Unitree Dex5-1P Tactile Hand For H1-H1-2 (Left) (DEX5-HANDL)

Dex5-1P Tactile Hand For H1/H1-2 (Left) B2 Series Accessories (Dex5-Handl) (DEX5-HANDL)

The Dex5-1P represents a significant advancement over earlier Unitree dexterous hands: it combines 20 total degrees of freedom (16 active joints and 4 passive joints), 94 tactile sensors distributed across the palm and fingertips, 1,000 Hz real-time communication providing position, torque, stiffness, and IMU data, and fully backdrivable joints that eliminate the rigid "stiff hand" feel common in gear-heavy dexterous hands. The hand weighs approximately 1,000 grams and measures 217.3 × 127.5 × 72.1 mm, with a payload capacity of 3.5 kilograms when grasping palm-down or 4.5 kilograms when grasping palm-left. The tactile sensors can detect pressure from as low as 10 grams to as high as 2.5 kilograms — a range spanning from the delicate contact of a thin paper sheet through the full grip force needed for sustained object transport. The hand operates at 24 to 60 VDC and includes a secondary algorithm support layer that grants researchers direct programmatic access to low-level sensor and motor data.

Five-Finger Anthropomorphic Layout

The Dex5-1P follows a human-inspired five-finger layout with an independently actuated thumb, index, middle, ring, and little finger. The thumb incorporates opposition kinematics enabling it to rotate across the palm to face the fingers — the mechanical prerequisite for precision pinch, lateral key grips, and the full complement of human grasping postures.

Ghostysky's technical review confirms: "20 Degrees of Freedom (DoF): The hand features 16 active joints and 4 passive joints. This includes a 4-DoF thumb and four fingers with lateral swing (±22°), allowing the hand to conform perfectly to curved or irregular surfaces."

Four-Finger Lateral Swing: ±22°

One of the most mechanically significant features of the Dex5-1P is the four-finger lateral swing capability at ±22°. Conventional robotic fingers flex only in the sagittal plane (opening and closing) — the Dex5-1P's fingers also abduct and adduct laterally, enabling the hand to spread its fingers wider for larger object grasp envelopes and to converge them for concentrated precision contact. The ±22° lateral range allows the fingers to conform to the curved surfaces of cylindrical, spherical, and irregularly shaped objects without requiring the whole hand to reorient.

This lateral swing is what enables the hand to wrap around a wine glass, conform to the shape of a tool handle, or spread for a stable grasp on a flat panel — grasping configurations that are impossible for hands with only flexion-extension finger DOF.

Replaceable Fingers: Field Maintenance and Customization

RobotShop's specification confirms: "20 Degrees of Freedom with Independent Finger Replacement." Each of the five fingers is independently replaceable — a design feature that reduces maintenance costs and downtime in research and deployment contexts where finger wear or damage from manipulation research would otherwise require complete hand replacement.

The replaceable finger design also enables task-specific customization: fingertips can be swapped for different material properties (high-friction pads for slippery objects, soft compliance pads for fragile items) without replacing the complete hand assembly. This modularity extends the Dex5-1P's effective lifespan and application versatility beyond what fixed-finger designs allow.

Ultra-Small Joint Gap Design

The Dex5-1P features an ultra-small joint gap design that prevents objects, clothing, or other materials from becoming caught in the finger joints during manipulation — a practical safety and reliability feature for deployment in human-shared environments. Robotic hands with larger joint gaps can catch human hair, thin fabric, or small components in their mechanisms, creating both safety risks and mechanical damage. The minimized joint gaps eliminate this failure mode.

Dimensions and Weight

Futurology.tech confirms the Dex5-1P dimensions as 217.3 × 127.5 × 72.1 mm with the note that "the final shipped version may vary." At approximately 1,000 grams, the hand adds 1 kilogram to each arm's distal mass — an increase over lighter grippers but consistent with the design expectations for a hand with 20 DOF, 94 sensors, and 12 self-developed micro force-controlled composite transmission joints.

Technology and Specifications

Full Dex5-1P Specifications

94 Tactile Sensors: Spatial Contact Mapping

Ghostysky confirms: "Equipped with 94 tactile pressure sensors distributed across the palm and fingertips. It can perceive pressure from as little as 10g up to 2.5kg, providing the sensory feedback necessary for Reinforcement Learning (RL) and AI-driven tasks."

With 94 sensors distributed across the five fingers and palm, the Dex5-1P provides a high-resolution spatial contact map of the full hand surface. At approximately 18 to 19 sensors per finger on average (though the distribution is concentrated at higher-contact-rate surfaces like fingertips and proximal phalanges), the sensor density enables:

Grasp quality assessment: The spatial pattern of active sensors reveals whether contact is well-distributed (stable grasp) or concentrated at edge contacts (unstable, slip-prone grasp).

Slip detection: The progressive change in the spatial contact pattern as an object begins to move relative to the hand — detectable milliseconds before the object actually slips free.

Object geometry inference: The shape of the contact patch on the palm and finger surfaces encodes information about the object's local geometry at each contact zone — enabling shape-from-touch estimation that supplements visual shape estimation.

