This paper presents a method for mapping translating, rotating icebergs with an autonomous underwater vehicle (AUV). To map an iceberg, the AUV first circumnavigates it, collecting multibeam sonar ranges and iceberg-relative Doppler sonar velocities from the submerged iceberg surface. The primary challenge is then to estimate the trajectory of the mapping vehicle in a noninertial reference frame attached to the moving iceberg. The collected multibeam ranges may then be projected from this trajectory to form a map of the iceberg's submerged surface. The approach of the method involves identifying the iceberg-frame locations of all the Doppler sonar measurements made during circumnavigation, allowing the AUV's iceberg-relative trajectory to be computed from those locations. The measurement locations are estimated simultaneously with the trajectory of the iceberg to be most consistent with the inertial-space positions, inertial-space velocities, distances between points on the iceberg surface as measured by the Doppler sonar, and alignment of multibeam ranges measured at the beginning and end of the circumnavigation. The measurements depend nonlinearly on the modeled positions and iceberg trajectory, and the paper presents a solution formulation that deals efficiently with the nonlinearity. By incorporating iceberg-relative vehicle velocity into the estimation, the method achieves two significant advances beyond prior work by the authors. First, and most significantly, the method adds ice-relative vehicle velocity measurements (e.g., using a Doppler velocity logger). This makes the method robust to common vehicle inertial navigation errors. Second, inclusion of iceberg-relative vehicle velocity data allows for the identification of a more general model of iceberg trajectory, making the method robust to changes in iceberg translation and rotation rates. Currently, no iceberg circumnavigation data sets are available that include iceberg-relative velocity from Doppler sonar. However, this paper includes results from simulated free-drifting icebergs, and experimental results from an AUV seafloor mapping dive. The simulation data provide a moving iceberg testbed with known ground truth for the mapping results. The seafloor data provide a qualitative verification that the method works with real vehicle data.
Read full abstract