Aiming at the high complexity of regression algorithms, the lack of unified classification standards and public data sets in the link quality estimation model of wireless sensor network, this paper proposed a link quality estimation method ELR-LQE(EWMA and linear regression based link quality estimation) based on EWMA and linear regression. Taking RSSI, LQI and SNR acquired from the physical layer, as well as PRR(packet reception ratio) as input metrics, it collected data in various experimental environments with different transmission power, contending conditions and deployment methods, and established a link quality estimation data set. It preprocessed the data by minimum interpolation and EWMA, which significantly improved the correlation between the input features of the regression model and the link quality. Compared with the existing methods, ELR-LQE is more adaptable to the network layer protocols, with lower complexity, thereby more suitable for implementation in wireless sensor network nodes with limited resources. The experimental results show that ELR-LQE has a high accuracy. Under a variety of experimental conditions, the average ME is 4.6×10^-2 and R² is 0.99.

[2022-02-16] Accepted Paper /doc/2021.12.0657/v1

[2022-07-05] Printed Article /doc/2021.12.0657/v2