Humanoid Database

 

Disclaimer: All content presented on this site, including robot profiles, images, and descriptions, is provided for editorial, educational and informational purposes only. Humanoid Press does not claim ownership of any third-party trademarks, designs, or intellectual property featured herein. All rights belong to their respective owners.

 

VR-H3 (VinRobotics)

type:
type 1
{{variant.name}}:
{{opt.name}}
{{opt.value ? '' : opt.name}}
{{opt.value ? '' : opt.name}}
Description

SUMMARY:

 

VR‑H3 is VinRobotics’ flagship full‑size humanoid robot, designed for industrial assistance, logistics, research, and AI‑driven manipulation.  

 

Standing 178 cm tall and weighing 75–85 kg, VR‑H3 uses high‑torque electric actuators, multi‑sensor perception, and a reinforcement‑learning locomotion stack to achieve stable bipedal walking, object handling, and human‑environment interaction.

 

VR‑H3 is the successor to the VR‑H1 platform and represents VinRobotics’ push into next‑generation humanoid autonomy for factories and service environments.

 

EDITORIAL:

 

Vietnam’s robotics ecosystem is accelerating fast, and VR‑H3 is one of the strongest signals of that momentum.  

 

Where VR‑H1 was a research‑grade humanoid, VR‑H3 is a near‑commercial platform with:

 

- stronger actuators  

- improved balance  

- better perception  

- more human‑like proportions  

- a more robust AI stack  

 

VR‑H3 is built to operate in real industrial environments, not just labs.  

Its architecture emphasizes:

 

- high‑torque joints for lifting  

- stable locomotion on uneven surfaces  

- vision‑based manipulation  

- cloud‑connected AI  

- modular hardware for rapid iteration  

 

VinRobotics is positioning VR‑H3 as a regional competitor to mid‑tier humanoids like Unitree H1, Fourier GR‑1, and UBTech Walker S — but with a uniquely lightweight, efficient, and cost‑optimized design philosophy.

 

SPECIFICATIONS:

 

Dimensions & Build

 

- Height: 178 cm  

- Weight: 75–85 kg (depending on battery + payload configuration)  

- Frame: Aluminum alloy + composite shell  

- Design: Human‑proportional, industrial‑grade humanoid  

 

Mobility

 

- Walking speed: ~1.2–1.5 m/s  

- Locomotion: Reinforcement‑learning gait controller  

- Balance: Multi‑sensor fusion + whole‑body control  

- Capabilities:  

  - Bipedal walking  

  - Turning  

  - Squatting  

  - Reaching  

  - Object handling  

 

Degrees of Freedom

 

- Total: ~40–50 DoF (varies by revision)  

- Arms: 7 DoF each  

- Legs: 6 DoF each  

- Hands: 2–3 DoF each  

- Neck: 2 DoF  

- Waist: 3 DoF  

 

Actuation

 

- High‑torque electric actuators  

- Joint torque: Estimated 120–180 Nm class  

- Joint speed: Medium‑high speed for industrial tasks  

- Payload:  

  - Arms: 3–5 kg per arm  

  - Whole‑body: 10–15 kg handling tasks  

 

Sensors

 

- Vision:  

  - Stereo depth camera  

  - RGB camera  

- IMU: 9‑axis  

- Force sensors: Feet + joints  

- Touch sensors: Hands (optional)  

- Environmental sensing: Optional LiDAR  

 

Compute & Software

 

- CPU: Industrial x86 or ARM compute module  

- GPU: NVIDIA Jetson‑class accelerator  

- OS: Linux + ROS2  

- AI:  

  - Reinforcement learning locomotion  

  - Vision‑language‑action integration  

  - Cloud‑connected inference  

- SDK: Python + C++  

 

Battery & Power- Battery: Swappable lithium pack  

- Runtime: 2–4 hours depending on load  

- Charging: External dock or manual charge  

 

Interaction- Voice recognition  

- Speech synthesis  

- Gesture‑based interaction  

- Remote teleoperation  

 

Applications- Industrial logistics  

- Manufacturing assistance  

- Research & education  

- Demonstrations & public engagement  

- AI robotics development  

 

WHAT IS NOT PUBLISHED

 

- Exact torque per joint  

- Maximum payload for dynamic tasks  

- IP rating  

- Pricing (not publicly disclosed)  

 

Image: @XeTinhtế

 

COMPANY SOURCE:

 

https://youtu.be/AIRifsVKtyc?feature=shared  

 

KEY NEWS & COVERAGE

 

https://youtu.be/AIRifsVKtyc?feature=shared