Role of intermediate nodes in Sparse Network Coding: characterization and practical recoding

Random Linear Network Coding (RLNC) has been proved to deliver improved performance in multicast sessions and robustness against packet losses over error-prone networks. One of its most relevant characteristics is that it allows intermediate nodes to recombine incoming data packets, the so-called recoding. Nevertheless, the computational resources required to decode the information has been questioned as one potential drawback of RLNC. Tunable Sparse Network Coding (TSNC) is a promising concept, which yields a trade-off between computational complexity and network performance. Although a limited TSNC recoding approach was proposed for smart grid applications, delivering a universal, practical and efficient algorithm for TSNC recoding remains elusive. This is critical to maintain the level of sparsity of the code end-to-end, which is crucial to attain the reduction in computational complexity promised by TSNC. In this paper we start by thoroughly characterizing the status at intermediate nodes, in order to better understand the possibilities of generating recoded packets of a given density. Based on this characterization and on the reduced echelon form of the decoding matrix, we propose an algorithm that provides a good trade-off between performance and computational complexity. A simulation-based study is carried out to assess the behavior of the proposed scheme, comparing it with other alternative solutions.