Teleoperation#
Overview#
Teleoperation allows for users to control a robot remotely in real time with a controller. One major use case for this is to collect demonstration data for use in training a model to perform tasks that require human dexterity or fine grained control. Teleoperation is key for collecting high quality, diverse real-world demonstrations.
Isaac ROS provides teleoperation support through two repositories:
Isaac ROS Teleop — a ROS 2 package that bridges Isaac Teleop XR headset data into the ROS 2 ecosystem.
Isaac ROS Physical AI — application-level bringup packages for deploying whole-body control and teleoperation on humanoid robots.
XR Teleoperation#
One method to perform teleoperation is using an XR headset, such as the PICO 4 Ultra to stream the operator’s hand poses to the robot in real time. The robot then mimics the poses of the operator using the whole-body controller.
Refer to Isaac ROS Physical AI for the package overview and launch arguments.