About
Description
Unitree Robotics — Smart Series
Unitree R1-A7 Smart B
Upper-Body Dual-Arm Humanoid Robot with Dex3-1 Three-Finger Dexterous Hands & Wrist RGB Cameras. 17 DOF, 7-DOF precision arms, 555 mm reach, Orin Nano 40 TOPS, stereo vision & full-stack open SDK. For advanced visual manipulation research.
17
Total DOF
7+7
Arm DOF
555
mm Reach
~2
kg Payload
40
TOPS AI
What's in This Configuration
The R1-A7 Smart B steps up from entry-level grippers to three-finger dexterous manipulation with integrated wrist vision. It features the same R1-A7 upper-body base with 7-DOF arms, 1-DOF waist (±150°), and 2-DOF head as Smart A, but replaces the parallel grippers with Dex3-1 three-finger dexterous hands and adds wrist-mounted RGB cameras on each arm. The Dex3-1 hands provide adaptive three-finger grasping — tripod pinches, power grasps, and precision holds — that parallel grippers cannot achieve. Wrist cameras deliver a gripper's-eye view for visual servoing, grasp verification, and hand-eye coordination. Powered by the NVIDIA Jetson Orin Nano 40 TOPS, Smart B handles visual perception, dexterous grasp planning, and arm control on board. At ~13 kg, it is a compact yet capable platform for visual manipulation research.
✓ R1-A7 Upper-Body Humanoid
Dual 7-DOF arms, 1-DOF waist (±150°), 2-DOF head. 835×357×190 mm compact footprint, ~13 kg. Crossed-roller bearings, low-inertia inner-rotor motors, hollow internal cabling.
✓ Dex3-1 Three-Finger Dexterous Hands
Adaptive three-finger dexterous hands pre-installed on both arms for tripod pinches, power grasps, precision holds, and in-hand manipulation.
✓ Wrist-Mounted RGB Cameras
RGB cameras integrated into each wrist for close-up visual feedback, grasp verification, and hand-eye coordination during manipulation.
✓ Orin Nano 40 TOPS
Onboard NVIDIA Jetson Orin Nano delivering 40 TOPS for stereo-wrist visual fusion, grasp planning, and AI model inference.
✓ Voice Interaction
4-array beamforming mics + dual 3 W speakers. Wake-word, multilingual TTS, LLM integration ready.
✓ Full-Stack Open SDK
Linux framework, ROS 2 native, open APIs for arms, audio, vision, lighting. Dual encoders per joint.
Dex3-1 — Three-Finger Dexterity for Adaptive Grasping
The Dex3-1 is a three-finger dexterous hand that bridges the gap between simple parallel grippers and complex five-finger anthropomorphic designs. With three independently actuated fingers, it can form tripod pinches for spherical objects, power grasps for cylindrical items, and precision holds for small or fragile parts. The third finger provides stability and adaptability that two-finger grippers lack — the hand can envelop irregular objects, apply distributed contact forces, and maintain grasp security even when the object shifts. On the R1-A7's 7-DOF arms, the Dex3-1 enables bimanual coordination studies: one hand holds a workpiece while the other manipulates it, or both hands cooperate to lift and orient larger objects. The hand's compact design keeps mass low at the wrist, preserving the arm's dynamic response and precision.
Wrist-Mounted RGB Cameras — See From the Gripper
Smart B adds wrist-mounted RGB cameras to both arms — providing a gripper's-eye view of the workspace. These cameras capture close-up visual detail that the head-mounted stereo pair cannot: texture, surface geometry, and grasp point quality at the moment of contact. Combined with the Dex3-1's adaptability, wrist vision enables visual servoing for precision grasp alignment, grasp verification (did the fingers close on the target or miss?), and in-hand tracking (monitoring object pose during manipulation). The Orin Nano 40 TOPS processes wrist streams alongside head stereo vision — running object detection and grasp pose estimation networks that would be impossible without dedicated wrist views. This two-scale visual hierarchy (wrist detail → head scene) is the foundation for visual dexterous manipulation research.
