About
Description
Unitree Robotics
Unitree R1-A7-D Ultimate H
Mobile-Base Dual-Arm Humanoid Robot with Tendon-Driven Dexterous Hands, 7-DOF Arms & Onboard Orin NX 100 TOPS. 20 DOF, 7-DOF Arms, MID360 LiDAR. The biomimetic dexterity platform for research-grade manipulation.
20
Total DOF
7+7
Arm DOF
120
Nm Torque
~2
kg Payload
100
TOPS AI
What's in This Configuration
The R1-A7-D Ultimate H is the biomimetic dexterity platform for research-grade manipulation. It pairs the R1-A7-D autonomous mobile base — complete with 7-DOF arms, MID360 LiDAR, stereo vision, and a removable chassis battery — with tendon-driven dexterous hands and the powerful NVIDIA Jetson Orin NX 100 TOPS compute module. With biomimetic cable-actuated architecture, antagonistic tendon routing, passive compliance via elastic coupling, and variable stiffness through co-contraction, the Ultimate H delivers human-like adaptive grasping without rigid gear-driven constraints. The 2.5× AI compute leap over the Smart series enables larger models, denser policies, and onboard training for the most demanding biomimetic manipulation research.
✓ R1-A7-D Mobile Base Body
Full upper-body humanoid on a wheeled autonomous chassis: dual 7-DOF arms, 1-DOF waist, 2-DOF head, 1-DOF pillar, 2-DOF base. 683×520×440 mm collapsed, ~32 kg with battery.
✓ Tendon-Driven Dexterous Hands
Biomimetic cable-actuated dexterous hands with antagonistic tendon routing, passive compliance, and variable stiffness. Pre-installed on both arms.
✓ MID360 LiDAR
360° surround LiDAR integrated into the mobile chassis for real-time SLAM, obstacle avoidance, and autonomous navigation.
✓ Orin NX 100 TOPS
Built-in NVIDIA Jetson Orin NX delivering 100 TOPS for processing stereo vision and tendon-driven hand control policies in real time.
✓ Full-Stack Open SDK
Linux framework, ROS 2 native, open APIs for arms, base, audio, vision, lighting. Full tendon-driven hand control API. Dual encoders per joint.
Tendon-Driven Dexterous Hands — Biomimetic Cable-Actuated Dexterity
The tendon-driven dexterous hands on the Ultimate H replicate the biomechanical architecture of the human hand through cable-actuated antagonistic routing. Unlike gear-driven or motor-in-joint designs, tendon transmission enables remote actuation — motors reside in the forearm, driving fingers via flexible cables — resulting in lighter, more compact end effectors with human-like finger proportions. The antagonistic tendon routing mimics the extensor-flexor pairs in human anatomy, enabling natural finger curling and extension. Passive compliance via elastic tendon coupling allows fingers to adapt to object geometry without active control — the hand passively conforms to irregular shapes, enabling robust grasps without precise geometric models. Variable stiffness via co-contraction lets the hand modulate rigidity from fully compliant (for fragile objects) to stiff (for forceful manipulation) by adjusting the tension balance between antagonistic cable pairs. This biomimetic approach yields grasps that are inherently adaptive, energy-efficient, and safe for human interaction — a fundamentally different paradigm from rigid robotic grippers.
Biomimetic Grasping — Human-Like Compliance & Dexterity
Ultimate H leverages biomimetic compliance to achieve grasps that rival human adaptability. The passive compliance via elastic tendon coupling means fingers automatically distribute contact forces across irregular surfaces — no explicit force control required. This enables power grips that envelop objects of unknown shape, precision pinches with delicate force modulation for small or fragile items, tripod grasps for spherical objects, lateral holds for thin or flat items, and in-hand manipulation with passive adaptation to changing contact geometry. The variable stiffness via co-contraction adds an active dimension: increasing antagonistic tension stiffens the hand for forceful tasks, while reducing tension maximises compliance for safe human-robot interaction. Unlike rigid gear-driven hands that require precise models and force feedback loops, the tendon-driven architecture achieves robust, adaptive grasping through mechanical intelligence — making the Ultimate H ideal for research in compliant manipulation, human-robot interaction, and embodied AI where biological plausibility matters.
