Unitree Quadrupeds

Home > Hardware > Quadrupeds > Unitree

---

Key Significance

• Pioneer in Quadruped Democratization: High-performance quadruped popularization at 1/20 ~ 1/2 price of Spot, research accessibility innovation
• Global Market Leader: High market share in global quadruped robot market (PanDaily)
• Research Ecosystem Building: Standard platform for academic research with official SDK, ROS2 support, active community
• Vertical Integration Manufacturing: Cost competitiveness through in-house R&D of core components (motors, reducers, sensors, batteries, etc.)
• VLA/Locomotion Research Core Platform: NVIDIA GR00T support, included in Open X-Embodiment dataset
• Wheeled-Legged Hybrid: Leading wheel+leg hybrid locomotion with Go2-W, B2-W
• Major Platform for RL Locomotion Research: Widely used for sim-to-real research due to affordable price and open SDK

Unitree Go2: The most popular quadruped platform for research

---

Overview

Item	Content
Manufacturer	Unitree Robotics
Headquarters	Hangzhou, China
Founded	2016 (Founder: Wang Xingxing)
Features	Value, research accessibility, ROS2 support
Market Position	Global quadruped market leader

Unitree Robotics is a Chinese robot company founded by Wang Xingxing in 2016. The company name “Unitree” is a compound of “Universe” and “Tree,” meaning a universal technology base branching out to various applications. Founder Wang Xingxing started quadruped research during graduate studies at Shanghai University, worked at DJI, then founded the company to lead the democratization of high-performance quadrupeds.

Note: Price and spec information is based on 2024-2025 official website and may vary by region and time.

---

Product Lineup

Unitree has a diverse quadruped lineup from consumer/research to industrial. Each model is optimized for specific use cases and price points.

Complete Lineup Comparison Table

Model	Release	Price	Weight	Speed	Payload	Runtime	Primary Use
Aliengo	2019	~$15,000	21.5kg	3.3m/s	5kg	2-4 hrs	Research/Dev
Go1 Air	2021	$2,700	12kg	2.5m/s	3kg	1 hr	Entry/Hobby
Go1 Pro	2021	$3,500	12kg	3.5m/s	5kg	1.5 hrs	Consumer
Go1 Edu	2021	~$8,500	12kg	3.7m/s	5kg	2 hrs	Research/Education
Go2 Air	2023	$1,600	15kg	2.5m/s	7kg	1-2 hrs	Next-gen Entry
Go2 Pro	2023	$2,800	15kg	3.5m/s	8kg	1-2 hrs	Consumer
Go2 Edu	2023	Inquiry	15kg	5m/s	8kg	2-4 hrs	Research/Education
Go2-W	2024	$4,500 ~ $14,000	18kg	2.5m/s	12kg	1.5-3 hrs	Hybrid
B1	2020	~$40,000	50kg	1.8m/s	5kg	2-4 hrs	Industrial
B2	2023	~$100,000	60kg	6m/s+	40kg	4-6 hrs	Industrial Premium
B2-W	2024	~$90,000	-	5.56m/s	40kg	50km range	Industrial Wheeled

Consumer/Research Series

Go1 Series (2021)

Go1 is the product that made Unitree famous worldwide. Starting at $2,700, it dramatically lowered the barrier to entry for quadruped research.

Model	Price	Weight	Max Speed	Payload	Runtime	Key Features
Go1 Air	$2,700	12kg	2.5m/s	3kg	1 hr	Entry level, basic sensors
Go1 Pro	$3,500	12kg	3.5m/s	5kg	1.5 hrs	Enhanced sensors, ultrasonic
Go1 Edu	~$8,500	12kg	3.7m/s	5kg	2 hrs	SDK/API access, Jetson

Go2 Series (2023)

Go2 is the successor to Go1, with significantly improved performance and sensors. The 4D LiDAR in particular greatly improves autonomous navigation capabilities.

Model	Price	Weight	Max Speed	Payload	Runtime	Key Features
Go2 Air	$1,600	15kg	2.5m/s	7kg	1-2 hrs	Next-gen entry
Go2 Pro	$2,800	15kg	3.5m/s	8kg	1-2 hrs	4G, ISS 2.0, 4D LiDAR
Go2 Edu	Inquiry	15kg	5m/s	8kg	2-4 hrs	Jetson Orin, foot force sensors
Go2-W	$4,500 ~ $14,000	18kg	2.5m/s	12kg	1.5-3 hrs	Wheel+leg hybrid

Research Series

Aliengo (2019)

Aliengo is Unitree’s first serious research-grade quadruped. Released before Go1, it was adopted by academic research communities first.