Vision-Wrist Fusion — Two-Scale Perception for Manipulation
Smart B processes two synchronized visual streams on the Orin Nano 40 TOPS: scene-level depth and color from the head binocular camera, and close-up RGB from the wrist-mounted cameras. The head stereo provides global context — where objects are, what the scene layout looks like, and where the arms can move without collision. The wrist cameras provide local detail — grasp point quality, finger-object contact geometry, and real-time grasp verification. Fusing these two scales enables research in active perception (moving the hand to get a better view before grasping), visual dexterous grasping (coordinating finger closure with visual feedback), and hand-eye coordination (learning policies that map visual input to precise arm and finger motions). The 40 TOPS and 8 GB RAM handle both perception streams, grasp planning, and arm control simultaneously — though researchers pushing large vision models may eventually want the headroom of the Orin NX 100 TOPS upgrade.
Dual 7-DOF Arms. Redundant Kinematics. Human-Like Dexterity.
The R1-A7 Smart B is engineered around two 7-DOF high-performance arms, each driven by low-inertia inner-rotor PMSM motors with crossed-roller and double-row ball bearings. With 555 mm reach and a ~2 kg payload per arm, it handles precise assembly, pick-and-place, and manipulation research with human-like fluidity. The extra two degrees of freedom per arm (compared to 5-DOF platforms) provide kinematic redundancy — the arm can maintain end-effector pose while reconfiguring its elbow to avoid obstacles, reach around corners, or optimize for natural motion. Dual encoders on every joint deliver closed-loop precision down to ±0.1 mm.
Waist Rotation — Expand Your Workspace
The 1-DOF waist joint provides ±150° of rotation, dramatically expanding the robot's reachable workspace without moving the base. In a seated or standing deployment, the waist allows the robot to turn and manipulate objects on either side, hand objects from one hand to the other across the body, and orient toward humans or workstations. Combined with 7-DOF arms, the R1-A7 achieves human-like reach envelopes that cover a full semicircle of desktop or countertop space — ideal for collaborative tasks, assembly workflows, and human-robot interaction studies.
Orin Nano 40 TOPS — AI Onboard for Visual Manipulation
The built-in NVIDIA Jetson Orin Nano delivers 40 TOPS of AI compute with a 6-core CPU and 8 GB RAM. For visual manipulation with three-finger hands and wrist cameras, this is sufficient to run real-time stereo depth estimation, wrist-based grasp pose detection, and arm trajectory planning — all on board. No external GPU server, no cloud dependency. The Orin Nano handles two-scale visual perception and arm control locally, making the R1-A7 Smart B a self-contained research platform for dexterous manipulation studies. Researchers working with larger vision-language models or multi-modal fusion may consider upgrading to the Orin NX 100 TOPS module for additional headroom.
Binocular Vision & AI Perception
A binocular camera head delivers a 146° × 124° field of view with a 60 mm stereo baseline. RGB streams at 1280×720@30 Hz and depth at 544×448@10 Hz give the R1-A7 true spatial awareness. Paired with an 8-core high-performance CPU + 10 TOPS NPU in the head module, plus the Orin Nano 40 TOPS body compute, it runs visual perception, object detection, and gesture recognition locally — no cloud required. Combined with wrist RGB cameras, the perception stack covers both scene-level and gripper-level visual understanding.
Voice Collaboration & Natural Interaction
Four-array beamforming microphones and dual 3 W speakers enable clear voice capture and synthesis in noisy labs or classrooms. The open audio framework supports custom wake words, multilingual TTS, and integration with LLMs for conversational command-and-control.
Compact. Lightweight. Desktop-Ready.
At just 835 × 357 × 190 mm and approximately 13 kg, the R1-A7 Smart B fits on a standard lab bench or classroom desk. The compact footprint leaves room for objects, tools, and sensors around the robot. Power via the included external supply — an optional upper-body Li-ion battery is available for ~1.5 hours of untethered operation.
Full-Stack Open Development
Unitree provides a mature, Linux-based robot development framework with open APIs for the underlying system, robotic arms, audio, lighting, and visual control. ROS 2 native support, dual encoders on every arm joint, and hollow internal wiring make the R1-A7 Smart B a clean slate for reinforcement learning, visual manipulation research, embodied AI, and human-robot interaction studies. The built-in Orin Nano 40 TOPS handles two-scale visual perception and arm control locally — with headroom to grow into tactile sensing, larger vision models, and more advanced end effectors as your research evolves.