Stereo Vision Perception — Pure Visual Manipulation
Ultimate H adopts a pure visual manipulation approach: the head-mounted binocular camera provides the sole sensing modality for grasp planning and execution. Without wrist cameras or tactile sensors, the robot relies on stereo vision alone to estimate object pose, plan biomimetic grasps, and monitor execution quality. This minimal-sensing maximum-dexterity philosophy challenges the robot to extract maximal information from vision — driving research in dense depth estimation, visual grasp affordance prediction, and vision-guided compliant grasping. The 100 TOPS Orin NX runs larger vision backbones for more accurate pose estimation and grasp quality prediction, with compute headroom for sim-to-real policy transfer and onboard reinforcement learning. The open vision API provides timestamp-synchronised stereo streams, supporting cutting-edge research in pure visual manipulation, vision-based embodied AI, and imitation learning from human videos where only visual demonstrations are available.
7-DOF Arms — Human-Like Redundancy & Reach
The R1-A7-D features seven degrees of freedom per arm — matching the kinematic complexity of the human upper limb. With a 555 mm reach (forearm + upper arm) and 120 Nm peak shoulder torque, these arms can reach around obstacles, maintain orientation while translating the end effector, and execute complex trajectories impossible for simpler 5- or 6-DOF manipulators. Crossed-roller and double-row ball bearings on every joint ensure smooth, precise motion with minimal backlash. Dual encoders per joint provide closed-loop accuracy for research-grade repeatability.
Orin NX 100 TOPS — 2.5× AI Performance for Biomimetic Research
The built-in NVIDIA Jetson Orin NX delivers 100 TOPS of AI compute — a 2.5× leap over the Smart series' 40 TOPS. For the Ultimate H, this extra headroom unlocks larger vision backbones for dense grasp affordance prediction and object pose estimation from stereo alone, onboard training of biomimetic control policies using reinforcement learning with tendon-driven dynamics, higher frame rates for visual servoing with compliant hands, and concurrent vision-language model inference for natural-language task specification. The Orin NX also accelerates dense depth estimation, visual-LiDAR fusion for navigation, and sim-to-real transfer — all while maintaining low-latency control loops for the coupled tendon dynamics of both hands. Pre-installed and thermally integrated with the chassis cooling system, the module runs entirely onboard with no external GPU server or cloud dependency.
Autonomous Mobile Base — Navigate, Then Manipulate
The R1-A7-D chassis adds 3 degrees of freedom to the upper body: a telescopic pillar for height adjustment and a 2-DOF wheeled base for omnidirectional navigation. Combined with the MID360 LiDAR and stereo camera head, the robot builds real-time maps, plans paths, and avoids obstacles autonomously — then stops precisely to manipulate. The removable lithium-ion battery in the chassis delivers approximately 1.5 hours of untethered operation. Please note: the upper body of this model does not accept a battery.
LiDAR + Stereo Vision — Complete Perception Stack
Ultimate H processes data from two complementary sensors: a MID360 360° LiDAR for metrically accurate mapping and long-range obstacle detection, and a binocular camera head (146°×124° FOV, 60 mm baseline) for colour vision, object recognition, and depth estimation. Fused on the Orin NX 100 TOPS, this dual-modality perception stack supports research in visual-LiDAR SLAM, semantic navigation, and active perception for mobile manipulation — with enough headroom to also run larger vision models for pure visual grasp planning.
Binocular Vision — Pure Visual Manipulation
A binocular camera head delivers a 146° × 124° field of view with a 60 mm stereo baseline for scene-level understanding. RGB streams at 1280×720@30 Hz and depth at 544×448@10 Hz. The R1-A7-D Ultimate H relies on stereo vision alone for all manipulation tasks — no wrist cameras, no tactile sensors — driving research in pure visual manipulation, vision-guided compliant grasping, and minimal-sensing maximum-dexterity approaches. Combined with the MID360 LiDAR, the robot offers a complete perception stack for research in vision-based embodied AI and biomimetic mobile manipulation — all processed on the 100 TOPS Orin NX.
Voice Collaboration & Natural Interaction
Four-array beamforming microphones and dual 3 W speakers enable clear voice capture and synthesis in noisy labs, warehouses, or classrooms. The open audio framework supports custom wake words, multilingual TTS, and integration with LLMs for conversational command-and-control — even while the base is in motion. The extra compute headroom of the Orin NX 100 TOPS allows running larger speech recognition and language models locally alongside biomimetic manipulation tasks.
