Convolutional codes for multi-shot network coding

Abstract

In this thesis, we aim to provide a general overview of the area of multi-shot codes for network coding. We will review the approaches and results proposed so far and present slightly more general definitions of rank metric block and convolutional codes that allows a wider set of rates than the definitions of rank metric codes that exist in the literature. We also present, within this new framework, the notion of column rank distance of a rank metric convolutional code. We investigate it properties and derive an upper-bound that allows us to extend the notions of Maximum Distance Profile and Strongly-Maximum Distance Separable convolutional codes to some rank metric codes analogues. We focused on the development of channel encoders as a mechanism that allows the recovery of the data lost during the transmission. We also concentrate on the construction of novel classes of MRD convolutional codes. In particular we aim at extending the constructions presented by Napp, Pinto, Rosenthal and Vettori, in order to increase the degree of the code and consequently it error correction capability. As alternative to rank metric convolutional codes, we present a novel scheme by concatenation of a Hamming metric convolutional code (as outer code) and a rank metric block code (as a inner code). The proposed concatenated code is defined over the base finite field instead of over several extension finite fields and pretend to reduce the complexity of encoding and decoding process and moreover use the more general definition of rank metric code in order to be more natural.