A fountain code-based fair scheme for multi-hop ad hoc networks

We propose an analytical cross-layer model to quantify tradeoffs between throughput and fairness. We study the effect of coding on the performance of the network while optimizing parameters that govern NETWORK, MAC and PHY layers. Indeed, the heart of our scheme is to code packets at MAC layer before transmitting them through the common lossy channel. To the best of our knowledge, we are the first to use Fountain codes at MAC layer. Indeed, coding packets is usually studied at APPLICATION or TRANSPORT layers. In standard IEEE 802.11, whenever a collision occurs the whole packet is lost and needs to be retransmitted. Our scheme uses an incremental combination of all previous copies of a backlogged packet to improve the decoding probability and therefore to correctly recover the original packet. Using Fountain coding increases the Jain´s Fairness index over the network. It indeed improves the throughput on paths that suffer from high collision probability. However unfortunately, it may decrease the throughput on the other paths. This is due to the injected redundancy in coded packets and its negative effect on reducing the channel idle period seen by concurrent users.