Erasure Floor Analysis of Distributed LT Codes

Hussain, İqbal; Xiao, Ming; Rasmussen, Lars K.

doi:10.1109/tcomm.2015.2451640

Cited by 8 publications

(1 citation statement)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…And to maximize the minimum-distance of variable-node degree, [13] proposed the Rescheduled distributed LT Codes, which can abate the error floor remarkably. [14] proposed a Hybrid Soliton Distribution Coding which engage the Robust Soliton Distribution (RSD) and the deconvolved Robust Soliton Distribution (DSD) [6], but it can only deal with the 2n-sources networks.…”

mentioning

confidence: 99%

Distributed LT Codes on Multi-hop Networks

Zhang¹,

Chen²,

Tang³

et al. 2017

Proceedings of the 2nd International Conference on Computer Engineering, Information Science &Amp; Application Technology (ICCI

View full text Add to dashboard Cite

Abstract. Distributed LT codes on binary erasure channels (BECs) have been widely investigated, but the number of sources was usually limited within ten. In the practical cases we often need to transmit information blocks from scores of sources to a common destination through one or more relays, which encourages us to reclaim the scenarios in which up to one hundred sources transmit information blocks to the destination. Obviously, in these scenarios, sources need to be re-organized to improve the system performance. Our propositions are Multi Distributed LT codes (MDLT) and Backtracking Distributed LT codes (BDLT). We investigate the performances of these codes by density evolution, and present the simulating results, compared with those Conventional Distributed codes (CDLT). IntroductionFountain codes [1] [2][3] are a novel forward error correction scheme for information transmission over Binary Erasure Channels (BEC). Unlike traditional ARQ schemes, fountain codes are able to adapt their rates on-the-fly, so they are also called rateless codes. In practical circumstances such as wireless networks, when the characteristics of the channels are unknown or time-varying, fountain codes can obtain much better performance than ARQ schemes because they need not repeat and request every erased packet. LT codes [2] are the first realization of fountain codes. Discarding Gaussian elimination decoding whose computational cost is O(k3) (where k is the number of packets in an original source block), LT codes utilize Belief Propagation decoding which can cut the computational cost down to O(k logk). But LT codes suffer from a relatively high error floor. To eliminate this problem Shokrollahi proposed a compound coding structure named Raptor codes [3], usually including a high-rate outer LDPC code and an inner LT code. Raptor codes could not only abate the error floor, but also reduced the computational cost to O(k).LT codes and Raptor codes are all suit well for point-to-point transmissions, but in the multicasting networks, we need multi sources transmitting to one destination, i.e. distributed LT codes. Distributed LT codes is the engagement of LT codes and Network Coding [4] [5]. Since the structure of Network Coding would destroy the degree distribution of LT codes, people must take some measures to overcome it. [6] firstly explore distributed LT codes by deconvolving the degree distribution of LT codes, named Robust Soliton Distribution (RSD), in order to obtain an approximate RSD at the destination. Considering the complexity of deconvolution, this approach supports only 2-sources networks and 4-sources networks. And-Or Tree provides a unifying, intuitive, and powerful framework for carrying out the analysis of several random including random loss-resilient codes, random k-SAT formulae using the pure literal rule, the greedy algorithm for matchings in random graphs, etc [7], thereafter Sejdinovic utilize it to analysis the asymptotic performance of Distributed LT codes [8]. And-Or Tree analysis provides a pre-view of ...

show abstract

mentioning

confidence: 99%