A Plane-based Map for Wheel-legged Rover Efficient Motion Planning during Planetary exploration

摘要

This paper proposes an algorithm for wheel-legged robot mapping in rocky terrain environments. The proposed plane-based mapping method employs two density-based machine learning algorithms to determine approximate flat surfaces from the point cloud. It models the terrain as an elevation map but sacrifices its elevation accuracy to detect planar surfaces in the scenery. The plane-based terrain modeling method is a submap of the whole environment. It maps the robot's desired area to pre-plane the following motion patterns, providing a solution for wheel-legged robots taking advantage of wheels and legs. In cooperation with an extra motion library, the wheel-legged robot's transit speed and moving efficiency could be accelerated.