Abstract:

This paper proposes Terrain-Aware Anisotropic A* (TA-A*), an A*-based path planning method for rough terrain. TA-A* preserves the standard A* search procedure and introduces a directiondependent transition cost computed from co-registered obstacle, elevation, and surface-modifier maps. Each directed move combines the 3D step length with three terrain-derived terms: a remapped surface-modifier factor, a pitch-related along-slope term, and a roll-related cross-slope term based on the local height gradient. Together, these terms define an anisotropic, directiondepend cost model in which the cost of moving from u to v may differ from the cost of moving from v to u.
Five planners (Dijkstra, A*, A* with terrain modifiers, A* with pitch/roll penalties, and TA-A*) are evaluated on six benchmark map sets. Because the used alghoritms use different cost functions, all generated path are evaluated using the TA-A* cost that allow comparision. The results show that TA-A* achieves the lowest cost where the largest improvements were observed on the most challenging terrains (dune-like and montain-pass terrain).