Improving Robustness of DASH Against Network Uncertainty

Bo Wang, Fengyuan Ren

Abstract: Most video players use adaptive bitrate (ABR) algorithms to ensure good quality-of-experience (QoE) across diverse network conditions. To balance conflicting QoE factors, state-of-the-art ABR algorithms select bitrate by optimizing a defined QoE function. However, this scheme relies on throughput prediction that is sensitive to network conditions, so the achieved QoE can be poor in unstable networks. In this paper, we propose a robust ABR algorithm called RBC to avoid the impact of prediction error on QoE. RBC controls the buffer occupancy within a safe range while maximizing the QoE, and employs an adaptive bitrate controller to ensure a good control performance in various network condition. Trace-driven experiments show that RBC achieves much less playback freezing and an improvement of 13.5% on average QoE over the best approach RobustMPC, which confirms the effectiveness of buffer control and adaptive controller in improving system robustness against network uncertainty.