Paper Image

Learning dexterous functional grasping of tools

Published on:

5 December 2023

Primary Category:


Paper Authors:

Ananye Agarwal,

Shagun Uppal,

Kenneth Shaw,

Deepak Pathak


Key Details

Uses vision model trained on internet images to locate grasp points

Trains policy in simulation to actually pick up objects at those points

Eigengrasps make training more stable and realistic

Works on various tools - hammers, drills, spatulas

Matches or exceeds human teleoperator performance

AI generated summary

Learning dexterous functional grasping of tools

This paper combines computer vision techniques and reinforcement learning in simulation to accomplish dexterous and functional grasping of tools and objects. It obtains grasp locations from internet images, then trains policies in simulation to execute the grasp. The method generalizes well to new objects.

Answers from this paper


No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up