A Roughness-based RRT for Mobile Robot Navigation Planning

Abstract This paper proposes a novel Rapidly exploring random tree algorithm on rough terrains (RRT-RT) for the purpose of outdoor mobile robot navigation. Differently from other RRTs adopted for rough terrains where finding a nearest neighbor from a new random state within the tree is based on Euclidian distance, the proposed algorithm uses a roughness based metric. The metric is defined by the help of the roughness based navigation function, RbNF, that represents an estimate of the roughness-to-go value from each terrain location to the goal position. Simulation study shows that the RRT-RT provides an effective way to explore more promising terrain regions in order to decrease the total roughness along the resulting path.