An inertial human upper limb motion tracking method for robot programming by demonstration

Abstract

This paper proposes an inertial human motion tracking for robot programming by demonstration (PbD). An original element called heading reset is proposed to catch the drift around gravity direction. It is based on a hypothesis made on the human arm motion during a task demonstration. It is used to overcome the non-use of the magnetometer due to magnetic disturbances from robotic environment. This element is implemented in an orientation estimation algorithm and compared with three other IMU algorithms and a commercial MARG algorithm. The human arm trajectory is estimated through three IMUs sensors directly set on the arm to estimate the segment orientation (hand, forearm and arm). A specific inertial-2-segment procedure is presented as well as a procedure to estimate the transformation from human reference frame to task frame, necessary for a PbD process. Experimental tests, using a robot as a reference, have been conducted to validate the different part of the method. The heading reset and the orientation algorithm show good results. The inertial-2-segment procedure is shown to be robust. Finally, experimental tests on a human arm and physical robot validate the complete method.