Hypertext Transfer Protocol (HTTP) adaptive streaming aims to achieve high Quality of Experience (QoE) in video streaming. However, when multiple clients stream videos, QoE fairness and overall QoE deteriorate due to the lack of consideration for multi-client environments and dynamically changing network environments. This paper proposes HTTP adaptive streaming based on edge collaboration using Reinforcement Learning (RL) in a multi-client environment. The proposed scheme trains policies based on RL to improve the QoE. In addition, an edge collaboration scheme has also been introduced to improve the overall QoE and QoE fairness by using a client reallocation strategy. In the proposed scheme, edge collaboration means reallocating clients to an edge network that can produce the best QoE performance. The proposed scheme ensures QoE fairness by modifying the existing RL-based adaptive streaming technique to be more suitable for a multi-client environment and creates a more robust adaptation policy for changes in network conditions. Further, due to edge collaboration, clients can maximize overall QoE by using the model that best suits their environment. The experimental results confirm that the proposed scheme leads to better overall QoE and QoE fairness performance than existing schemes in a multi-client environment.

超文本传输​​协议 (HTTP) 自适应流媒体旨在实现视频流媒体的高质量体验 (QoE)。然而，当多个客户端流式传输视频时，由于缺乏对多客户端环境和动态变化的网络环境的考虑，QoE 公平性和整体 QoE 会恶化。本文提出了在多客户端环境中使用强化学习（RL）的基于边缘协作的 HTTP 自适应流。所提出的方案基于 RL 来训练策略以提高 QoE。此外，还引入了边缘协作方案，通过使用客户端重新分配策略来提高整体 QoE 和 QoE 公平性。在所提出的方案中，边缘协作意味着将客户端重新分配到可以产生最佳 QoE 性能的边缘网络。所提出的方案通过修改现有的基于强化学习的自适应流技术以更适合多客户端环境来确保 QoE 公平性，并针对网络条件的变化创建更稳健的自适应策略。此外，由于边缘协作，客户可以通过使用最适合其环境的模型来最大限度地提高整体 QoE。实验结果证实，在多客户端环境中，所提出的方案比现有方案具有更好的整体 QoE 和 QoE 公平性性能。