Unitree G1 Edu Pro A U9 Humanoid Robot (G1EDU-U9)

Unitree G1 Edu Pro A U9 Humanoid Robot (G1EDU-U9)

The G1EDU-U9's confirmed hardware  Dex3-1 three-finger hands with tactile sensors at 37 total DOF — positions it at an intermediate tier between the 35-DOF U10 (BrainCo Revo 2 Basic five-finger, no tactile) and the 41-to-43-DOF Ultimate tier configurations (U3 through U8). At 37 DOF, the U9 sits at what BotInfo.ai calls the "manipulation threshold" — the 35 to 37 DOF range where manipulation research becomes substantively enabled  while retaining the three-finger Dex3-1 tactile configuration of the U4 in a slightly less articulated body structure.

G1 Foldable Frame: 132 cm, 35 kg, Portable

The G1 EDU U9 uses the standard G1 physical frame. Standing dimensions are 1320 x 450 x 200 mm with a folded configuration of 690 x 450 x 300 mm for transport. Body mass is approximately 35 kilograms with battery. The foldable design enables single-operator transport and storage in standard equipment cases — consistently practical across all G1 EDU configurations.

The aluminum alloy structural components, ABS/PC blend exterior panels, and full joint hollow internal wiring design are standard G1 family elements. The 9,000 mAh quick-release battery provides approximately two hours of operational runtime.

37 Degrees of Freedom: Manipulation Threshold with Tactile Hands

RobotShop's confirmed U9 specification states: "The G1 features a total of 37 DoF, with each leg having 6 DoF, a waist joint adding 1 DoF, and each arm featuring 5 DoF for intricate task execution." Adding bilateral Dex3-1 hands (7 DOF each, 14 total) to this 23-DOF body reaches the 37-DOF total.

Body and hand DOF distribution:

• Legs: 6 DOF per leg (bilateral, 12 total) — full hip-knee-ankle articulation for stable bipedal locomotion and athletic behaviors
• Waist: 1 DOF (yaw ±155°) — torso rotation
• Arms: 5 DOF per arm (bilateral, 10 total) — shoulder-elbow-wrist control without the additional wrist yaw of the 7-DOF Ultimate arm
• Dex3-1 Hands with Tactile: 7 DOF per hand (thumb: 3, index: 2, middle: 2), bilateral, 14 total

Dex3-1 Three-Finger Hands with Tactile Sensors

The G1 EDU U9 is described as "Equipped with 2 Dex3-1 force-controlled 3-finger dexterous hands (with tactile sensor)." The Dex3-1 is Unitree's proprietary three-finger hand with the confirmed kinematic distribution: thumb (3 DOF), index (2 DOF), middle (2 DOF), providing 7 active degrees of freedom per hand.

The tactile sensor arrays in the U9's Dex3-1 hands are the configuration's primary research differentiator over simpler G1 EDU body tiers: the distributed tactile sensing enables contact-aware manipulation that joint torque estimation alone cannot provide. The Dex3-1 with tactile sensors enables:

Grasp quality assessment: Mapping contact pressure distribution across the three fingers to evaluate whether a grasp is stable — a spatial pattern that indicates if contact is distributed appropriately for the object's geometry or concentrated in a way that risks slip.

Slip detection: Detecting the change in tactile sensor readings that precedes full object sliding — the pre-slip shear and pressure redistribution that occurs before a drop event becomes inevitable.

Force-regulated grasping of fragile objects: Using tactile feedback to regulate grip closure force in real time, preventing excessive grip pressure on pharmaceutical packaging, glass containers, or delicate electronic components.

Tactile-based contact verification: Confirming that the hand has made the intended contact pattern with a target object before initiating a manipulation sequence, reducing failures from partial or misaligned grasps.

Force-position hybrid control with contact feedback: Operating the Dex3-1's force-position control mode informed by the distributed tactile readings rather than relying solely on motor current for contact force estimation.

Technology and Specifications

G1 EDU U9 (G1EDU-U9) Full Specifications

100-TOPS AI Compute for On-Device Research

RobotShop confirms the U9 is "Built-in 100 TOPS computing unit." At 100 TOPS, the compute platform provides GPU-accelerated inference for tactile grasp policy networks, visual manipulation models, and the UnifoLM multimodal AI system. For the U9's 37-DOF body with 5-DOF arms, the 100-TOPS compute provides the AI capability needed for manipulation research while the 5-DOF arm limits the action space slightly compared to the 7-DOF Ultimate tier.

