Sonar-based iceberg-relative AUV navigation

AUVs have been operating under the ice for years. All of these systems have relied on combinations of dead-reckoning using inertial measurements, acoustic transponder networks, and/or velocity measurements from a Doppler velocity logger (both seafloor-relative and ice-relative) for navigation and control. These existing systems can be very accurate for operation under a stationary ice sheet, but they cannot provide ice-relative navigation accounting for the full motion of free-floating icebergs (especially rotation). Further, while some of these AUVs have collected sonar images of the underside of the ice, none has used these data for navigation. This paper explores the extension of sonar-based terrain-aided navigation techniques to enable an AUV to localize its position with respect to a moving and rotating iceberg. Terrain-navigation techniques provide drift-free position estimates with respect to mapped terrain and have been demonstrated for aircraft, missile, and numerous underwater vehicle applications. The availability of terrain-aided navigation would enable an AUV to return to sites of interest for sampling and serial observations. In particular, this paper presents an approach to developing maps of the underside of icebergs that would be sufficient to enable autonomous localization and navigation for AUVs. The viability of this approach is demonstrated using data collected from a sideways-looking multibeam sonar system mounted on the R/V Nathaniel B. Palmer in Antarctica, June 2008. During data collection, the ship completed approximately 400 degrees of circumnavigation of a small (<1 nmi
²
) free-floating iceberg. Hence, data from the beginning and end of the experiment overlap the same section of the iceberg. These data are used to estimate parameters in a simple iceberg motion model, and the iceberg-relative ship track is then recovered by subtracting the estimated iceberg motion from the measured GPS track of the ship. Projection of the measured sonar ranges from the iceberg-relative ship track yields a self-consistent iceberg map, up to the accuracy of the estimated iceberg motion.