Tightly-coupled Visual-DVL-Inertial Odometry for Robot-based Ice-water Boundary Exploration

Underwater robots, like Autonomous Underwater Vehicles (AUVs) and Remotely Operated Vehicles (ROVs), are promising tools for the exploration and study of the under-ice environment and the ecosystems that thrive there. However, state estimation is a well-known problem for robotic systems, especially, for the ones that travel underwater. In this paper, $w$e present a tightly-coupled multi-sensors fusion framework to increase localization accuracy that is robust to sensor failure. Visual images, Doppler Velocity Log (DVL), Inertial Measurement Unit (IMU) and Pressure sensor are integrated using a Multi-State Constraint Kalman Filter (MSCKF) for state estimation. Besides, a modified keyframe-based clone marginalization and a new DVL-aided feature enhancement method are presented to further improve the localization performance. The proposed method is validated in the under-ice environment on Lake Michigan, USA, and the results are cross-compared with 10 other different sensor fusion setups. Overall, the integration of keyframe enabled and DVL-aided feature enhancement yielded the best performance with a Root-mean-square error of less than 2 m compared to the ground truth path over a total traveling distance of about 200 m.