The standard G1 EDU research workflow applies: train policies in Isaac Sim or MuJoCo using the G1 URDF, apply domain randomization, deploy to the 100-TOPS compute unit via unitree_sdk2, and iterate with real-world data collection through VR teleoperation for imitation learning refinement.

UnifoLM-VLA-0 and Open-Source Manipulation Baseline

The March 2026 open-source release of UnifoLM-VLA-0 (built on Qwen2.5-VL-7B, covering 12 manipulation task categories) provides U9 buyers with an immediately deployable manipulation policy baseline. For the U9's Dex3-1 tactile configuration, relevant task categories include those involving fragile object handling, pick-and-place with grasp verification, and tool use where tactile feedback provides contact confirmation.

Full SDK: Python, C++, ROS 2

Complete secondary development support through Python, C++, and ROS 2 SDK access enables the full research development workflow on the U9. ROS 2 integration via Unitree's SDK2 (CycloneDDS) connects the U9 with the full ROS 2 manipulation ecosystem — MoveIt2 for manipulation planning with the Dex3-1's 7-DOF hand configuration, tactile data integration for feedback control, Nav2 for navigation, and rosbag2 for synchronized multi-modal dataset recording (joint state, tactile sensor data, camera streams).

Applications and Use Cases

Tactile Manipulation Research at the Manipulation Threshold

The U9's 37 DOF with Dex3-1 tactile hands places it at BotInfo.ai's "35–37 = manipulation threshold" tier — the DOF level where manipulation research becomes substantively enabled. The Dex3-1 tactile hands provide the contact sensing needed for research on grasp quality assessment, slip detection, and force-regulated grasping of fragile objects, in a body configuration that may be more accessible in price than the full 7-DOF Ultimate tier while providing equivalent hand capability.

AI Learning Programs with Tactile Sensing

OpenELAB's explicit positioning of the U9 for "AI learning" reflects the configuration's appropriateness for curriculum programs that include tactile sensing as a study topic. The Dex3-1 tactile hands enable students and researchers to work with contact sensing in the robot manipulation curriculum — developing control strategies that use tactile data beyond what joint torque estimation provides.

Fragile Object Handling in Service and Healthcare Robotics

Research on robotic handling of fragile objects — pharmaceutical packaging, medical supplies, laboratory glassware — benefits from the U9's Dex3-1 tactile sensing for force regulation. For healthcare robotics research programs where the robot must handle medications or medical equipment without damaging them, the tactile feedback from the Dex3-1 arrays enables the force-regulated control that positions research requires.

Physical HRI with Contact Force Measurement

For physical human-robot interaction research where the robot's contact force during handshaking, object handover, or physical collaboration must be measured and regulated, the U9's Dex3-1 tactile arrays provide contact pressure data that enables force regulation during physical contact with a human hand. This is particularly relevant for research where the robot must adjust grip force based on detecting the soft compliance of human skin contact versus the rigid resistance of an object.

Loco-Manipulation at the DOF Threshold

Combining bipedal locomotion with manipulation at the 37-DOF level — where the 5-DOF arm and 1-DOF waist provide meaningful reach and body coordination without the full Ultimate tier kinematics — enables loco-manipulation research where the tactile feedback from the Dex3-1 hands is the primary research focus rather than the full arm workspace coverage.

Comparison: U9 in Context

The U9 occupies a unique position: it provides the Dex3-1 tactile sensing of the U4 in a 5-DOF arm, 1-DOF waist body — the same DOF body structure as the U10 (35 DOF) with BrainCo five-finger hands, but with the Dex3-1 three-finger tactile configuration instead. Research programs where 5-DOF arms cover their reach requirements and Dex3-1 tactile sensing is the primary research hardware can consider the U9 as potentially more cost-accessible than the U4 while providing equivalent tactile capability.

Advantages and Benefits

Dex3-1 Tactile Sensing at the 5-DOF Arm Body Tier: The U9 provides Dex3-1 tactile hand capability — contact pressure mapping, slip detection, force regulation — in a body configuration that may be priced below the 7-DOF arm Ultimate tier, making tactile manipulation research accessible to programs with tighter equipment budgets.

