Rough-Terrain Robot Motion Planning based on Obstacleness

Various approaches have been proposed to tackle the rough-terrain robot motion planning task to date. Depending on the application at hand, a different algorithm is selected to generate a motion plan. Nevertheless, they all have the fundamental characteristics of the rough terrain in common. Based on this notion, a unifying formalism is presented which can be used to represent application-dependent properties of the terrain to the motion planner. The approach is based on the characteristic difficulty associated with each configuration of the robot on the rough terrain. This concept is formalised resulting in the definition of "obstacleness" as generic degree of presence of an obstacle. A novel rough-terrain motion planning algorithm is presented which makes use of the obstacleness measure. The approach is based on a bi-directional rapidly exploring random tree (RRT) algorithm. The exploration of configuration space by the RRTs is biased towards regions of low navigational difficulty to compute trajectories in easily traversable terrain. Preliminary test results obtained with a fictional planetary rover in a rigid body dynamics simulation are also presented