以演化計算之機制與工具對LT Codes進行改良之研究

Abstract

噴泉碼(digital fountain)是近年來被提出的一種錯誤糾正碼類別。它最重要的一項性質為無固定編碼率(ratelessness)，其特色是不需要一個固定的編碼率而編碼封包可以無限制地產生。對於任何想要接收資料的接收端，只要收集一定數量的編碼資訊即可完整解出原始資料，且不需要顧慮收到的封包順序。其相對應的好處是接收端不需要回傳訊息的傳輸通道來處理封包遺失的問題，因此同時面對多個傳輸品質不同的接收端也可以用同一套編碼程序。這個特性使噴泉碼十分適合做為網路廣播系統或是需要在異質網路環境間傳輸資料的應用。盧比變換碼(Luby Transform Codes)是第一個真正可實用的噴泉碼。它於2002年被提出，許多的現實應用便開始嘗試以噴泉碼作為解決方案，也因此對於盧比變換碼的改良研究也變成一個很熱門的議題。本論文整合兩個不同的研究領域，試圖以演化計算裡的工具與機制對盧比變換碼進行改良。文中探討了一系列的研究工作，包含盧比變換碼的最佳化與如何設計更優良的編碼機制。這些研究都針對同一個目標–期望以噴泉碼提供更穩定優良的傳輸方案。本論文的各項成果將有助於提升噴泉碼的應用品質，並推進相關領域的研究。

Digital fountain is a novel category of forward error correction proposed in the last decade. The most important characteristic of digital fountain is ratelessness which means that unlimited encoding codewords can be generated on the fly and a fixed code rate is not necessary. Any receiver interested in the message could reconstruct it while a sufficient amount of codewords were received without care of the order. No extra communication is required for handling lost packages, even though the channel status of each receiver is different. The feature is a critical reason for that digital fountain codes are good solutions for broadcasting application and transmission in heterogeneous network. Luby Transform (LT) codes are the first practical implementation of digital fountain. Since LT codes were introduced, lots of applications were proposed based on the code. Thus, to improve the performance of LT codes becomes an important issue. The thesis integrates two research fields and makes study on improving LT codes via the methods in evolutionary computation. A series of works are presented including the optimization of LT codes and the developing of advanced coding schemes. All these works aim to an identical objective that is to provide better solution for reliable communications. The results in the thesis would contribute to the applications of digital fountain codes. We also expect that the effort would be help for further achievement in the research field.