Upgrade Path — Grow With Your Research
Smart B is a visual dexterous manipulation platform ready to grow:
- End Effectors — Upgrade to Dex3-1 Tactile for force-aware grasping, BrainCo Revo 2 for anthropomorphic manipulation, LinkerBot O6 for high-DOF dexterity, or tendon-driven hands for biomimetic research.
- Tactile Sensing — Integrate force-sensitive fingertips for contact-aware grasping and slip detection.
- Large Vision-Language Models — Run VLA policies and multi-modal foundation models locally with the Orin NX 100 TOPS upgrade.
- Compute Module — Swap the Orin Nano for the Orin NX 100 TOPS module to unlock 2.5× AI performance, 16 GB RAM, and dual DLA cores.
Designed For
Visual Manipulation
Hand-eye coordination, visual servoing, and grasp verification using wrist RGB cameras combined with head stereo vision and 7-DOF redundant arms.
Dexterous Grasping
Adaptive three-finger grasping for irregular, spherical, and delicate objects with tripod pinches, power grasps, and precision holds.
Active Perception
Moving the hand and head to optimize visual views before grasping, and using wrist cameras for real-time grasp quality assessment.
Embodied AI & Bimanual Coordination
RL and imitation learning for dexterous policies, two-hand coordination, and multi-modal sensor fusion on a compact desktop platform.
Technical Specifications
What's in the Box
- Unitree R1-A7 Smart B Humanoid Robot (upper body)
- Dex3-1 three-finger dexterous hands (pre-installed, both arms)
- Wrist-mounted RGB cameras (pre-installed, both wrists)
- NVIDIA Jetson Orin Nano 40 TOPS compute module (pre-installed)
- External power supply & AC cable
- Optional: upper-body lithium-ion battery pack
- 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
Description
Unitree Robotics — Smart Series
Unitree R1-A7 Smart B
Upper-Body Dual-Arm Humanoid Robot with Dex3-1 Three-Finger Dexterous Hands & Wrist RGB Cameras. 17 DOF, 7-DOF precision arms, 555 mm reach, Orin Nano 40 TOPS, stereo vision & full-stack open SDK. For advanced visual manipulation research.
17
Total DOF
7+7
Arm DOF
555
mm Reach
~2
kg Payload
40
TOPS AI
What's in This Configuration
The R1-A7 Smart B steps up from entry-level grippers to three-finger dexterous manipulation with integrated wrist vision. It features the same R1-A7 upper-body base with 7-DOF arms, 1-DOF waist (±150°), and 2-DOF head as Smart A, but replaces the parallel grippers with Dex3-1 three-finger dexterous hands and adds wrist-mounted RGB cameras on each arm. The Dex3-1 hands provide adaptive three-finger grasping — tripod pinches, power grasps, and precision holds — that parallel grippers cannot achieve. Wrist cameras deliver a gripper's-eye view for visual servoing, grasp verification, and hand-eye coordination. Powered by the NVIDIA Jetson Orin Nano 40 TOPS, Smart B handles visual perception, dexterous grasp planning, and arm control on board. At ~13 kg, it is a compact yet capable platform for visual manipulation research.
✓ R1-A7 Upper-Body Humanoid
Dual 7-DOF arms, 1-DOF waist (±150°), 2-DOF head. 835×357×190 mm compact footprint, ~13 kg. Crossed-roller bearings, low-inertia inner-rotor motors, hollow internal cabling.
✓ Dex3-1 Three-Finger Dexterous Hands
Adaptive three-finger dexterous hands pre-installed on both arms for tripod pinches, power grasps, precision holds, and in-hand manipulation.
✓ Wrist-Mounted RGB Cameras
RGB cameras integrated into each wrist for close-up visual feedback, grasp verification, and hand-eye coordination during manipulation.
✓ Orin Nano 40 TOPS
Onboard NVIDIA Jetson Orin Nano delivering 40 TOPS for stereo-wrist visual fusion, grasp planning, and AI model inference.
