Exploiting Spatiotemporal Correlation for Wireless Networks under Interference

This paper starts from an empirical observation on the existence of spatiotemporal correlation among nearby wireless links within short time intervals. The phenomenon due to correlated interference has become pervasive with densely deployed wireless devices, causing potential errors in existing popular metrics built upon the assumption of link independence. To this end, we propose correlated ETX (cETX), which generalizes the widely-adopted ETX to maintain the accuracy under correlated links. To the best of our knowledge, this is the first work to introduce a unified metric embracing both temporal and spatiotemporal correlations. As a generalized metric, the highlight of our work is the broad applicability and effectiveness-extensive evaluations on ZigBee (802.15.4) and Wi-Fi (802.11b/g/n) testbeds deployed in a lab, corridor, and on a bridge reveal that: simply replacing ETX with cETX: 1) cuts down the error by 62.1%-70.2% and 2) saves averages of 22% and 37% communication cost in three unicast and nine broadcast protocols, respectively, under only 0.7% additional overhead.