Adaptive sampling of thermoclines with Autonomous Underwater Vehicles

Abstract

Autonomous Underwater Vehicles (AUVs) are routinely being used to provide the scientific community with detailed ocean data at very reasonable costs. In typical operations, AUVs are programmed to follow pre-defined geo-referenced trajectories, while collecting the relevant information about the underwater environment, with a clear separation between navigation and payload sensors. Under the adaptive sampling paradigm, the AUVs are able to interpret some of the payload data in order to change the sampling pattern and concentrate measurements in the regions of interest. In this paper, we describe an implementation of such paradigm, in which a small sized AUV is able to process CTD data, in real time, and change depth in order to maintain tracking of the thermocline region. We demonstrate the developed algorithms with data from field experiments in a dam reservoir, which show a very good performance, even in very shallow waters with hardly detectable features. The implementation ensures the safety of the AUV, by resuming to standard yo-yo patterns if the thermocline is not detected.