Lightweight, Fluctuation Insensitive Multi-Parameter Fusion Link Quality Estimation for Wireless Sensor Networks

Abstract

Accurate and agile link quality estimation is essential for wireless sensor networks. Using the mapping models between physical layer parameters and packet reception ratio, link quality can be estimated with advantages of high agility and low overhead. However, existing estimators based on physical layer parameters fail to utilize link quality information carried by different physical layer parameters efficiently and effectively and fail to effectively solve the problem that physical layer parameters fluctuate greatly, which makes them difficult to describe link conditions really. In this study, a lightweight, fluctuation insensitive multi-parameter fusion link quality estimator is proposed. Two physical layer parameters, Signal-to-Noise Ratio and Link Quality Indicator are preprocessed by exponential weighted Kalman filtering to get more stable estimation values. Then, these two parameters are fused using lightweight weighted Euclidean distance to fully utilize link quality information carried by them. On this basis, link quality is estimated quantitatively with the mapping model of the fused parameter and packet reception ratio, which is constructed by logistic regression. Experimental results show that the proposed estimator could reflect link quality more realistically. Compared with similar estimators, estimate error of the proposed one is reduced by 18.32% to 60.11% under moderate and bad links with large fluctuations, by 1.42% to 83.43% under sudden changed links, and by 16.64% to 65.61% under a long-time link. More importantly, computation overhead of the proposed estimator is equivalent to that of single-parameter estimators, but much less than other multi-parameter fusion estimators. Compared with the later, computation overhead is reduced by 72.36% to 95.61%.

Link quality estimation based on over-sampling and weighted random forest

“…Literature [17] proposed a lightweight, fluctuation insensitive multi-parameter fusion link quality estimator, which have characteristics of high flexibility and low overhead. Signal-to-Noise Ratio and Link quality indicator are preprocessed by exponential weighted Kalman filtering.…”

Light Gradient Boosting Machine-Based Link Quality Prediction for Wireless Sensor Networks

“…Effectiveness of LightGBM-LQE. In order to verify the effectiveness of our estimation model, we compare the performance of LightGBM-LQE with the link quality estimation model based on wavelet neural network (WNN-LQE) [30], the model based on support vector classification (SVC-LQE) [31], and the model based on lightweight fluctuation (LFI-LQE) [32] and take accuracy and recall as the evaluation index. The comparison results are shown in Figures 7 and 8.…”

A Link Quality Estimation Method for Wireless Sensor Networks Based on Deep Forest

“…Sun [12] uses a wavelet neural network to estimate the link quality, and it provide a guarantee for the development of the WSN routing protocol by analyzing whether link quality meet the communications standard. Liu [13] uses the lightweight weighted Euclidean distance to fuse SNR and LQI. Then, the link quality estimation model is constructed by logistic regression to reflect the link quality.…”

“…To further verify the estimation ability of the link quality estimation model SCForest-LQE, we conduct more comparison experiments, and the results are shown in Figures.11-13. The gcForest-based model (gcForest), the random forest-based model [19] (RFC), the wavelet neural network-based model [13] (WNN-LQE), the naive Bayesbased model [11] (NB), the stacked autoencoder-based model [20] (LQE-SAE) and the lightweight, fluctuation insensitive multi-parameter fusion-based model [14] (LFI-LQE) are chosen to compare with the proposed estimator.…”