Galaxea

Introduction / Overview

Galaxea AI (Galaxea Dynamics, 星海图) builds robots the way modern embodied-AI research actually works: as a full stack — hardware, teleoperation, learning-based control, and open datasets engineered together rather than bolted on. The company's centerpiece is the R1 series of dual-arm wheeled humanoids: full-size mobile manipulators whose vertical operating range spans from the floor to 2 meters — a workspace envelope covering ground-level bins to top shelves that most humanoids simply cannot reach at both extremes. Built from Galaxea's own component families — the lightweight, high-dynamic, force-controlled A-series arms (A1/A1X), the G-series grippers, a 4-DoF articulated torso, and vector-steering wheeled chassis — the R1 platforms pair mechanical reach with the perception and compute embodied AI demands: NVIDIA Jetson AGX Orin (200 TOPS) across the range, with the flagship R1 Pro carrying seven HD cameras, 360° LiDAR, wrist cameras, and integrated force sensing.

The series is tiered with unusual clarity. The R1 Pro leads: 26 degrees of freedom, dual 7-axis arms with 10 kg arm load, sub-millimeter arm repeatability, and a 6-DoF omnidirectional chassis. The standard R1 carries 24 DoF with dual 6-axis A1 arms and G1 grippers on the three-wheel vector chassis. The R1 Lite opens the range at 23 DoF with a 1.7 m reach ceiling and two-finger grippers. All three pair with the R1-T isomorphic teleoperation platform — full-joint mapping at millimeter precision and millisecond response — turning every robot into a demonstration-data engine. Around the hardware, Galaxea publishes open-world datasets, ROS2 stacks, and learning-based control integrating vision-language-action models and diffusion policies — the reason academic groups researching whole-body household manipulation have explicitly selected the platform. For labs, embodied-AI teams, and industrial pilots, Galaxea is the mobile-manipulation stack that arrives research-ready.

Design and Features

The R1 Architecture — Reach as a Feature

• Floor-to-2-meter workspace — the R1's defining envelope: the 4-DoF torso and arm geometry let it pick from the ground and place at 2 meters, covering the full vertical span of shelves, carts, counters, and bins in human environments (~700 mm horizontal radius).
• Wheeled by design — vector-steering and (on the Pro) 360° omnidirectional chassis deliver the stability, endurance, and indoor agility that make wheeled humanoids consistently higher-throughput than bipeds in real manipulation work — maneuvering tightly through labs, pilot factory zones, and service spaces.
• Modular component families — the same A-series arms and G-series grippers sold standalone compose the R1, meaning spares, upgrades, and custom configurations stay within one coherent ecosystem.

The Three Tiers

• R1 Pro — 26 DoF; dual 7-axis arms with 10 kg arm load and 64 cm reach; sub-millimeter repeatability; integrated force sensors and wrist camera; 7 HD cameras plus 360° LiDAR; speech interaction; isomorphic and VR teleoperation; 6-DoF omnidirectional chassis — the flagship for demanding manipulation and perception-heavy research.
• R1 — 24 DoF; dual 6-axis A1 arms with 70 cm reach and G1 grippers; full floor-to-2 m range on the three-wheel vector chassis — the balanced core platform.
• R1 Lite — 23 DoF; dual 6-axis arms, 70 cm reach, 1.7 m ceiling; two-finger gripper; dual central cameras; isomorphic teleoperation — the accessible entry for education and cost-sensitive labs.

All models: NVIDIA Jetson AGX Orin 32GB, 8-core CPU, 200 TOPS GPU.

A-Series Arms and G-Series Grippers

The A1/A1X arms are Galaxea's force-controlled foundation: lightweight, highly dynamic manipulators with high payload-to-weight ratios and joint performance suited to teaching, research, and mobile integration — safe, compliant interaction as a design principle. The G1 grippers complete dexterous and coordinated bimanual grasping.

R1-T — Teleoperation as Data Pipeline

The R1-T isomorphic remote-operation platform maps an operator's joints to the robot's at millimeter precision with millisecond response — the demonstration-collection front end for imitation learning, and the practical teleoperation interface for deployment.

The Open Full Stack

• Learning-based control — vision-language-action models and diffusion policies mapping perception and natural-language prompts to action.
• Open-world datasets — published data resources bootstrapping training and cross-task transfer.
• ROS2 workflows and documentation — standard research integration, with the open models-datasets-docs combination explicitly aimed at reproducibility and lowering lab adoption barriers.

Technology and Specifications

Choosing Within the Range

Confirm current datasheets before specifying — configurations evolve, and datasheets are available on request.

Why Full-Stack Wins in Embodied AI

Research time dies in integration: mismatched arms, undocumented drivers, absent datasets, hand-rolled teleoperation. Galaxea's proposition is that hardware, teleop, learning stack, and data ship as one designed system — the reason published academic work has selected the R1 family for whole-body manipulation requirements like bimanual coordination, stable navigation, and end-effector reachability.

