- Fingers you can swap Each finger is a self-contained module held by six screws and one connector, so a damaged finger is replaced while it goes for repair.
- Built to survive untrained policies Shadow designed DEX-EE to withstand repeated impacts, high forces and abrasion from untrained neural networks, with a long mean time between failures and a short repair time.
- Camera-based tactile fingertips A stereo camera pair tracks about 500 markers inside each fingertip's silicone skin and streams 1280 × 480 video at 50 fps.
- 155 sensor channels per finger Joint angles at 0.022°, tendon forces at 4 mN, acceleration, angular rate and motor health all report at 1,000 Hz.
- N+1 tendon drive at 10 kHz Five motors drive each finger's four joints, and each motor closes a 10 kHz force loop for active compliance.
- 180°/s joints, 8 N fingertips Each of the 12 joints moves at up to 180°/s, and each fingertip presses with at least 8 N.
- EtherCAT, ROS and an ISO flange The hand connects over EtherCAT with full ROS integration and mounts with four M6 screws on an ISO 9409-1-50-4-M6 pattern.
Description
The Shadow DEX-EE is a three-finger dexterous robotic hand from Shadow Robot Company, developed in collaboration with Google DeepMind for machine-learning research. It is built to keep running through the trial-and-error phase of machine learning, when an untrained policy drives the hand into repeated impacts, high forces and abrasion. Each of its three fingers is a self-contained module with four joints, 12 degrees of freedom in all, and a damaged finger comes off with six screws and one connector while it goes for repair.
DEX-EE is aimed at labs running reinforcement learning, imitation learning and autonomous grasping on real hardware, where uptime matters more than human proportions: at 350 mm long, 165 × 160 mm across and 4.1 kg, it is larger than a human hand. Five motors in each finger base drive the four joints through tendons in an N+1 arrangement, each closing a 10 kHz force loop for active compliance, and the host runs joint position or torque control at 1 kHz. Joints move at up to 180°/s and each fingertip presses with at least 8 N.
Every finger carries 155 sensor channels: four joint sensors at 0.022°, a 3-axis accelerometer and gyroscope, five tendon force sensors at 4 mN, and motor velocity, current, temperature and clutch-slip readings, all at 1,000 Hz. The fingertip is a camera-based tactile sensor in which a stereo pair watches about 500 markers inside the silicone skin and streams 1280 × 480 video at 50 fps, with all three fingertips sharing one USB 3.0 link. Optional modules add 14 and 22 three-axis taxels on the middle and proximal phalanges. Motor temperature reduces power gradually instead of cutting out, and tendon odometry shows when a tendon is due for replacement.
The standard DEX-EE spaces its fingers evenly around the base; the DEX-EE Chiral offsets the third finger like a thumb for teleoperation and bimanual work, as a left hand, a right hand or a pair, and the configuration is confirmed with the quotation. The hand mounts on an ISO 9409-1-50-4-M6 flange, connects over EtherCAT with full ROS integration and runs from a 48 V, 200 W supply. For human-like kinematics with five fingers, see the Shadow Dexterous Hand (DH), or compare our full range of dexterous robotic hands.
Key Capabilities
- Fingers you can swap: Each finger is a self-contained module held by six screws and one connector, so a damaged finger is replaced while it goes for repair.
- Built to survive untrained policies: Shadow designed DEX-EE to withstand repeated impacts, high forces and abrasion from untrained neural networks, with a long mean time between failures and a short repair time.
- Camera-based tactile fingertips: A stereo camera pair tracks about 500 markers inside each fingertip's silicone skin and streams 1280 × 480 video at 50 fps.
- 155 sensor channels per finger: Joint angles at 0.022°, tendon forces at 4 mN, acceleration, angular rate and motor health all report at 1,000 Hz.
- N+1 tendon drive at 10 kHz: Five motors drive each finger's four joints, and each motor closes a 10 kHz force loop for active compliance.
- 180°/s joints, 8 N fingertips: Each of the 12 joints moves at up to 180°/s, and each fingertip presses with at least 8 N.
