In Situ Analysis for Intelligent Control

We report a pilot study on in situ analysis of backscatter data for intelligent control of a scientific instrument on an Autonomous Underwater Vehicle (AUV) carried out at the Monterey Bay Aquarium Research Institute (MBARI). The objective of the study is to investigate techniques which use machine intelligence to enable event-response scenarios. Specifically we analyse a set of techniques for automated sample acquisition in the water-column using an electro-mechanical "Gulper", designed at MBARI. This is a syringe-like sampling device, carried onboard an AUV. The techniques we use in this study are clustering algorithms, intended to identify the important distinguishing characteristics of bodies of points within a data sample. We demonstrate that the complementary features of two clustering approaches can offer robust identification of interesting features in the water-column, which, in turn, can support automatic event-response control in the use of the Gulper.