A Particle Filter SLAM Approach to Online Iceberg Drift Estimation From an AUV

Abstract

Using autonomous underwater vehicles (AUVs) for mapping the underwater topography of sea-ice and icebergs, or detecting keels of ice ridges, is foreseen as an enabling technology in future Arctic offshore operations. This paper presents a method for online iceberg drift estimation using a Simultaneous Localization and Mapping (SLAM) approach using an AUV with a multi-beam echosounder (MBE) during such survey/monitoring operations. Iceberg drift is affected by wind, current, and Coriolis forces. This can be hard to predict, making automated mapping of icebergs difficult. The method proposed in this paper estimates the iceberg’s pose using a particle filter, where each particle uses extended information filters to estimate the topography of the iceberg. A grid map is used to store the iceberg topography, and distributed particle mapping is used to avoid expensive copy operations during particle resampling. The proposed method is verified through a simulation study, using a 6 DOF AUV model, an MBE sensor model, and an iceberg topography taken from the PERD iceberg sightings database. The method is able to provide a georeferenced iceberg position, thus, estimating the iceberg’s drift trajectory. A topography estimate of the iceberg, corrected for iceberg drift, is also generated. Furthermore, the algorithm estimates the iceberg drift velocity, as well as the relative iceberg-AUV pose, for use in future iceberg mapping guidance algorithms. The simulation study illustrates the performance of the method, and a short execution time analysis is presented to illustrate the method’s real-time potential.