Autonomous underwater vehicle-based adaptive sensing of natural oil seeps

An increasingly critical mission for autonomous underwater vehicles (AUVs) will be providing data for response to anthropogenic disasters like oil spills. These types of events present three primary challenges to conventional AUV operation: a priori information (e.g. from aerial surveys) often has little relation to underwater conditions, the features of interest are patchy in space, occupying a small percentage of the total sampled volume, and spatial extent is variable with time, subject to currents and physical forcing from the environment. To truly map out the complex and 4-dimensional features of oil in an ocean environment is also made difficult by sensing limitations (e.g. biology in the water column aliasing with oil in the water column in acoustic and fluorometric data) that continue to make expensive, time consuming water sampling the gold standard. This work discusses an exploration of multi-AUV and sensor-adaptive techniques for mapping of natural oil seeps conducted in coastal waters off of Santa Barbara, California, in August 2019. With up-looking echosounder data to see full water column scattering, fluorometer for chlorophyll, FDOM and scattering measurements, ordinary cameras, a holographic camera for micro-scale imaging, and AUV-based water gulpers for collecting water samples, a 2-vehicle, multi-sensor approach was tested for locating areas of interest and capturing water column properties. Autonomous techniques using adaptation constrained by environmental sensing and operator instructions were used to improve information gain and conduct AUV-based water sampling in high-concentration areas. The data from this experiment highlights the difference of the Santa Barbara oil seeps to fronts and plumes discussed in existing AUV adaptive sampling literature, and provides insight into design of future autonomous adaptation for sensing and sampling of similar small-scale oceanographic features.