CtrlK
The Galaxea R1 Pro is a full-size wheeled dual-arm [[humanoid robot]] developed by [[Galaxea AI]] (also known as Galaxea Dynamics), a Chinese [[embodied AI]] and [[robotics]] startup founded in September 2023. The R1 Pro is the premium variant of the Galaxea R1 series, which also includes the standard [[Galaxea R1]] and the compact [[Galaxea R1 Lite]]. It features 26 [[degrees of freedom]], dual 7-DOF Galaxea A2 robot arms, integrated force sensors, and an [[NVIDIA]] Jetson AGX Orin 32GB computing platform with 200 TOPS of AI inference capability. Designed for industrial collaboration, research, and service applications, the R1 Pro combines a human-like upper body with a wheeled omnidirectional mobile base to provide both dexterous manipulation and reliable indoor mobility.
The R1 Pro was announced in January 2025 alongside the R1 Lite, joining the original R1 model that had been released in 2024. As of early 2026, the R1 Pro is commercially available through Galaxea Dynamics' online store at a price of approximately $69,999 USD, with a lead time of roughly 15 days from order placement.[1][2]
Galaxea AI was founded in September 2023 by a team of scientists and engineers with roots at [[Tsinghua University]], [[Stanford University]], the University of Southern California, and MIT. The company is headquartered in Beijing, China.
The company was co-founded by four individuals. CEO Gao Jiyang studied at Tsinghua University as an undergraduate and earned his Ph.D. in computer vision from the University of Southern California. He previously worked at [[Waymo]] on perception systems and at Momenta on autonomous driving production systems. Co-founder and Co-Chief Science Officer Xu Huazhe holds a Ph.D. from UC Berkeley and conducted postdoctoral research at the Stanford Vision and Learning Lab before becoming a tenure-track assistant professor at Tsinghua University, where he leads the Tsinghua Embodied AI Lab (TEA Lab). Additional co-founders include Hang Zhao, who holds a Ph.D. from MIT, and Tianwei Li, a former Senior Director at Momenta.[3][4]
Galaxea AI has raised over $500 million in cumulative funding as of April 2026. Early investment rounds in 2024 were backed by Baidu Ventures, GSR Ventures, and IDG Capital. In 2025, the company raised over $100 million across its A4 and A5 rounds at a $700 million valuation, with backing from Capital Today, Meituan's Long-Z Investments, Ant Group, IDG Capital, and GL Ventures. The company's Series B round in February 2026 brought in approximately $144 million at a $1.4 billion valuation, and a Series B+ round in April 2026 raised approximately $291 million at a $2.9 billion valuation, with participation from nearly 20 institutional investors including Lens Technology and several state-backed capital funds.[5][6][7]
The company serves more than 40 clients, including Huawei Cloud, [[Volkswagen]], Haier, [[Samsung]], [[ByteDance]], Physical Intelligence, Stanford University, and MIT. Its stated mission is to build "embodied intelligence at a global scale: 10 billion robots for 10 billion people."[4]
The R1 Pro follows a design philosophy that pairs a human-like upper body with a wheeled mobile base rather than bipedal legs. This approach prioritizes practical deployability in structured indoor environments such as factories, warehouses, and laboratories over the terrain versatility offered by bipedal humanoid competitors like [[Tesla Optimus]], [[Unitree Robotics]]' H1, and [[UBTECH Robotics]]' Walker. Wheels provide greater stability, simpler control, and more mature locomotion technology, allowing the R1 Pro to focus its engineering complexity on upper-body manipulation rather than balance and walking.
The R1 Pro sits at the top of the R1 product line. Where the standard R1 uses 6-DOF Galaxea A1 arms and the R1 Lite serves primarily as a data collection platform, the R1 Pro upgrades to 7-DOF Galaxea A2 arms with integrated force sensing. This additional degree of freedom per arm gives the R1 Pro greater kinematic flexibility, allowing the arms to reach around obstacles and maintain more comfortable joint configurations during complex manipulation tasks.
