COS 290-1 - Autonomous detection and adaptive sampling of marine biological features with the Chiton imaging system

Abstract: Ocean dynamics constantly influence the presence, abundance, and density of marine organisms. The resulting rapidly fluctuating, spatially heterogeneous biological distributions make measurement and monitoring an on-going challenge. Finding and persistently sampling zones of elevated biological activity in the ocean is particularly vexing; these features are ephemeral and highly localized, challenging most conventional observation systems. Concurrent advances in three areas – in situ imaging systems, automated image analysis, and autonomous sampling platforms – promise to help scientists better understand how these ubiquitous structures shape community ecology. Here we describe a proof-of-concept experiment to modify the behavior of a long range Autonomous Underwater Vehicle (LRAUV) in real-time based on data collected by the new seven-camera Chiton imaging array. As the LRAUV begins a mission, Chiton starts to collect images of organisms and the ambient particle field immediately in front of the vehicle. When Chiton detects a salient feature – an elevated density of krill in this experiment – it transmits candidate regions of interest acoustically to an autonomous surface support vehicle tracking the LRAUV while it is at depth. Chiton then sends summary counts of possible krill detections to shore via satellite once the LRAUV returns to the surface. Analysis of all image data once Chiton is collected suggests the system returned plausible, but imperfect, event detections and recorded several apparent migrations. This early experiment suggests many avenues for autonomous targeted sampling both on the individual vehicle and with networks of instruments. The types of adaptive strategies enabled by supervised autonomy promise to illuminate cryptic facets of the pelagic community, expand monitoring coverage, and provide new understanding of physical-biological interactions.