Modeling and improving CSMA unfairness in Mobile Ad Hoc Networks

In this paper we study and model the unfairness in Carrier Sense Multiple Access (CSMA) Mobile Ad hoc Networks. We use a Markov model to represent the functioning of our network. The Markov states correspond to the set of nodes which transmit simultaneously. This model allows the steady state to be simply computed when the network is overloaded. When the CSMA back-off is small compared with the duration of a packet, we show that the most probable states are those containing the maximum number of stations, which corresponds to a maximum number of simultaneously transmitting nodes. In other words, the nodes in the maximum sets of the graph inferred by the CSMA rule are favored. We study this model with several examples of ad hoc networks. We compare the predictions of this model with simulation results and observe that the matching is good. We show how this model can be extended to situations where there are hidden nodes. We also show that our model can be used to adjust the back-off windows of the network nodes in order to enhance fairness.