Learning-BEB: Avoiding Collisions in WLAN
Jaume Barcel√≥, Boris Bellalta, Cristina Cano, Miquel Oliver · 2008
Abstract —Random access protocols have been the mechanismof choice for most WLANs, thanks to their simplicity anddistributed nature. Nevertheless, these advantages come at theprice of sub-optimal channel utilization because of empty slotsand collisions. In previous random access protocols, the stationstransmit on the channel without any clue of other stations’intentions to transmit. In this article we provide a framework tostudy the efficiency of channel access protocols. This frameworkis used to analyze the efficiency of the Binary Exponential Backoffmechanism and the maximum achievable efficiency that can beobtained from any completely random access protocol. Then wepropose Learning-BEB (L-BEB).L-BEB is exactly the same as legacy BEB, with one exception:L-BEB chooses a deterministic backoff value after a successfultransmission. We call this value the virtual frame size ( V ). Thissubtle modification significantly reduces the number of collisions.It can be observed that, as the system runs, the number of colli-sions is progressively reduced. Thus we conclude that the systemlearns. Further, if the number of contending stations is equal orlower than