Volatility-Aware Channel Sensing with Commodity 802.11 Hardware

Information about channel quality is essential for channel selection and channel hopping in wireless ad-hoc networks. In this paper, we present a sophisticated channel quality metric, its implementation on commodity 802.11 hardware, and results of real experiments in our testbed. The metric has the following properties. First, collection of channel busy times as raw data exploits customary hardware mechanisms, yielding accurate values. Second, the accuracy of raw data is further improved by determining and subtracting the effect of internal traffic of overlapping channels. Third, corrected raw data are aggregated, making the metric more stable. Fourth, an exponential weighted moving average is applied, providing for adaptivity. Fifth and finally, several candidates to measure the volatility of raw data in a moving window regarding downward fluctuations are defined and applied, yielding channel quality metrics that are more conservative without being overly pessimistic. We have implemented each of these steps and present experimental results.