A Gripper-like Exoskeleton Design for Robot Grasping Demonstration

Learning from demonstration (LfD) is a practical method for transferring skill knowledge human demonstrator to robot. Several studies have shown the effectiveness of LfD in robotic grasping tasks improve success rate and accelerate development new tasks. A well-designed device can effectively represent motion transfer skills robots. In this paper, an improved gripper-like exoskeleton with data collection system proposed. First, we present mechatronic details its motion-tracking system, considering manipulation flexibility acquisition requirements. We then capabilities which collects position, pose displacement gripper on exoskeleton. The collected further processed by processing software. Next, describe principles Gaussian mixture model (GMM) regression (GMR) robot learning, are used raw demonstrations motions. experiment, optimized trajectory was learned multiple reproduced results show that GMR complemented GMM able learn smooth trajectories noise.