基於深度學習技術於無細胞大規模多輸入多輸出系統做最小化功率分配;Minimizing Power Allocation in Cell-Free Massive MIMO Systems Using Deep Learning Techniques

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Communication Engineering > Electronic Thesis & Dissertation > Item 987654321/93006

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/93006

Title:	基於深度學習技術於無細胞大規模多輸入多輸出系統做最小化功率分配;Minimizing Power Allocation in Cell-Free Massive MIMO Systems Using Deep Learning Techniques
Authors:	陳柄富;Chen, Bing-Fu
Contributors:	通訊工程學系
Keywords:	無細胞大規模多輸入多輸出;深度學習;功率分配;cell free massive multiple-input multiple-output;deep learning;power allocation
Date:	2023-11-16
Issue Date:	2024-09-19 16:38:16 (UTC+8)
Publisher:	國立中央大學
Abstract:	無細胞大規模MIMO系統被視為5G和6G的重要技術之一，具有相當潛力。該系統不同於傳統蜂巢式網路架構，其中包含一個中央控制器和眾多的無線存取點(AP)，分佈在覆蓋的範圍內。每個無線存取點皆搭載多根的服務天線，形成獨特的結構，通過相干聯合傳輸同時為覆蓋範圍內的所有用戶進行服務。一個實際的挑戰是如何在受限於限制功率下，選擇用戶的設備並進行適當的功率分配，以便可以在所有用戶設備獲得適當的傳輸功率下並且不超過所限制的功率。本文考慮了在無細胞大規模多輸入多輸出(MIMO)系統下行鏈路中的功率分配問題。我們基於最大比(MR)預編碼技術下完成最小化問題，並使用基於大規模衰落(LSF)參數的啟發式功率分配作為DNN的預處理輸入。我們使用隨機分布的存取位置和使用者設備用作訓練數據去訓練DNN。將訓練後的神經網路做比較後可以證明此方法運行時間上的優勢。;The concept of a cell-free massive MIMO system is a promising technology, marked as a pivotal element for 5G and 6G advancements. Contrary to the conventional cellular arrangement, the cell-free massive MIMO system involves a central control unit and an array of wireless access points (APs) dispersed across the coverage zone. Each of these APs is equipped with an extensive set of service antennas. This architecture facilitates coherent joint transmission, capable of concurrently servicing all user devices within the coverage area. However, a practical hurdle lies in efficiently selecting user devices and distributing power within predetermined limits to ensure optimal transmission without breaching stipulated power constraints. This study tackles the power allocation predicament within the downlink facet of a cell-free massive MIMO system. We tackle this problem using the maximum ratio (MR) precoding technique and formulate the minimization problem. We utilize a heuristic power allocation technique relying on large-scale fading (LSF) coefficients as an initial input for the deep neural network (DNN). Training data for the DNN is generated using a random distribution of access points and user iii devices. After training, we compare the performance of the trained neural network to demonstrate its advantages in terms of runtime.
Appears in Collections:	[Graduate Institute of Communication Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	41	View/Open

社群 sharing

Loading...