Allegro Hand

16-DOF dexterous robot hand by Wonik Robotics. 4 fingers, each with 4 independent joints — position-controlled with per-joint torque sensing. A workhorse for manipulation research: grasping, in-hand manipulation, and dexterous assembly. ROS2 driver included, mounts to UR5e, UR10e, Kinova, and custom frames.

16
Degrees of Freedom
4
Fingers (4 DOF each)
~1.5 kg
Hand Weight
333 Hz
Control Loop Rate

Full Hardware Specs

VendorWonik Robotics (formerly SimLab)
Total DOF16 (4 fingers x 4 joints each)
Finger Configuration3 fingers + 1 opposable thumb
Actuation TypePosition-controlled, brushless DC motors
Torque SensingPer-joint torque feedback
Control Loop Rate333 Hz (via EtherCAT/CAN)
Weight~1.5 kg (hand unit)
Finger Tip Force~5 N per finger
Joint Range (typical)~90–135 deg per joint
InterfaceEtherCAT / CAN bus
SDKROS / ROS2 (allegro_hand_ros2), Python API
Compatible ArmsUR5e, UR10e, Kinova Gen3, custom mounts
Hand VariantsRight hand, Left hand
IP RatingIP20 (indoor use)
Operating Temp0 – 40 C
Power Supply24 V DC

Dexterous Hand Comparison

How the Allegro Hand stacks up against Shadow Hand and Wuji Hand across the dimensions that matter for manipulation research.

Feature Allegro Hand Wuji Hand Shadow Hand
VendorWonik RoboticsWuji RoboticsShadow Robot
DOF162024
Fingers455
ActuationPosition + torque sensingTendon-driven, brushless DCTendon-driven, pneumatic option
TactileNone stock (add-on possible)768 taxels (included)BioTac (add-on)
Control Rate333 Hz30–100 Hz1 kHz
Weight~1.5 kg~420 g~4.2 kg
Price (approx.)From $15,000Contact for quote (~$16K)$80,000+
ROS2 SupportOfficial driver (allegro_hand_ros2)Community bridge nodeOfficial ROS2 stack
Best ForTorque-intensive manipulation, graspingTeleop data collection w/ rich sensingHigh-fidelity human-like research
Wuji Hand Details → Compare All Hardware →

Built for Manipulation Research

The Allegro Hand's 16 DOF and per-joint torque sensing make it a practical choice for labs running grasping, contact-rich manipulation, and policy learning experiments.

Dexterous Grasping

Four independently actuated fingers handle power grasps, precision pinches, and tripod grips. Per-joint torque feedback lets controllers detect contact and regulate force without external sensors.

In-Hand Manipulation

Reorienting objects within the hand — rolling, pivoting, regrasping — requires high DOF and fast control loops. The Allegro Hand runs at 333 Hz, giving policies tight feedback for fine-grained finger coordination.

Imitation Learning Data

Record joint positions and torque feedback at 333 Hz. Export to JSONL, HDF5, or MCAP for ACT, Diffusion Policy, or custom imitation learning pipelines via the SVRC data platform.

Compatible Robot Arms

The Allegro Hand mounts to any arm with a standard tool flange. Below are tested and supported configurations available through SVRC.

Universal Robots UR5e / UR10e

Standard UR tool flange adapter available. EtherCAT cable routes through the arm's cable management channel. Most common lab configuration — quick to set up, well-documented.

Kinova Gen3 / Gen3 Lite

Custom adapter plate connects to Kinova's tool plate. ROS2 driver runs alongside Kinova's kortex_driver for synchronized joint recording. SVRC has tested this combination at our Mountain View lab.

Custom & Fixed Mounts

Standard 4-bolt ISO 9283 flange pattern. For custom frames or benchtop rigs, SVRC can fabricate or source adapter brackets. Contact us with your arm model and we will confirm compatibility.

ROS2 & Python SDK

Wonik Robotics maintains the official allegro_hand_ros2 package. Below is the minimal setup to get joint state publishing and position control running.

ROS2 Driver (Humble / Iron)

# Clone and build the official ROS2 driver cd ros2_ws/src git clone https://github.com/Wonikrobotics-git/allegro_hand_ros2 # Build cd .. && colcon build --packages-select allegro_hand_ros2 # Launch (right hand, EtherCAT) ros2 launch allegro_hand_ros2 allegro_hand.launch.py hand:=right # Published topics: # /allegroHand/joint_states sensor_msgs/JointState (16 joints) # /allegroHand/torque_cmd std_msgs/Float64MultiArray

Python — Read Joint Positions

import rclpy from rclpy.node import Node from sensor_msgs.msg import JointState class AllegroReader(Node): def __init__(self): super().__init__('allegro_reader') self.create_subscription( JointState, '/allegroHand/joint_states', self.cb, 10 ) def cb(self, msg): print(f"Positions: {list(msg.position)}") # 16 floats

Research Applications

Policy Learning & RL

16-dimensional joint space and per-joint torque observations feed directly into standard RL frameworks. MuJoCo and Isaac Gym URDF models are available for sim-to-real transfer experiments.

Contact-Rich Assembly

Torque sensing lets the hand detect when a finger contacts a surface, enabling peg-in-hole insertion, bolt tightening, and deformable object handling without external F/T sensors.

Bimanual Systems

Run two Allegro Hands — one on each arm — for bimanual manipulation experiments. Both hands register in the same SVRC platform session for synchronized data capture.

Community

Have a question about Allegro Hand integration or want to share your setup?

SVRC Forum → Book a Call →

Pricing & Availability

Purchase
From $15,000
Buy in SVRC Store
Demo / Questions
30-min call
Book on Calendly

Ships worldwide from Mountain View, CA. Pricing is for right or left hand unit. Adapter plates, cables, and integration services quoted separately. Talk to our team.

Need richer sensing?

Wuji Hand — 20-DOF with 768-point Tactile Map

5 fingers, 20 DOF, 768-taxel pressure grid, 6-axis IMU, EMF sensing, and 30 Hz JSONL streaming. Purpose-built for teleoperation data collection.

Wuji Hand Details Compare Hands

Ready to Add Dexterity to Your Robot?

Allegro Hand — 16 DOF, torque sensing, ROS2-ready. From $15,000 at SVRC.