Advanced Data Processing for Communication-constrained Underwater Domain

Abstract

In many practical underwater sensor networks covering a large area, only a small portion of the data may be collected in a certain time interval due to limited capacity of acoustic communications underwater. This posed a great challenge for situation awareness applications where the data for the entire area is needed. In this paper, a joint sensor selection and data recovery scheme is proposed to address the challenge. Specifically, a small portion of sensors are selected using the independent thinning principle from Stochastic Geometry to transmit their data to the surface station. Then Total Variation Inpainting is applied to recover the entire data of the whole area from the available data. The proposed scheme offers a promising solution in situations where only small amount of data could be captured due to the difficult underwater environment. It also saves time because computing at the surface station for data reconstruction takes much less time comparing to data collection. The simulation results using both synthetic data and real data demonstrate the effectiveness of the proposed method.