✓ Voice Interaction
4-array beamforming mics + dual 3 W speakers. Wake-word, multilingual TTS, LLM integration ready.
✓ Full-Stack Open SDK
Linux framework, ROS 2 native, open APIs for arms, audio, vision, lighting. Dual encoders per joint.
Dex3-1 — Three-Finger Dexterity for Adaptive Grasping
The Dex3-1 is a three-finger dexterous hand that bridges the gap between simple parallel grippers and complex five-finger anthropomorphic designs. With three independently actuated fingers, it can form tripod pinches for spherical objects, power grasps for cylindrical items, and precision holds for small or fragile parts. The third finger provides stability and adaptability that two-finger grippers lack — the hand can envelop irregular objects, apply distributed contact forces, and maintain grasp security even when the object shifts. On the R1-A7's 7-DOF arms, the Dex3-1 enables bimanual coordination studies: one hand holds a workpiece while the other manipulates it, or both hands cooperate to lift and orient larger objects. The hand's compact design keeps mass low at the wrist, preserving the arm's dynamic response and precision.
Wrist-Mounted RGB Cameras — See From the Gripper
Smart B adds wrist-mounted RGB cameras to both arms — providing a gripper's-eye view of the workspace. These cameras capture close-up visual detail that the head-mounted stereo pair cannot: texture, surface geometry, and grasp point quality at the moment of contact. Combined with the Dex3-1's adaptability, wrist vision enables visual servoing for precision grasp alignment, grasp verification (did the fingers close on the target or miss?), and in-hand tracking (monitoring object pose during manipulation). The Orin Nano 40 TOPS processes wrist streams alongside head stereo vision — running object detection and grasp pose estimation networks that would be impossible without dedicated wrist views. This two-scale visual hierarchy (wrist detail → head scene) is the foundation for visual dexterous manipulation research.
Vision-Wrist Fusion — Two-Scale Perception for Manipulation
Smart B processes two synchronized visual streams on the Orin Nano 40 TOPS: scene-level depth and color from the head binocular camera, and close-up RGB from the wrist-mounted cameras. The head stereo provides global context — where objects are, what the scene layout looks like, and where the arms can move without collision. The wrist cameras provide local detail — grasp point quality, finger-object contact geometry, and real-time grasp verification. Fusing these two scales enables research in active perception (moving the hand to get a better view before grasping), visual dexterous grasping (coordinating finger closure with visual feedback), and hand-eye coordination (learning policies that map visual input to precise arm and finger motions). The 40 TOPS and 8 GB RAM handle both perception streams, grasp planning, and arm control simultaneously — though researchers pushing large vision models may eventually want the headroom of the Orin NX 100 TOPS upgrade.
Dual 7-DOF Arms. Redundant Kinematics. Human-Like Dexterity.
The R1-A7 Smart B is engineered around two 7-DOF high-performance arms, each driven by low-inertia inner-rotor PMSM motors with crossed-roller and double-row ball bearings. With 555 mm reach and a ~2 kg payload per arm, it handles precise assembly, pick-and-place, and manipulation research with human-like fluidity. The extra two degrees of freedom per arm (compared to 5-DOF platforms) provide kinematic redundancy — the arm can maintain end-effector pose while reconfiguring its elbow to avoid obstacles, reach around corners, or optimize for natural motion. Dual encoders on every joint deliver closed-loop precision down to ±0.1 mm.
Waist Rotation — Expand Your Workspace
The 1-DOF waist joint provides ±150° of rotation, dramatically expanding the robot's reachable workspace without moving the base. In a seated or standing deployment, the waist allows the robot to turn and manipulate objects on either side, hand objects from one hand to the other across the body, and orient toward humans or workstations. Combined with 7-DOF arms, the R1-A7 achieves human-like reach envelopes that cover a full semicircle of desktop or countertop space — ideal for collaborative tasks, assembly workflows, and human-robot interaction studies.
