A Landmark-based Motion Planner for Rough Terrain Navigation

: In this paper we describe a motion planner for a mobile robot on a natural terrain. This planner takes into account placement constraints of the robot on the terrain and landmark visibility. It is based on a two-step approach: during the first step, a global graph of subgoals is generated in order to guide the robot through the landmark visibility regions; the second step consists in planning local trajectories between the subgoals, satisfying the robot placement constraints. 1. Introduction Autonomous navigation on natural terrains is a complex and challenging problem with potential applications ranging from intervention robots in hazardous environments to planetary exploration. Mobility in outdoors environments has been demonstrated in several systems [16, 7, 8, 3]. The adaptive navigation approach currently developped at LAAS within the framework of the EDEN experiment [3] demonstrates fully autonomous navigation in a natural environment, gradually discovered by the robot. The a..