Cloud Model-Based Link Quality Prediction Method For Wireless Sensor Networks

Wireless sensor network is a self-organizing network formed by nodes deployed in the monitoring area, which has been widely used in many fields. Due to the complexity environment, the link quality is not reliable. The obtaining of link quality information in advance which provides reference for the upper routing protocol to select link can greatly reduce the number of data retransmission so that we can reduce energy consumption. Establishing a comprehensive and accurate link quality prediction mechanism is very critical to improve the reliability of the whole network communication and to prolong network life. This thesis is supported by the National Natural Science Fund. It focuses on link quality prediction method for WSN, aiming to establish a link quality prediction method with high accuracy.This thesis analyzes link characteristics, including link irregularity, asymmetry, and volatility. Analyzing the status of the art on link quality prediction method, we classify these methods. This thesis analyzes deeply the advantages and disadvantages that the physical layer parameters and link layer parameters are used as the prediction model parameters, and analysis the correlation between parameters. A novel model, Cloud Model, is proposed to predict link quality. Adaptive gauss cloud transformation is applied to cluster the link quality parameter. An apriori algorithm is applied to mine the association rules from the physical layer parameters and link layer parameters which has been clustered. Then cloud reasoning method is proposed to predict link quality. And considering the short-term and long-term in the two time dimensions, we propose the link quality prediction method based on time-serial prediction with cloud models.Considering link quality easy influenced by environment, we collect the link quality samples from multiple pairs of nodes at different scenarios, including Indoor corridor,wood and road.Compared with BP neural network prediction method in two cases of stable and unstable link quality, the cloud reasoning method achieves higher accuracy.Then compared with the method with Window Mean Exponentially Weighted,time-serial with cloud models method can predict stability meantime are more effcient in their percetion of link change as experiment results show.