The R1 Pro stands 1,700 mm (170 cm) tall in its standard upright posture. The wheeled chassis measures 675 mm in length, 636 mm in width, and 346 mm in height. The total system weight is 96 kg including the battery.[8]
The robot's operating height range extends from ground level up to 2,000 mm (200 cm), enabled by the torso's lift mechanism. This wide vertical workspace allows the R1 Pro to pick items off the floor and place them on high shelves without repositioning.
The R1 Pro has 26 total degrees of freedom distributed across four subsystems:[8]
| Subsystem | DOF | Description |
|---|---|---|
| Chassis | 6 | Three-wheel vector steering providing Ackermann, translation, and spinning motion modes |
| Torso | 4 | Lift, tilt, and swivel; waist yaw range of +/- 175 degrees, hip pitch range of -105 to 90 degrees |
| Each Arm (Galaxea A2) | 7 | 7-DOF per arm (14 total for both arms) |
| Each Gripper (Galaxea G1) | 1 | Parallel gripper with 0 to 100 mm stroke (2 total) |
The R1 Pro is equipped with two Galaxea A2 robot arms and two Galaxea G1 parallel grippers. The A2 arm is a 7-DOF unit, one degree of freedom more than the A1 arm used in the standard R1 and R1 Lite. This extra joint provides superior kinematic redundancy, enabling the arm to reach the same endpoint through multiple joint configurations and to navigate around obstacles more effectively.
Key arm performance metrics include:[8]
| Parameter | Value |
|---|---|
| Working radius (arm only) | 620 mm |
| Working radius (with gripper) | 860 mm |
| Arm weight (per arm) | 7.8 kg |
| Rated single-arm payload (at 0.6 m) | 3.5 kg |
| Maximum single-arm payload (at 0.6 m) | 5 kg |
| Maximum dual-arm payload | 10 kg |
| Maximum end-effector linear velocity | 7.5 m/s |
| Maximum end-effector acceleration | 10 m/s squared |
| Repeat positioning accuracy | +/- 0.5 mm |
The Galaxea G1 gripper provides 100 N of rated gripping force with a stroke range of 0 to 100 mm. The arms can alternatively be fitted with dexterous hands for tasks requiring more anthropomorphic finger manipulation.[2][8]
The torso assembly provides four degrees of freedom that expand the robot's workspace and enable naturalistic upper-body movement. The waist yaw motor delivers a range of +/- 175 degrees with a rated torque of 108 Nm and a maximum torque of 180 Nm. The hip pitch mechanism provides a range from -105 to 90 degrees. Combined with the torso's vertical lift capability, these joints allow the R1 Pro to reach from ground level up to 2 meters high while maintaining stability.[8]
The R1 Pro's chassis uses a three-wheel vector steering design with 360-degree omnidirectional movement capability. The chassis accepts speed commands in three simultaneous directions (x, y, and angular velocity), enabling smooth translational sliding, spinning in place, and Ackermann-style steering for forward travel. The maximum travel speed is approximately 5.4 km/h (1.5 m/s).[8][9]
The R1 Pro runs on a 48V lithium-ion battery pack with the following specifications:[8]
| Parameter | Value |
|---|---|
| Rated voltage | 48 V |
| Capacity | 35 Ah |
| Energy | 1,680 Wh |
| Chemistry | Lithium-ion |
| Battery management system | Supported |
| Approximate runtime | 2 to 3 hours |
The R1 Pro is powered by an [[NVIDIA Jetson]] AGX Orin 32GB module, providing substantial onboard AI compute for perception, planning, and control tasks.[8]
| Parameter | Value |
|---|---|
| CPU | 8-core, 2.2 GHz |
| AI inference performance | 200 TOPS |
| Memory | 32 GB LPDDR5 |
| Storage | 1 TB SSD |
| Camera interfaces | 8x GMSL |
| Network | 4x Gigabit Ethernet (M12 connectors), M.2 Wi-Fi (AP mode) |
| Ports | USB 3.0, HDMI, Ethernet |
