Network coding for bit error recovery in IEEE 802.11 mesh networks

  • Authors:
  • Mathias Kurth;Ulf Hermann;Anatolij Zubow;Jens-Peter Redlich

  • Affiliations:
  • Humboldt University, Berlin, Germany;Humboldt University, Berlin, Germany;Humboldt University, Berlin, Germany;Humboldt University, Berlin, Germany

  • Venue:
  • ICC'09 Proceedings of the 2009 IEEE international conference on Communications
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

Opportunistic routing (OR) relies on links of intermediate quality, i.e. packet losses are common. However, the reasons for packet losses are manifold, e.g. a received packet may contain corrupted bits. According to traditional approaches, the receiver discards the whole frame in such a case. In this paper, we present measurements from an indoor IEEE 802.11 wireless mesh network (WMN), which indicate that corrupted frames still contain a significant amount of correct data, which can be utilized. In particular, corrupted frames are common for intermediate quality links. Bit errors tend to occur in proximity, i.e. they are bursty. Furthermore, bit errors are uncorrelated across different receivers in most cases. Based on our observations, we propose a HARQ scheme for OR called Hybrid ARQ with Limited Fragmentation (HALF). It operates on a hop-by-hop manner and requires only local knowledge. Due to the bursty nature of bit errors, we are dividing frames into fragments with additional error detection. Using random linear network codes, the sender transmits incremental redundancy until one of its receivers is able to decode all fragments and therefore sends an acknowledgement packet. However, the partial information at all other receivers is not lost. Instead, to increase the throughput further, it is also used in subsequent forwarding rounds along the multi-hop route. We implemented a prototype of our protocol to evaluate its performance. With the help of detailed simulations, we analyzed the reasons why HALF significantly outperforms traditional approaches like DSR.