Home
/
Robots
/
AgiBot G1

AgiBot G1

AgiBot G1 is a flexible, AI-powered humanoid robot with up to 43 DoF, dexterous hands, and 2-hour battery life, excelling in dynamic mobility and precise manipulation.
Software Type
Open Source
Software Package
Deep reinforcement learning and imitation learning frameworks. Open SDK for secondary development (G1 EDU). UnifoLM (Unitree Robot Unified Large Model) for collaborative AI development. Cloud-based monitoring and update services.
Actuators
Equipped with high-torque actuators capable of delivering up to 120 Nm, the G1’s joints enable smooth, powerful, and precise movements essential for dynamic locomotion and dexterous manipulation.
Compiute
Powered by onboard AI computing modules (optional NVIDIA Jetson Orin in EDU version), supporting real-time sensor fusion, deep learning inference, and complex motion planning.
Sensors
Features Intel RealSense D435 depth cameras, LIVOX-MID360 3D LiDAR for 360-degree environment mapping, and optional tactile sensor arrays on the dexterous hand for precise force feedback.
Max Op. time
120
mins
16000

Robot Brief

AgiBot G1 is a highly agile and flexible humanoid robot developed by Unitree Robotics, designed to push the boundaries of humanoid mobility and manipulation. Standing approximately 1.3 meters tall and weighing about 35 kg, G1 offers between 23 and 43 degrees of freedom depending on the version, providing a wide range of fluid, human-like motions. Powered by deep reinforcement learning and advanced AI, it can perform dynamic movements such as standing up, folding, kicking, and maintaining balance when pushed. The robot features a modular design with options for a force-controlled dexterous hand (Dex3-1) that enables precise manipulation tasks from smashing walnuts to opening soda bottles and welding. Equipped with Intel RealSense D435 depth cameras and LIVOX-MID360 3D LiDAR for 360-degree environmental perception, G1 understands and interacts with its surroundings effectively. It supports up to 2 hours of battery life with quick battery swaps and folds compactly for portability. The G1 EDU version enhances capabilities with additional joints, tactile sensors, and an optional NVIDIA Jetson Orin module for high-performance computing and secondary development. The robot is suitable for research, education, entertainment, industrial, and collaborative human-robot applications.

Use Cases

The AgiBot G1 performs a broad spectrum of tasks requiring agility and dexterity, including autonomous walking and running at speeds up to 2 m/s, dynamic balancing, and complex object manipulation with its dexterous hand. It can handle delicate and heavy objects with high precision, perform welding, and interact naturally with humans through voice commands and facial recognition. Its AI-driven control system enables continuous learning and adaptation to new tasks and environments, making it ideal for research, industrial automation, education, and entertainment.

Industries

  • Research & Education: Provides a versatile platform for robotics development and AI experimentation.
  • Industrial Automation: Supports precision tasks such as welding, assembly, and material handling.
  • Entertainment: Performs dynamic movements and interactive demonstrations.
  • Healthcare: Assists with patient interaction and light-duty support.
  • Human-Robot Collaboration: Enhances workplace safety and productivity with adaptive interaction.

Specifications

Length
450
mm
Width
200
mm
Height (ResT)
690
mm
Height (Stand)
-
1320
mm
Height (Min)
690
mm
Height (Max)
1320
mm
Weight (With Batt.)
-
kg
Weight (NO Batt.)
-
35
kg
Max Step Height
-
mm
Max Slope
+/-
-
°
Op. Temp (min)
-
°C
Op. Temp (Max)
-
°C
Ingress Rating
-
No items found.

Intro

The AgiBot G1 stands about 130 cm tall and weighs approximately 35 kg. It features:

  • Between 23 to 43 degrees of freedom depending on configuration
  • Maximum joint torque up to 120 Nm for powerful and precise movements
  • Optional Dex3-1 force-controlled dexterous hand for fine manipulation
  • Intel RealSense D435 depth camera and LIVOX-MID360 3D LiDAR for 360° perception
  • Lightweight and foldable design for portability and ease of deployment
  • Advanced AI control based on deep reinforcement learning enabling continuous skill improvement
  • Multi-modal interaction including voice command recognition, facial recognition, and expressive LED face mask.

Connectivity

  • WiFi connectivity (2.4 GHz and 5 GHz bands)
  • USB ports for peripheral integration
  • Bluetooth for short-range communication
  • Cloud connectivity for remote updates and data processing.

Capabilities

  • Dynamic locomotion including walking, running (up to 2 m/s), jumping, and balancing under external forces
  • Complex manipulation such as opening bottles, welding, and handling fragile or heavy objects
  • Real-time environmental perception with 3D LiDAR and depth sensing for obstacle avoidance and navigation
  • Adaptive learning through simulation and reinforcement learning for evolving task performance
  • Human-robot interaction via voice commands, facial recognition, and expressive feedback
  • Compact folding for easy transport and deployment.