AILARON will image, process, analyze, classify plankton-based imagery to enable intelligent onboard targeted sampling on an autonomous underwater vehicle (AUV). The end outcome will be a novel mobile robotic tool for upper water-column microbial biology. In particular, this highly interdisciplinary project will have substantial impact on the study of microbial time-series in ongoing studies of the changing planet. It will accelerate the time consuming process in asking -who is there- from traditional methods of obtaining water samples, storage, transport, microscopy and analysis while leveraging advances in imaging and robotics hitherto not accomplished before. The proposed project targets fundamental research in the fields of Robotic Vision, Machine Learning, Artificial Intelligence (AI) and Control Theory. In particular, we propose to couple Robotic Vision based upper water-column sampling and Machine Learning techniques for exploration with a marine robotic vehicle to autonomously target specific microbiological taxa. The AUV will use a novel camera system to image microorganisms in the photic zone (upper 75 meters), process imagery in-situ, categorize and classify based on Deep Learning, generate a probability density map in X,Y and Z planes, and use an advanced AI based controller to return to the most highly correlated hotspots with respect to species of interest. The entire processing chain will be embedded and guided by a human expert (as needed) via a communication link to shore to potentially alter her sampling preferences dynamically making the vehicle to adapt on-the-fly. The project funds one doctoral and one post-doctoral fellow and is organized in 4 work packages: WP1: Imaging and Sensor Fusion WP2: Robotic Vision and Machine Learning WP3: Flow Estimation and Mapping WP4: Mapping with onboard AI Planning and Control Our primary experimental site will be the Trondheimfjord.
Project leader: Annette Stahl
Institution: Institutt for teknisk kybernetikk