Energy-Efficient Approximate Data Collection and BP-Based Reconstruction in UWSNs

Abstract

Approximate data collection that has been extensively studied in Terrestrial Wireless Sensor Networks (TWSNs) can be leveraged in underwater scenarios. However, it is challenging to balance between energy cost and data quality because timely quality feedback strategies applicable in TWSNs may not be suitable for underwater scenarios constrained by long acoustic delay. Faced with high-frequency packet failure, we first formulate the problem of selecting the minimum sensing node set into a minimum m-dominating set problem that is known to be NP-hard, and then propose a heuristic Center-based Active Sensor Selection (CASS) algorithm for approximate data collection with the consideration of node correlation and residual energy. With the computing ability of the cloud, Belief Propagation (BP) is utilized to infer missing data. Evaluation based on real-world datasets shows our proposed approximate collection strategy can reduce 60% more energy cost with little accuracy loss.