Item	Spec
Release	2019
Price	~$15,000 (Edu version)
Weight	21.5kg
Size	650mm x 320mm x 600mm
Max Speed	3.3m/s
Payload	5kg
Joint Torque	40N.m
Runtime	2-4 hours
Computing	NVIDIA Jetson TX2/Xavier NX
Features	EtherCAT motor communication, C++ SDK

Why Aliengo was used in research:

• Only affordable research-grade quadruped before Go1
• Early platform for reinforcement learning sim-to-real research
• Adopted by famous labs including UC Berkeley, ETH Zurich
• Experimental platform for various locomotion algorithm papers

Industrial Series

B1 (2020)

B1 is Unitree’s first large industrial quadruped.

Item	Spec
Release	2020
Price	~$40,000
Weight	50kg
Size	1080mm x 460mm x 650mm
Max Speed	1.8m/s
Payload	5kg
Joint Torque	180N.m
Runtime	2-4 hours
Water/Dust	IP54
Operating Temp	-20C ~ 55C

B2 Series (2023)

B2 is the successor to B1, offering industry-leading performance.

Model	Price	Weight	Max Speed	Payload	Runtime	Key Features
B2	~$100,000	60kg	6m/s+	40kg(moving)/120kg(stationary)	4-6 hrs	Industrial, IP67
B2-W	~$90,000	-	5.56m/s (20km/h)	40kg	50km range	12-inch wheels

---

Go2 Detailed Specs

Unitree’s most popular research quadruped.

Physical Specifications

Item	Spec
Size (standing)	70cm x 31cm x 40cm
Size (sitting)	76cm x 31cm x 20cm
Weight	~15kg (with battery)
Materials	Aluminum alloy + engineering plastic

Performance

Item	Spec
Max Speed	5m/s (Edu basis)
Climb Angle	30-40 degrees
Stair Height	15-16cm
Joint Torque	~45N.m
Battery	8000mAh (standard) / 15000mAh (extended)

Sensor System

Item	Spec
4D LiDAR	360 x 90 degree hemispherical perception, min 0.05m
Camera	HD wide-angle
IMU	Built-in
Foot Force Sensors	Edu models

Computing (Edu Model)

Version	Computing Module	AI Performance
Edu Standard	NVIDIA Jetson Orin Nano 8GB	40 TOPS
Edu Plus	NVIDIA Jetson Orin NX 16GB	100 TOPS

---

B2 Detailed Specs

Industrial high-performance quadruped.

Physical Specifications

Item	Spec
Size (standing)	1098mm x 450mm x 645mm
Size (sitting)	880mm x 460mm x 330mm
Weight	~60kg (with battery)

Performance

Item	Spec
Max Speed	6m/s+
Jump Distance	1.6m+
Joint Torque	360N.m
Climb Angle	45 degrees+
Obstacle Clearance	40cm
Stairs	20-25cm continuous climb

Load and Durability

Item	Spec
Stationary Load	120kg+
Moving Load	40kg+
With 20kg Load	4 hours, 15km+
Unloaded Continuous	5 hours, 20km+
Battery	45Ah (2250Wh), 58V

Environmental Adaptation

Item	Spec
Water/Dust	IP67
Operating Temp	-20C ~ 55C

Computing

Item	Spec
Base CPU	Intel Core i5/i7
Expansion Options	Jetson Orin NX (up to 3)
Sensors	3D LiDAR, depth camera x2, optical camera x2
Interfaces	1000M Ethernet x4, USB3.0 x4

---

Wheeled Hybrid (Go2-W, B2-W)

Unitree’s innovative wheel+leg hybrid robots.