Force distribution balancing: During grasping, the distribution of forces across the 94 sensors can be used to balance grip force across the hand, preventing stress concentration at any single contact point that could damage fragile objects.

1,000 Hz Communication Rate: Real-Time Tactile Control

RobotShop's specification confirms "1000Hz communication rate" providing "real-time perceptual feedback — including position, torque, stiffness, and IMU data."

The 1,000 Hz communication rate means the hand provides 1,000 complete sensor state updates per second — one every millisecond. For comparison, the fastest human reflexive responses to unexpected contact take approximately 50 to 80 milliseconds. At 1,000 Hz, the Dex5-1P's control system receives 50 to 80 sensor state updates within the same time window as a single human reflex response — enabling closed-loop control algorithms to detect and respond to contact events at a rate that is essentially imperceptible at human observation time scales.

This communication rate is what enables reinforcement learning policies to use tactile data as part of their observation at real-time interaction speeds — the policy receives updated tactile observations every millisecond, enabling truly reactive tactile manipulation behavior rather than merely detecting contact events after the fact.

Hollow-Cup Motors and Micro-Force Transmission

Blue Skies Drone Shop's product description confirms: "Hollow-Cup Motors and Micro-Force Transmission achieve high power density, ultra-low joint damping, and lightning-fast reflexes."

Hollow-cup motors (also called coreless motors) eliminate the iron core from the motor's rotor, substantially reducing the rotor mass and inertia. Lower rotor inertia means the motor can accelerate and decelerate faster — enabling the finger's reflexive responses to contact events that occur at millisecond time scales. The "lightning-fast reflexes" characterization refers to the ability to transition from a closing motion to a compliant hold when contact is detected, or to tighten grip instantaneously when slip begins.

RobotShop also confirms the 12 self-developed micro force-controlled composite transmission joints that "enable robot to achieve tactile proprioception" — a term indicating that the transmission joints are designed to sense their own force state (proprioception) rather than relying only on dedicated force sensors for contact feedback.

Backdrivability: The Anti-Stiff-Hand Design

OpenELAB's Dex5-1 listing notes: "With 16 active joints + 4 passive, Dex5-1 mimics human hand dynamics with remarkable precision. Each finger joint supports smooth backdrivability, eliminating the 'stiff hand' effect and enabling fluid motion — perfect for reinforcement learning (RL) and adaptive control."

Backdrivability is the property that allows external forces to move the joint without the motor actively resisting — the motor can be back-driven by contact forces, yielding to object geometry rather than rigidly maintaining position. This is how human finger joints work: when you press against an object, your finger yields while maintaining a regulated contact force, rather than rigidly maintaining its commanded angle.

For robotic manipulation, backdrivability is essential for two capabilities. First, safe physical human-robot contact: a backdrivable hand yields when it contacts a human, preventing the rigid impact forces that non-backdrivable hands impose. Second, robust grasping of irregular objects: a backdrivable hand conforms to object geometry naturally during closing, rather than requiring the control algorithm to perfectly predict the contact angle for every object shape.

Applications and Use Cases

Reinforcement Learning and Imitation Learning Research

OpenELAB's research guide states: "The 1000Hz communication rate ensures that control algorithms receive timely sensory updates, enabling closed-loop control strategies that react dynamically to changing contact conditions." For RL research on dexterous manipulation — learning policies that use tactile sensor readings as observations — the Dex5-1P's sensor density, communication rate, and secondary algorithm access provide the data quality and latency that reinforcement learning control loops require.

Precision Object Handling and Fragile Grasping

Ghostysky describes service robotics applications: "Tasks requiring 'soft touch,' such as folding clothes, handling glassware, or opening doors." The 10-gram minimum pressure detection and full backdrivability enable the Dex5-1P to handle objects that impose strict force limits — glass containers, pharmaceutical packaging, fragile electronics components, and biological specimens — with the contact sensitivity needed to prevent damage.

Dexterous Manipulation Demonstrations

The hand has been demonstrated performing several landmark dexterous manipulation tasks that have attracted significant attention in the robotics community: solving a Rubik's cube through coordinated in-hand manipulation, handling fragile glass objects with regulated grip force, and performing precise finger coordination tasks that demonstrate the 20-DOF kinematic range.

Ghostysky notes: "Engineered for the Unitree H2 and H1 series, it enables robots to perform delicate tasks — from solving Rubik's cubes to handling fragile objects — with a level of agility that mimics the human hand."

Human-Robot Interaction and Physical Collaboration

The backdrivable joints, 94 tactile sensors, and 1,000 Hz real-time feedback make the Dex5-1P suitable for HRI research involving physical contact — handshakes, object handovers, collaborative manipulation, and physical guidance. The hand can sense human touch at the 10-gram detection threshold and regulate its response forces to match the compliance of a human hand grip.

Advanced Robotics Training and STEM Education

Blue Skies Drone Shop's product description identifies: "Whether you're developing advanced RL algorithms, training tomorrow's robotics talent, or tackling precision assembly and handling, the Dex5-1P brings unmatched tactile control and adaptability to your H1 platform."

