應用於低延遲多天線偵測之選區平行搜尋法

Abstract

無線通訊中使用了多天線傳輸系統改善了通道容量與資料傳送的可靠度，最大相似演算法是已知的最佳解，但難實作在高維度的調變上，針對這問題，分散式k最佳演算法可以運用在高維度的調變中，且複雜度不會受調變點數的影響，更重要的是，總體的節點拜訪數量是降低的，然而它卻犧牲了平行架構的可能性，導致其搜索時間間距過長。 在這篇論文裡，我們改良分散式K演算法，推薦一個具有高輸出率、固定複雜度的選區平行搜尋法(Electing Parallel Search)，去解決分散式K演算法搜索時間間距過長的問題。將候選人分割成不同的選區，每塊選區各自挑選自己的候選人，達到平行運作的目的，最後根據條件把候選人交換，補償其效能。與原始的分散式k最佳演算法相比，可以降低在搜索每層最佳子點所需時間。 實作於IEEE 802.11n的通訊平台上，提供4T4R和8T8R在高維度64QAM、256QAM調變，在符合TGN-E所規範的通道模型中與分散式K演算法進行模擬，如果維持PER在0.08在4T4R 64QAM 誤差為0.1dB模擬結果、在8T8R 256QAM則是0.27dB內，與原始分散式k最佳演算法相比，改善了平行度不佳的問題且搜索時間只有原來的1/6，因此，選區平行搜尋法為多輸入多輸出系統提供了具有低複雜度、接近效能最佳化的偵測演算法。

MIMO-OFDM is widely adopted for recently wireless communication systems for enhancing capacity and reliability. The maximum likelihood (ML) is known to be best solution; however, it’s impractical to implement on high modulation order. Distributed K-best (DKB) can work on high modulation order and its cost no longer depends on modulation mode. More importantly, total comparing operation is decreased. However, it lacks of the parallelism. We propose the Electing Parallel Search that enhances the parallelism of DKB. Divide different comparison range to each election district and select different winners from each electoral district. After this step, swap survivors between neighboring electoral districts according to their PED, which let different election work dependently. Through simulation in IEEE 802.11n platform with TGN channel E in 64QMA/256QAM, it indicates performance loss of proposed algorithm is less than 0.1dB (4T4R 256QAM), 0.27dB (8T8R 256QAM) based on PER 0.08. The latency of DKB is larger than that of EP Search. Hence, the proposed algorithm provides a near-optimal solution and shortly latency for MIMO system.