Overview

Motion retargeting is the process of mapping motions between different kinematic structures. This stack implements a simple, yet effective motion retargeting method that is used in the context of a unilateral teleoperation system for transferring upper body motions from a human operator to a robot in an online fashion.

The operator's motion is captured using a Kinect device. In order to capture as much of the original motion's expressiveness as possible, we are making use of multiple end-effectors, which are currently the wrists, elbows, torso and head (caveat: the openni_tracker does not capture head/wrist orientation). Then, we adapt the captured motion to the robot's kinematic structure, which can differ significantly in terms of number of degrees of freedom and body proportions. Finally, an inverse kinematics computation produces a joint space reference for the robot to follow. This reference takes into account technological and geometric constraints like joint position and velocity limits, as well as self-collisions.

Installation

Groovy version is not ready yet.

Tutorials/Examples

The tutorials page contains a tutorial for testing the pipeline using your Kinect or a provided bag file of an example captured operator motion.

Documentation

See individual packages for details.

Report a Bug

https://github.com/pal-robotics/reem_teleop/issues