The computing unit runs [[Linux]] with [[ROS]] 2 Humble as the primary middleware framework. A low-noise local air cooling system with two air ducts (82 cubic meters per minute total airflow) maintains thermal stability during intensive computation.[8]
The R1 Pro carries a comprehensive array of sensors for perception, navigation, and manipulation:[8]
| Sensor | Count | Specifications |
|---|---|---|
| Head cameras (monocular) | 2 | 1920x1080 at 30 fps, 118-degree horizontal x 62-degree vertical FOV, 120 mm baseline |
| Wrist cameras (depth, optional) | 2 | 1280x720 at 30 fps, 87-degree horizontal x 58-degree vertical FOV, 0.2 to 3 m range |
| Chassis cameras (monocular) | 5 | 1920x1080 at 30 fps |
| LiDAR | 1 to 2 | 360-degree coverage, 905 nm wavelength, 0.1 m minimum range |
| Force sensors | Integrated | Built into the arms for force-controlled manipulation |
| IMU | Yes | Inertial measurement unit for pose estimation |
The integrated force sensors are a distinguishing feature of the R1 Pro compared to the standard R1, which does not include them. These sensors enable compliant force-controlled interaction with the environment, which is critical for tasks involving delicate objects, safe human proximity, or precision assembly.[2]
The R1 Pro uses a low-noise local air cooling system with two air ducts providing a combined airflow of 82 cubic meters per minute. This system keeps the computing platform, motors, and other heat-generating components within operating temperature ranges during extended use.[8]
The R1 Pro sits alongside the standard R1 and the R1 Lite in Galaxea's product lineup. Each model targets different use cases while sharing a common design language.
| Specification | R1 | R1 Pro | R1 Lite |
|---|---|---|---|
| Total DOF | 24 | 26 | 23 |
| Arm DOF (per arm) | 6 | 7 | 6 |
| Arm model | Galaxea A1 | Galaxea A2 | Galaxea A1X |
| Torso DOF | 4 | 4 | 3 |
| Standing height | 170 cm | 170 cm | ~98 cm |
| Maximum operating height | 200 cm | 200 cm | 170 cm |
| Weight (with battery) | 96 kg | 96 kg | 55 kg |
| Working radius (arm) | 70 cm | 62 cm (86 cm with gripper) | 60 cm |
| Rated single-arm payload | 3.5 kg | 3.5 kg | 3 kg |
| Maximum single-arm payload | 5 kg | 5 kg | 5 kg |
| Maximum dual-arm payload | 10 kg | 10 kg | 10 kg |
| Battery capacity | 1,680 Wh | 1,680 Wh | 720 Wh |
| Computing platform | NVIDIA Jetson AGX Orin 32GB | NVIDIA Jetson AGX Orin 32GB | Intel Core i9-12900HK |
| Head cameras | 1 stereo depth camera | 2 monocular cameras | 1 binocular camera |
| Wrist cameras | 2 depth cameras | 2 depth cameras (optional) | 2 monocular depth cameras |
| Chassis cameras | 5 | 5 | None |
| LiDAR | 1 (optional 2nd) | 1 to 2 | 1 |
| Force sensors | No | Integrated | No |
| Teleoperation modes | Isomorphic + VR | Isomorphic + VR | Isomorphic only |
| Speech interaction | Yes | Yes | Not specified |
| Maximum speed | 5.4 km/h | 5.4 km/h | 5.4 km/h |
| Approximate price (USD) | ~$28,000 | ~$70,000 | ~$40,000 |
| Primary use case | General-purpose research and industry | Industrial collaboration and complex manipulation | Data collection for AI training |
The R1 Pro's key advantages over the standard R1 are its 7-DOF arms (providing one additional degree of freedom per arm for enhanced dexterity), integrated force sensors for compliant manipulation, and its focus on precision industrial tasks. Compared to the R1 Lite, the R1 Pro offers significantly greater height, a more powerful computing platform (Jetson Orin vs. Intel NUC), comprehensive sensor coverage, and VR teleoperation support.[2][8][10]
The R1 Pro supports two teleoperation modes: isomorphic control and VR-based control. Both modes serve dual purposes as direct operational control mechanisms and as tools for collecting training data for [[machine learning]] models.
