Toward Autonomous Sampling and Servicing with the Ranger Dexterous Manipulator

Abstract

enable autonomous manipulation, robotic systems must accurately identify grasping targets and must plan and robustly execute trajectories between them. This paper describes an extension of the University of Maryland’s Ranger manipulator system to autonomously identify, locate, and retrieve known objects suspended within the manipulator’s workspace. A machine vision system identifies target locations with a stereo camera pair rigidly axed to the Ranger body, and then a trajectory planner generates smooth joint-space motions to the target while avoiding workspace singularities. Experimental results demonstrate Ranger’s ability to reliably retrieve the target with a variety of positions and lighting conditions. After the target has been grasped, a final scripted action sequence is executed to stow the manipulator.