結合通道估計之混合式波束成形最佳空間預編碼應用於大規模多輸入多輸出正交分頻多工系統

Abstract

本論文延續本實驗室過去所提出之波束成形選擇空間預編碼(Beamforming-selection spatial precoding, BSSP)並延伸到了在第五代行動通訊技術(5th generation wireless systems, 5G)中常見之大規模多輸入多輸出正交分頻多工系統(massive MIMO-OFDM)。我們結合了在5G中常見之混和式毫米波架構(hybrid mmWave structure)與經過灰狼演算法(Grey wolf optimizer, GWO)進行參數最佳化後之BSSP預編碼來節省分頻雙工Frequency-division duplexing, FDD）傳輸中之負擔(overhead)並節省硬體實現時之成本。在本篇論文中體現了過去所提出的多段式最佳化及透過離散灰狼最佳化(discrete grey wolf optimization, DGWO)進行部分連接(partially connected)最佳化在不重疊子陣列(non-overlapped subarray, NOSA)架構下應用於MIMO-OFDM時不僅可以比起在全連接(fully connected)架構時射頻訊號鏈(radio frequency chain, RF chain)連接較少的天線以簡紹功率損耗，更可以擁有更佳之系統平均位元錯誤率(average approximated bit error rate,AA-BER)。另外在本論文中亦考慮四種不同的常見天線陣列設置(antenna array configuration)，透過最佳化平均互消息(average mutual information, AMI)來比較在何種配置及天線位置擺設下可以得到最佳之系統表現。最後，本論文考慮了在通道狀態資訊(channel state information, CSI)不足時，透過了壓縮感知(compressed sensing)中的正交匹配追蹤(orthogonal matching pursuit, OMP)算法和應用張量計算之正交匹配追蹤(tensor-based OMP,T-OMP)算法，分別在全數位和混合式架構下進行了通道估計(channel estimation)。在考量此種狀況後，配合既有的誤差(error effects)影響通道和加入準正交空時編碼(Quasi-orthogonal spacetime block code, QOSTBC)通訊，使得本系統得以發展為更加強健(robust)之預編碼算法。

The hybrid massive multi-input-multi-output orthogonal frequency division multiplexing(MIMO-OFDM) system is known for its high throughput and high transmission data rate, and is frequently used in the 5th generation (5G) wireless systems. However, the training overhead accompanied with the system is substantial due to the massive number of the antennas. In this paper, we consider the MIMO-OFDM system with quasi-orthogonal space-time block code (QOSTBC) transmission in the presence of mutual coupling effect (MCE) and spatial correlation effect (SCE). Applying the beamforming-selection spatial precoding (BSSP) method along with the multi-stage optimization technique to decrease the overhead and in the meanwhile ascending the performance of the bit error rate (BER). We also consider the common-used partially connected non-overlapped subarray (NOSA) structure, which is used to reduce the number of the radio frequency (RF) chains to bring down the power consumption of the entire system. Similar to the optimization criterion, we adopt the discrete grey wolf optimization (DGWO) algorithm to optimize the selected antennas for each channel. Finally, we consider the case of imperfect channel state information (CSI) and employ the traditional vector orthogonal matching pursuit (OMP) and tensor-based OMP (TOMP) algorithms to tackle the channel estimation issue.