DRL-Based Adaptive Algorithm Selection for Massive MIMO Resource Allocation

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Communication Engineering > Electronic Thesis & Dissertation > Item 987654321/83864

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/83864

Title:	DRL-Based Adaptive Algorithm Selection for Massive MIMO Resource Allocation
Authors:	周彥丞;Chou, Yen-Cheng
Contributors:	通訊工程學系
Keywords:	強化學習;資源分配;多天線陣列;結合排程及預編碼;多視角影片應用
Date:	2020-08-20
Issue Date:	2020-09-02 17:17:13 (UTC+8)
Publisher:	國立中央大學
Abstract:	大規模多輸入多輸出天線陣列已經被視為一個在第五代通訊系統中新興的解決方案。大規模多輸入多輸出天線陣列透過在傳送端使用大量的服務天線以及操作在時分多工上，突破了當前實現上的限制。額外的天線有助於將能量集中在比較小的空間區域，從而極大的提高吞吐量及傳輸效能。然而在未來5G通訊中，資料流量將迅速的成長，特別是超高品質的多媒體，使得5G通訊中資源分配系統的效能越來越重要。隨著人工智慧的發展，也越來越多的研究使用機器學習來解決通訊網路的問題。因此，在本文我們提出了一種基於深度強化學習(DRL) 模型的適應性算法選擇架構，擴展成為解決在大規模多輸入多輸出天線陣列環境中多視角影片(MVV)的天線分配及影像合成(AAVS)問題的UMVS方法。在新的架構中，結合了兩種特殊的方法來輔助原本的UMVS方法，使得分配系統能表現得更全面。此外，為了整合大規模多輸入多輸出天線陣列的資源分配問題，本文還提出一個結合使用者排程和混合預編碼的問題，並使用深度強化學習模型來解決。除了使用深度強化學習模型，我們還將在深度強化學習模型中的動作(Action)設計為組件化的動作(Componentized Action)，使得資源分配系統更加的靈活。;Massive multiple-input multiple-output (MIMO) is one of the essential technologies for 5G networks and beyond. At the same time, the traffic demand grows rapidly, largely due to emerging high-quality multimedia applications, and requires next-generation resource management approaches. In this work, we propose adaptive models selecting MIMO resource allocation methods with Deep Reinforcement Learning (DRL) on Multi-View video (MVV) multicasting and cross-layer resource allocation problems. Based on the Utility-based Multi-View Synthesis (UMVS) algorithm with theoretically-proven performance, the proposed solution adaptively complements UMVS with other representative methods. Besides, to integrate the resource allocation problem across protocol layers, the joint user scheduling and hybrid precoding problem also investigated with a componentized action model, including fundamental allocation methods. Simulations show more users in the MIMO system are satisfied under DRL-based allocation schemes.
Appears in Collections:	[Graduate Institute of Communication Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	256	View/Open

社群 sharing

Loading...