Collapsible. Mobile. Ready to Deploy.
At 683 × 520 × 440 mm collapsed (1323 mm elevated), the R1-A7-D Ultimate H navigates standard doorways, elevators, and lab aisles. The telescopic pillar lowers the upper body for transport and raises it for manipulation. At approximately 32 kg with battery, it is a self-contained mobile research platform — no external compute rack, no tether cables during operation.
Full-Stack Open Development
Unitree provides a mature, Linux-based robot development framework with open APIs for the underlying system, robotic arms, mobile base, audio, lighting, and visual control. ROS 2 native support, dual encoders on every arm joint, hollow internal wiring, and low-inertia, high-speed inner-rotor permanent magnet synchronous motors make the R1-A7-D Ultimate H a clean slate for reinforcement learning, biomimetic manipulation research, tendon-driven control, embodied AI, and autonomous navigation studies. The built-in Orin NX 100 TOPS handles perception, planning, vision processing, and large model inference locally without external hardware, with enough headroom for the most demanding sim-to-real and imitation learning scenarios with compliant hands.
Upgrade Path — Grow With Your Research
Ultimate H is a biomimetic dexterity platform. Expand capability as your research evolves:
- End Effectors — Swap to Dex1-1 V2 for simple parallel gripping, Dex3-1 for three-finger adaptive manipulation, BrainCo Revo 2 for anthropomorphic dexterity, or LinkerBot O6 for high-DOF gear-driven precision.
- Wrist Cameras — Add wrist-mounted RGB cameras for close-up visual feedback of tendon-driven grasp execution and visual servoing.
- Tactile Sensing — Integrate force-sensitive arrays for contact-aware biomimetic manipulation and multi-modal sensorimotor learning.
- Large VLM — Deploy vision-language models for natural-language task specification and scene understanding. The 100 TOPS Orin NX has the headroom to run the largest VLMs entirely onboard alongside biomimetic control.
Designed For
Tendon-Driven Mechanism Research
Study cable-actuated dynamics, antagonistic control, variable stiffness, and passive compliance in a full mobile manipulation platform.
Compliant Manipulation
Adaptive grasping of irregular, fragile, or deformable objects through mechanical intelligence and elastic coupling.
Human-Robot Interaction
Safe contact-rich interaction with inherently compliant hands, physical handover, and shared workspace collaboration.
Embodied AI & Imitation Learning
Learning biomimetic manipulation policies from human demonstrations, sim-to-real transfer, and reinforcement learning on 100 TOPS.
Vision-Guided Manipulation
Pure visual manipulation research — dense grasp affordance prediction, pose estimation, and vision-guided compliant grasping without extra sensors.
Advanced University Research
Cutting-edge research in biomimetic robotics, tendon-driven control, and mobile sensorimotor intelligence with full open-source APIs.
Technical Specifications
What's in the Box
- Unitree R1-A7-D Ultimate H Humanoid Robot (mobile base)
- Tendon-driven dexterous hands (pre-installed, both arms)
- NVIDIA Jetson Orin NX 100 TOPS compute module (pre-installed)
- Removable chassis lithium-ion battery pack
- External power supply & AC cable
- MID360 LiDAR module (pre-installed in chassis)
- Binocular vision head module (pre-installed)
- 4-array microphone + dual speaker module (pre-installed)
- Standard dual-hand remote controllers
- Quick-start guide & safety manual
- Secondary development documentation access
R1-A7-D Configuration Guide
Every configuration shares the same R1-A7-D mobile base with 7-DOF arms — the difference is the compute module and end effector. Choose the one that matches your research goals:
* Specifications are based on manufacturer data and may vary slightly by production batch. Optional wrist cameras, tactile sensors, compute modules, and alternative end effectors sold separately.
Unitree® is a registered trademark of Unitree Robotics. Images are for reference; actual product appearance may differ.
Users must maintain a sufficient safety distance. This is a civilian robot — comply with local laws and regulations.
Description
Unitree Robotics
Unitree R1-A7-D Ultimate H
Mobile-Base Dual-Arm Humanoid Robot with Tendon-Driven Dexterous Hands, 7-DOF Arms & Onboard Orin NX 100 TOPS. 20 DOF, 7-DOF Arms, MID360 LiDAR. The biomimetic dexterity platform for research-grade manipulation.
