Unitree G1 Competition Humanoid Robot

G1-Comp G1 Comp Series Products (G1-Comp) (G1-Comp)

In practical terms, “G1-Comp” refers to a G1-class humanoid tuned and packaged for competitive locomotion and perception, integrating software elements such as a visual recognition API (YOLO-based), simulation-to-real workflows, and SDK tooling aimed at fast iteration. Unitree’s official G1-Comp page highlights YOLO11-based stadium recognition, a reinforcement-learning framework (unitree_rl_gym), and a dedicated RoboCup SDK development guide as key components of the ecosystem.

Although it shares core humanoid architecture with the broader Unitree G1 family, the G1-Comp branding is closely tied to competition use cases—especially robotic soccer, where robots must balance agility, stability, perception, and control responsiveness under unpredictable contact, acceleration, and game pacing.

Design and Features

Competition-focused locomotion and stability

Unitree describes the G1-Comp as “lightweight and agile” with a steady gait and “industry-leading motion control systems,” aimed at maintaining performance even with distractions and dynamic match conditions. This framing aligns with the needs of competitive robotics, where a humanoid must reliably perform repeated cycles of:

• rapid start–stop motion

• turns and lateral steps

• posture recovery after disturbance

• fast tracking toward targets (such as a ball)

In many competition formats, stability under contact and repeated motion resilience are as important as raw speed, since falls and recovery delays can decide matches.

Vision recognition built around YOLO11 detection

A central feature highlighted for G1-Comp is real-time perception for stadium environments. Unitree’s official description states the robot uses the YOLO11 network algorithm for “rapid identification of the stadium environment and accurate positioning.”

For robotic soccer contexts, this style of detection-driven perception is commonly used to locate and track:

• the ball

• goal areas

• field boundaries and landmarks

• other robots and obstacles

Unitree’s page also describes a Visual Recognition API with YOLO11-based target detection and a “rich stadium information interface,” reflecting an application-layer design meant to simplify development for teams building competition behaviors.

Reinforcement-learning workflow and simulation tool support

G1-Comp is marketed with support for modern training and deployment workflows using simulation-to-real pipelines. Unitree highlights unitree_rl_gym as a reinforcement learning framework and explicitly references integration with simulation tools such as Isaac Gym and MuJoCo, supporting the process from Sim2Sim training through Sim2Real deployment.

This matters for competition robotics because teams often develop locomotion and strategy behaviors in simulation at scale, then transfer learned policies to the real platform—reducing the cost and risk of physical trial-and-error.

Developer ecosystem (RoboCup SDK and documentation)

Unitree’s official G1-Comp page references a RoboCup SDK development guide, consistent with the product’s competition identity. In addition, Unitree lists G1-Comp as a distinct “Application Plan” item in its download center, implying dedicated documentation or resources aligned to the competition variant.

Technology and Specifications

Because “G1-Comp” is often sold through different channels, published specifications can vary by listing and configuration. A detailed product entry describing the Unitree G1 Comp provides a representative snapshot of hardware and compute characteristics, including:

Mechanical structure and degrees of freedom

The G1 Comp is described as a humanoid robot with 25 degrees of freedom, including 6 per leg, 5 per arm, and additional mobility at the waist and head.
These degrees of freedom (DoF) are foundational for competition agility because they support balance recovery and rapid posture transitions.

Size, weight, and foldable transport format

A commercial specification summary states that, when standing, G1 Comp measures approximately 1320 × 450 × 200 mm, weighs around 35 kg with battery, and can fold down to about 690 × 450 × 300 mm for transport.
Foldability is frequently emphasized as a practical advantage for teams traveling to events or moving between training facilities.

Joint torque and payload handling

Competition performance is strongly influenced by lower-body torque. One published listing cites powerful joint motors with up to 120 N·m at the knee and an arm load capacity of around 3 kg.
Separate seller listings for the G1-Comp similarly highlight a 120 N·m knee torque figure and ~3 kg arm payload.