Position in the Market

Among wheeled mobile manipulators, the R1 series sits between the premium Astribot S1 (speed-precision showcase) and industrial platforms like the DexForce W1 Pro (shift-endurance manufacturing) — differentiated by its floor-to-2 m workspace, three-tier accessibility, component-ecosystem modularity, and the most research-native open stack in the class.

Applications and Use Cases

Embodied-AI Research

The design center: VLA and diffusion-policy research, whole-body manipulation studies, and benchmark work on a platform with published datasets and academic citations already behind it.

Manipulation Data Collection

R1-T isomorphic teleoperation turns R1 fleets into demonstration engines for imitation-learning pipelines and foundation-model training.

Industrial Prototyping and Logistics

Bin picking, line feeding, kitting, and parcel handling pilots — force-controlled arms and mobile reach applied to the material-flow tasks wheeled manipulators serve best.

Laboratory Automation and Service

Floor-to-2 m reach across benches, shelves, and carts suits lab logistics and indoor service scenarios in human-built spaces.

Education

The R1 Lite plus standalone A1 arms give robotics programs an affordable, ROS2-native curriculum ladder within one ecosystem.

Advantages / Benefits

• The workspace envelope — genuine floor-to-2-meter reach few platforms in any class match.
• Three tiers, one ecosystem — Lite to Pro on shared components, software, and teleoperation.
• Research-native by construction — open datasets, ROS2, VLA/diffusion stacks, and academic adoption already in print.
• Force-controlled safety — compliant A-series arms engineered for interaction, not just actuation.
• Data pipeline included — R1-T isomorphic teleoperation closes the demonstration-to-policy loop.
• Component modularity — arms and grippers sold standalone keep configurations and repairs in-house.

For buyers researching where to buy Galaxea robots, or comparing R1 Pro, R1, and R1 Lite prices and costs against the Astribot S1 and other mobile manipulators, the evaluation should weigh workspace envelope, tier fit, stack openness, teleoperation capability, and ecosystem coherence — the dimensions where Galaxea's full-stack design leads.

FAQ

What is Galaxea?

Galaxea AI (Galaxea Dynamics) is an embodied-AI robotics company building the R1 series of dual-arm wheeled humanoids, the force-controlled A-series arms and G-series grippers, and the R1-T teleoperation platform — unified by open datasets, ROS2 stacks, and learning-based control.

What is the Galaxea R1?

The R1 is a full-size dual-arm wheeled humanoid combining two 6-DoF A1 arms with G1 grippers, a 4-DoF torso, and a vector-steering chassis — operating from floor level to 2 meters of height with a ~700 mm horizontal radius, on NVIDIA Jetson AGX Orin compute.

What is the difference between the R1 Pro, R1, and R1 Lite?

The R1 Pro (26 DoF) adds dual 7-axis arms with 10 kg arm load, sub-millimeter repeatability, seven HD cameras with 360° LiDAR, force sensing, and an omnidirectional chassis; the R1 (24 DoF) is the balanced core; and the R1 Lite (23 DoF, 1.7 m ceiling, two-finger grippers) is the accessible entry — all sharing 200 TOPS Orin compute.

What is the R1-T?

The R1-T is Galaxea's isomorphic remote-operation platform: full-joint mapping between operator and robot at millimeter precision with millisecond response — serving both teleoperated deployment and demonstration-data collection for imitation learning.

Why does Galaxea use wheels instead of legs?

Deliberate throughput logic: wheeled bases deliver the stability, endurance, and indoor agility that let mobile manipulators outperform bipeds at real tasks — while the torso and arm geometry preserve the floor-to-2 m workspace legs would otherwise provide.

Is the Galaxea R1 good for research?

Exceptionally — with published open-world datasets, ROS2 workflows, VLA and diffusion-policy integration, and documented selection by academic groups for whole-body household-manipulation research requiring bimanual coordination and broad reachability.

What are the Galaxea A1 arms?

The A1 and A1X are lightweight, high-dynamic, force-controlled 6-DoF manipulators with high payload-to-weight ratios — sold standalone for teaching, research, and integration, and forming the R1 series' arm foundation.

How much do Galaxea robots cost?

The R1 series spans an accessible entry tier (R1 Lite) through the professional R1 to the flagship R1 Pro, with A-series arms available for component budgets. Contact a distributor for current pricing, tier guidance, datasheets, and delivery timelines for your region.

Summary

Galaxea built the mobile manipulator the research community was assembling piecemeal: dual force-controlled arms reaching from the floor to two meters, an omnidirectional base, 200 TOPS of onboard compute, isomorphic teleoperation feeding imitation-learning pipelines, and an open stack of datasets, ROS2 workflows, and VLA-ready control — tiered from the classroom-friendly R1 Lite to the sub-millimeter, LiDAR-equipped R1 Pro, on component families (A1 arms, G1 grippers) sold openly themselves. For labs, data teams, and industrial pilots, it is the rare platform where the integration work arrives done. For anyone looking to buy a Galaxea R1 Pro, R1, R1 Lite, or A1 arm, compare their costs against the Astribot S1 and other mobile manipulators, or plan a research, data-collection, or logistics deployment, this guide provides the essential foundation.