- 學位論文
針對無蜂巢式通訊系統之適應性下行預編碼演算法之設計
Design of Adaptive Downlink Precoding Algorithms for Cell-free Communication Systems
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
隨著現代通訊系統持續的發展，除了追求更高的傳輸速度外，服務品質的一致性也是一直被關注的主題。為了解決傳統蜂巢式系統(cellular system)在細胞邊緣(cell edge)的使用者服務品質低落的問題，無蜂巢式系統(cell-free system)這個名詞開始受到討論，其顧名思義，就是將原有的蜂巢架構移除，讓各個基地台透過協作的方式服務所有使用者，提供更一致的服務品質。而沒有了蜂巢架構的區隔下，基地台對非目標使用者的干擾就需要靠適當的預編碼(precoding)來解決。然而現有的預編碼演算法都欠缺一個有效的機制讓系統可以因應不同的情境進行不同的預編碼，因此本論文針對現有的預編碼演算法進行改良，並提出新的演算法讓基地台在下行時能根據不同環境進行適應性的預編碼，最後進一步透過計算資料傳輸與矩陣運算時的能量成本，從能耗的角度來比較不同的預編碼演算法。 本論文的第二章中，我們會簡介所採用的散射通道模型，以及模擬時的系統參數及情境設定。另外除了會介紹預編碼的數學概念之外，也會介紹分散式與集中式的預編碼，以及兩者分別會用什麼演算法來進行最基本的下行預編碼。 第三章將會介紹本論文提出的分散式預編碼演算法，我們首先基於最基礎的共軛波束成型(Conjugate Beamforming)預編碼進行改良，讓演算法在維持其低運算量的特性下，同時讓基地台能夠抑制對非目標使用者的干擾。相較傳統的預編碼作法，提出的方法運算量下降66%，並且在鏡面式反射環境下提升7%的平均頻譜效率。接下來，由於此提出的演算法在環境中散射情形較嚴重時，預編碼效果會下降，因此我們進一步針對這項缺點做改良，此時我們提出了一個能夠判斷環境中散射程度高低的指標，並且根據這個指標讓基地台能夠適應性的進行準確的切換，得出的分散式可適性預編碼演算法可以在不同環境下皆有最佳的預編碼效果。 在第四章中，為了進一步縮小所提出的分散式預編碼與完全集中式的預編碼在頻譜效率上的差距，此時基於第三章中提出的演算法，加入一定比例的集中式預編碼運算，並提出一個讓系統能因應當下運算忙碌程度進行調整的預編碼流程，讓中央處理器把集中式運算的資源分配給最需要的子頻帶(subband)，系統此時便可以在有限的運算量下提升最多的頻譜效率。最後，為了以一個共同的基準來比較不同的預編碼演算法，我們將所有在下行傳輸時的功耗累計起來，計算經過能源成本標準化的頻譜效益，並且採用不同製程GPU的能耗統計不同製程下的結果。根據得出的數值，提出的預編碼演算法在採用28 nm、40 nm製程的GPU時，可以比現有演算法提升12.2%的能源成本標準化頻譜效益；而採用更先進的製程下，也能比現有演算法提升高達49%。
並列摘要
With the continuous development of modern communication systems, in addition to the pursuit of higher transmission speed, the consistency of service quality is also a topic of constant concern. The term cell-free system has been discussed to solve the problem of low QoS for users at the cell edge of the traditional cellular system. Since the cellular architecture is removed, all access points will serve all users collaboratively, providing uniform QoS to all users. Without the cellular structure, the interference to the unintended users needs to be solved by appropriate precoding. However, the existing precoding algorithms lack an effective mechanism to allow the system to perform different precoding according to different situations. Therefore, this paper modifies the current precoding algorithms and proposes new algorithms to allow the access points to perform precoding based on different situations adaptively. Finally, different precoding algorithms are compared from the energy consumption perspective by calculating the energy cost of data transmission and matrix operation. In the second chapter of this paper, we will introduce the scattering channel model used, the system parameters, and the scenario settings for the simulation. In addition to introducing the mathematical concept of precoding, distributed and centralized precoding will also be introduced, including the most basic downlink precoders. The third chapter will introduce the distributed precoding algorithms proposed in this paper. We first modify the most basic precoder, conjugate beamforming, so that the algorithm can maintain its low computational complexity and simultaneously perform nulling to unintended users. The proposed precoder has 66% less complexity and 7% more spectral efficiency in a specular environment than conventional precoders. Next, since the proposed algorithm has a performance degradation with a severe scattering in the environment, we further improve this shortcoming. We then propose an indicator that can judge the degree of scattering in the environment. And according to this indicator, the access points can adaptively perform accurate switching, and the obtained distributed adaptive precoder has the best precoding effect in different environments. In Chapter 4, a proportion of centralized precoding is added to narrow the gap between the proposed distributed precoder and fully centralized precoder in spectral efficiency. We also propose a precoding process that allows the system to adjust according to the current computing busyness, allowing the central unit to allocate the computing resources to the most needed subbands, leading to maximum spectral efficiency improvement. Finally, to compare different precoding algorithms with the same benchmark, we accumulate all the power consumption during downlink transmission and calculate the energy-normalized spectral efficiency. We also use the energy consumption of GPUs with different processes to calculate the results under different processes. According to the obtained values, the proposed precoding algorithm can increase the energy-normalized spectrum efficiency by 12.2% compared to the existing algorithms when using GPUs with 28 nm and 40 nm processes. When using more advanced processes, it has a 49% improvement compared to current methods as well.
參考文獻
[1] W. Jiang, B. Han, M. A. Habibi, and H. D. Schotten, ‘‘The road towards 6G: A comprehensive survey,’’ IEEE Open J. Commun. Soc., vol. 2, 884 pp. 334–366, 2021.
[2] M. Latva-Aho and K. Leppanen, "Key drivers and research challenges for 6G ubiquitous wireless intelligence," 6G Flagship, Univ. Oulu, Finland, Tech. Rep., 2019.
[3] R. Irmer, ‘‘Coordinated multipoint: Concepts, performance, and field trial results,’’ IEEE Commun. Mag., vol. 49, no. 2, pp. 102–111, Feb. 2011.
[4] Özlem Tugfe Demir; Emil Björnson; Luca Sanguinetti, Foundations of User-Centric Cell-Free Massive MIMO , now, 2021.
[5] R. W. Heath, N. González-Prelcic, S. Rangan, W. Roh, and A. M. Sayeed, “An overview of signal processing techniques for millimeter wave MIMO systems,” IEEE J. Sel. Topics Signal Process., vol. 10, no. 3, pp. 436–453, Apr. 2016.