Скачать или смотреть Real-time teleoperation of industrial robots with the motion capture suit Perception Neuron @TUM ICS

Bachelor's thesis in Engineering Science at Instutite for Cognitive Systems, Technical University of Munich.

Author: Björn Lütjens, advised by Dr. Emmanuel Dean‐Leon and Prof. Gordon Cheng

The Perception Neuron system creates a moving human skeleton model from the worn motion capture suit (based on IMU data). The demonstrated software system send the motion data from the suit's Windows API over rosserial to ROS. A ROS node on an Ubuntu machine receives the motion data and publishes it on ROS tf as a skeleton (visible on the screen in the background). The tf skeleton is transformed via inverse kinematics into commanding six degree-of-freedom robot (UR10) joint angles. The commanded joints angles are checked on safety regulations and published to real and simulated UR10. All nodes run at 60Hz.

The video was shot when I first tried out the code on a real robot. So, I don't necessarily look very relaxed. Later I have gained more convidence in the safety regulations and one could potentially use the system with unexperienced first-time operators, too.

The code is available on https://github.com/blutjens/perc_neur...