A Review of Learning-Based Methods for Adaptive Video Streaming Over HTTP

Adaptive video streaming offers enhanced Quality of Experience (QoE) by dynamically adjusting the video quality to match network conditions and device capabilities. However, employing effective video streaming systems is becoming more challenging as user demands for high quality and low latency grow. The surge in video traffic is not only straining network resources but also causing a decline in video quality. To address these challenges, machine learning algorithms leverage data-driven techniques to optimize video delivery, improve QoE, and reduce network congestion. Although several surveys exist on the role of machine learning in adaptive streaming, there remains a lack of an up-to-date review that comprehensively covers the use of machine learning throughout all the stages of video streaming and explores practical deployment challenges and opportunities. This survey addresses this gap by systematically categorizing and analyzing recent research on learning techniques applied to adaptive video encoding, bandwidth optimization, and quality adaptation. Emerging trends and open challenges are identified, providing researchers with timely insights into how learning algorithms can be used to shape the future of adaptive streaming systems.