基于自适应压缩感知的分簇式网络数据收集

Clustered Network Data Collection Based on Adaptive Compressive Sensing

分簇式传感器网络中基于混合压缩感知的数据收集方法可以有效减少数据传输量并均衡网络负载.但是,固定采样率因不能考虑信号的稀疏度在时间和空间上的变化,导致低采样率难以保证信号的重构质量,而高采样率又会造成资源浪费.针对此问题,基于数据线性程度分析提出了一种采样率自适应调整的分簇式网络数据收集方法.首先,Sink节点以簇为单位分析当前采样时刻与上一采样时刻重构数据的线性程度,以掌握数据变化趋势;然后,根据分析结果计算网络在下一采样时刻所需采样率及簇头节点所需观测值维数;最后利用数据传输树实现对簇头节点观测数目的自适应调节.仿真结果表明,与基于固定采样率的网络数据收集方法相比,该方法能够在增加少量反馈观测值维数所需的数据传输量的情况下有效提高压缩数据的重构精度.

The data collection method based on hybrid compressive sensing(CS) in the clustered sensor networks can reduce the number of data transmissions and balance the traffic load throughout networks.However,this method uses a fixed sampling rate which could not take the variation of natural signal sparsity in time and space into account.Consequently,low sampling rate can hardly guarantee the quality of signal reconstruction,while high sampling rate will cause the waste of resources.To solve this problem,a clustered network data collection method based on the data linearity analysis is proposed,and the sampling rate was adaptively adjusted.First,the Sink node analyzed the linearity of the reconstructed data between the current sampling time and the previous sampling time of each clusters,and grasped the trend of data changed.Then,according to the analysis result,the sampling rate required by the network at the next sampling time and the required observation dimension of the cluster head nodes are calculated.Finally,the number of adaptive observations of the cluster head node was transferred by data transmission tree.The experimental results show that the proposed adaptive sampling method can effectively improve the reconstruction accuracy with a small amount of data required to feed back the observation dimension than the CS data collection method based on fixed sampling rate.