Deep Belief Networks-based Link Quality Prediction Mechanism For Wireless Sensor Networks

Wireless sensor network is a self-organizing multi-hop network,which including a large number of sensor nodes.These sensor nodes are limited in energy,communications,computing and storage.And the nodes are usually deployed the complexity and various interference sources environment which can lead to the quality of communication links between nodes are not reliable.On the one hand,link quality prediction mechanism is the basis of topology control,routing protocol and mobile management,on the other hand is for the improvement of network data throughput,prolong network lifetime and reduce energy consumption and so on all play a crucial role.The research is aiming at research on wireless sensor networks link quality prediction.This thesis analyzes link characteristics,the status of the art on link quality prediction method,the prediction model parameters,and deep learning.According to the characteristics of link quality and combining with the advantage of the deep learning in prediction.We propose a deep belief networks(DBN)-based link quality prediction model.This model is consist of evaluate level and predict level.The model using support vector machine(SVM)to evaluate the link quality first and get the link quality level;According to the history of continuous link quality level combine the DBN regression theory to prediction the next unit time of link quality;At the end of the model,we added a predictor which names “Softmax”.According to the mean square error performance evaluation index to determine the DBN model structure.According to the practical application of WSNs,we collect the link quality samples from nodes in different scenarios,including indoor corridor,grove and road.The experiment results show that in this paper,deep belief networks-based link quality prediction model whether compared with prediction model based on BP neural network,and compared with prediction model based on Logistic,has better prediction accuracy in various scenarios.