37-DOF Total: Full Manipulation Research Enablement: At BotInfo.ai's "35–37 = manipulation threshold," the U9's 37 DOF enables meaningful manipulation research — grasping, transport, basic assembly, and physical interaction — with the tactile sensing layer that makes its results generalizable to contact-aware manipulation.

100-TOPS On-Device AI for Tactile Policy Research: The built-in 100-TOPS compute enables neural network-based tactile grasp policy inference on-device, supporting research on learned tactile manipulation without cloud compute dependency.

Dual Encoders Per Joint for Position Fidelity: The dual encoder standard extends to all 37 joints of the U9, providing accurate joint state feedback that supports the precise manipulation control the Dex3-1 tactile hands enable.

Full SDK and ROS 2 for Complete Research Workflow: Complete programmatic access and ROS 2 integration enables the standard G1 EDU research workflow across the U9's 37-DOF body and Dex3-1 tactile hands.

12-Month Warranty (24-Month EU/UK at OpenELAB): Standard institutional warranty coverage through authorized distributors.

Frequently Asked Questions (FAQ)

What is the Unitree G1 EDU U9 (G1EDU-U9)? The Unitree G1 EDU U9 (G1EDU-U9) is an upper-tier research configuration of the Unitree G1 humanoid robot, listed by RobotShop as the "G1 EDU Ultimate E (U9)" and by OpenELAB as an "EDU Kit for AI Learning and Research Labs." It features 37 total degrees of freedom (6 DOF per leg, 1 DOF waist, 5 DOF per arm, plus Dex3-1 hand DOF), two Dex3-1 force-controlled three-finger dexterous hands with integrated tactile sensor arrays, a built-in 100-TOPS AI compute unit, LIVOX MID-360 3D LiDAR, Intel RealSense D435i, dual encoders per joint, and full secondary development support through Python, C++, and ROS 2 SDK.

What makes the G1 EDU U9 different from the G1 EDU U4 (Ultimate B)? Both the U4 and U9 feature Dex3-1 three-finger hands with tactile sensors. The key difference is the body DOF configuration: the U4 (EDU Ultimate B, $66,277) has 43 total DOF with 7-DOF arms and 3-DOF waist — the full Ultimate tier body. The U9 has 37 total DOF with 5-DOF arms and 1-DOF waist — a more constrained arm kinematics at a potentially lower price. Research programs where 5-DOF arms cover their grasp approach requirements can consider the U9 as a cost-accessible alternative to the U4's tactile configuration.

Is the G1 EDU U9 the same as the "G1 EDU Pro A U9"? Based on verified distributor listings as of April 2026, the G1EDU-U9 is listed by RobotShop as "G1 EDU Ultimate E (U9)" — not "G1 EDU Pro A U9." OpenELAB refers to it as "G1edu U9 — EDU Kit for AI Learning and Research Labs." The "Pro A" naming does not appear in verified authorized distributor documentation. Buyers should confirm the specification (37 DOF, Dex3-1 with tactile sensors, 100-TOPS compute) when ordering regardless of the naming convention used by their distributor.

What research is the G1 EDU U9 most suitable for? The U9 is well-suited for tactile grasp policy learning and testing, fragile object handling with force regulation, slip detection research, physical HRI studies requiring contact force measurement, healthcare robotics prototyping with delicate object handling, and AI learning programs that include tactile sensing as a curriculum topic. The 5-DOF arm configuration makes it more suitable for research where grasp approach direction is constrained to the 5-DOF workspace; the 7-DOF Ultimate tier provides greater arm workspace coverage for programs requiring unconstrained approach direction.

Summary

The Unitree G1 EDU U9 (G1EDU-U9) — listed by RobotShop as the "G1 EDU Ultimate E (U9)" with 37 degrees of freedom, Dex3-1 three-finger hands with tactile sensors, 100-TOPS AI compute, and dual encoders per joint — provides the Dex3-1 tactile hand configuration of the U4 in a 5-DOF arm, 1-DOF waist body structure at the "manipulation threshold" DOF tier. For research programs studying tactile grasp quality assessment, fragile object handling, slip detection, or physical HRI force regulation where 5-DOF arm kinematics covers their manipulation approach requirements, the U9 provides the tactile sensing research capability at a potentially lower price point than the 7-DOF arm U4. Available through RobotShop globally and OpenELAB with institutional warranty coverage, the U9 extends the Dex3-1 tactile manipulation research option across the full G1 EDU configuration range.