下世代無線裝置通訊技術挑戰與其解決方案

Abstract

隨著無線通訊技術的持續進步，頻譜效率以及資料傳輸率比起以往已有了長足的進步。然而，除了持續提升現今行動寬頻系統資料以因應資料流量的爆炸性成長，許多基於新型態裝置的嶄新服務如物聯網(IoT)、近程通訊(Proximity Service)以及無線前行/回程網路(Wireless Fronthaul/Backhaul)預期也將在下世代通訊系統中扮演重要的角色。這些新興應用各有其特定的效能需求，為了將之付諸實現，有幾種關鍵技術在近年來開始受到矚目。本論文中，我們將關注於幾樣代表性關鍵技術的現行技術挑戰，並提出新的解決方案。這些關鍵技術包含巨量機器間通訊(Massive M2M)、裝置間通訊(D2D)以及全雙工系統(FDR)。 有效率的描述與傳送排程資訊可避免大量的信令虛耗，為使下世代巨量機器通訊網路能支援巨量裝置的關鍵之一。為了達到此目標，我們提出了一新穎的描述法，簡稱為SRD，來有效率的描述一個正交分頻多重存取(OFDMA)訊框下的時頻資源分配。SRD首先將一個OFDMA訊框分割成多個矩形，稱之為方形集。其中每個方形集內的資源單位皆屬於相同的資源配置。我們證明了每個方形集的其中一角座標(如左上角)以及這些方形集間的一線性順序關係足以還原整個被方形集切割的訊框，意即僅需傳送經過特定排序後的方形集左上角座標即足以表示一個訊框內的資源分配情形。我們同時亦提出了排序方形集與重組訊框的演算法，其時間複雜度皆與方形集總數呈線性。從模擬實驗中，我們驗證了SRD確實相較於現行系統所使用的方法能帶來顯著的容量提升，且這些提升在有大量使用者的系統中會更加明顯。 在第二個主題中，我們探討與巢狀系統結合的裝置間通訊網路下的時間不對齊問題。我們發現避免時間不對齊係為使裝置間通訊系統能以合理通訊品質運作的關鍵之一。否則，嚴重的區塊間干擾(IBI)以及載波間干擾(ICI)將會發生，這將使得多個裝置間通訊配對同時傳收資料的使用情境近乎無法運作。為了解決時間不對齊問題，我們提出了一種新的方法：簡稱為JDWM。JDWM可藉由平移裝置間通訊裝置接收端的離散傅立葉轉換窗口來避開時間不對齊所造成的區塊間干擾與載波間干擾。其中離散傅立葉轉換窗口的平移值以取樣時間為單位，稱之為跳躍值。我們證明了可以完全避開所有區塊間干擾與載波間干擾的合適跳躍值存在的充分條件，並提出一演算法使裝置間通訊裝置接收端可藉由現行LTE-A系統上行資料通道所使用的參考訊號來估測合適跳躍值。在模擬實驗中，我們證實了相較於傳統的端對端時間同步，JDWM確實可有效的避開因時間不對齊所造成的干擾問題，並達到與理想全域同步十分接近的封包錯誤率。 在第三個主題中，我們探討全雙工裝置中自我干擾的消除方法。一般而言，自我干擾與所需訊號往往有著相當大的能量差距(如100 dB)，所以在裝置的數位區域中直接進行全雙工傳輸的自我干擾消除幾乎是不可行的。有鑑於此，我們提出了一新的演算法：簡稱為PUM ，來調整多路類比自干擾消除濾波器的路徑增益值。PUM使用一基於比較剩餘自我干擾能量量測值來更新增益值的機制來調整類比自我干擾消除濾波器的路徑增益值。經過足夠次數的更新後，PUM可逼近可達的最小自我干擾能量剩餘值。除此之外，PUM適用於多路徑通道環境，且不需要事先取得通道資訊，這些特性可有效簡化PUM的實作。模擬結果證實在頻寬20 MHz的系統下，PUM可達到距離雜訊下限僅5.5 dB的剩餘SI能量。如此低的剩餘SI能量足以在裝置的數位區域中以相對簡易的方式量測並進一步消除。 總結而言，在本論文中，我們探討了下世代裝置通訊中的數個關鍵性技術議題並提出新的解決方案。藉由分析與模擬實驗等方法，我們相信我們提出的方法不僅有效，同時也具有可行性，可做為克服這些技術難題的候選解法。