20
Total DOF
7+7
Arm DOF
120
Nm Torque
~2
kg Payload
100
TOPS AI
What's in This Configuration
The R1-A7-D Ultimate H is the biomimetic dexterity platform for research-grade manipulation. It pairs the R1-A7-D autonomous mobile base — complete with 7-DOF arms, MID360 LiDAR, stereo vision, and a removable chassis battery — with tendon-driven dexterous hands and the powerful NVIDIA Jetson Orin NX 100 TOPS compute module. With biomimetic cable-actuated architecture, antagonistic tendon routing, passive compliance via elastic coupling, and variable stiffness through co-contraction, the Ultimate H delivers human-like adaptive grasping without rigid gear-driven constraints. The 2.5× AI compute leap over the Smart series enables larger models, denser policies, and onboard training for the most demanding biomimetic manipulation research.
✓ R1-A7-D Mobile Base Body
Full upper-body humanoid on a wheeled autonomous chassis: dual 7-DOF arms, 1-DOF waist, 2-DOF head, 1-DOF pillar, 2-DOF base. 683×520×440 mm collapsed, ~32 kg with battery.
✓ Tendon-Driven Dexterous Hands
Biomimetic cable-actuated dexterous hands with antagonistic tendon routing, passive compliance, and variable stiffness. Pre-installed on both arms.
✓ MID360 LiDAR
360° surround LiDAR integrated into the mobile chassis for real-time SLAM, obstacle avoidance, and autonomous navigation.
✓ Orin NX 100 TOPS
Built-in NVIDIA Jetson Orin NX delivering 100 TOPS for processing stereo vision and tendon-driven hand control policies in real time.
✓ Full-Stack Open SDK
Linux framework, ROS 2 native, open APIs for arms, base, audio, vision, lighting. Full tendon-driven hand control API. Dual encoders per joint.
Tendon-Driven Dexterous Hands — Biomimetic Cable-Actuated Dexterity
The tendon-driven dexterous hands on the Ultimate H replicate the biomechanical architecture of the human hand through cable-actuated antagonistic routing. Unlike gear-driven or motor-in-joint designs, tendon transmission enables remote actuation — motors reside in the forearm, driving fingers via flexible cables — resulting in lighter, more compact end effectors with human-like finger proportions. The antagonistic tendon routing mimics the extensor-flexor pairs in human anatomy, enabling natural finger curling and extension. Passive compliance via elastic tendon coupling allows fingers to adapt to object geometry without active control — the hand passively conforms to irregular shapes, enabling robust grasps without precise geometric models. Variable stiffness via co-contraction lets the hand modulate rigidity from fully compliant (for fragile objects) to stiff (for forceful manipulation) by adjusting the tension balance between antagonistic cable pairs. This biomimetic approach yields grasps that are inherently adaptive, energy-efficient, and safe for human interaction — a fundamentally different paradigm from rigid robotic grippers.
Biomimetic Grasping — Human-Like Compliance & Dexterity
Ultimate H leverages biomimetic compliance to achieve grasps that rival human adaptability. The passive compliance via elastic tendon coupling means fingers automatically distribute contact forces across irregular surfaces — no explicit force control required. This enables power grips that envelop objects of unknown shape, precision pinches with delicate force modulation for small or fragile items, tripod grasps for spherical objects, lateral holds for thin or flat items, and in-hand manipulation with passive adaptation to changing contact geometry. The variable stiffness via co-contraction adds an active dimension: increasing antagonistic tension stiffens the hand for forceful tasks, while reducing tension maximises compliance for safe human-robot interaction. Unlike rigid gear-driven hands that require precise models and force feedback loops, the tendon-driven architecture achieves robust, adaptive grasping through mechanical intelligence — making the Ultimate H ideal for research in compliant manipulation, human-robot interaction, and embodied AI where biological plausibility matters.