Compute and sensor suite

A detailed product description indicates the robot’s computing unit is based on an 8-core CPU system with NVIDIA Jetson Orin, supported by sensors including 3D LiDAR, a depth camera, and a microphone array.
For competition tasks, this combination supports fast perception, state estimation, and voice/audio interaction experiments, while also enabling AI workflows.

Connectivity, battery life, and update model

The same source describes Wi-Fi 6, Bluetooth 5.2, around 2 hours of battery life, and OTA upgrades, along with a one-year warranty as standard.
OTA updates are notable for competition teams, since firmware and behavior tuning often continue through the season.

Applications and Use Cases

Robotic soccer and RoboCup-style competitions

Unitree positions G1-Comp as a platform designed for match environments, highlighting motion control robustness and YOLO-based stadium recognition.
In soccer-style competitions, typical tasks include:

• ball search and tracking

• fast approach and alignment

• kick execution and follow-through

• obstacle avoidance and collision resilience

• localization on a marked field

The G1-Comp’s stated emphasis on agility, stability, and visual recognition APIs directly maps to these requirements.

Motion-intensive research and dynamic control benchmarking

Beyond soccer, G1-Comp is marketed as suitable for motion-intensive challenges where rapid biped locomotion and recovery behaviors are essential. Seller summaries describe it as built for “motion-intensive challenges” and engineered for elite simulation and control workflows.

Simulation-to-real (Sim2Real) training programs

Unitree’s official messaging around simulation integration (Isaac Gym, MuJoCo) and end-to-end training workflows supports the use of G1-Comp in research programs that train locomotion or perception policies in simulation before deployment.

Education and team-based development

Competition humanoids are frequently adopted by universities, robotics clubs, and labs because they offer a structured challenge environment (rules, fields, metrics) that accelerates learning. Product descriptions also frame G1 as useful for teaching, prototyping, and algorithm development.

Advantages / Benefits

Purpose-built for competitive environments

Unitree’s G1-Comp messaging is explicitly centered on competitions, emphasizing match-ready motion control and robustness.
This specialization can reduce development effort for teams whose primary objective is robotic soccer performance.

Faster perception-to-action loop through vision APIs

The inclusion of a YOLO11-based recognition pipeline and a “stadium information interface” can shorten the time required to build usable perception modules for competition play.

Reinforcement learning workflow support

The availability of unitree_rl_gym and simulation integrations supports modern development approaches where locomotion and policy optimization are learned and refined rapidly.

Portable, foldable humanoid form factor

A foldable design can be a meaningful operational advantage for competition teams transporting hardware repeatedly across training sites and events.

FAQ Section 

What is Unitree G1-Comp?

Unitree G1-Comp is a competition-oriented version of Unitree’s G1 humanoid robot platform, designed for robotic soccer and motion-intensive challenges, with match-focused motion control and stadium recognition features.

How does Unitree G1-Comp work?

G1-Comp combines a humanoid body with multi-joint mobility and a perception + control stack that includes YOLO11-based visual recognition, simulation-to-real training workflows, and SDK tools aimed at competition development.

Why is Unitree G1-Comp important?

G1-Comp is important because it provides a purpose-built humanoid foundation for competition robotics, helping teams develop faster with integrated recognition APIs, reinforcement learning workflows, and match-oriented motion design.

What are the benefits of Unitree G1-Comp?

Benefits include competition-oriented motion control, YOLO11 stadium recognition support, Sim2Real workflow tooling, and a foldable humanoid design that can simplify transport and testing.

Summary

Unitree G1-Comp (G1-Comp) is a competition-focused humanoid robot configuration designed for robotic soccer environments and motion-intensive challenges, emphasizing agile locomotion, robust motion control, and stadium perception features. Unitree’s official materials highlight YOLO11-based recognition, reinforcement-learning tooling (unitree_rl_gym), and simulation-to-real deployment workflows intended to accelerate development for competitive match play. Published technical summaries describe a foldable humanoid form factor around 35 kg, with a multi-DoF body, high knee torque figures, and a compute/sensor stack suitable for real-time robotics control. Together, these characteristics position G1-Comp as a specialized platform for teams seeking a practical humanoid foundation for RoboCup-style competitions and dynamic robotics research.