Imissing Data Estimation Methods In Wireless Sensor Networks

Wireless sensor network consists of sensor nodes,aggregation nodes and management nodes.These nodes have limited energy and communication capability generally.Sensor nodes sense the data of the detected objects and communicate with other nodes within its communication range.The sensed data will be sent to the sink node consequently.Data loss often occurs because of node failure which presents a challenge for sensor data processing.Data evaluation is estimating the missing data based on the characteristics of the data which can effectively solve the problem of the data loss.Data estimation method relies on the properties of data object.Compared with general data,sensor data has many inherent features such as data uniqueness,discrete relevance,temporal and spatial correlation.There have been many research results for the missing data estimation.However,there are some limitations and deficiencies when these schemes are applied in wireless sensor networks directly.The reason is that they do not consider the characteristics of the sensor data.If they are used directly to estimate the missing sensor data,complexity is high and accuracy is low.Therefore,we should consider fully the characteristics of the sensor data and develop the efficient assessment algorithms.Based on the structural characteristics of the network,the wireless sensor network is divided into single and double networks.We will study missing data evaluation algorithms for these two network structures.As for wireless sensor networks in a single layer,we consider the relevance of the sensor data in time and space.We propose space and time correlation algorithm.This algorithm estimates the missing data by the two dimensions of time and space.Firstly,it selects data collection which has higher temporal and spatial correlation as the basis of sample collection.Secondly,the spatial evaluation value is determined by the space correlation analysis.Finally,temporal evaluation is determined by distinguishing corresponding data weights at the time order.Spatial and temporal estimations make the final assessment of the missing data.Considering the difference between 2-tiered wireless sensor network and the traditional networks,we propose the relay node-centric estimation algorithm.Firstly,frequent items are mind according to the dependence of the sensed data in each cluster.Estimation is determined by these mining laws.Secondly,according to the similarity of the association rules,associated sensor nodes are selected.Different weight is assigned to the associated nodes to obtain the evaluation value of the missing data.Finally,integrate cluster and inter-cluster assessment values to get the estimation results.Then a complete set of sensor data is obtained.Through simulation experiments compared with classical similar algorithms,we prove that the two assessment algorithms have high evaluation accuracy for different types of data in different data loss probabilities.