802.11G WLAN modeling for real time packet communication

This work proposes a statistical model of packet arrival and loss for wireless indoor environments in infrastructure mode. Usually, a Markov model with two states (Gilbert-Elliott, 1965) is used to model random processes of erroneous bits or packets. If we consider an equivalent model without memory, i.e. a random variable representing the burst length, then the Gilbert-Elliott model has a geometric distribution. We propose a statistical model for burst lengths of received and lost real time packets based on the logarithmic series distribution. We make use of a goodness-of-fit chi-square test to prove that our model surpasses the Gilbert-Elliott for 802.11g channels. Despite our model has been tested in a particular wireless scenario, it could also model other communications systems.