基於頻譜之無線網路分散式聚落化與鄰近裝置之偵現

Abstract

裝置對裝置(Device-to-Device, D2D)通訊最近成為備受關注的主題。 D2D通訊相較於現今的透過大蜂巢網路(macro-cellular network, MCN)的基站之通訊架構有許多優點，像是它可以提供D2D鏈路較高的傳輸速率，較短暫的延遲，較低的發射功率及干擾，更有助於減輕MCN的負載。然而，如何在現存網路下支援大量D2D通訊且不會影響既有系統之通訊品質仍缺乏實體層上可行的方案。本論文所要解決的問題即是：在不透過既有4G長期演進技術(LTE)網路協助的前提下，如何發現鄰近的裝置並建立空中介面(air interface)。

我們提出了一套利用各自估計的頻譜分散式聚落化（clustering）且自行聚集配對建立實體層通訊鏈路的方法。這套方法是基於相鄰裝置應會感測到相似的頻譜的前提。由於裝置若要比較彼此所感測的頻譜勢必要交換甚多資訊，且同時多個頻譜資訊的交換若無集中式的協調將不可避免的造成彼此的干擾。我們所提的方案能同時解決這些問題，不但大量降低頻譜資訊交換所需頻寬，也讓裝置之間的辨認與鏈路的建立變得簡易可行。頻譜資訊的壓縮我們先透過對子載波與頻譜高度的量化將其轉變成向量，復將這些向量樣視為受雜訊干擾的里德-所羅門(RS)碼字(codeword)來加以解碼。RS碼優異的距離性能讓我們很方便區隔不同的碼字，鑒於前述相鄰裝置應偵測到類似的頻譜，這些裝置的頻譜向量間的小差異，可視為RS碼字上的雜訊。因此鄰近裝置將所估測的頻譜經RS解碼器後，即可獲得相同的聚落名稱（即codeword），並將解出之RS碼字對應到一跳頻探詢信號(probing signal)。相鄰裝置若要知道是否有附近裝置之詢問只要量測相對應的時頻區段之接收能量即可判斷。為防止相鄰裝置同時傳送相同的探詢信號，我們更利用到RS碼之循環特性，讓個別裝置將跳頻信號做隨機循環移動(random cyclic shift)已降低碰撞機率。我們經由電腦模擬證實，假如裝置與現有MCN之用戶位置依照互相獨立之博松點過程（Poisson point processes）分佈，則我們所提方法的確能在沒有經由MCN即可支援數量龐大的D2D通訊。

Direct short range device-to-device (D2D) communication enjoys many a advantage with respect to conventional cellular systems: high transmission rates, lower propagation delays, reduced transmit power and interference level. Moreover, direct D2D links offload traffics from existing macro-cellular networks (MCN) and support emergency communications when the core network fails.

To realize the above-mentioned advantages, a device or user equipment (UE) must be able to discover other devices or UEs in the proximity and establish links with them. The purpose of this thesis is to present feasible and efficient distributed solution to these two critical issues for a D2D communication system underlaying an MCN.

We propose a clustering scheme which generating a probing frequency-hopped (FH) sequence based on the device’s sensed spectrum. Such scheme is based on the presumption that devices in close proximity would observe similar spectrums. Hence if they can exchange the spectrum information the devices with similar spectra should lie in the same neighborhood and be grouped into the same cluster. Our scheme includes a novel spectrum compressing method that not only compresses the spectrum information but converts it into a FH probing sequence with embedded automatically-generated cluster ID to discover if there exit devices in the neighborhood. The main idea is to model the sensed spectrum, which is represented by a real vector denoting the spectral heights at the subcarriers of interest, as a corrupted nonbinary codeword so that similar spectra are different noise-corrupted versions of the same codeword while dissimilar spectra are originated from different codewords. As Reed-Solomon (RS) codes are maximum distance separable (MDS), they are perfect candidate code to model the sensed spectra and to resolve dissimilar spectra.

We first quantize the spectrum vector into an M-ary vector and decode it into a legitimate systematic RS codeword. The non-binary RS codeword is then randomly cyclic-shifted and used to generate an FH sequence with the information symbols as the cluster (neighborhood) ID. Such a design allows multiple simultaneous probing and clustering with minimum collision. It also naturally fits into an interference avoiding adaptive spread spectrum transmission scheme.

We verify the feasibility and efficiency of our solution through computer simulations, assuming that devices and MCN UEs are located according to mutually independent Poisson point processes (PPP) with different densities.