Reliable, Low Overhead Link Quality Estimation for 802.11 Wireless Mesh Networks

We propose QUEST (QUality ESTimation), a new method that accurately estimates IEEE 802.11 wireless link quality with no in-band signaling overhead. Existing link quality estimation methods either are based on hello exchanges by fixing or varying transmission rates or rely on the history (e.g., delivery ratio) of previously sent data packets in a per-rate/-neighbor manner. QUEST on the other hand, is based on a delivery ratio vs. SNR (Signal to Noise Ratio) relation, called profile, that is managed offline. QUEST estimates the target link quality in terms of delivery ratio by performing profile lookup for any incoming messages including broadcast hello, beacon, data packets, etc. Therefore, it does not depend on a designated protocol to obtain the delivery ratio. Instead, in QUEST, the per-rate/-neighbor management of link quality is achieved by profile lookup. We perform testbed experiments to achieve the profile and also unravel two major bugs in MadWifi driver, widely employed by many researchers to build an 802.11-based system. Utilizing the large database of transmitter and receiver traces with an indigenously developed tool, we study the impact of altering the averaging time period on the profile for different transmission rates.