Go2-W

Item	Spec
Size	70cm x 43cm x 50cm
Weight	~18kg
Tires	7-inch pneumatic
Max Speed	2.5m/s
Obstacle Clearance	70cm
Climb Angle	35 degrees

B2-W

Item	Spec
Tires	12-inch pneumatic (225mm)
Wheel Torque	50N.m
Max Speed	5.56m/s (20km/h)
Range	50km with 40kg load
Price	~$90,000

---

Boston Dynamics Spot Comparison

Item	Unitree Go2 Edu	Unitree B2	Boston Dynamics Spot
Price	~$8,000	~$100,000	~$75,000
Weight	15kg	60kg	32kg
Payload	8kg	40kg	14kg
Runtime	2-4 hrs	4-6 hrs	1.5 hrs
Max Speed	5m/s	6m/s+	1.6m/s
Water/Dust	-	IP67	IP54
SDK	Open	Open	Limited
ROS Support	Official ROS2	Official ROS2	Community

Unitree’s Key Advantages

• Price: Significantly cheaper than Spot, advantageous for research-scale deployment
• Speed: B2 at 6+ m/s is top speed class among industrial quadrupeds
• Runtime: B2 up to 6 hours for extended operation
• Openness: Research-friendly with official SDK, ROS2 support
• Cost Structure: Competitive pricing through China manufacturing and vertical integration (36Kr)

---

SDK and Development Environment

Unitree maintains a research-friendly open SDK policy, a key differentiator from competitors.

Official SDK Overview

+----------------------------------------------------------+
|                Unitree SDK Architecture                   |
+----------------------------------------------------------+
|                                                          |
|   Application Layer    Python Script / C++ App / ROS2 Node    |
|        |                                                |
|   SDK Layer     unitree_sdk2 / unitree_ros2            |
|        |                                                |
|   Communication Layer    CycloneDDS (DDS Middleware)    |
|        |                                                |
|   Hardware       Go2 / B2 / H1 Controller               |
|                                                          |
+----------------------------------------------------------+

Official SDKs

SDK	Language	Description	Supported Robots
unitree_sdk2	C++/Python	CycloneDDS-based communication	Go2, B2, H1
unitree_sdk2_python	Python	unitree_sdk2 Python bindings	Go2, B2, H1
unitree_ros2	C++/Python	ROS2 environment development package	Go2, B2, H1
unitree_rl_lab	Python	Isaac Lab-based reinforcement learning	Go2, B2
unitree_IL_lerobot	Python	LeRobot framework modification	Go2
unitree_sdk	C++	Legacy SDK (Go1, Aliengo)	Go1, Aliengo

Python SDK Usage Example

# unitree_sdk2_python basic usage
from unitree_sdk2py.go2.robot import Go2Robot

robot = Go2Robot()
robot.set_mode("sport")
robot.move(vx=0.5, vy=0.0, vyaw=0.0)  # Forward movement
robot.stand_up()
robot.sit_down()

C++ SDK Features

Feature	Description
Low-level Control	Direct joint torque/position/velocity control (Edu models)
High-frequency Control	Up to 1000Hz control loop
Sensor Access	IMU, joint state, foot contact sensors
Safety Features	Emergency stop, torque limits

ROS2 Support

Unitree provides the most robust ROS2 support in the industry.

Item	Content
DDS Compatibility	Uses CycloneDDS 0.10.2, direct ROS2 communication possible
Official Package	unitree_ros2 (GitHub)
Supported Versions	ROS2 Humble (LTS), Jazzy
Simulation	Gazebo Harmonic, Ignition Fortress

ROS2 Topics and Services

+----------------------------------------------------------+
|              unitree_ros2 Main Topics/Services            |
+----------------------------------------------------------+
|                                                          |
|  Topics:                                                  |
|    /go2/joint_states        - Joint state (12 DoF)        |
|    /go2/odom                - Odometry                    |
|    /go2/imu                 - IMU data                   |
|    /go2/cmd_vel             - Velocity command            |
|    /go2/lidar               - 4D LiDAR point cloud        |
|    /go2/camera/color        - RGB image                  |
|                                                          |
|  Services:                                                |
|    /go2/sport_mode          - Sport mode switch           |
|    /go2/stand_up            - Stand up                    |
|    /go2/sit_down            - Sit down                    |
|    /go2/emergency_stop      - Emergency stop              |
|                                                          |
+----------------------------------------------------------+

