隨著網路流量快速的成長，根據Cisco報告指出2022年時的流量將會是2017年8倍的，而且影像類型的應用所產生的流量占整體的80%。另外第五代通訊的來臨，由於系統的靈活度以及可調性，第五代通訊系統將會允許更多種極端類型的應用服務，例如高速傳輸的高畫質影片、極低延遲的即時串流服務或混合型的線上虛擬實境遊戲，因此下一代的行動通訊即將面臨更壅塞以及更複雜的網路環境。因此第五代通訊的資源排程問題會相對過往的複雜很多，許多論文致力於提出最佳的資源排程方式，研究結果顯示絕大數的機器學習方式都更適合傳統的排程方式。而有一部分文章除了使用機器學習的方式外更加入了環境的分析協助機器學習做出更佳的排程決策。本篇文章，運用了Google DeepMind所開發之DDPG機器學習演算法在多基地台網路環境中排程資源，並透過邊緣雲端伺服器對所有基地台下達資源分配的指令。另外，我們加入了分組的機制，並透過DIBR以及MBMS的技術提高頻寬的使用效率。從實驗結果可知我們的機器學習方法相較於單一的排成方式雖然不能達到最高的傳輸速率，但是在使用者滿足的方面我們提出的學習方法可以達到最高的滿足率。

The traffic flow grows rapidly in recent years. According to Cisco report, the network traffic of 2022 will be more than 8 times of 2017, and the traffic flow generated by image-type application will occupy up to 80% of total traffic flow. In addition the 5G system’s characteristic of flexibility and adjustability let the 5G system allow to serve more different application with different request and characteristic, e.g., The ultra-high video, video streaming and online interactive virtual reality gaming. So that next generation mobile communication system is going to face a higher congestion and higher complexity network environment. Therefore, the resource scheduling problem is more complicate than the past. Many papers devoted to proposing the best resource scheduling scheme. The results show the machine learning are more suitable for the 5G network. There are some papers analyze the network environment to assist the machine learning to make better scheduling decision. In this article, the DDPG, a machine learning algorithm developed by Google DeepMind is used to schedule resources in a multi-base network environment, and allocate resources to all base stations through the edge cloud server. In addition, we have added a grouping mechanism to improved bandwidth usage efficiency through DIBR and MBMS technologies. From the experimental results, we can see that our machine learning method cannot achieve the highest transmission rate compared with the single arrangement method, but the learning method proposed by the user can achieve the highest satisfaction rate.

[1] J. Park, P. Popovski, and O. Simeone. Minimizing latency to support vr social interactions over wireless cellular systems via bandwidth allocation. IEEE Wireless Communications Letters, 7(5):776–779, Oct 2018.

[2] Lin Wang, Lei Jiao, Ting He, Jun Li, and Max M¨uhlh¨auser. Service entity placement for social virtual reality applications in edge computing. IEEE INFOCOM 2018 - IEEE Conference on Computer Communications, pages 468–476, 2018.

[3] M. Chen, W. Saad, and C. Yin. Virtual reality over wireless networks: Quality-of-service model and learning-based resource management. IEEE Transactions on Communications, 66(11):5621–5635, Nov 2018.

[4] J. Guo, X. Gong, J. Liang, W. Wang, and X. Que. An optimized hybrid unicast/multicast adaptive video streaming scheme over mbms-enabled wireless networks. IEEE Transactions on Broadcasting, 64(4):791–802, Dec 2018.

[5] Y.-C. Chou S.-C. Tseng, Z.-W. Liu and C.-W. Huang. Radio resource scheduling for 5g nr via deep deterministic policy gradient. IEEE International Conference on Communications Workshops (ICC WS), 2019.

[6] Timothy P. Lillicrap, Jonathan J. Hunt, Alexander Pritzel, Nicolas Heess, Tom Erez, Yuval Tassa, David Silver, and Daan Wierstra. Continuous control with deep reinforcement learning, 2015.

[7] M. Chen, W. Saad, and C. Yin. Resource management for wireless virtual reality: Machine learning meets multi-attribute utility. In GLOBECOM 2017 - 2017 IEEE Global Communications Conference, pages 1–7, Dec 2017. 31

[8] M. Chen, W. Saad, C. Yin, and M. Debbah. Echo state transfer learning for data correlation aware resource allocation in wireless virtual reality. In 2017 51st Asilomar Conference on Signals, Systems, and Computers, pages 1852–1856, Oct 2017.

[9] Jar-Ferr Yang, Shih-Cheng Chang, and Chin-Yun Chen. Computation reduction for motion search in low rate video coders. IEEE Transactions on Circuits and Systems for Video Technology, 12(10):948–951, Oct 2002.

[10] J. Lee, D. Yang, Y. Chen, and W. Liao. Efficient multi-view 3d video multicast with depth-image-based rendering in lte-advanced networks with carrier aggregation. IEEE Transactions on Mobile Computing, 17(1):85–98, Jan 2018.

[11] Y. Mori, N. Fukushima, T. Fujii, and M. Tanimoto. View generation with 3d warping using depth information for ftv. In 2008 3DTV Conference: The True Vision - Capture, Transmission and Display of 3D Video, pages 229–232, May 2008.

[12] 3GPP TS 23.501. System Architecture for 5G System. Technical report.

[13] Ali E Abbas. Constructing multiattribute utility functions for decision analysis. In Risk and Optimization in an Uncertain World, pages 62–98. INFORMS, 2010.

[14] Ali E Abbas. Wimax forum, mobile wimax – part i: A technical overview and performance evaluation. August. 2006.

[15] Raj Jain. Channel models: A tutorial. In WiMAX forum AATG, pages 1–6, 2007.