Item

Abstract

Sea ice retreat as a consequence of climate change leads to increasing shipping activities within polar waters, as newly opened shipping routes can be much shorter than the established ones. Consequently the demand for time and fuel is strongly reduced. However, navigation in polar waters is still challenging and even dangerous, e.g. because of fast changing ice conditions or unknown bathymetry. Even with having access to the proper earth observation data like radar images, ice classifications, or ice charts, manoeuvring in polar waters is not trivial and requires trained staff as well as expert knowledge. To provide navigational assistance in polar regions, we develop a system that provides route suggestions based on earth observation data, given ship characteristics, bathymetry and drift / weather models. Using these models, ice classifications derived from earth observation data are interpolated in time to gain high-resolution knowledge about the changing ice conditions. Solving the resulting 3-dimensional route optimization problem using an A* algorithm is inefficient when applied to long-distance routes, because large datasets offer too many possible combinations of connecting waypoint candidates. To overcome this issue, preprocessing steps and further modification of the A* algorithm are investigated. The preprocessing techniques reduce the number of waypoint candidates by creating a 'road map' , while keeping important information about small features in the ice. Investigated variants of the A* algorithm include e.g. weighting methods and an anytime implementation. Both approaches and their combination are evaluated in terms of efficiency and reasonability.