Simulation Environments

Simulator	Support Level	Features
NVIDIA Isaac Gym	Official	Large-scale parallel RL training
NVIDIA Isaac Lab	Official	Next-gen Isaac, Omniverse integration
Gazebo Harmonic	Official	ROS2 standard simulator
MuJoCo	Community	High physics accuracy
PyBullet	Community	Lightweight RL experiments
Webots	Community	Educational

Community Resources

Project	Description	Link
unitree-go2-ros2	CHAMP controller + Ignition Fortress	GitHub
go2_robot	Universidad Rey Juan Carlos IRLab	GitHub
Go2 MCP Server	LLM natural language command control	GitHub
go2_ros2_ws	Complete ROS2 workspace	GitHub
unitree_legged_sdk	Legacy Go1 SDK	GitHub

Development Environment Setup Summary

# Ubuntu 22.04 + ROS2 Humble recommended
# 1. Install ROS2 Humble
sudo apt install ros-humble-desktop

# 2. Install CycloneDDS
sudo apt install ros-humble-rmw-cyclonedds-cpp
export RMW_IMPLEMENTATION=rmw_cyclonedds_cpp

# 3. Clone Unitree ROS2 package
cd ~/ros2_ws/src
git clone https://github.com/unitreerobotics/unitree_ros2

# 4. Build and run
cd ~/ros2_ws && colcon build
source install/setup.bash
ros2 launch unitree_ros2 go2_bringup.launch.py

---

VLA Research Applications

Unitree quadrupeds have established themselves as core platforms for VLA (Vision-Language-Action) and locomotion research.

NVIDIA GR00T Support

+----------------------------------------------------------+
|              NVIDIA GR00T Ecosystem                       |
+----------------------------------------------------------+
|  Isaac Lab    →   GR00T N1 Training   →   Unitree Deploy  |
|  (Simulation)     (Foundation Model)    (Real Robot)       |
+----------------------------------------------------------+
|  - Unitree Go2 simulation environment provided            |
|  - Unitree B2 simulation environment provided             |
|  - Used for cross-embodiment policy transfer research     |
+----------------------------------------------------------+

• GR00T N1 Series Supported Platform: NVIDIA Isaac Lab provides Unitree robot simulation environments
• Isaac Lab/Omniverse Simulation: High-quality Unitree URDF and simulation environments provided
• Cross-embodiment Learning: Used for humanoid-quadruped policy transfer research
• Jetson Orin Equipped: NVIDIA hardware integration for onboard AI inference

Open X-Embodiment

Unitree robot data is included in the large-scale robot dataset project led by Google DeepMind.

Item	Content
Dataset Scale	22 robot types, 1M+ episodes
Unitree Included Data	Go1, Go2 locomotion data
Models Used	RT-X, Octo, OpenVLA, etc.
Contributing Institutions	UC Berkeley, Stanford, CMU, etc.

Research Applications:

• RT-X Model: Generalization ability learning across various robots
• Octo: Open-source general robot policy
• OpenVLA: Vision-Language-Action model locomotion extension

Locomotion Research

Unitree is a widely used platform for reinforcement learning-based locomotion research.

Key Research Areas

Area	Description	Representative Research
Sim-to-Real Transfer	Policy transfer from simulation to real robot	ETH Zurich, MIT
Terrain Adaptation	Adaptive walking on various terrains	UC Berkeley
Blind Locomotion	Walking with proprioception only without vision	CMU
Parkour	Dynamic movements like obstacle jumping	ETH RSL
Multi-skill Learning	Various skills with single policy	Stanford

sim-to-real Research Pipeline

+----------------------------------------------------------+
|                sim-to-real Research Flow                  |
+----------------------------------------------------------+
|                                                          |
|  [Isaac Gym/Lab]  →  [RL Policy Training]  →  [Domain Random] |
|        |                  |                  |          |
|   Unitree URDF      PPO/SAC etc        Terrain/Friction/Mass     |
|        |                  |                  |          |
|  -------------------------------------------------------- |
|                          |                              |
|              [Unitree Go2/B2 Deployment]                 |
|                          |                              |
|              Zero-shot performance in real environment   |
|                                                          |
+----------------------------------------------------------+

Representative Research Papers

Paper	Institution	Robot	Key Contribution
Learning Quadrupedal Locomotion	ETH Zurich	Aliengo	Large-scale sim-to-real
Legged Robots that Keep on Learning	UC Berkeley	Go1	Online adaptation learning
Walk These Ways	MIT	Go1	Various walking styles
Extreme Parkour	CMU	Go1	Parkour motion learning
DreamWaQ	KAIST	Go1	World model-based

Why Go2 is Popular in Research

Go2 Edu is currently the most popular research quadruped.