Orin Nano 40 TOPS — AI Onboard for Visual Manipulation
The built-in NVIDIA Jetson Orin Nano delivers 40 TOPS of AI compute with a 6-core CPU and 8 GB RAM. For visual manipulation with three-finger hands and wrist cameras, this is sufficient to run real-time stereo depth estimation, wrist-based grasp pose detection, and arm trajectory planning — all on board. No external GPU server, no cloud dependency. The Orin Nano handles two-scale visual perception and arm control locally, making the R1-A7 Smart B a self-contained research platform for dexterous manipulation studies. Researchers working with larger vision-language models or multi-modal fusion may consider upgrading to the Orin NX 100 TOPS module for additional headroom.
Binocular Vision & AI Perception
A binocular camera head delivers a 146° × 124° field of view with a 60 mm stereo baseline. RGB streams at 1280×720@30 Hz and depth at 544×448@10 Hz give the R1-A7 true spatial awareness. Paired with an 8-core high-performance CPU + 10 TOPS NPU in the head module, plus the Orin Nano 40 TOPS body compute, it runs visual perception, object detection, and gesture recognition locally — no cloud required. Combined with wrist RGB cameras, the perception stack covers both scene-level and gripper-level visual understanding.
Voice Collaboration & Natural Interaction
Four-array beamforming microphones and dual 3 W speakers enable clear voice capture and synthesis in noisy labs or classrooms. The open audio framework supports custom wake words, multilingual TTS, and integration with LLMs for conversational command-and-control.
Compact. Lightweight. Desktop-Ready.
At just 835 × 357 × 190 mm and approximately 13 kg, the R1-A7 Smart B fits on a standard lab bench or classroom desk. The compact footprint leaves room for objects, tools, and sensors around the robot. Power via the included external supply — an optional upper-body Li-ion battery is available for ~1.5 hours of untethered operation.
Full-Stack Open Development
Unitree provides a mature, Linux-based robot development framework with open APIs for the underlying system, robotic arms, audio, lighting, and visual control. ROS 2 native support, dual encoders on every arm joint, and hollow internal wiring make the R1-A7 Smart B a clean slate for reinforcement learning, visual manipulation research, embodied AI, and human-robot interaction studies. The built-in Orin Nano 40 TOPS handles two-scale visual perception and arm control locally — with headroom to grow into tactile sensing, larger vision models, and more advanced end effectors as your research evolves.
Upgrade Path — Grow With Your Research
Smart B is a visual dexterous manipulation platform ready to grow:
- End Effectors — Upgrade to Dex3-1 Tactile for force-aware grasping, BrainCo Revo 2 for anthropomorphic manipulation, LinkerBot O6 for high-DOF dexterity, or tendon-driven hands for biomimetic research.
- Tactile Sensing — Integrate force-sensitive fingertips for contact-aware grasping and slip detection.
- Large Vision-Language Models — Run VLA policies and multi-modal foundation models locally with the Orin NX 100 TOPS upgrade.
- Compute Module — Swap the Orin Nano for the Orin NX 100 TOPS module to unlock 2.5× AI performance, 16 GB RAM, and dual DLA cores.
Designed For
Visual Manipulation
Hand-eye coordination, visual servoing, and grasp verification using wrist RGB cameras combined with head stereo vision and 7-DOF redundant arms.
Dexterous Grasping
Adaptive three-finger grasping for irregular, spherical, and delicate objects with tripod pinches, power grasps, and precision holds.
Active Perception
Moving the hand and head to optimize visual views before grasping, and using wrist cameras for real-time grasp quality assessment.
Embodied AI & Bimanual Coordination
RL and imitation learning for dexterous policies, two-hand coordination, and multi-modal sensor fusion on a compact desktop platform.
Technical Specifications
What's in the Box
- Unitree R1-A7 Smart B Humanoid Robot (upper body)
- Dex3-1 three-finger dexterous hands (pre-installed, both arms)
- Wrist-mounted RGB cameras (pre-installed, both wrists)
- NVIDIA Jetson Orin Nano 40 TOPS compute module (pre-installed)
- External power supply & AC cable
- Optional: upper-body lithium-ion battery pack
- 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
Unitree
by
Unitree
Unitree R1-A7 Smart B Humanoid Robot
- Regular price
- $16,390.00
- Sale price
- $16,390.00
- 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
