Estimating heterogeneous traffics in CSMA networks at runtime

The efficient utilization of CSMA/CA network has been shown to heavily depend on the knowledge on the traffic loads of other nodes in the network. In a purely distributed network, the on-going traffic loads can only be estimated by exploiting locally available information. However, existing estimation algorithms are mainly focused on traffic estimation of homogeneous networks and when extending to heterogeneous networks the existing techniques are not applicable. In this paper, a novel heterogeneous Markov chain model is proposed to relate the two types of local information, namely clear channel assessment (CCA) result and acknowledgment (ACK) packet, to two variables characterizing the on-going traffic loads, namely the aggregate transmission rates and aggregate traffic loads. An improved particle filtering technique named Autoregressive Moving Average Particle Filter (ARMA-PF) is used to perform traffic estimation. It has been shown that the proposed method is accurate in characterizing the traffic loads and outperforms the existing methods.