Flow Field Estimation in Underwater Vehicle Navigation Using Sporadic Image Observations

Underwater flow field estimation is a crucial step in marine robot missions, as the stability and safety of Autonomous Underwater Vehicles (AUVs) are heavily influenced by the ocean currents. This paper introduces a novel learningbased algorithm, called the Vision-based Motion Tomography (VMT), which enables AUVs to estimate unknown flow fields using only sporadically available visual sensor data. Simulation results show that the estimated flow map generated by the proposed algorithm converges to its corresponding ground truth.