Overview
DEX-EE is an advanced robotic hand designed for deep-learning research into dexterous robot manipulation. Built to be extremely robust and survive long, harsh training sessions, it can maintain fine dexterity and capture copious sensor data. Created in collaboration with Google DeepMind, it features high-speed sensor networks, tactile sensing channels, stereo camera-based fingertip tactile sensors, and is integrated with the Robot Operating System (ROS) for supporting long-running reinforcement learning experiments.
Detailed specifications
Motion & kinematics1
- Dof
- 12
Sensors1
- Sensor Suite
- Each finger contains 155 individual sensor channels including 5 tendon force sensors, 5 motor encoders, 4 joint angle sensors, 3 inertial measurement units, 36 total taxels in middle and proximal tactile sensors, plus 50 FPS 640x480 pixel stereo video stream from distal tactile sensor. Additional current, voltage, and temperature monitoring for all motors.
Compute1
- Ros Compatible
- true
Other19
- Height Mm
- 165
- Price Tier
- 40-80K
- Applications
- dexterous robot manipulation, deep-learning research, extended machine learning experiments, AI experiments
- Sub Category
- dexterous hand
- Api Available
- true
- Datasheet Url
- Preview PDF online
- Max Speed M S
- Up to 180 º/s per joint (approximately 3.14 rad/s or 0.055 m/s assuming ~10 cm lever arm).[3][4]
- Pricing Model
- purchase
- Force Limiting
- true
- Model Variants
- DEX-EE Chiral
- Company Country
- GB
- Deployment Notes
- Designed for long-duration AI experiments with modular fingers and robust construction. Features fail-safes and shutdown routing to reduce failure and downtime. Built to survive harsh training sessions with minimal maintenance training required.
- Youtube Video Id
- upi6c9dmJpM
- Industries Served
- research
- Software Platform
- EtherCAT bus fully integrated into ROS (Robot Operating System).[1][3][5][6]
- Training Required
- advanced_>5days
- Availability Status
- available
- Programming Interface
- code_python,code_ros
- Additional Information
- Designed specifically for long-running reinforcement learning experiments with emphasis on hardware reliability and uptime.,Modular finger units allow user replacement, reducing downtime.,High-bandwidth sensing: position, force, inertial measurements, and hundreds of tactile channels per finger.,Camera-based optical tactile sensors at the fingertip with video output over USB 3.0.,DEX-EE Chiral variant has a human-like thumb offset for teleoperation and bimanual tasks.,Active compliance via 10 kHz force control in motor units.,Resistant to repeated impacts and aggressive interaction from untrained policies.,Patent-pending N+1 tendon actuation redundancy for reliability.,Target audience: research labs and industry teams in dexterous manipulation, ML, and imitation learning.,No teaching pendant or drag-and-drop programming; advanced programming via Python/ROS expected.
Reviews for DEX-EE
Loading reviews…