From human motion capture to industrial robot imitation

Abstract

Making a robot duplicate human operator gestures is a difficult task. To achieve this goal, we have defined a methodology and have started designing software tools to simplify and automate this process. The first step is the motion acquiring of the tool or of the object manipulated by the human subject. The recorded motion is then processed to be transformed as the trajectory of the Tool Center Point (TCP) of a robotic arm-hand set whose joint architecture and kinematics are different from the human arm ones. All robot axis coordinates are computed by solving the inverse kinematics model. We demonstrated the feasibility of our approach by experimentally having a robot duplicate the movement of a tool motioned by an operator. The 3D motion capture system used is based on passive marker technology which does not perturb the human gesture. After being processed, the tool trajectory is used to generate a list cyclic positions controls synchronously applied by a real-time controller to the robot motor drives. The robotic set-up consists of a newly designed 4-finger hand embedded on an industrial 6-axis serial robot provided with an innovative open robotic drive.