The ongoing evolution of wireless techniques gains great progress in improving the spectrum efficiency and data rate. However, enhancing the achievable data rate to accommodate the explosive growing of data traffic is not the only goal in the next generation communication system. Besides the mobile broadband data service, numbers of brand-new applications are developed based on a varieties of new-type devices, such as Internet of Things (IoT), proximate services, and wireless front/backhaul, and they are supposed to also play important roles in the near future. To realize these emerging applications with diverse requirements, several new essential technologies gain attentions in recent years. In this thesis, we aim to solve the critical challenges of three of the representative ones, namely, massive machine-to-machine networks, device-to-device (D2D) communication, and full-duplex radio (FDR) systems. To enable an M2M network supporting massive number (e.g. more than one thousands) of devices, efficient description and conveying method of the scheduling information to scheduled users is essential to avoid the tremendous increasing of signaling overhead. To do this, we proposed a new efficient description approach, namely, Sorted-Rectangular Description (SRD), to describe the time/frequency resource allocations in an OFDMA frame. SRD first partitions a frame into rectangular polygons, namely rectangular bursts (RB), in which each RB contains the resource units belongs to the same assignment. We showed that an RB-partitioned frame is sufficient to be reconstructed by the top-left coordinates of each RB in a specific sorted order. Moreover, the sorting and reconstruction algorithms with linear time complexity in the total number of RBs was also proposed. Simulation results showed that, when it was compared with existing methods for conveying the scheduling information, SRD could provide dramatic improvement of the capacity in a system with large number of users. Concerning the deployment of cellular overlaying D2D networks, we found that mitigating the timing misalignment (TM) among nearby D2D devices is critical to allow more than one D2D source devices to convey data to their D2D sink devices simultaneously. Otherwise, severe interference problems including inter-block interference (IBI) and inter-carrier interference (ICI) might occur and make acceptable reception quality impossible. To cope with this problem, we proposed a new approach to avoid the interferences at the receiver side via shifting the Discrete Fourier Transform (DFT) window by a number of sample durations, namely, jumping value. We proved the sufficient condition for the existence of a proper jumping value which could ideally mitigate the interferences. An algorithm for estimating the proper jumping value via the existing reference signal structure of LTE-A D2D data channel is also developed. Simulation results shown that our proposed JDWM (Jumping DFT Window Method) approach could provide much more effective interference mitigation than that of traditional end-to-end timing synchronization methods and achieve a packet error rate quite close to that under perfect global timing synchronization. In the study of the final topic, we aim to solve the problem of mitigating the strong self-interference (SI) for full-duplex radio (FDR) devices. In general, achieving SI cancellation at the digital domain in the initial stage is very difficult since the transmitted signal from the transmit chain of an FDR device could be dramatically stronger (e.g. larger than 100 dB) than that of the received desired signal. Thus, we proposed a new algorithm, called, probe-and-update method (PUM), to tune the gains of a multi-tap analog SI cancellation filter. PUM performs a novel gain updating procedure based on the comparison of two consecutive measurements of the residual SI power. Via sufficient number of updates, it can be proved that PUM could approach the minimum average residual SI power. Moreover, PUM is applicable in multipath environment and no channel knowledge is needed in advance, which considerably facilitate the implementation. Simulation results showed that PUM could achieve a reasonably low residual SI power down to 5.5 dB of the noise floor in a 20-MHz system. The remained SI is sufficiently low to be measured and canceled in the digital domain. In summary, several new methods are developed to provide solutions for some of the essential problems known in the next generation device based communication system in this thesis. By our analysis and simulations, we believe the proposed methods are both effective and practical, and can be the possible solutions for overcoming these critical challenges.