“Toward Digital-Twin-Assisted Data Acquisition in Underwater Acoustic Sensor Networks”

By Ivan Spajić, Lars Michael Kristensen, Volker Stolz

Ivan’s paper at ANT 2025 got accepted!

He will be presenting our work not only on underwater sensors, mind you, shortly. Grab our open access preprint here or at our institutional repo!

Abstract:

Data transfer from underwater acoustic sensor networks (UASNs) to cloud services is challenging due to the severe bandwidth constraints of acoustic communication and energy consumption constraints of battery-powered sensor devices. To mitigate this problem, various forms of time series compression algorithms, both lossless and lossy, have been proposed in the literature. These proposals, however, focus on large time series whose in-memory buffering is not feasible in a UASN context. Furthermore, especially in oceanographic data acquisition, buffering large time series sacrifices the Age of Information. To address this gap, we evaluate a modern lossy compression algorithm, Mix-Piece, on small-sized time series. Motivated by our findings, we have developed the Custom-Piece algorithm, an adaptation of Mix-Piece that yields higher data compression at the expense of increased execution times. Our experimental evaluation of the Mix-Piece/Custom-Piece algorithms on real-world oceanographic data shows that compression ratios are highly dependent on the configuration of the algorithms due to the emerging characteristics of the underlying data sets. To facilitate practical use of the Mix-Piece/Custom-Piece algorithms, we outline how a digital twin may dynamically run simulations on available past data to support end-users in developing sensor configurations according to data acquisition requirements.