search

UT-Austin-RobIn / telemoma

TeleMoMa: A Modular and Versatile Teleoperation System for Mobile Manipulation
44 stars 7 forks source link

How to extend to other 6-DoF Arm #3

Open GayStarc opened 1 week ago

GayStarc commented 1 week ago

Hi! Your work is so great. But I have some questions about how to extend to other 6-DoF arm real robot. Why you don't try Fetch Robot in real world ? Is it because that such structure of robot arm is hard to mobile manipulate? I now want to try to mobile manipulate a robot similar to Fetch Robot. Thank you for your assistance in advance!

ShivinDass commented 4 days ago

Hello @GayStarc, We currently don't have support for Fetch Robot in real world because we didn't have access to one but it's easy to extend Telemoma to your own robot.

To set up Telemoma on your own robot, the easiest way would be to create a gym-like wrapper around the functionalities of your robot (refer to HSRGym for a real world example and FetchEnv for a sim example).

Specifically, you would need to define,

  1. A function that returns the base, torso and end-effector poses as observations (see here for an example on a real HSR)
  2. A function that receives the actions from the Telemoma's human-interfaces and sends appropriate controls to the robot's joints to take those actions (see here for an example on a real HSR). The actions for the base are linear and angular velocities and for the end-effector are 6DoF cartesian deltas.

There is also some more info provided here about the observation and action spaces to help set up telemoma on your own robot.

Let me know if you get stuck or have any other questions. If you end up setting Telemoma up on your fetch, feel free to open a PR and we would be happy to incorporate it into Telemoma for anyone looking to work with Fetch in the future.