- EtherCAT, ROS and an ISO flange: The hand connects over EtherCAT with full ROS integration and mounts with four M6 screws on an ISO 9409-1-50-4-M6 pattern.
This model is offered on a quotation basis. Configuration, accessories, integration scope and delivery schedule all change the final figure, so pricing is confirmed per project rather than listed.
Buying through Robotix Market
This unit is sourced new through established distribution channels and ships worldwide with the manufacturer warranty. Technical support is coordinated with the manufacturer's own service network. Deployment scope, configuration, and commissioning responsibilities are agreed as part of the quotation — on-site installation is not included in the product price.
Features
Specs
| Brand | Shadow Robot Company |
|---|---|
| Model | DEX-EE |
| Configuration | Three fingers spaced evenly around the hand base; DEX-EE Chiral offsets the third finger like a thumb (left, right or pair) |
| Fingers | 3, each a self-contained, user-replaceable module (six screws and one connector) |
| Degrees of freedom | 12 (4 per finger) |
| Finger joint ranges | J0 base abduction/adduction −50° to 50°; J1 proximal −80° to 45°; J2 middle 0–80°; J3 distal −30° to 85° |
| Actuation | Tendon-driven; 5 motors per finger in an N+1 arrangement, housed in the finger base |
| Joint speed | Up to 180°/s |
| Fingertip force | At least 8 N |
| Size | 350 mm long, 165 × 160 mm wide and high |
| Weight | 4.1 kg (three-finger hand); 1.2 kg per finger |
| Control loops | Tendon force at 10 kHz in each motor unit; joint position and joint torque at 1 kHz on the host |
| Sensor channels | 155 per finger, plus distal tactile video |
| Joint sensing | 4 per finger, 1,000 Hz, 0.022° resolution |
| Inertial sensing | 3-axis accelerometer and 3-axis gyroscope per finger, 1,000 Hz |
| Tendon force sensing | 5 per finger, 1,000 Hz, 4 mN resolution |
| Motor monitoring | Velocity, current, temperature and clutch slip, 1,000 Hz |
| Distal tactile sensor | Stereo camera pair inside a silicone skin tracking about 500 markers (up to 2,180); 1280 × 480 video at 50 fps per fingertip |
| Middle and proximal tactile sensors (option) | 14 and 22 three-axis Hall-effect cells per finger, 12-bit, 100 Hz |
| Communication | EtherCAT (100 Mbps) to the host; one USB 3.0 link for all fingertip video |
| Software | Full ROS integration |
| Arm mounting | 4 × M6 on a 50 mm PCD (ISO 9409-1-50-4-M6) |
| Power | 48 V, 200 W supply |
| Health monitoring | Motor power reduced gradually at high temperature; tendon odometry to plan tendon replacement |
Industries
Frequently Asked Questions
It is offered by quotation — the configuration (DEX-EE or DEX-EE Chiral, left, right or a pair, optional finger tactile modules), quantity and delivery are confirmed per project.
Three fingers with four joints each, 12 degrees of freedom in all: a side-to-side joint at the base and three flex and extend joints along the finger, driven by five motors per finger.
Each fingertip is a camera-based sensor: a stereo camera pair tracks markers inside the silicone skin and streams 1280 × 480 video at 50 fps, sensing light touches and forces of several tens of newtons. Optional modules on the middle and proximal phalanges add 14 and 22 three-axis taxels per finger.
It is larger than a human hand: 350 mm long and 165 × 160 mm across, at 4.1 kg with 1.2 kg per finger. Each fingertip presses with at least 8 N and the joints move at up to 180°/s.
It mounts with four M6 screws on an ISO 9409-1-50-4-M6 flange and takes three cables: 48 V power, EtherCAT to the host and USB 3.0 for the fingertip video. The software is fully integrated with ROS.
DEX-EE spaces its three fingers evenly around the base for balanced grasping and a simpler kinematic model. DEX-EE Chiral moves the third finger down and to the side like a thumb, for teleoperation, imitation learning and bimanual work, and comes as a left hand, a right hand or a pair.
Recently Viewed
You haven't viewed any products yet.
