H1 Robot Tutorial

Humanoid Robot H1 Typing	Humanoid Robot H1 Playing Piano
Humanoid Robot H1 Playing Tennis	Humanoid Robot H1 Boxing

Important

Please note that the content showcased here represents the individual achievements of Humanoid AI and Omni Human-to-Humanoid, and there has been no collaboration between them and our organization. The videos and gifs included on this page are intended strictly for educational and demonstrative purposes.

The provided tutorial will assist you with setting up and operating your H1 bi-pedal robot. The tutorial topics are listed in the left column and presented in the suggested reading order.

Insights from Leading Researchers

We are excited to highlight the innovative work presented on Humanoid AI and Omni Human-to-Humanoid. The researchers on Humanoid AI have developed a comprehensive system enabling humanoid robots to learn motion and autonomous skills from human data. Their approach involves training low-level policies in simulation through reinforcement learning using extensive human motion datasets. These policies allow humanoid robots to shadow human movements in real-time using only an RGB camera. The collected data is then used for supervised behavior cloning, enabling robots to autonomously perform various tasks such as wearing shoes, unloading objects, and interacting with other robots with impressive success rates. On the other hand, the work on Omni Human-to-Humanoid has resulted in OmniH2O, a versatile system for whole-body humanoid teleoperation and autonomy. By using kinematic poses as a universal control interface, OmniH2O allows humans to control humanoids through VR, verbal instructions, and RGB cameras. They have also integrated advanced models like GPT-4 to enable full autonomy. Their RL-based sim-to-real pipeline and the release of the OmniH2O-6 dataset highlight their contributions to humanoid skill learning.

Humanoid Robot H1 Brush Teleoperation	Humanoid Robot H1 Verbal Instructions
Humanoid Robot H1 Squat & Pick Basket	Humanoid Robot H1 GPT4o-Autonomous Greeting

Warm Tips for Unitree Humanoid Robot Development and Usage

To align with the public’s preference for natural human-like movements, it’s important to adhere to these guidelines, particularly when filming humanoid robots:

• Strive for upright or nearly upright knee joint positions when developing leg movement programs.

• Reduce step frequency and minimize stationary stepping.

• Keep the feet slightly close together during walking to maintain natural movement.

These tips are intended to enhance the authenticity of humanoid robot movements, especially in video documentation.

Component Overview

The H1 robot comprises upper and lower bodies, each featuring multiple degrees of freedom. The single arm includes joints for body-shoulder, shoulder, upper arm, and elbow movements, while the single leg features joints for hip, leg, knee, and ankle motions. Additionally, the waist section offers a waist joint. In total, the robot boasts 19 degrees of freedom, enabling precise motion and posture control through 19 joint motors.

Electrical Interfaces

Located on the right side of the H1 robot, the electrical interfaces facilitate connections to various components such as body joint motors, sensor peripherals, and Ethernet ports. This design simplifies debugging, troubleshooting, and secondary development processes.

On-board Computers

The H1 robot comes equipped with a “Motion Control Computing Unit” and optionally a “Development Computing Unit” (additional expansion unit available).

Component	Motion Control Computing Unit (PC 1)	Development Computing Unit (PC 2, PC 3)
Model	i5-1235U	i7-1255U / i7-1265U
Number of Cores	10	10
Number of Threads	12	12
Max Turbo Frequency	4.40 GHz	4.70 GHz & 4.80 GHz
Memory	8GB	16GB & 32GB
Memory Type	LPDDR5 5200 MT/s (dual-channel)	LPDDR5 5200 MT/s (dual-channel)
Cache	12 MB Intel® Smart Cache	12 MB Intel® Smart Cache
Storage	500GB or more	500GB or more
Graphics Processing Unit (GPU)	None	Intel® Iris® Xe Graphics eligible
Max Dynamic Frequency of GPU	1.20 GHz	1.20 GHz
Gauss and Neural Accelerator	3.0	3.0
Intel® Deep Learning Boost	Yes	Yes
Intel® Adaptix™ Technology	Yes	Yes
Intel® Hyper-Threading Technology	Yes	Yes
Instruction Set	64-bit	64-bit
OpenGL	4.6	4.6
OpenCL	3.0	3.0
DirectX	12.1	12.1
MCU IP Address	192.168.123.162	n/a
Auxiliary PC IP Address	192.168.123.163	UnitreeXXXX
Lidar IP Address	192.168.123.120	n/a

The “Motion Control Computing Unit” is dedicated to Unitree motion control programs and is not available for external use. Developers can only utilize the “Development Computing Unit” for secondary development. Initial username and password can be obtained by contacting tech support. PC3 “Development Compute Unit” is an optional configuration, with the address being 192.168.123.163. The CPU module may be upgraded to a more advanced version with performance meeting or exceeding the specifications mentioned above.

Lidar

The H1 robot’s head incorporates the MID-360 Lidar (IP: 192.168.123.120), which offers superior environmental perception capabilities. Employing omnidirectional, full-angle scanning technology, it delivers real-time and precise environmental data, facilitating accurate recognition and measurement of surrounding objects.

Depth Camera

Equipped with the D435i depth camera, the H1 robot’s head boasts exceptional visual perception capabilities. This enables more precise perception and understanding of the surrounding environment, allowing for accurate spatial perception and obstacle detection. Consequently, the robot can interact intelligently with its environment and adapt to various scenarios effectively.

M107 Joint Motor

The H1 robot utilizes the self-developed M107 joint motor by Unitree, known for its outstanding performance. Featuring a maximum torque of 360N.m and a maximum tension of 10000N (under equivalent conditions of a 3.5cm force arm), this motor’s hollow axle design makes it lightweight and compact. Additionally, equipped with dual encoders, it provides more accurate position and velocity feedback, meeting the requirements of high-precision control.