在混合自動重傳機制中設計 位元交錯編碼調變及位元交錯編碼調變迭代解碼之 適應性多重傳輸位元映射規則

Abstract

在文獻中已經指出，相較於使用同一位元映射規則於同一封包的多次傳輸， 改變多次傳輸的位元映射規則可以提供更高的吞吐量。根據位元交錯編碼調變的 通道容量分析，重傳時的位元映射規則必須隨著所操作的通道訊雜比的改變而改變。因此我們提出了使用基因演算法找出對每個訊雜比而言最好的位元映射規則。而當位元交錯編碼調變迭代解碼用在混合自動重傳機制時，由外來資訊轉換圖的分析也可得知，對同一編碼方式而言，每個不同的訊雜比也應該有不同的位元映射規則。因此我們在無限長度封包與無限迭代次數的假設下，針對每個訊雜比設計出最佳的位元映射規則。而在實際有限長度封包的應用下，我們畫出了解映射器與解碼器轉換圖相對於平均值的變動，同時建議在解映射器與解碼器的轉換圖之間保留一些空間來設計有限長度封包的位元映射規則。

It has been shown in the literature that varying bit-to-symbol mappings for multiple transmissions of the same packet provides better throughput performance compared to using the same bit-to-symbol mapping throughout all the transmissions. The analysis of BICM capacity suggests that the retransmission mapping scheme should be adaptive to the operating SNR. Hence we propose using a genetic algorithm to find the optimal mapping for each SNR. When BICM with iterative decoding (BICM-ID) is applied in HARQ systems, the analysis of EXIT-chart also suggests that the suitable mapping for each SNR for the same code should be different. Therefore, we find optimal mapping for each SNR under the assumption of infinite block length and unlimited iteration number. In the real application of finite block length, we plot the variation of demapper and decoder transfer curve relative to the averaged one and suggest that margins between the demapper and the decoder transfer curve should be preserved for the finite block length design.