DEX-EE (Shadow Robot × Google DeepMind)

A rugged, three-fingered research dexterous hand co-developed by Shadow Robot and Google DeepMind.

DEX-EE is a dexterous hand developed by the UK's Shadow Robot Company over roughly five years of iteration, at the request of Google DeepMind's robotics team. It doesn't try to look like a human hand: it has three fingers, each with 4 degrees of freedom for 12 total, is about half again as large as a human hand, and weighs about 4.1 kg. The design priorities are durability and sensitivity, since reinforcement learning needs to try and fail repeatedly on the real hardware, and an ordinary dexterous hand breaks too easily under that kind of abuse. Each fingertip carries an optical tactile sensor with hundreds of sensing points, the knuckles also carry multi-dimensional tactile sensing, and position, force, and IMU data are all streamed back at high speed. A later version, DEX-EE Chiral, moves the third finger to the side so it opposes the other fingers like a human thumb, making it easier for human teleoperation and imitation learning; it's offered in left-hand, right-hand, and two-handed sets.

ExampleDeepMind uses DEX-EE as a platform for real-world learning research, letting a policy grasp and manipulate objects repeatedly on the real hardware to collect data.

Also called

DEX-EE, DEX-EE Chiral

Related

Dexterous Hand · Shadow Robot Company · Shadow Dexterous Hand · Google DeepMind · Tactile Sensor · Dexterous Manipulation

Sources

DEX-EE Series - Shadow Robot
Our latest advances in robot dexterity - Google DeepMind

As of

2026-09