Fountain Codes and Their Applications: Comparison and Implementation for Wireless Applications

Abstract

Fountain codes are rateless erasure codes that have initially been designed for multicast and streaming applications. They provide dynamic code rates which are defined by the channel conditions. Subsequently, fountain codes have also demonstrated their benefits for transmissions over wireless channels. With fountain codes, feedback channel deployment can be reduced which could reduce collisions in half-duplex systems. Contrary to previous research on fountain codes that focused on defining the theoretical bound of code rate and complexity, in this paper, we would like to put them into practice by extending their performance evaluation in two other aspects. First, we would like to investigate the implementation process of fountain codes in a real-world deployment including header analysis, packet layout investigation as well as implementation cost. Next, contrary to previous works that aim at defining code rate after decoding completion, in this work, we would like to focus on the partial decoding capabilities of fountain codes. This metric could be used for performance evolution for transmissions over degraded channel conditions found in the wireless applications. We discussed and compared in this paper several types of fountain codes including LDPC, LT, Raptor, etc.