CARVIEW

MOTORHOMES

Select Language

HTTP/2 200 server: GitHub.com content-type: text/html; charset=utf-8 last-modified: Thu, 04 Dec 2025 17:56:12 GMT access-control-allow-origin: * strict-transport-security: max-age=31556952 etag: W/"6931cb3c-5e14" expires: Mon, 29 Dec 2025 08:02:48 GMT cache-control: max-age=600 content-encoding: gzip x-proxy-cache: MISS x-github-request-id: 2437:3946E9:870376:97C7E1:6952334E accept-ranges: bytes age: 0 date: Mon, 29 Dec 2025 07:52:49 GMT via: 1.1 varnish x-served-by: cache-bom-vanm7210027-BOM x-cache: MISS x-cache-hits: 0 x-timer: S1766994769.891859,VS0,VE202 vary: Accept-Encoding x-fastly-request-id: 489e231c2baff90f52f18fdac60032d3a14e72f4 content-length: 5562 Learning Dexterous Manipulation Skills from Imperfect Simulations

Learning Dexterous Manipulation Skills from Imperfect Simulations

Elvis Hsieh^*, Wen-Han Hsieh^*, Yen-Jen Wang^*, Toru Lin, Jitendra Malik, Koushil Sreenath^†, Haozhi Qi^†

^* Equal contribution (listed in alphabetical order). ^† Equal advising.

UC Berkeley

pdf arxiv code (sim-to-real) code (skill-teleop)

* Full fastening and screwdriving videos are shown below.

TLDR:

We learn rotational motion in a simplified simulator, use it for skill-based teleoperation to collect multisensory data, and train a policy that performs screwdriving and nut-bolt fastening in the real world.

Stage 1: Simulation

We train a transferable rotational motion policy in simulator.

Simulation objects

Simulation Training

Screwdriver (Sim2Real Failure Mode)

Bolt-nut (Sim2Real Failure Mode)

Stage 2: Teleoperation

We use the trained rotational motion policy for skill-based teleoperation to collect multisensory data.

Screwdriver

Bolt-nut