Advantages and Benefits

94 Tactile Sensors for Research-Grade Contact Mapping: The 94-sensor distribution provides spatial resolution far exceeding most commercial robotic hand tactile systems, enabling the contact quality data that manipulation AI research requires for generalizable policy development.

1,000 Hz Real-Time Feedback for Closed-Loop Tactile Control: The millisecond update rate enables control algorithms that react to contact events at speeds imperceptible to human observers — essential for the reactive manipulation quality that dexterous tasks require.

Full Backdrivability for Natural Contact Compliance: Backdrivable joints provide human-like contact compliance that eliminates the rigid "stiff hand" feel of gear-heavy designs, enabling both safe physical HRI and robust grasping of irregularly shaped objects.

Replaceable Individual Fingers for Reduced Maintenance Cost: Per-finger replacement capability substantially reduces the cost of maintaining the Dex5-1P over its operational life — individual finger wear or damage requires replacing only the affected finger, not the complete hand assembly.

Secondary Algorithm Access for Research Customization: Full low-level access to sensor and motor data enables researchers to implement custom control architectures, novel tactile processing algorithms, and specialized manipulation policies without being limited to Unitree's default control layer.

10g to 2.5 kg Pressure Range for Delicate and Robust Grasping: The 250:1 pressure sensing range spans from barely-detectable surface touch through firm functional grasps — enabling a single hand to address the full spectrum of object compliance in research and service contexts.

Comparison: Dex5-1P vs. Dex3-1 for H1/H1-2

The Dex5-1P provides substantially richer tactile sensing, higher DOF, and more sophisticated mechanical design than the Dex3-1 — reflecting its targeting of the H1/H1-2 full-scale research humanoid rather than the compact G1 platform. The 94 sensors versus the Dex3-1's optional arrays and the 20 DOF versus 7 DOF represent qualitatively different capability tiers.

Frequently Asked Questions (FAQ)

What is the Unitree Dex5-1P Tactile Hand (Left) — DEX5-HANDL? The Unitree Dex5-1P Tactile Hand (Left), product code DEX5-HANDL, is the left-handed five-finger dexterous end-effector designed for the Unitree H1 and H1-2 humanoid robots. It features 20 total degrees of freedom (16 active, 4 passive), 94 tactile pressure sensors per hand detecting 10g to 2.5 kg, 1,000 Hz real-time perceptual feedback covering position, velocity, torque, IMU, and pressure data, fully backdrivable joints eliminating the "stiff hand" effect, four-finger lateral swing of ±22°, individually replaceable fingers, hollow-cup motors with micro-force transmission, and secondary algorithm access for full research-level sensor and motor control. Dimensions are 217.3 × 127.5 × 72.1 mm at approximately 1,000 grams.

What is the difference between the Dex5-1 (Left) and Dex5-1P (Right)? The Dex5-1 and Dex5-1P are the left-hand and right-hand variants of the same hand system, sharing identical specifications in all dimensions — DOF count, tactile sensor count, communication rate, payload capacity, dimensions, and weight. The only difference is handedness: the Dex5-1 is configured as a left hand (DEX5-HANDL) and the Dex5-1P is configured as a right hand (DEX5-HANDR) for bilateral installation on the H1 or H1-2. Both are required for bimanual manipulation research on the H1/H1-2 platform.

What makes the Dex5-1P different from other robotic hands? The Dex5-1P distinguishes itself through three simultaneous capabilities that are individually rare and together unique in the commercial humanoid accessory market: (1) 94 tactile sensors with 10g to 2.5 kg pressure detection across a single hand, (2) 1,000 Hz real-time communication providing perceptual feedback every millisecond, and (3) fully backdrivable joints across all 20 DOF, eliminating the stiff-hand contact behavior common in gear-heavy robotic hands. The individually replaceable fingers add a fourth practical advantage in field maintenance contexts.

Is the Dex5-1P compatible with robots other than the H1 and H1-2? The Dex5-1P is specifically optimized for the H1 and H1-2 with matching mechanical mounting, power requirements, and software integration. Ghostysky's review also notes compatibility with the Unitree H2 series. For the Unitree G1, Unitree recommends the Dex3-1 as the appropriate hand. OpenELAB notes that the USB 2.0 interface could theoretically interface with other systems, but the mechanical mounting and software integration are tailored for Unitree H1-series platforms.

Summary

The Unitree Dex5-1P Tactile Hand (Left) — DEX5-HANDL — is the most sensorially capable dexterous end-effector in Unitree's product lineup, combining 20 degrees of freedom (16 active, 4 passive), 94 tactile sensors per hand with 10-gram minimum pressure detection, 1,000 Hz real-time perceptual feedback, fully backdrivable joints, four-finger lateral swing of ±22°, individually replaceable fingers, hollow-cup motor actuation, and full secondary algorithm access for H1/H1-2 humanoid researchers. For research teams deploying the H1 or H1-2 on manipulation research, reinforcement learning, physical HRI, or precision dexterous task development, the Dex5-1P provides the tactile sensing density and mechanical compliance quality that makes the difference between a robot that can hold objects and one that can genuinely sense and respond to the contact physics of manipulation with human-like sensitivity.