Accurate WiFi packet delivery rate estimation and applications
Muhammad Owais Khan, Lili Qiu · 2016
The signal-to-noise ratio (SNR) is the gold standard metric for capturing wireless link quality, but offers limited predictability. Recent work shows that frequency diversity causes limited predictability in SNR, and proposes effective SNR. Owing to its significant improvement over SNR, effective SNR has become a widely adopted metric for measuring wireless channel quality and served as the basis for many recent rate adaptation schemes. In this paper, we first conduct trace driven evaluation, and find that the accuracy of effective SNR is still inadequate due to frequency diversity and bursty errors. While common wisdom says that interleaving should remove the bursty errors, bursty errors still persist under the WiFi interleaver. Therefore, we develop two complementary methods for computing frame delivery rate to capture the bursty errors under the WiFi interleaver. We then design a new interleaver to reduce the burstiness of errors, and improve the frame delivery rate. We further design a rate adaptation scheme based on our delivery rate estimation. It can support both WiFi and our interleaver. Using extensive evaluation, we show our delivery rate estimation is accurate and significantly out-performs effective SNR; our interleaver improves the delivery rate over the WiFi interleaver; and our rate adaptation improves both throughput and energy.