Effective Ground Mapping for Autonomous Excavation

Abstract

The accurate and detailed sensing information of the excavated ground is a prerequisite for successful autonomous excavation. For this issue, our study proposes efficient methodologies for ground mapping that can generate a reconstructed map to deal with sensor occlusion and a 5D map beyond a normal 3D map to provide the various ground information at excavation sites. For the map reconstruction, exteroceptive and proprioceptive data were merged for occlusion areas. The 5D map was created using the information of 3D geometry, Lidar's intensity, and ground force index. Experimental tests using a mini excavator were carried out to validate the proposed methodologies.