Victoria Preston

Her research goal is to enable intelligent mobile scientific observatories for robust and autonomous long-term monitoring of complex, dynamic environments. She is particularly motivated by geochemical contexts in which a robot may be tasked with mapping greenhouse gases or toxic chemicals in aquatic or atmospheric systems. Victoria’s SM thesis (2019) presented a novel robotic decision-making framework for maximum-seeking in potentially transient environments, and her PhD thesis (2022) presented a full autonomy framework for detecting, predicting, and sampling deep ocean hydrothermal plumes with state-of-the-art autonomous underwater vehicles. Through this work, key challenges on data efficiency in learning, inference over high-dimensional states, and planning in spatiotemporal systems have been key themes. Currently, her research is focused on developing a data-efficient probabilistic models to learn spatiotemporal dynamics from sparse robotic measurements that support active scientific exploration in some of the most extreme environments on Earth.