基于空间结构的无线传感器网络缺失值估计方法

Spatial Structure Based Missing Value Estimation for Wireless Sensor Networks

为了保证无线传感器网络中数据的完整性,针对感知数据在传输中存在数据丢失或者数据异常的问题,提出一种基于空间结构的无线传感器网络缺失值估计方法。矩阵补全方法是估计缺失值的有效方法,但目前尚未考虑到数据间的连续性而导致估计误差较大。本文结合传感器节点数据具有时空相关性的特点,通过增加正则化项的方式来约束矩阵补全的解空间对数据进行缺失值估计。仿真时采用伯克利英特尔实验室的传感器数据,通过Matlab软件对模型进行测试并分析仿真结果。实验结果表明:该算法对连续多个数据的缺失值估计效果理想,估计误差始终保持在较低水平。

To address the problems of data loss or data anomalies in the transmission of sensed data,the missing value estimation algorithm based on spatial structure was proposed to guarantee the integrity of data in wireless sensor networks.The matrix complement method is an effective method for estimating missing values.However,the continuity of data has not yet taken into account the large estimation error.This paper combines the characteristics of spatio-temporal correlation of sensor node data and constrains the matrix complementation solution space to estimate missing data by adding regularization terms.The Berkeley Intel lab’s sensor data streams were adopted in simulation as empirical data,the simulation results were estimated with model by Matlab software.Simulation result proved that algorithm achieved ideal effect in estimating consecutive data,and estimation error stayed at a lower level.