適用於物聯網應用之壓縮感知上行多用戶偵測系統設計

Compressive Sensing-Based Uplink Multi-User Detection Scheme for Internet-of-Things (IoT) Applications

劉佳琛(Jiachen Liu)

指導教授 ： 吳安宇

國立臺灣大學/電機資訊學院/電子工程學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.10289

摘要

未來數年內，全球物聯網設備的數量將會達到數十甚至上百億。在如此巨量的增長下，物聯網網路的可擴展性將會受到嚴峻的挑戰。物聯網通訊的大部分均屬於上行通訊，這是由於網路中的用戶基本都是功能、結構簡單的傳感器設備，必須將數據傳送至網路的中心節點才能做進一步處理與分析。因此，中心節點在進行多用戶偵測時，要如何高效、準確的偵測出被傳送的資訊就成為了一個突出的議題。
傳統上，多用戶偵測主要依靠分配給各用戶不同的展頻碼來進行。這些展頻碼互不相同且相關性極低，所以被加總接收到的訊號與不同展頻碼間的相關值高低便直接說明了相應的用戶是否在活躍狀態。然而隨著用戶數量的激增，生成、儲存與使用這些性質特殊展頻碼進行運算的成本會迅速增長。為了因應這一問題，有研究者為多用戶偵測引入了壓縮感知這一技術。壓縮感知利用原多用戶訊號的稀疏性，允許在展頻係數低於總用戶數時仍能夠較成功的進行多用戶偵測。然而，已有的壓縮感知多用戶偵測系統仍然存在可擴展性不足、偵測運算成本過高、準確性有待提高等問題。
我們提出一種以多測量矢量模型為基礎的壓縮感知多用戶偵測系統。多測量矢量壓縮感知可將物聯網中存在的用戶稀疏性與傳輸突發性結合起來，滿足其偵測時對列稀疏性（而非單矢量模型中的單位稀疏性）的要求，進一步提高系統的偵測準確度。並且，我們所提出的系統相較單矢量模型的壓縮感知偵測系統，可將模型中的取樣矩陣尺寸大大減小，以降低運算成本並提高偵測速度。此外，我們還提出了一種延遲值偵測方法並將之與前述多用戶偵測系統結合起來，使得所提出的系統可二段式的偵測在物聯網問題中更可能出現的非同步多用戶傳輸。

關鍵字

壓縮感知 ； 多用戶偵測 ； 物聯網

並列摘要

The scale of Internet-of-Things (IoT) devices is estimated to reach tens of billions in the coming few years. With such a level of increase, the user scalability of IoT networks is expected to be seriously challenged. Most of the IoT communication is uplink, as users are usually simple sensor devices depending on central gateway units to process the collected data. Therefore, the gateway unit is required to efficiently and accurately detect the transmitted symbols in the Multi-User Detection (MUD) process.
Traditionally, MUD relies on highly-uncorrelated spreading sequences allocated to the users. As the number of users soars, generating, storing and computing with more (and thus longer) spreading sequences will be excessively costly in the context of IoT networks. To change that, researchers have introduced Compressive Sensing (CS) to exploit user sparsity commonly seen in IoT applications. However, these CS-MUD schemes had to tread lightly when trading-off between detection speed and accuracy.
We propose a reformulated detection scheme based on the much more efficient Multiple Measurement Vector Compressive Sensing (MMV-CS) model. By combining the features of user sparsity and sporadicity in IoT networks, the joint sparsity condition required by MMV-CS is perfectly met in our design to improve detection accuracy. The proposed scheme uses a much smaller sampling matrix for detection, thus boosting the detection speed. Moreover, a delay detection scheme with relatively low overhead is introduced to enhance the MMV-CS-MUD scheme. The resulting two-stage detection scheme is able to efficiently cope with unsynchronized user transmissions, which is commonly expected for low-power IoT networks.

並列關鍵字

compressive sensing ； multi-user detection ； internet-of-things

參考文獻

[1] A. Zanella, et al., “M2M massive wireless access: challenges, research issues, and ways forward,” in 2013 IEEE Globecom Workshops, Atlanta, GA, 2013, pp. 151-156.

[2] H. Kopetz, Real-Time Systems. New York: Springer, 2011.

[3] F. Monsees, et al., “Sparsity aware multiuser detection for machine to machine communication,” in 2012 IEEE Globecom Workshops, Anaheim, CA, 2012, pp. 1706-1711.

[4] C. Bockelmann, H. Schepker and A. Dekorsy, “Compressive sensing based multi-user detection for machine-to-machine communication,” Transactions on Emerging Telecommunications Technologies, vol. 24, no. 4, pp. 389-400, 2013.

[5] S. Verdú, Multiuser Detection. Cambridge: Cambridge University Press, 1998.