State-of-the-art analysis and protocols in wireless mesh networks typically assume an independent packet loss channel for each receiver of a transmission. Although this is usually transparent for single-path protocol design, this assumption may severely degrade the performance of opportunistic and/or multi-path routing approaches as well as network coding (NC) subgraph selection problems (routing in NC). This paper proposes simple channel models to incorporate the effect of correlation between receivers in a parametric fashion and supports them with a measurement campaign that leverages various commercial devices and network conditions. Finally, we illustrate the effect of correlation in the design and performance of the PlayNCool network coding protocol introduced by the authors in prior work. Our results show that the modified PlayNCool protocol decreases the overhead by a factor of 2.5x.
展开▼
