AdaBoost-based Link Quality Prediction Mechanism For Wireless Sensor Networks

Wireless sensor networks is a network composed of a large number of inexpensive miniature sensors.It is centered on transmission data,and networked through selforganizing and multi-hop and its topology is dynamic.The sensor nodes in the network communicate through electromagnetic waves and the battery capacity of the power supply is limited,which is prone to influence by multipath propagation,environment noise and signal interference,resulting in link instability.Therefore,if high-quality link could be chose through link quality prediction,the successful data transmission rate will be improved and the network lifetime will be extended.This subject comes from the National Natural Science Foundation of China and studies the method of link quality prediction in wireless sensor networks.The thesis analyzed the characteristics of wireless sensor network links,introduced the existing link quality prediction methods at home and abroad,and then proposed an Ada Boost-based(Adaptive Boosting)link quality estimation model.The model used Density-Based Spatial Clustering of Noise(DBSCAN)to cluster the physical layer parameters and determined the link quality level according to the packet reception rate in the cluster.And then adopted support vector classifier as Ada Boost individual learner to construct mapping relationship between physical layer parameters and link quality levels.In the light of the link quality characteristics on time-serial,support vector regression is used as individual regression of Ada Boost regression.Taking the link quality level of the historical time slice as the input and the link quality level of the next time slice as the output,a link quality prediction model based on Ada Boost is established to predict the link quality level at the next moment.Five campus scenes including campus squares,corridors,campus groves,campusparking lots and student laboratories are selected to experiment and collect link qualitydata.The collected data of link quality under each scene is denoised and standardized.The simulation results show that the Ada Boost model with support vector classifier ismore accurate than that of the Ada Boost model which adopts the decision tree as theindividual learner.Compared with the Dynamic Bayesian-based link quality predictionmodel,the Ada Boost-based link quality prediction model can predict the link qualitymore accurately.