用於鄰近裝置偵現與自主性聚落化之具本地資訊及驗證機制的展頻ALOHA規約

Abstract

鄰近裝置偵現(proximity discovery)在建立無線感測(wireless sensor)或隨建即 連(Ad hoc)網路中有著重要的作用，亦是在短距離裝置對裝置(Device-to-Device, D2D)和機器對機器(Machine-to-Machine, M2M)通訊中的關鍵技術之一。隨著全 球行動數據的流量快速增長，D2D 通訊近來備受關注。有別於利用大蜂巢網路 (macro-cellular network, MCN)的大型基地台之通訊架構，D2D 通訊提供了許多 優勢也滿足各種應用需求。例如提供較高的傳輸速率，低時間延遲，低發射功 率和低干擾，更重要的是有助於分流基地台與核心網路(MCN)的負載，提升頻 譜使用效率。然而，在現行通訊網路環境下，考慮不影響既有系統的通訊品質 且支援大量D2D 之通訊，仍有許多技術與應用上的挑戰，也缺少實體層上的實 際方案。本篇論文所要探討與解決的問題為：在不藉由既有4G 長期演進技術 (LTE)網路的協助為前提下，如何建立一套自主性系統來偵測鄰近的裝置，進一 步配對並建立連線，且利用數學分析來提升模擬效能以及評估整體系統設計的 效應。 在我們的設計架構中，基於鄰近裝置有較大機率會感測到相似頻譜，提出了 裝置與裝置之間利用偵測頻譜資訊來達到自行聚集，進一步建立配對鏈路的實體 層方法。在無協調者的情況且同時多個裝置之間感測頻譜資訊的交換，我們使用 展頻分段ALOHA(spread slotted ALOHA)的系統來降低訊號間的干擾和碰撞，且 ii 此展頻碼結合裝置的本地資訊。我們利用一套頻譜資訊壓縮機制，透過對頻譜高 度量化壓縮後得到一組向量，將向量視做受雜訊影響的里德-所羅門(RS)碼字 (codeword)後來加以利用其特性解碼，RS 碼優異的距離特性能讓我們很方便的區 隔不同的碼字，且每一組RS 碼字對應到一跳頻探詢信號(probing signal)，有相同 RS 碼字的裝置即被歸為同一聚落(cluster)，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)來降低碰撞機率。鄰近裝置透過時頻區段上的能量偵測即可 知道附近是否有裝置詢問，藉由裝置間探測、回覆和確認的機制來尋找鄰近裝置 並建立連線。另外，除了找尋和本身相同聚落(cluster)的裝置外，我們使用RS 碼 錯誤和抹除錯誤(Error-and-Erasure)的特性來找到鄰近但其感測頻譜和本身不相 似的額外裝置，大幅提升了在一個範圍內所能找到的裝置數與系統效能。此外， 我們更進一步的提出系統效能評估的數學分析式，降低整體系統效能評估因為多 重變數造成的複雜度，也經由電腦的分析式模擬來加以證實此系統的可靠度。

Proximity discovery refers to the ability for a device to find a access station or other communication devices in the neighborhood. It is an essential learning capability of an intelligent wireless device and is needed in setting up a wireless sensor or ad hoc network autonomously. Direct short distance device-to-device (D2D) or machine-to- machine communications also require that a device/machine be equipped with such a capability. Clustering is useful in a network whose nodes spread randomly over a wide area and direct node-to-node communication may be difficult or energy-inefficient for some. Clustering with cluster heads also simplifies intra-network communications and enables relaying messages over a long range. We develop a distributed device discovery and clustering scheme which requires lit- tle or no assistance from an existing network. The basic idea of our design is to group the devices with similar local information–in this case the spectrum sensed–thereby en- suring the locality or proximity is preserved. As the probing and probed devices are not aware the existence of each other and no multiple devices transmission coordination is available for initial information exchanges, we adopt a noncoherent (slotted) spread Aloha protocol to avoid phase coherent detection need and mitigate potential collisions. The spectrum spreading is accomplished by using local information aware frequency- hopping (FH) patterns. The FH patterns are derived by compressing the sensed spec- trum through quantization and Reed-Solomon (RS) encoding. Although RS-coded FH patterns possess excellent distance property for pattern identification, we include device- specific signature information in the probing FH waveform, an extra FH step, and an additional verification mode in our protocol to reduce the collision probability, enhance the detection probability and avoid false detection. Using error-and-erasure correcting capability of RS code, we present an efficient multi-pattern detection scheme to sepa- rate similar but non-identical probing signals. Detailed initial link setup and clustering procedure are given and extensive simulation resulted are given to verify the feasibility and efficiency of the proposed protocol.