[Paper Review] A Holistic Survey of Wireless Multipath Video Streaming

Samira Afzal, Vanessa Testoni+3 more|arXiv (Cornell University)|Jun 14, 2019

Advanced Wireless Network OptimizationEngineering192 references17 citations

TL;DR

This paper presents a comprehensive survey of multipath wireless video streaming, analyzing techniques across the protocol stack with a focus on scheduling, path selection, and error protection to improve Quality of Experience (QoE). It identifies key trade-offs, proposes a taxonomy, and highlights open challenges in compatibility, cross-layer design, and QoE-aware optimization using machine learning.

ABSTRACT

Most of today's mobile devices are equipped with multiple network interfaces and one of the main bandwidth-hungry applications that would benefit from multipath communications is wireless video streaming. However, most of the current transport protocols do not match the requirements of video streaming applications or are not designed to address relevant issues, such as delay constraints, networks heterogeneity, and head-of-line blocking issues. This survey provides a holistic literature review of multipath wireless video streaming, shedding light on the different alternatives from an end-to-end layered stack perspective, unveiling trade-offs of each approach, and presenting a suitable taxonomy to classify the state-of-the-art. Finally, we discuss open issues and avenues for future work.

Motivation & Objective

• To address the growing demand for high-quality wireless video streaming in bandwidth-constrained, heterogeneous networks.
• To identify limitations in existing transport protocols that fail to handle delay, heterogeneity, and head-of-line blocking in video streaming.
• To provide a unified taxonomy and layered analysis (application, transport, network, cross-layer) of multipath video streaming techniques.
• To evaluate scheduling strategies, path selection, and error protection mechanisms for improved QoE and resilience.
• To highlight open research challenges, including client/network compatibility, middlebox traversal, and QoE-driven optimization using machine learning.

Proposed method

• Conducts a holistic literature review of 40+ works on multipath wireless video streaming, focusing on data plane solutions.
• Classifies approaches based on protocol stack layer (application, transport, network, cross-layer) and key features like packet scheduling and path selection.
• Analyzes scheduling functions including packet selection, protection schemes (e.g., FEC, retransmission), and path-aware decision-making.
• Evaluates performance using QoS metrics (e.g., loss rate, startup delay) and QoE indicators (e.g., PSNR, playback fluency, quality switching).
• Examines compatibility requirements: client-only, server-client, or infrastructure changes, and middlebox traversal capabilities.
• Proposes future research directions, including QoE-aware scheduling using machine learning models trained on user feedback or QoS measurements.

Experimental results

Research questions

• RQ1 How do different multipath scheduling strategies across the protocol stack impact video quality and QoE in wireless networks?
• RQ2 What are the key trade-offs between performance gains, implementation complexity, and compatibility in multipath video streaming solutions?
• RQ3 To what extent can cross-layer design improve scheduling decisions by integrating information from multiple protocol layers?
• RQ4 How do existing approaches handle head-of-line blocking, packet loss, and bandwidth variations in heterogeneous wireless environments?
• RQ5 What are the open challenges in deploying multipath video streaming at scale, particularly regarding standardization, compatibility, and QoE optimization?

Key findings

• Multipath transmission improves video quality by up to 3 dB in PSNR and reduces loss rates by up to 64.14% compared to single-path streaming in bursty environments.
• RTRA and MPRTP achieve up to 50% better bandwidth utilization and up to 50% reduction in startup delay and playback interruptions in dynamic bandwidth scenarios.
• Cross-layer scheduling outperforms single-layer approaches by integrating buffer state, RTT, and available bandwidth for better decision-making.
• MPLOT achieves over 50% goodput improvement in high-loss mesh networks by leveraging multiple TCP paths.
• Approaches like Apostolopoulos et al. reduce initial PSNR degradation from 12–15 dB (single path) to only 1.5–7 dB in lossy networks.
• Only a minority of surveyed works evaluate QoE using end-user feedback or perceptual metrics, highlighting a gap in user-centric performance evaluation.