming

This letter proposes a novel rate adaptation framework for dynamic adaptive streaming where three factors-available bandwidth, buffer state, and residual battery status, jointly determine the bit rate selected. We model the rate adaptation problem as a Markov Decision Process (MDP) with a goal to optimize video streaming experience as measured by video playback quality and buffer occupancy. Using simulation study, we show that our approach, termed as 3B-ARA, reduces rebuffering rate by over 20% while delivering a comparable average video rate and maintaining an average of 93% buffer occupancy as compared to other existing rate adaptation approaches.

I. Introduction

The dynamics of a video streaming system under time-varying network conditions, serve as a challenge in designing a robust rate adaptation algorithm. An efficient client-based rate adaptation algorithm is critical in ensuring the quality of user experience during video streaming. Despite several proposals, previous studies have shown limitations of state-of-the-art rate adaptation solutions [1]. While some solutions are too aggressive, others are too conservative to the dynamics of the network conditions.