Ground robot path planning on 3D mesh surfaces using Bi-directional RRT based on local regions

Abstract

The increasing application of ground robots requires efficient path planning algorithms in three-dimensional (3D) environments containing non-spherical topology. Path planning on surface meshes is possible, however, expensive computation of geodesics is required. To reduce the length and, hence, cost of the geodesics, a growing submesh based on local regions is used. Rapidly-exploring Random Trees (RRT) with local regions are computed and compared with the bi- directional variant, based on RRT-Connect. Results show that RRT-Connect with local regions reduces the computational burden for mesh-based path planning.