Stereo Vision Perception — Pure Visual Manipulation
Ultimate H adopts a pure visual manipulation approach: the head-mounted binocular camera provides the sole sensing modality for grasp planning and execution. Without wrist cameras or tactile sensors, the robot relies on stereo vision alone to estimate object pose, plan biomimetic grasps, and monitor execution quality. This minimal-sensing maximum-dexterity philosophy challenges the robot to extract maximal information from vision — driving research in dense depth estimation, visual grasp affordance prediction, and vision-guided compliant grasping. The 100 TOPS Orin NX runs larger vision backbones for more accurate pose estimation and grasp quality prediction, with compute headroom for sim-to-real policy transfer and onboard reinforcement learning. The open vision API provides timestamp-synchronised stereo streams, supporting cutting-edge research in pure visual manipulation, vision-based embodied AI, and imitation learning from human videos where only visual demonstrations are available.
7-DOF Arms — Human-Like Redundancy & Reach
The R1-A7-D features seven degrees of freedom per arm — matching the kinematic complexity of the human upper limb. With a 555 mm reach (forearm + upper arm) and 120 Nm peak shoulder torque, these arms can reach around obstacles, maintain orientation while translating the end effector, and execute complex trajectories impossible for simpler 5- or 6-DOF manipulators. Crossed-roller and double-row ball bearings on every joint ensure smooth, precise motion with minimal backlash. Dual encoders per joint provide closed-loop accuracy for research-grade repeatability.
Orin NX 100 TOPS — 2.5× AI Performance for Biomimetic Research
The built-in NVIDIA Jetson Orin NX delivers 100 TOPS of AI compute — a 2.5× leap over the Smart series' 40 TOPS. For the Ultimate H, this extra headroom unlocks larger vision backbones for dense grasp affordance prediction and object pose estimation from stereo alone, onboard training of biomimetic control policies using reinforcement learning with tendon-driven dynamics, higher frame rates for visual servoing with compliant hands, and concurrent vision-language model inference for natural-language task specification. The Orin NX also accelerates dense depth estimation, visual-LiDAR fusion for navigation, and sim-to-real transfer — all while maintaining low-latency control loops for the coupled tendon dynamics of both hands. Pre-installed and thermally integrated with the chassis cooling system, the module runs entirely onboard with no external GPU server or cloud dependency.
Autonomous Mobile Base — Navigate, Then Manipulate
The R1-A7-D chassis adds 3 degrees of freedom to the upper body: a telescopic pillar for height adjustment and a 2-DOF wheeled base for omnidirectional navigation. Combined with the MID360 LiDAR and stereo camera head, the robot builds real-time maps, plans paths, and avoids obstacles autonomously — then stops precisely to manipulate. The removable lithium-ion battery in the chassis delivers approximately 1.5 hours of untethered operation. Please note: the upper body of this model does not accept a battery.
LiDAR + Stereo Vision — Complete Perception Stack
Ultimate H processes data from two complementary sensors: a MID360 360° LiDAR for metrically accurate mapping and long-range obstacle detection, and a binocular camera head (146°×124° FOV, 60 mm baseline) for colour vision, object recognition, and depth estimation. Fused on the Orin NX 100 TOPS, this dual-modality perception stack supports research in visual-LiDAR SLAM, semantic navigation, and active perception for mobile manipulation — with enough headroom to also run larger vision models for pure visual grasp planning.
Binocular Vision — Pure Visual Manipulation
A binocular camera head delivers a 146° × 124° field of view with a 60 mm stereo baseline for scene-level understanding. RGB streams at 1280×720@30 Hz and depth at 544×448@10 Hz. The R1-A7-D Ultimate H relies on stereo vision alone for all manipulation tasks — no wrist cameras, no tactile sensors — driving research in pure visual manipulation, vision-guided compliant grasping, and minimal-sensing maximum-dexterity approaches. Combined with the MID360 LiDAR, the robot offers a complete perception stack for research in vision-based embodied AI and biomimetic mobile manipulation — all processed on the 100 TOPS Orin NX.
Voice Collaboration & Natural Interaction
Four-array beamforming microphones and dual 3 W speakers enable clear voice capture and synthesis in noisy labs, warehouses, or classrooms. The open audio framework supports custom wake words, multilingual TTS, and integration with LLMs for conversational command-and-control — even while the base is in motion. The extra compute headroom of the Orin NX 100 TOPS allows running larger speech recognition and language models locally alongside biomimetic manipulation tasks.
Collapsible. Mobile. Ready to Deploy.
