Optimal Any-Angle Pathfinding on a Sphere

Abstract

Pathfinding in Euclidean space is a common problem faced in robotics and computer  games. For long-distance navigation on the surface of the earth or in outer space however,  approximating the geometry as Euclidean can be insufficient for real-world applications  such as the navigation of spacecraft, aeroplanes, drones and ships. This article describes an any-angle pathfinding algorithm for calculating the shortest path between point pairs  over the surface of a sphere. Introducing several novel adaptations, it is shown that Anya  as described by Harabor & Grastien for Euclidean space can be extended to Spherical  geometry. There, where the shortest-distance line between coordinates is defined instead by a great-circle path, the optimal solution is typically a curved line in Euclidean space.  In addition the turning points for optimal paths in Spherical geometry are not necessarily  corner points as they are in Euclidean space, as will be shown, making further substantial  adaptations to Anya necessary. Spherical Anya returns the optimal path on the sphere,  given these different properties of world maps defined in Spherical geometry. It preserves all primary benefits of Anya in Euclidean geometry, namely the Spherical Anya algorithm always returns an optimal path on a sphere and does so entirely on-line, without any  preprocessing or large memory overheads. Performance benchmarks are provided for several  game maps including Starcraft and Warcraft III as well as for sea navigation on Earth  using the NOAA bathymetric dataset. Always returning the shorter path compared with  the Euclidean approximation yielded by Anya, Spherical Anya is shown to be faster than  Anya for the majority of sea routes and slower for Game Maps and Random Maps.