Data Reconstruction in Wireless Sensor Networks From Incomplete and Erroneous Observations

Many basic scientific works use wireless sensor networks (WSNs) to collect environmental data and use the observations for scientific research. The completeness and accuracy of the collected environmental observations determine the reliability of the research results. However, due to the inherent characteristics of WSNs, data loss, and data error usually occur during the process of data collection. Therefore, it is necessary to design an effective method to reconstruct the environmental data from the incomplete and erroneous observations. In this paper, we propose a novel data reconstruction scheme via temporal stability guided matrix completion. First, based on the low-rank feature of sensory environmental data, we formulate the data reconstruction problem as a matrix completion with structural noise. We also introduce a constraint about short-term stability to the matrix completion problem for further reducing the reconstruction error. We then, design an algorithm based on the block coordinate descent method and the operator splitting technique to solve the problem. Finally, simulation results on real sensory data sets show that the proposed approach not only significantly outperforms existing solutions in terms of reconstruction accuracy but also can recognize the sensor nodes with erroneous sensory data.