At 683 × 520 × 440 mm collapsed (1323 mm elevated), the R1-A7-D Ultimate H navigates standard doorways, elevators, and lab aisles. The telescopic pillar lowers the upper body for transport and raises it for manipulation. At approximately 32 kg with battery, it is a self-contained mobile research platform — no external compute rack, no tether cables during operation.
Full-Stack Open Development
Unitree provides a mature, Linux-based robot development framework with open APIs for the underlying system, robotic arms, mobile base, audio, lighting, and visual control. ROS 2 native support, dual encoders on every arm joint, hollow internal wiring, and low-inertia, high-speed inner-rotor permanent magnet synchronous motors make the R1-A7-D Ultimate H a clean slate for reinforcement learning, biomimetic manipulation research, tendon-driven control, embodied AI, and autonomous navigation studies. The built-in Orin NX 100 TOPS handles perception, planning, vision processing, and large model inference locally without external hardware, with enough headroom for the most demanding sim-to-real and imitation learning scenarios with compliant hands.
Upgrade Path — Grow With Your Research
Ultimate H is a biomimetic dexterity platform. Expand capability as your research evolves:
- End Effectors — Swap to Dex1-1 V2 for simple parallel gripping, Dex3-1 for three-finger adaptive manipulation, BrainCo Revo 2 for anthropomorphic dexterity, or LinkerBot O6 for high-DOF gear-driven precision.
- Wrist Cameras — Add wrist-mounted RGB cameras for close-up visual feedback of tendon-driven grasp execution and visual servoing.
- Tactile Sensing — Integrate force-sensitive arrays for contact-aware biomimetic manipulation and multi-modal sensorimotor learning.
- Large VLM — Deploy vision-language models for natural-language task specification and scene understanding. The 100 TOPS Orin NX has the headroom to run the largest VLMs entirely onboard alongside biomimetic control.
Designed For
Tendon-Driven Mechanism Research
Study cable-actuated dynamics, antagonistic control, variable stiffness, and passive compliance in a full mobile manipulation platform.
Compliant Manipulation
Adaptive grasping of irregular, fragile, or deformable objects through mechanical intelligence and elastic coupling.
Human-Robot Interaction
Safe contact-rich interaction with inherently compliant hands, physical handover, and shared workspace collaboration.
Embodied AI & Imitation Learning
Learning biomimetic manipulation policies from human demonstrations, sim-to-real transfer, and reinforcement learning on 100 TOPS.
Vision-Guided Manipulation
Pure visual manipulation research — dense grasp affordance prediction, pose estimation, and vision-guided compliant grasping without extra sensors.
Advanced University Research
Cutting-edge research in biomimetic robotics, tendon-driven control, and mobile sensorimotor intelligence with full open-source APIs.
Technical Specifications
What's in the Box
- Unitree R1-A7-D Ultimate H Humanoid Robot (mobile base)
- Tendon-driven dexterous hands (pre-installed, both arms)
- NVIDIA Jetson Orin NX 100 TOPS compute module (pre-installed)
- Removable chassis lithium-ion battery pack
- External power supply & AC cable
- MID360 LiDAR module (pre-installed in chassis)
- Binocular vision head module (pre-installed)
- 4-array microphone + dual speaker module (pre-installed)
- Standard dual-hand remote controllers
- Quick-start guide & safety manual
- Secondary development documentation access
R1-A7-D Configuration Guide
Every configuration shares the same R1-A7-D mobile base with 7-DOF arms — the difference is the compute module and end effector. Choose the one that matches your research goals:
* Specifications are based on manufacturer data and may vary slightly by production batch. Optional wrist cameras, tactile sensors, compute modules, and alternative end effectors sold separately.
Unitree® is a registered trademark of Unitree Robotics. Images are for reference; actual product appearance may differ.
Users must maintain a sufficient safety distance. This is a civilian robot — comply with local laws and regulations.
Unitree
by
Unitree
Unitree R1-A7-D Ultimate H Humanoid Robot
SKU:
R1-A7-D-Ultimate-H
- Regular price
- $14,690.00
- Sale price
- $14,690.00
- Regular price
-
Shipping calculated at checkout.
Providing the best new Tech Experience, Energy Independence and Robotics Ways of Optimisation
Providing the best new Tech Experience, Energy Independence and Robotics Ways of Optimisation
Packaging note:
Equipment is being supplied in a manufacturer packaging
