A Data Collection Scheme Using Transformer Autoencoder in Wireless Sensor Networks

Data collection is one of the key operations in wireless sensor networks (WSNs). An efficient data collection scheme is designed using a transformer autoencoder for WSNs. During the model training stage, the model is trained using historical monitoring data to obtain an optimized measurement matrix and reconstruction matrix. In the data collection stage, sensor nodes compress the data using the measurement matrix and enable efficient data transmission via a clustered network. Finally, the sink node utilizes the reconstruction matrix to restore the original data. Simulation and verification analysis show that the proposed data reconstruction algorithm outperforms existing algorithms such as OMP, DAE, SSDAE-CS, and ISTA-1DNet in terms of reconstruction accuracy and reconstruction efficiency at a compression ratio of 30%.