Multi-Timeslots Data Collection With Low Rank and Modified Second-Order Horizontal Total Variation for Wireless Sensor Networks

Decreasing the number of data gathered is the most highly effective way to decrease the power consumption for wireless sensor networks. Compressed Data Gathering, as it known to all, is a data collection method in wireless sensor networks, but it cannot achieve sparse sensing as all data need to be sensed and then transmitted in most practical applications. At the same time, it has been shown the effectiveness of the total variation and low rank constraints in data restoration. In order to enhance the accuracy of data recovery and decrease energy cost in wireless sensor networks, we propose a Multi-Timeslots Data Collection scheme, which includes two aspects: Structure Random Sparse Sampling method and data restoration algorithm with Low Rank and Modified Second-Order Horizontal Total Variation Constraints. By adopting the proposed sampling method, the number of data sensing and transmission is greatly reduced, thereby prolong the network lifetime. We fully exploit temporal stability and low rank characteristics of wireless sensor networks data, and build a temporal-stability based nuclear norm regularization minimization model. Meanwhile, we apply the alternating direction method to solve the problem. The simulation results present that the proposed sampling method has a corresponding enhancement effect on the matrix-completion based data restoration algorithms. In terms of recovery precision, the proposed scheme outperforms the state-of-the-art methods for different types of data in the network. Moreover, with the compression ratio increasing, the proposed scheme can still exactly recover the lost data and the advantages become increasingly obvious.