상세 보기
Deep Reinforcement Learning for Adaptive Bitrate Streaming over Wireless Networks: Advances, Challenges, and Open-Source Simulation
- Dang, Thanh Thien-An;
- Tran, Anh-Tien;
- Dao, Nhu-Ngoc;
- Cho, Sungrae
SCOPUS
0초록
Adaptive Bitrate (ABR) streaming is the de facto mechanism for delivering video over heterogeneous and time-varying wireless networks. However, volatile bandwidth, interference, mobility, and resource contention complicate bitrate selection and buffer control, often degrading users’ Quality of Experience (QoE). Traditional rule-based ABR algorithms struggle to generalize across diverse conditions. Recent work demonstrates that Deep Reinforcement Learning (DRL) can jointly reason over network dynamics and player status to optimize QoE, including quality, rebuffering, smoothness, and latency. This paper surveys DRL-based ABR advances from 2022 to 2025 in wireless contexts and emphasizes the role of open and reproducible research using ns-3 for network simulation and FFmpeg for video processing and objective quality assessment (e.g., PSNR, SSIM, VMAF). We first formalize the DRL problem for ABR (state, action, and reward design), then discuss how ns-3 models cellular and Wi-Fi environments and how FFmpeg enables practical bitrate ladders, decoding, trace generation, and automated quality evaluation. We synthesize findings across recent works on access-point-assisted ABR, meta-RL bitrate guidance, cross-layer optimization for real-time communication, fairness-aware multiuser bandwidth allocation at the mobile edge, and throughput prediction for ABR. We highlight reported gains, modeling choices, and, critically, open-source availability where indicated. We conclude by outlining open challenges—generalization, QoE modeling, and real-world deployment—and advocating for open-source benchmarks, datasets, and truly end-to-end ns-3+FFmpeg pipelines to accelerate impactful and reproducible research in DRL-driven ABR for wireless video streaming.
키워드
- 제목
- Deep Reinforcement Learning for Adaptive Bitrate Streaming over Wireless Networks: Advances, Challenges, and Open-Source Simulation
- 저자
- Dang, Thanh Thien-An; Tran, Anh-Tien; Dao, Nhu-Ngoc; Cho, Sungrae
- 발행일
- 2025
- 유형
- Conference Paper
- 저널명
- International Conference on ICT Convergence
- 페이지
- 675 ~ 680
- 언어
- ENG
- 출판사
- IEEE Computer Society
- 발행국가
- 미국
- 분량
- 6 페이지
- ISSN
- E 2162-1241
P 2162-1233