A Time Series Analysis-based Link Quality Prediction Algorithm and Its Application to Reliable Routing

It has been proved that many popular link quality prediction algorithms perform much more poorly when they are applied to real environments. To address this problem, a low-complexity, on-line link quality prediction mechanism is constructed which can be applied to reliable routing protocol design. The algorithm does not require the use of any deterministic propagation model or mobility model; it relies simply on Received Signal Strength (RSS) to estimate link quality. The nodes monitor and store the links signal strength values to their communicating neighbors in order to obtain a time series of RSS measurements. Local linear kernel smoothing method and moving window local polynomial prediction method in time series analysis are introduced to construct an efficient link quality prediction mechanism, which provides significant cross-layer information for the design of mobility-adaptive proactive route repair scheme. Simulation results indicate that the proposed algorithm can provide high accuracy of link quality prediction, resulting in improved route stability and network performance.