Galaxea offers the R1-T Isomorphic Remote Operation Platform, a scaled-down physical replica of the R1 Pro that enables operators to control the robot through direct body movement mapping. The isomorphic system provides full-joint mapping between the operator's movements and the robot's kinematics, achieving millimeter-level precision with millisecond-level response times.[2][11]
This approach has several advantages over other teleoperation methods. The operator's arms naturally stay within reachable postures, avoiding inverse kinematics failures. No retargeting is needed between the operator's body dimensions and the robot's kinematic chain. Force feedback travels from the robot back to the operator, enabling tactile awareness of the manipulation environment.
The R1-T hardware connects to the host computer via CAN bus and USB-CAN interfaces. Default chassis velocities during teleoperation are 0.2 m/s for linear motion and 0.6 rad/s for angular motion. Both wired and wireless connections are supported, with wired connections recommended for stability.[11]
The R1 Pro also supports VR-based teleoperation using a VR headset and controllers. This mode maps the operator's hand and head movements to the robot's arms and camera viewpoint, providing an immersive remote control experience. VR teleoperation is useful in scenarios where physical proximity to an isomorphic control platform is not practical, such as controlling robots at remote sites or operating multiple units from a centralized station.[12]
[[Teleoperation]] plays a central role in Galaxea AI's approach to embodied AI development. Rather than relying primarily on [[reinforcement learning]] in simulation, the company emphasizes real-world data collection through teleoperated demonstrations paired with [[imitation learning]]. CEO Gao Jiyang has stated that "at least for the next three years, reinforcement learning with simulators will not lead to the ultimate goal" for upper-limb manipulation tasks. The teleoperated data collected by R1 Pro and R1 Lite units feeds directly into the training pipeline for Galaxea's AI models.[4]
The R1 Pro runs ROS 2 Humble on Ubuntu Linux. Galaxea provides a comprehensive ROS 2 SDK (the Galaxea ATC SDK) with modules for joint control, arm pose control, torso speed control, chassis control, and pose estimation. The SDK includes drivers for all onboard actuators and sensors, along with interfaces for navigation, SLAM, and teleoperation. Documentation covers hardware setup, software installation, teleoperation tutorials, autonomous navigation, and demo guides.[13][14]
The navigation subsystem uses multi-sensor fusion combining [[LiDAR]] point clouds, IMU data, and camera feeds for accurate localization. The system supports global and local path planning with obstacle avoidance functionality.[9]
Galaxea AI has developed a proprietary family of AI models called G0 that serve as the intelligence layer for the R1 series. The G0 system uses a dual-system [[Vision-Language-Action]] (VLA) architecture:[15]
The two systems run asynchronously at different frequencies, with System 2 operating at a slower planning cadence and System 1 executing actions at higher frequency for responsive real-time control.
