Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/20487
Title: | 5G行動網路結合機器學習之無線資源管理演算法 Radio Resource Management Algorithms with Machine Learning for 5G Mobile Networks |
Authors: | Li-Sheng Chen 陳立勝 |
Advisor: | 郭斯彥(Sy-Yen Kuo) |
Keyword: | 第五代行動網路,增強型行動寬頻,新無線電,大規模機器類型通訊,無線資源管理,自適應調變與編碼,重傳,機器學習,反向傳播人工神經網路,k-最近鄰居法, 5G,enhanced mobile broadband (eMBB),new radio (NR),massive machine type communications (mMTC),radio resource management (RRM),adaptive modulation and coding (AMC),repetition,machine learning,back propagation artificial neural network (BP-ANN),k nearest neighbor (KNN), |
Publication Year : | 2020 |
Degree: | 博士 |
Abstract: | 在第五代(5G)行動網路增強型行動寬頻 (enhanced mobile broadband, eMBB) 與新無線電 (new radio, NR) 中,自適應調變與編碼 (adaptive modulation and coding,AMC) 是一種重要的無線資源管理 (radio resource management, RRM) 技術。AMC 為依無線通道狀態與品質,能自動調整無線鏈結的調變、編碼方式與其相關參數,提昇無線鏈結傳輸品質,並增強頻譜利用效率。而在收發的傳輸過程中會有通道估計的誤差,無線通道的訊號對干擾加雜訊比 (signal to interference plus noise ratio,SINR) 量測也會有誤差。當發送端所收到與估算的通道狀態與其品質資訊可能不精確,會提高資料傳輸的錯誤率,也會降低頻譜利用效率。因此本研究探討這樣的問題,並提出解決與改進方案,針對5G eMBB與 NR 系統分別提出的基於機器學習的AMC 演算法,將通道狀態與訊號品質量測誤差、干擾、HARQ重傳次數..等作為學習模型的特徵,從而對5G eMBB與 NR 系統的通道狀態與訊號品質 進行較精確的預測,以機器學習式提出更優良的AMC演算法,從實驗模擬的結果來看,所提的演算法可大幅提高系統的傳輸效率和頻譜利用率。 另外,在 5G大規模機器類型通訊(massive machine type communications, mMTC) 應用情境中,訊號品質較差或覆蓋範圍要求較大的終端設備需要使用更多重傳 (Repetition) 來補償額外的訊號衰減。終端設備的Repetition次數過多會浪費寶貴的無線資源,而Repetition次數不足會導致接收端的資料接收失敗。根據訊號品質選擇合適的Repetition次數相當的重要。因此,本研究提出了在5G mMTC應用情境的增強型機器類型通訊 (enhanced machine type communication, eMTC)系統RRM技術機器學習式重傳演算法,以有效提高整體網路傳輸效率。模擬結果呈現本研究所提出的學習式Repetition演算法效率更高,能夠有效地提高成功傳輸率,資源利用率,並能降低非必要的Repetition次數和平均能源消耗。 Adaptive modulation and coding (AMC) is a link adaptation technique based on the physical layer of fifth-generation (5G) enhanced mobile broadband (eMBB) and new radio (NR) systems. The basic principle of AMC is that under constant transmission power, AMC ensures the quality of wireless link transmission and maximizes spectral efficiency through reasonable link transmission modulation and coding rate adaptation. Under poor channel conditions, low modulation and coding rates are selected; by contrast, under favorable channel conditions, high modulation and coding rates are employed. This maximizes the transmission efficiency and ensures communication quality. During the selection of a suitable modulation and coding rate, the transmission rate must be adjusted according to channel changes, thus maximizing the transmission capability of channels. Because of equipment limitations in the receiving end, channel estimation algorithms cannot be used to acquire ideal solutions, and estimation errors are inevitable. Simultaneously, the measurement of SINR of the channel also has some deviation. Consequently, parameter information returned from the receiving end may be inaccurate, which may impede the reliability of data transmission and reduce spectrum efficiency. Therefore, these problems concerning conventional AMC must be addressed. This study proposed the AMC with learning approaches for 5G eMBB and NR systems. The algorithms addresses problems regarding reduced efficiency in conventional AMC caused by channel and noise estimation errors. In particular, these errors and channel quality (interference and HARQ retransmissions etc.) are used as features for machine learning, which allows eMBB and NR systems to accurately estimate the SINR and ensure that the estimated value approximates the value determined through ideal channel estimation. Accordingly, the transmission reliability and spectrum efficiency of such systems can be enhanced. The simulation results indicated that the proposed algorithms could achieve ideal performance and a superior bit error rate compared with a scheme of AMC technology with a lookup table, and furthermore, a lower bit error rate could improve the transmission reliability and spectral efficiency in 5G eMBB systems. In 5G massive machine type communication (mMTC) scenario, user equipment with poor signal quality requires numerous repetitions to compensate for the additional signal attenuation. However, an excessive numbers of repetitions consume additional wireless resources, decreasing the transmission rate and increasing energy consumption. An insufficient number of repetitions prevents the successful deciphering of data by receivers, leading to a high bit error rate. The present study developed adaptive repetition algorithms with machine learning to substantially increase network transmission efficacy for the enhanced machine type communication (eMTC) system in 5G mMTC scenario. The simulation results showed that the proposed repetition with learning approach effectively improved the probability of successful transmission, resource utilization, the average number of repetitions, and average energy consumption. It is therefore more suitable than the common lookup table is for the eMTC system in 5G mMTC scenario. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/20487 |
DOI: | 10.6342/NTU202004174 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 電機工程學系 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
U0001-2608202018093200.pdf Restricted Access | 2.41 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.