On Channel State Inference and Prediction Using Observable Variables in 802.11b Networks

Performance of cross-layer protocols that recommend the relay of corrupted packets to higher layers can be improved significantly by accurately inferring/predicting the bit error rate (BER) in the packets. Here, inference refers to estimating the BER in an already received packet, while prediction specifically refers to anticipating the BER in a future packet. This paper presents a measurement based study of 802.11b WLANs that analyzes the utility of observable variables in channel state inference (CSI) and prediction (CSP). The first part of the paper investigates the utility of SSR and background traffic (BT) intensity ρ as observable side-information for CSI. Our results show significant utility for SSR indications and the feasibility of utilizing ρ for additional improvements. The second part of the paper utilizes the proposed CSI mechanism for BER prediction (i.e., the CSP aspect). We observe that the BER of temporally adjacent packets are correlated and thus BER of a current packet, by itself or on the basis of link-specific temporal correlation model, can provide efficient prediction. However since BER of a packet is not an observable variable, the above prediction mechanism cannot be realized practically without significant processing overheads. Thus we propose to estimate the channel state for the current packet using observable sideinformation and this estimate is subsequently used as an input to a BER based predictor. Our analysis and simulations based on an extensive set of actual 802.11 traces show that the proposed methods can provide accurate CSI/CSP under a variety of realistic channel conditions