G0 is trained on the Galaxea Open-World Dataset, comprising over 500 hours of teleoperation data collected across more than 150 distinct tasks in 50 real-world scenes. Training follows a three-stage curriculum: cross-embodiment pre-training, single-embodiment pre-training on R1 hardware, and task-specific post-training.[15]
In January 2026, Galaxea released G0Plus, an updated VLA model for multi-task robot manipulation, along with G0Tiny, a 250-million-parameter lightweight model optimized for edge deployment on the R1 Pro's Jetson Orin computing platform. The G0 models are open-sourced through Galaxea's GalaxeaVLA repository on GitHub and model weights are available on [[Hugging Face]].[16]
The R1 Pro's primary commercial applications include bin picking, line feeding, kitting, parcel handling, and assembly tasks in factory and warehouse environments. Its 10 kg dual-arm payload, force-controlled interaction, and +/- 0.5 mm repeat positioning accuracy make it suitable for tasks requiring both strength and precision. The integrated force sensors enable safe human-robot collaboration in shared workspaces, a critical requirement for industrial deployments where robots operate alongside human workers.[2][4]
Universities and research laboratories use the R1 Pro as a testbed for learning-based control, dexterous manipulation, and human-robot interaction research. The platform's ROS 2 compatibility, open-source VLA models, comprehensive documentation, and NVIDIA Jetson computing platform make it accessible for academic use. Notable research clients include Stanford University and MIT. The R1 Pro has been used in the BEHAVIOR Robot Suite, a benchmark for real-world whole-body manipulation of everyday household activities.[4][17]
The R1 Pro can perform repetitive tasks such as restocking shelves, delivering items, and organizing inventory in hospitality and retail environments. Its autonomous navigation capabilities, combined with speech interaction features, allow it to interact with people in customer-facing settings. The wheeled design is well-suited to the flat floors and structured layouts typical of retail stores, hospitals, and office buildings.[2]
Although the R1 Pro is currently targeted primarily at commercial and research customers, Galaxea AI envisions eventual deployment of humanoid robots in home environments. The G0 VLA models have been tested on household tasks including cleaning, bed-making, and object retrieval. CEO Gao Jiyang has predicted that humanoid robots capable of cooking and cleaning will enter homes within less than a decade.[4]
The Galaxea A2 is the 7-DOF robot arm used exclusively in the R1 Pro. It represents an upgrade over the 6-DOF Galaxea A1 arm found in the standard R1, adding one additional degree of freedom for enhanced kinematic redundancy. Each A2 arm weighs 7.8 kg and has a working radius of 620 mm. The arms use force-controlled actuation for compliant interaction, enabling safe collaboration with humans and gentle handling of fragile objects.[8]
The G1 is a self-developed parallel gripper included with all R1 variants. It provides 100 N of gripping force with a 0 to 100 mm stroke range. The two-finger design is suitable for a wide variety of grasping tasks, from picking small components to handling larger objects.[8]
The R1 Pro's arm endpoints can be swapped from grippers to dexterous hands for applications requiring anthropomorphic finger manipulation. At CES 2026, Galaxea showcased the DEXO dexterous hand, which features 0.1 N high-precision tactile sensing and 1 kg fingertip force per finger. Experiments have been conducted using 22-DOF dexterous robot hands on the R1 Pro platform.
Galaxea also produces the A1 (6-DOF) and A1XY (6-DOF dual-configuration) standalone robot arms for desktop research applications. The A1 is the arm model integrated into the standard R1, while the A1XY serves as a lightweight tabletop development platform.
The Galaxea R1 Pro is listed on the Galaxea Dynamics online store at $69,999 USD, with educational discounts and bulk-order pricing available upon request by contacting product@galaxea-dynamics.com. The standard lead time is approximately 15 days from order confirmation. The R1 Pro is the most expensive model in the R1 series, reflecting its upgraded 7-DOF arms, integrated force sensors, and enhanced sensor suite compared to the standard R1 (approximately $28,000 USD) and the R1 Lite (approximately $40,000 USD).[1][2]
The R1 Pro competes in the growing market for wheeled humanoid and mobile manipulation platforms. Within China, competitors include [[Unitree Robotics]]' G1 and H1 bipedal humanoids, [[UBTECH Robotics]]' Walker, [[Agibot]]'s A2 series, and various platforms from companies such as Galbot and LimX Dynamics. In the international market, comparable platforms include Mobile ALOHA from Stanford and Google DeepMind, and dual-arm mobile manipulators from companies like Agility Robotics.
Galaxea AI differentiates the R1 Pro through several factors: its emphasis on real-world data collection and VLA-based AI models (rather than simulation-heavy approaches), its comprehensive open-source software ecosystem, and its commercially available teleoperation platform for seamless data collection. The company's rapid fundraising trajectory and large client base (over 40 organizations) suggest strong commercial traction in both the Chinese and international markets.[4][5]
Galaxea AI has outlined several directions for the evolution of the R1 platform and its broader product line: