On Modeling of 802.11b Residue Errors

摘要

Errors that are observed at the MAC layer of arnwireless protocol are known as residue errors due to the failure ofrnthe physical layer to correct them. In this paper, based onrnexperimentation with actual 802.11b Wireless LANs (WLANs), itrnis shown that the 802.11b residue error process is a highly non-stationaryrnprocess, and hence error models derived on the basisrnof observations at one time/environment may not be applicablernat/in another. Thus MAC-layer modeling/prediction techniquesrnneed to adapt on the basis of some Channel State Informationrn(CSI). Consequently, in this work, we characterize the variationrnin the behavior of the residue error process in accordance to thernradio-link Signal to Silence Ratio (SSR) Indications (SSRI). Thernresidue error process is treated as a piecewise stationary process,rnwhere each piece corresponds to a PHY/MAC packet. Inrnparticular, we construct a unified Markov framework to obtainrnSSR aware Markov models for residue errors. Further, it isrnshown that the proposed SSR aware Markov frameworkrnprovides significant performance benefits and can characterizernthe residue error traces even across varied WLAN setups. Thisrnleads to the design of non-homogenous Markov models that canrnvary the transition matrix on the basis of using SSR indication asrnside-information. We deploy such a non-homogenous Markovrnmodel and show that it can characterize the error processrnsignificantly better than Markov models that are derived byrnignoring SSR values.