+----------------------------------------------------------+
|               Why Go2 is Preferred for Research           |
+----------------------------------------------------------+
|                                                          |
|  1. Price Accessibility                                   |
|     - 1/10 the price of Spot ($75K)                      |
|     - Multi-robot experiments possible (swarm research)   |
|     - Low replacement cost if damaged                     |
|                                                          |
|  2. Open SDK                                              |
|     - Official Python, C++, ROS2 support                 |
|     - Low-level joint control access (Edu version)       |
|     - Active community packages                           |
|                                                          |
|  3. Sufficient Computing Power                            |
|     - Jetson Orin (40-100 TOPS)                          |
|     - Onboard RL policy execution possible               |
|     - Vision-based navigation support                     |
|                                                          |
|  4. Proven Hardware                                       |
|     - Reliability verified with thousands sold           |
|     - Easy parts replacement                              |
|     - Global technical support                            |
|                                                          |
|  5. Rich Simulation Environments                          |
|     - Official Isaac Gym/Lab support                     |
|     - Gazebo, MuJoCo URDF provided                       |
|     - PyBullet community models                           |
|                                                          |
+----------------------------------------------------------+

LLM/VLM Integration Research

Recently, research combining large language models with Unitree robots is active.

Project	Description
Go2 MCP Server	Robot control via natural language with Claude/GPT
SayPlan + Unitree	Long-term task plan execution
LLM-Planner	Convert natural language commands to robot actions
VLM Navigation	Vision-language model-based navigation

Teleoperation Data Collection

Method	Equipment	Use
Joystick Control	PS4/Xbox controller	Basic movement data
App Control	Unitree official app	Consumer control
VR Teleop	Quest/Vive	Research precision control
Motion Capture	Xsens/Noitom	Human motion imitation

---

Research Popularity

Academic Research Adoption

Unitree quadrupeds are widely adopted by universities and research institutions worldwide due to affordable prices and open SDK:

• Price Accessibility: Research-scale deployment possible at 1/10 Spot price
• Multi-robot Experiments: Affordable price supports swarm robot research
• Educational Use: Popular as robotics education platform

Major Customers

• NVIDIA (since 2020, 36Kr)
• Google (since 2020, 36Kr)
• Major university research labs worldwide

---

Company History

Year	Event
2013	Wang Xingxing starts quadruped research at Shanghai University
2016	XDog development, Unitree Robotics founded
2017	Laikago unveiled (named after Laika)
2020	NVIDIA, Google customers secured, profitable (36Kr)
2021	Go1 released, consumer quadruped democratization
2023	Go2 released, B2 announced, H1 humanoid unveiled
2024	Go2-W released
2025	Hong Kong IPO under consideration (per reports)

Investment Status

• Recent Round: Series C-IV (2025 basis)
• Major Investors: Alibaba, Tencent, Beijing State-owned Capital Operation and Management Company, etc. (PitchBook)

---

Application Areas

Research & Education

• Locomotion algorithm research
• Reinforcement learning/imitation learning experiments
• Robotics education

Industrial (B2/B2-W)

• Facility inspection (power plants, chemical plants)
• Construction site monitoring
• Hazardous environment exploration
• Logistics automation

Consumer (Go1/Go2)

• Tech enthusiasts/developers
• Prototyping
• Entertainment

---

References

Official Resources

• Unitree Go2 Official Page
• Unitree B2 Official Page
• Unitree Go2-W Official Page
• Unitree GitHub
• Unitree SDK2 Documentation

Technical Reviews

• IEEE Spectrum - Unitree Go2
• IEEE Spectrum - Go1 Launch Review

Market Analysis

• Unitree vs Boston Dynamics Comparison
• The Rise of Unitree Robotics

ROS2 Community

• unitree_ros2 GitHub
• ROS Discourse - Go2 Jazzy Release

---

See Also

• Hardware List
• Boston Dynamics Spot
• ANYbotics ANYmal
• Unitree Humanoids
• GR00T